fate and tracking of engineered nanomaterials in aqueous environments › bitstream › handle ›...

FATE AND TRACKING OF ENGINEERED

NANOMATERIALS IN AQUEOUS ENVIRONMENTS

by Julie Lynn Bitter

A dissertation submitted to Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy

Department of Chemistry Johns Hopkins University

Baltimore, Maryland March 2014

© 2014 Julie L. Bitter All Rights Reserved

ii

Abstract

Engineered nanomaterials are incorporated into over 1600 commercially available

products on the market today, increasing the likelihood of nanoparticle release during a

nano-containing product’s life cycle. To determine any potential risks associated with

nanoparticle release, it is important to develop a detailed understanding of their behavior

in different aquatic systems. Towards this goal, this thesis focuses on the use of various

microscopic and spectroscopic techniques to study how the physical and chemical

properties of different nanoparticles affect their behavior in bulk aquatic media and near

environmental surfaces under a variety of conditions.

Suspensions of oxidized multiwalled and single walled carbon nanotubes (O-

MWCNTs/O-SWCNTs) were used to examine nanoparticles in bulk aquatic

environments. The effect of ultraviolet (UV) radiation on their colloidal stability was

investigated because UV light is used in drinking and waste water treatment to destroy

harmful pathogens; however, its effect on engineered nanomaterials remains unclear.

Results have shown that absorption of 254nm light causes colloidal O-MWCNTs to

become unstable and aggregate from a loss of surface oxygen by a photodecarboxylation

mechanism. There were observed removal and changes in functional group densities at

the O-MWCNT surface. The same mechanism was exhibited by O-SWCNTs; however,

whereas this transformation proceeds in the absence of any significant mass loss or

changes to the O-MWCNT structure, O-SWCNTs were effectively mineralized by the

UV radiation.

To examine nanoparticles near environmental surfaces, video microscopy was

used to track individual and ensemble averages of silica microspheres and micron-sized

iii

gold rods under various aquatic conditions above silicate surfaces. Using image analysis

algorithms and theoretical calculations, accurate and quantitative measurements of weak

(kT-scale) particle-surface interactions, diffusion behavior, and stability were obtained.

Hydrodynamic interactions provided evidence that silica “gel-layers”, which decrease in

thickness with increasing ionic strength, gave rise to anomalous colloidal stability of

silica microspheres seen over a range of solution conditions. These interactions were also

probed by examining the position dependent translational and rotational diffusion of gold

rods through slit pores and model 2-dimensional porous media. Theoretical calculations

were found to fit experimental rod trajectories well, exhibiting an ionic strength mediated

particle-surface separation dependence of the translational diffusion.

Advisor: Dr. D. Howard Fairbrother

Co-Advisor: Dr. Michael A. Bevan

Reader: Dr. John P. Toscano

iv

Acknowledgements

First I need to thank my advisor, Dr. Howard Fairbrother, for taking me into his

group six years ago. He gave me the opportunity to study chemistry in the environment,

which has been an interest to me since I was a child. Though difficult at times, he

encouraged me to pursue the most quality results possible in each project I undertook. I

thank him for the passion he tackled every obstacle, and for opening doors of

opportunity.

I also need to thank my co-advisor, Dr. Michael Bevan, for taking me under his

tutelage even though I was not an engineer by training. I am grateful for his patience with

my questions and ability to deconstruct difficult concepts. I would not have been able to

achieve what I have without his help.

My journey would not have been as fulfilling without the people who have gone

through it with me, and I need to thank the former and current lab members of both the

Fairbrother group: Dr. Justin Gorham, Dr. Joshua Wnuk, Dr. Billy Smith, Dr. Kevin

Wepasnick, Dr. Samantha Rosenberg, Jin Yang, Michael Barclay, David Goodwin,

Miranda Gallagher, Ronald Lankone, and Julie Spencer; and Bevan group: Dr. Daniel

Beltran, Dr. Tara Edwards, Gregg Duncan, Julia Swavola, Brad Rupp, Yuguang Yang,

Xiaoqing Hua, and Anna Coughlan. Without their assistance and guidance I would not

have been successful.

I owe great thanks to my undergraduate research advisor, Dr. Michael Sigman,

without whose support I never would have pursued graduate school in the first place. His

constant encouragement and support from academia was much appreciated.

v

Lastly, to my parents, Ralph and Annette Bitter; my brother and sister-in-law,

Ralph and Jennifer Bitter; and my boyfriend, Timothy Tivvis. I want to thank them for

their undying love and support, especially in this last year when it has definitely been the

most trying. Be it the late phone calls after a bad day or sharing the happiness associated

with publishing my first paper, they stuck next to me through it all and I am forever

grateful for the hugs and the tears.

vi

Table of Contents

Page

ABSTRACT ...................................................................................................................... ii

ACKNOWLEDGMENTS ................................................................................................ iv

TABLE OF CONTENTS .................................................................................................. vi

LIST OF TABLES ...........................................................................................................xiv

LIST OF FIGURES .........................................................................................................xvi

LIST OF SCHEMES...................................................................................................... xxii

1. INTRODUCTION........................................................................................ ................1

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

1.1.1 Natural and Engineered Nanomaterials ...........................................1

1.1.2 Importance of Nanoparticle Size and Shape ....................................2

1.1.3 Engineered Nanoparticles in Consumer Products ............................4

1.1.4 Release and Transformation of Engineered Nanoparticles in the Environment ..........................................................................................5

1.1.5 Toxicity of Engineered Nanoparticles .............................................7

1.1.6 Methods to Characterize and Study Nanoparticles ........................ 10

1.2 Significance and Objectives .......................................................................... 11

1.3 Summary and Dissertation Outline ............................................................... 12

1.4 References ..................................................................................................... 14

PART I: INTERACTIONS OF COLLOIDAL PARTICLES IN AQUEOUS SUSPENSIONS

2. EXPERIMENTAL SET-UP AND PARAMETERS FOR UV STUDIES ................. 21

2.1 Chemicals and Materials ............................................................................... 21

2.1.1 Chemicals ....................................................................................... 21

2.1.2 Materials ........................................................................................ 22

vii

2.2 Carbon Nanotubes ......................................................................................... 23

2.2.1 Oxidized Multiwalled Carbon Nanotubes (O-MWCNTs) ............. 23

2.2.2 Oxidized Single Walled Carbon Nanotubes (O-SWCNTs) .......... 24

2.3 UV Irradiation Apparatus ............................................................................. 25

2.3.1 Rayonet Photochemical Reaction Chamber ................................... 25

2.3.2 Calibration of the Light Intensity ................................................... 26

2.4 Instrumentation ............................................................................................. 29

2.4.1 UV-Visible Spectroscopy .............................................................. 29

2.4.2 Dynamic Light Scattering (DLS) .................................................. 30

2.4.3 Zeta Potential ................................................................................. 31

2.4.4 X-ray Photoelectron Spectroscopy (XPS) .................................... 32

2.4.5 Chemical Derivatization (CD) ...................................................... 33

2.4.6 Transmission Electron Microscopy (TEM) .................................. 37

2.4.7 Raman Spectroscopy ...................................................................... 37

2.4.8 Total Inorganic Carbon (TIC) ....................................................... 39

2.4.9 Near-Infrared Fluorescence Spectroscopy (NIRF) ....................... 40

2.5 Oxidized CNT (O-CNT) Suspensions .......................................................... 44

2.5.1 Preparation of Stock O-CNT Suspensions ..................................... 44

2.5.2 Preparation of Experimental O-CNT Suspensions ........................ 45

2.6 UV Irradiation of O-CNT Suspensions......................................................... 46

2.6.1 Large Batch Volumes .................................................................... 46

2.6.2 Small Sample Volumes .................................................................. 48

2.7 References ..................................................................................................... 49

3. PHOTOCHEMICAL TRANSFORMATIONS OF OXIDIZED CARBON NANOTUBES AS A RESULT OF EXPOSURE TO UVC IRRADIATION ................. 51

3.1 Introduction ................................................................................................... 52

3.2 Experimental ................................................................................................. 56

3.2.1 O-CNTs .......................................................................................... 56

3.2.2 Chemicals ....................................................................................... 57

3.2.3 Preparation of O-CNT Suspensions ............................................... 58

3.2.4 UVC Irradiation ............................................................................. 59

3.2.4.1 Large Batch Volumes ..................................................... 60

viii

3.2.4.2 Small Sample Volumes ................................................... 61

3.2.5 Calibration of UVC Light Intensity ............................................... 62

3.2.6 Characterization of O-MWCNT Powders ..................................... 63

3.2.6.1 Chemical Characterization .............................................. 63

3.2.6.2 Structural Characterization ............................................. 64

3.3 Results and Discussion of O-MWCNTs ....................................................... 65

3.3.1 Visual Effect of UVC Irradiation ................................................... 65

3.3.2 Effects of Water Quality Parameters ............................................. 67

3.3.3 Chemical Transformations to O-MWCNTs Caused by UVC Irradiation .................................................................................................. 69

3.3.4 Structural Transformations to O-MWCNTs .................................. 78

3.3.5 Phototransformations of Oxidized Carbon Based Nanomaterials ........................................................................................... 82

3.3.6 Environmental Implications ........................................................... 84

3.4 Results and Discussion of O-SWCNTs ........................................................ 85

3.4.1 Visual Effect of UVC Irradiation ................................................... 85

3.4.2 Chemical Transformations to O-MWCNTs Caused by UVC Irradiation .................................................................................................. 87

3.4.3 Structural Transformations to O-SWCNTs ................................... 88

3.5 Conclusions ................................................................................................... 91

3.6 Acknowledgements ....................................................................................... 92

3.7 Supplemental Information ............................................................................ 92

3.7.1 UV-Visible Spectroscopy of O-MWCNT Suspensions ................. 92

3.7.2 Choosing a Suitable Buffer ............................................................ 94

3.7.3 Calibration of Light Intensity via Actinometry ............................. 96

3.7.4 Water Quality Parameters .............................................................. 99

3.7.5 Chemical Characterization of Large Volume Irradiation Experiments .............................................................................................100

3.7.6 Measurement of Total Inorganic Carbon (TIC) and Calculation of CO2 Evolved .......................................................................................101

3.7.7 Oxic versus Anoxic Conditions ....................................................104

3.8 References ....................................................................................................105

ix

PART II: INTERACTIONS OF PARTICLES AND SURFACES IN AQUATIC ENVRIRONMENTS

4. THEORY ...................................................................................................................112

4.1 Solution and Surface Chemistry ..................................................................112

4.1.1 Solution Chemistry .......................................................................112

4.1.2 Surface Chemistry .........................................................................114

4.2 Colloidal and Surface Interactions of Spherical Particles ............................115

4.2.1 Net Potential Energy Interactions .................................................115

4.2.2 Gravitational Body Forces ............................................................116

4.2.3 Electrostatic Repulsion .................................................................116

4.2.4 The Derjaguin Approximation ......................................................118

4.2.5 van der Waals Attraction ..............................................................120

4.2.6 Steric Repulsion ............................................................................123

4.3 Diffusion Modes of Spherical Particles .......................................................124

4.3.1 Diffusion of Spheres near a Flat Surface ......................................124

4.3.2 Diffusion of Spheres through Obstacles .......................................126

4.4 Colloidal and Surface Interactions of Rod-Shaped Particles .......................128

4.4.1 Net Potential Energy Interactions .................................................128

4.4.2 Gravitational Body Forces ............................................................128

4.4.3 Electrostatic Repulsion derived from the Derjaguin Approximation .........................................................................................129

4.4.4 Electrostatic Repulsion from the Linear Superposition Approximation .........................................................................................130

4.5 Diffusion Modes of Rod-Shaped Particles ....................................................131

4.5.1 Diffusion of Rods in the Bulk .........................................................131

4.5.2 Diffusion of Rods near a Flat Surface .............................................133

4.5.3 Diffusion of Rods between Two Parallel Surfaces .......................134

4.5.4 Diffusion of Rods through Obstacles ............................................136

4.6 References ....................................................................................................137

5. EXPERIMENTAL SET-UPS AND PARAMETERS FOR MICROSCOPY STUDIES .........................................................................................................................138

5.1 Chemicals and Materials ..............................................................................138

x

5.1.1 Chemicals ......................................................................................138

5.1.2 Materials .......................................................................................138

5.2 Colloids ........................................................................................................140

5.2.1 Silica Microspheres .......................................................................140

5.2.2 Gold Rods .....................................................................................141

5.2.2.1 Estimation of Surface Potential .....................................142

5.3 Preparation of Samples ................................................................................142

5.3.1 One-wall Cells ..............................................................................142

5.3.2 Confined Cells ..............................................................................143

5.3.3 Porous Media ................................................................................144

5.4 Microscopy Techniques ...............................................................................145

5.4.1 Bright Field ...................................................................................145

5.4.2 Dark Field .....................................................................................147

5.4.3 Total Internal Reflection ...............................................................148

5.5 Image and Data Analysis .............................................................................152

5.6 References ....................................................................................................153

6. ANAMOLOUS SILICA COLLOID STABILITY AND GEL LAYER MEDIATED INTERACTIONS.......................................................................................154

6.1 Introduction ..................................................................................................154

6.2 Theory ..........................................................................................................156

6.2.1 Potential Energy Profiles ..............................................................156

6.2.2 Diffusivity Profiles........................................................................160

6.3 Materials and Methods .................................................................................161

6.3.1 Colloids and Surfaces ...................................................................161

6.3.2 Ensemble Total Internal Reflection Microscopy ..........................161

6.3.3 Diffusivity Landscape Analysis ....................................................162

6.4 Results and Discussion ................................................................................163

6.4.1 Example Deviation from DLVO Theory ......................................163

6.4.2 Interaction Potentials vs. Ionic Strength (at fixed pH = 10) .........164

6.4.3 Hydrodynamic Interactions vs. Ionic Strength (at fixed pH = 10) ..................................................................................................167

xi

6.4.4 Role of “Gel” Layer in van der Waals, Electrostatic, and Steric Potentials ..................................................................................................168

6.4.5 Fitting DLVO and Steric Interactions in the presence of a “Gel” Layer ..............................................................................................171

6.4.6 Do Inferred Gel Layer Properties Make Sense? ...........................173

6.4.7 Potentials and Stability vs. Ionic Strength and pH .......................176

6.5 Conclusions ..................................................................................................180

6.6 Appendix ......................................................................................................181

6.7 Acknowledgements ......................................................................................182

6.8 Supplemental Information ...........................................................................183

6.8.1 Solution Chemistry .........................................................................183

6.8.2 pH and Ionic Strength Dependent Surface Potentials .....................185

6.8.3 Potential Energy Profiles at Various Solution Conditions ..............186

6.8.4 Comparison of hm Derived from Different Gel Layer Scenarios ..................................................................................................188

6.9 References ....................................................................................................189

7. DIFFUSION OF MICRON-SIZED GOLD RODS ACROSS SILICATE SURFACES AND THROUGH SLIT PORES ................................................................193

7.1 Introduction ..................................................................................................193

7.2 Theory ..........................................................................................................198

7.2.1 Potential Energy Profiles ..............................................................198

7.2.2 Bulk Diffusion Modes...................................................................201

7.2.2.1 Bulk Translational Diffusion Coefficients .....................202

7.2.2.2 Bulk Rotational Diffusion Coefficients .........................202

7.2.3 Interfacial Diffusion Modes ..........................................................203

7.2.3.1 Interfacial Translational Diffusion Coefficients ............204

7.2.3.2 Interfacial Rotational Diffusion Coefficients .................204



7.3.2 Ensemble Video Microscopy ........................................................207

7.3.3 Image Analysis..............................................................................207

7.3.4 Measuring System Noise ..............................................................210

xii

7.3.5 Calculations of Position-Dependent Diffusivities ........................210


7.4.1 Measuring the Mean Squared Displacement for Translational and Rotational Diffusion ..........................................................................212

7.4.2 Effects of Ionic Strength on Diffusion Coefficients .....................215

7.4.3 Comparing Diffusion Coefficients from Experiment and Theory ......................................................................................................218

7.5 Conclusions ..................................................................................................220


7.7 Supplemental Information ...........................................................................221

7.7.1 Numerical Calculation of the Electrostatic Repulsion between a Rod in a Parallel Configuration and a Wall ..........................................221

7.7.2 Stokesian Dynamics Simulations ..................................................223

7.7.2.1 Approximate diffusivities for cylindrical rod above wall ...............................................................................................224

7.7.3 Evaluating Stuck Particle Behavior ..............................................225

7.8 References ....................................................................................................226

8. COLLOIDAL ROD DIFFUSION THORUGH MODEL 2-DIMENSIONAL POROUS MEDIA ............................................................................................................231

8.1 Introduction ..................................................................................................231

8.2 Theory ..........................................................................................................236

8.2.1 Diffusion of Spheres through Obstacles .......................................237

8.2.2 Diffusion of Rod-Shaped Particles ...............................................238

8.2.3 Diffusion of Rods through Obstacles ............................................238



8.3.2 Dark Field Microscopy .................................................................241

8.3.3 Image Analysis..............................................................................241


8.4.1 Tracking Rod-Shaped Particles through Porous Media ................243

8.4.2 Deciphering Calculated Mean Squared Displacements ................244

8.4.3 Comparing the Effect of Silica Area Fraction ..............................245

xiii

8.5 Conclusions ..................................................................................................247


8.7 References ....................................................................................................246

CURRICULA VITA ........................................................................................................251

xiv

List of Tables

Page Chapter 3

Table 3.1 – Mass loss experienced by O-MWCNTs from NanoLab Inc. and Cheaptubes as a result of UVC induced aggregation under different solution conditions .......................................................................................................................... 81

Table 3.2 – Mass loss experienced by O-SWCNTs from Southwest Nanotechnologies as a result of UVC induced aggregation at pH 7 ................................. 88

Table S3.1 – Calculation of the quantum flux for 16, 8, 6, 4, 2, and 0 lamps. Flux is determined by actinometric measurements performed with potassium ferrioxalate. The quantum flux listed for 16 lamps appears twice to indicate that which was actually measured during the actinometry experiment, and what the likely flux is based on extrapolation via linear regression of the data from 0 – 8 lamps ....................... 98

Table S3.2 – XPS measurements performed on various O-MWCNTs before and after irradiation with 254nm UVC light for various O-MWCNTs under oxic or anoxic conditions at pH 10. Only the total oxygen percent is shown, the percent carbon is neglected, but the %C + %O = 100%. For example, if the %O = 7.5%, the carbon peak result was 92.5%. The numbers in parentheses show the percentage of carboxylic acid groups that were measured before and after irradiation. The asterisk (*) indicates that the experiment was performed at pH 7 instead of pH 10 .....................101

Chapter 6

Table 6.1 – Constants used in theoretical fits..................................................................160

Table 6.2 – Experimental parameters for each pH and ionic strength condition examined. The column labeled as “-1 (#3)” is the steric decay length from a net potential fit based on Case 3 in Fig. 3, and “-1 (#4)” is the steric decay length from a net potential fit based on Case 4. The columns labeled as hm-U are most probable particle-wall separations obtained from potential energy profile fits, and hm-D is the most probable height from diffusivity profile fits. Dashes indicate cases without a steric contribution, and “x”s indicate irreversibly deposited particles where potential energy and diffusivity profiles could not be measured ....................................................179

Table 6.3 – Constants used in to fit the Hamaker function from Lifshitz theory ............182

Table S6.1 – Constants used in theoretical fits ...............................................................184

xv

Chapter 7

Table 7.1 – Constants used in theoretical fits..................................................................205

Table 7.2 – Measured values used in the fitting of one-wall experimental data .............217

Table 7.3 – Measured values used in the fitting of two-wall experimental data .............217

Table 7.4 – Calculated values of the most probable and average heights from the Derjaguin and linear superposition approximations (Derjaguin/LSA) used to find diffusion coefficients .......................................................................................................219

xvi

List of Figures

Page Chapter 2

Figure 2.1 – Chemical derivatization reactions using fluorinated reagents to tag carbonyl, hydroxyl, and carboxyl functional groups ........................................................ 35

Figure 2.2 – Fluorescence emission spectra for pristine SWCNTs from South West Nano Technologies with excitation wavelength 638nm ................................................... 42



Figure 2.5 – Diagram of different roll up vectors/chiralities (m,n) for SWCNTs describing the different configurations that exist (armchair vs. zigzag). The more blue that is filled in each hexagon equates to a more abundant that species, therefore the (6,5) species is the most abundant and is completely blue ......................................... 44

Chapter 3

Figure 3.1 – Visual effects of UVC irradiation on oxidized multiwalled CNTs from NanoLab, Inc. under anoxic conditions at pH 10 and 3mM Na+. No observable change is apparent over the first 18 hours; afterwards aggregation and settling are observed. Starting O-MWCNT concentration is approximately 5mg/L ........................... 65

Figure 3.2 – Change in absorbance (filled red circles) and particle size (open blue squares) plotted as a function of UVC irradiation time for oxidized multiwalled CNTs under anoxic conditions at pH 7 and 3mM Na+ under radiation with 8 UVC lamps. The shaded region indicates the time where visible aggregation of CNTs was observed ............................................................................................................................ 67

Figure 3.3 – Absorbance profiles for oxidized multiwalled CNTs under anoxic conditions as a function of ionic strength at constant pH 7 (A) and pH at constant ionic strength 3mM Na+ (B) plotted as a function of UV irradiation time ....................... 68

Figure 3.4 – XPS results for oxidized multiwalled CNTs in absence of any irradiation (solid red line) and after UVC-induced aggregation and settling had occurred (dashed blue line). The solution conditions in these experiments were 3mM NaCl and pH 10. Figures 4A and 4B show the C(1s) and O(1s) envelopes before and after irradiation. The distribution of oxygen-containing functional groups (C) determined by chemical derivatization illustrates how UVC irradiation changes the concentration of different oxygen-containing functional groups .................. 69

xvii

Figure 3.5 – Absorbance (A) and particle size (B) profiles for oxidized multiwalled CNTs at pH 7 and 12mM NaCl exposed to different light intensities, plotted as a function of irradiation time. The dashed lines indicate t1/2 (A) where the absorbance reaches half of its initial value, and t630nm (B) where the particle size reached ~630nm. Kinetic data is plotted as a log-log function of t1/2 (C) or t630nm (D) versus light intensity (I) ............................................................................................................... 73

Figure 3.6 – Absorbance (A) and particle size (B) profiles for oxidized multiwalled CNTs at pH 7 and 12mM NaCl. plotted as a function of UVC irradiation time, conducted under anoxic (nitrogen purged) or oxic (oxygen purged) conditions .............. 75

Figure 3.7 – Low (top row) and high (bottom row) resolution TEM micrographs of O-MWCNTs before and after UVC irradiation at pH 7 under anoxic conditions ............ 79

Figure 3.8 – Raman spectroscopy showing the effects of UVC irradiation on O-MWCNTs purchased from NanoLab, Inc. and Cheaptubes. Results are shown for O-MWCNTs before irradiation and after UVC-induced aggregation under oxic and anoxic conditions at pH 10 ............................................................................................... 80

Figure 3.9 – Visual effects of UVC irradiation on single walled CNTs from Southwest Nanotechnologies oxidized with 40% nitric acid. Irradiation was performed under ambient conditions at pH 7. No observable aggregation is observed for the first 10 days, as only the color of the suspension lightens over the course of this time. Starting O-SWCNT concentration is approximately 13.6mg/L ........ 86

Figure 3.10 – Change in absorbance (filled red circles) and particle size (open blue squares) plotted as a function of UVC irradiation time for oxidized single walled CNTs under ambient conditions at pH 7 under radiation with 16 UVC lamps ................ 87

Figure 3.11 – NIRF signal for pristine SWCNTS compared to differently oxidized SWCNT under excitation wavelengths (A) 638nm, (B) 691nm, and (C) 782nm. Suspension concentration was slightly varied as a result of centrifugation to remove bundling ............................................................................................................................ 90

Figure 3.12 – NIRF results for lightly oxidized SWCNT control suspension versus exposures to 8 UVC lamps for 10, 30, and 60min under excitation wavelengths (A) 638nm, (B) 691nm, and (C) 782nm. Suspension concentration was 10mg/L at pH 10....................................................................................................................................... 91

Figure S3.1 – UV-Vis absorbance spectra from 200 – 450nm of oxidized multiwalled CNTs at pH 7 and 12mM NaCl, purged with nitrogen and measured as a function of irradiation time. Absorption maximum (λ = 264nm) corresponds to the π→ π* transition in the conjugated sidewall ring structure. The dashed line indicates the irradiation wavelength of 254nm, and the solid line indicates the wavelength at which measurements were taken (350nm). Inset shows the full spectra ranging from 200 – 900nm ................................................................................... 93

Figure S3.2 – UV-Vis absorbance spectra from 200 – 900nm for the individual constituents that make up an O-MWCNT suspension. The inset shows the region from 200 – 220nm to show the increase displayed in the absorbance profiles of the 3mM phosphate buffered water and the 12mM NaCl solution. These contributions

xviii

can be seen in the profiles of the experimental O-MWCNT suspension from Figure S1 ...................................................................................................................................... 94

Figure S3.3 – Absorbance measurements for O-MWCNTs from NanoLab, Inc. under anoxic conditions using two common buffers to keep the suspension stable at pH 4 ................................................................................................................................... 95

Figure S3.4 – Calibration curves (A) and the calculated quantum flux (B) for various lamp intensities measured with the ferrioxalate actinometry experiments .......... 97

Figure S3.5 – Comparison of absorbance measurements for a control/dark sample and a sample exposed to UVC radiation at pH 7 and 3mM Na+ ...................................... 99

Figure S3.6. – Particle size measurement profiles for oxidized multiwalled CNTs under anoxic conditions as a function of ionic strength (A) and pH (B) plotted as a function of UV irradiation time .......................................................................................100

Figure S3.7 – CO2 measurements from irradiated and dark control samples performed at pH 7 ............................................................................................................103

Figure S3.8 – The change in total dissolved oxygen plotted as a function of irradiation time measured from the same experiment as shown in Figure 6 ...................105

Chapter 5

Figure 5.1 – Schematic representation of dark field set-up. Adapted from Hu et. al.5 ....................................................................................................................................148

Figure 5.2 – Internal reflection of a laser as predicted by Snell’s Law ..........................149

Figure 5.3 – Schematic representation of TIRM set-up, inset shows exponential decay of evanescent wave with a spherical particle scattering light. Adapted from Wu and Bevan.7 ...............................................................................................................151

Chapter 6

Figure 6.1 – Example of disagreement between ensemble TIRM measured particle-wall potential energy profile (points) and DLVO theory (red solid line) for 2μm SiO2 in [NaCl]=20mM at pH=10. Addition of a steric potential to the DLVO potentials produces a net potential prediction (blue dashed line) in better agreement with the depth of the secondary minimum and produces an energy barrier consistent with the particles’ observed stability ...............................................................................164

Figure 6.2 – Ensemble TIRM measurements of (A) potential energy profiles, U(h), and (B) diffusivity profiles, D(h), for 2.1μm SiO2 at pH = 10 with [NaCl] = 0.1 – 100mM. The color scheme for lines and points indicates [NaCl] given in the legend in (A). In (A), the points are measured data from an equilibrium analysis of particle

xix

trajectories using Equation 6.1, solid lines indicate DLVO potentials only (Equation 6.3 with UE+UV), and dashed lines indicate DLVO plus a short range steric contribution (Equation 6.3 with UE+UV+US). In (B), the points are measured data from a non-equilibrium analysis of particle trajectories using Equation 6.13, solid lines are fits to theoretical predictions from Equation 6.14, and error bars are explained in the Methods section .....................................................................................166

Figure 6.3 – Schematics and predicted potentials (for pH = 10, [NaCl] = 80mM in Figure 6.2A) based on various cases for including SiO2 gel layers. In the schematics and predictions, h is the separation between the outer edges of the SiO2 gel layers (i.e. the H2O/SiO2 gel interfaces), and L is the separation between the inner edges of the SiO2 gel layers (i.e. the SiO2 gel/bulk interface). See text for detailed explanation of each case, but in brief: (top-to-bottom) (1) the typical configuration with no gel layers considered in the DLVO theory, (2) gel layers of mostly SiO2 composition, (3) gel layers of mostly H2O composition that are permeable to fluid flow, (4) gel layers of mostly H2O composition that are impermeable to fluid flow. The potentials are color coded as: electrostatics (red), van der Waals (blue), steric (yellow), and net (green) ..................................................................................................170

Figure 6.4 – Steric decay length, γ-1, (left) and gel layer thickness, Δ, (right) vs. [NaCl]/mM at pH=10 from fits in Figure 6.2A based on models for Cases 3 (red triangles) and 4 (blue circles) in Figure 6.3 .....................................................................172

Figure 6.5 – Summary of whether DLVO theory fit measured potentials and the degree of particle stability vs. solution pH and [NaCl]. Points indicate: (1) robust levitation, accurately modeled by DLVO theory (green circles), (2) robust levitation, modeled by DLVO + steric repulsion (green triangles), (3) slow deposition of particles, levitated particles are modeled by DLVO + steric repulsion (yellow inverted triangles), and (4) irreversible deposition (red squares) .......................176

Figure 6.6 – Hamaker functions for two silica half spaces vs. separation and medium ionic strength. Points were computed from the Lifshitz theory in Eq. 6.9 for salt concentrations of 0.01 mM (blue), 0.1 mM (pink), 1 mM (green), 10 mM (red), 100 mM (black), as well as an infinite salt case computed by neglecting the n=0 term in Eq. 6.9 (black triangles). The infinite salt case was fit by Eq. 6.18 (solid black line), which was used in Eq. 6.17 to capture all other salt concentrations (dashed lines) ...................................................................................................................182

Figure S6.1 – Empirical fit to literature data33 and model34 for quartz surface potential vs. pH and ionic strengths of 1mM (green squares), 10mM (red triangles), and 100mM (black circles). The lines are fits to the data given by Eq. (S6.8) ................185

Figures S6.2 – Ensemble TIRM measurements of potential energy profiles, U(h), at pH 7 with same format as pH = 10 data and fits in Fig. 6.2A. The ionic strengths range from [NaCl] = 0.1 – 20mM with exact values reported in Table 6.2 .....................186

Figures S6.3 – Ensemble TIRM measurements of potential energy profiles, U(h), at pH 5.5 with same format as pH = 10 data and fits in Fig. 6.2A. The ionic strengths range from [NaCl] = 0.1 – 20mM with exact values reported in Table 6.2 .....................187

xx

Figures S6.4 – Ensemble TIRM measurements of potential energy profiles, U(h), at pH 4 with same format as pH = 10 data and fits in Fig. 6.2A. The ionic strengths range from [NaCl] = 0.1 – 5mM with exact values reported in Table 6.2.......................187

Figure S6.5 – The most probable separation at the potential energy minimum, and where the sum of the forces equal zero, for the potential energy profiles at pH = 10 in Fig. 6.2A. The x-axis shows estimates of hm from DLVO (closed symbols) and non-DLVO (open symbols) fits of Eq. 6.3 to the U(h) data in Fig. 6.2A for both Cases 3 (red triangles) and 4 (blue circles), and the y-axis shows estimates of hm from fits of Eq. 6.14 to the D(h) data in Fig. 6.2B. The x-axis is labeled as Lm and hm since L is the hydrodynamic separation scale in Case 3 and h is the hydrodynamic separation scale in Case 4. A 1:1 line shows when the two measurements are equivalent ...........................................................................................188

Chapter 7

Figure 7.1 – Cropped (120 pixel x 120 pixel) and scaled 2x: the original experimental image (A), an inverted version of the same image (B), the result after the thresholding algorithm has been performed on the inverted version of the image (C), and the original image now with marked centers and end points (D). Also included are plots of the center of mass position (E) and angular rotation (F) as a function of time to illustrate how the tracking algorithm works .....................................208

Figure 7.2 – Comparison of the schematics used for experiments and calculations, including all pertinent scales and variables .....................................................................211

Figure 7.3 – Mean squared displacement data plotted as a function of time for the translational (A and C) and rotational (B and D) trajectories of varying length colloidal rods in an open system (top) and confined system (bottom). Closed and open symbols represent actual experimental data acquired from the tracking algorithms and solid lines represent the best fit linear regression to the first five data points in each set ..............................................................................................................213

Figure 7.4 – Translational (A and C) and rotational (B and D) diffusion coefficients determined from the fits to mean squared displacement data for each ionic strength condition examined is plotted as a function of rod aspect ratio for colloidal rods in an open system (top) and confined system (bottom). The highest (hmax, dash), lowest (hmin, dot-dot-dash), and best fit (hm, solid) lines are plotted for each data set ....216

Figure 7.5 – Measured diffusion coefficients from one-wall experiments ratioed to the calculated diffusion coefficients from values of hm obtained from Derjaguin and linear superposition approximations at various ionic strength conditions ................219

Figure S7.1 – Histograms of the various length (A) and theta (B and C) values tracked for a rod that was irreversibly bound to the surface. These measurements help to inform on the extent of noise originating from the system as well as the tracking algorithms used ..................................................................................................226 Chapter 8

xxi

Figure 8.1 – Different concentrations of porous media resulting in increasing area fractions of (A) 0.048, (B) 0.085, (C) 0.114, (D) 0.156, (E) 0.210, and (F) 0.245 ..........242

Figure 8.2 – Plotted trajectories of three differently sized gold rods maneuvering through porous media with an area fraction 0.085. Green = 3.9μm, Blue = 3.4μm, and Pink = 4.5μm .............................................................................................................244

Figure 8.3 – Mean squared displacement calculations at (A) 600ms, (B) 60sec, and (C) 600sec for the three gold rods maneuvering through porous media with an area fraction 0.085. Green = 3.9μm, Blue = 3.4μm, and Pink = 4.5μm ..................................245

Figure 8.4 – Mean squared displacement calculations at (A) 600ms, (B) 60sec, and (C) 600sec for the gold rods of similar sizes (4.5 – 4.9μm) maneuvering through porous media with area fractions 0.0 (black filled circles), 0.048 (red open squares), 0.085 (yellow filled triangles), 0.114 (green open upside down triangles), and 0.156 (blue filled diamonds) ......................................................................................................246

xxii

List of Schemes

Page Chapter 3

Scheme 3.1 – Mechanism for the photodecarboxylation of 5H-dibenzo[a,d] cyclohepten-5-carboxylic acid in water proceeding through a carbanion intermediate upon irradiation with 254nm light based on work by McAuley et. al.47 ............................................................................................................................... 76 Scheme 3.2 – Proposed pathway for photodecarboxylation and subsequent aggregation of O-MWCNTs through the removal of carboxylic acid functional groups ................................................................................................................................ 77

1

Chapter 1:

Introduction

1.1 Background

1.1.1 Natural and Engineered Nanomaterials

A colloid is defined as one substance microscopically dispersed throughout

another substance, and includes examples such as blood cells in plasma or soot particles

in air to create smoke. Colloids are made up of particles in sizes of up to 1μm in

diameter, thereby automatically encompassing the separate group of nanoparticles. A

nanoparticle (NP) is defined as a particle having at least one dimension between 1 and

100nm. Natural forms of NPs have existed in the environment for as long as the Earth

has, and are generated by a wide variety of geological and biological processes. Natural

sources of NPs include volcanic dust, natural waters, soils, and sediments.1 These

colloids exist in a variety of forms, most noticeably as inorganic species such as clays,

mica, iron oxides (hematite, magnetite), manganese, and silicates, and additionally as

dissolved organic species like humic and fulvic acids, as well as biopolymer materials

excreted by bacteria.2 Organisms living in an ecosystem have evolved in an environment

that contains any number of these natural NPs, and have adapted to live among them and

clear them from their organ systems. While there is evidence that some natural NPs can

be toxic (e.g., volcanic ash), a new problem is posed by the introduction of unfamiliar

engineered NPs to an ecosystem.1

Engineered NPs are synthesized in a laboratory using a variety of methods and a

myriad of chemical precursors to make particles that contain transition metals that are

2

wholly metallic (gold, Au; silver, Ag; platinum, Pt) or oxidized (titanium, TiO2; iron,

Fe2O3; zinc, ZnO), semiconductors (silicon, SiO2), organic (fullerenes (C60), nanotubes

(CNTs), polymers), or any combinations of these groups. These methods have been

developed and fine-tuned so that NPs can be specifically created to yield a particular size

or shape. Interestingly, the hydrophobicity/hydrophilicity or electronic charge of a NP

can be tailored by adhering layers of additional chemicals. Such a high interest in NPs is

garnered by the vast assortment of particles that can be created by varying the

composition of cores, shells, and layers used during synthesis. However, some

engineered NPs contain chemicals in concentrations toxic to most organisms or exist in

forms that do not occur naturally. They also possess the ability to persist long after

disposal depending on how the surface has been artificially modified.

1.1.2 Importance of Nanoparticle Size and Shape

The size and shape of a NP can greatly influence its properties. This is very

important because depending on the field of study different properties are necessary to

achieve a desired outcome. For example, all types of NPs, in a variety of sizes and

shapes, have been used in polymer composites for energy and optics applications.

Polymer films made into composite materials can be prepared using graphene oxide (GO)

nanosheets,3 AuNPs,4 CNTs,5 fullerenes, and a host of semiconducting and metal oxide

nanoparticles6 for micro and nanoelectronic devices ranging from light emitting diodes to

transistors and photovoltaics. Depending on the particle, sometimes the tensile strength of

polymer nanocomposites can be increased by larger particles better than smaller ones, or

strength through alignment can be provided by elongated rods better than better than

spheres and triangles.7

3

Alignment is very important in optics research when creating devices that use

liquid crystals, where the alignment of metallic or semiconducting NPs is controlled

using electric fields. Acharya et. al. showed that anisotropic particles clearly displayed a

greater advantage over spheres because their inherent dipole allows them to more easily

align with the field, enhancing ordering and performance by tuning the aspect ratio of the

anisotropic particles in the liquid crystals.8 Conversely, a variety of available NP shapes

have been taken advantage of in the field of nonlinear optics where NPs have become

attractive for use as chemical and biological sensors. Depending on the type of sensor,

NPs made of Au, Ag, copper (Cu), metal oxides, and quantum dots can be grown and

modified at their surfaces to tailor the desired properties to a target cell.9 A cornucopia of

nanomaterials can also be found in the field of energy research. Energy storage and

conversion have been the two main goals of energy research, and groups have been able

to use a variety of semiconductors,10, 11 metals,12 and organic NPs13 to create

supercapacitors, dye-sensitized solar cells, electrodes, and filters for remediation.

Two important fields of research where shape has been key are biology/medicine

and electronics. Research has found that certain types of anisotropic particles have given

better results over spheres. For example, Roohani-Esfahani et. al. found that the use of

needle shaped hydroxyapatite nanoparticles improved the strength, interconnectivity, and

porosity of biphasic calcium phosphate scaffolds to most closely simulate natural bone.

The second and third best choices, showing about 2/3 and 1/2 the strength of the needle

composites, were spheres and rod-shaped particles respectively.14 CNTs have been one of

the most commonly studied anisotropic nanoparticles in research for the past 20 years.

Outside of their use to increase the structural integrity of thin films15-17, CNTs have been

4

of special interest to researchers in the electronics field. Single walled CNTs (SWCNTs)

can have metallic or semiconducting properties depending on the orientation of their

sidewall structure. Therefore, electrons and ion species can be carried by these NPs,

which makes them effective field effect transistors, electrical switches, and sensors.18-21

1.1.3 Engineered Nanoparticles in Consumer Products

Currently, nanomaterials are found in more than 1600 products on the market.22

Evaluation of that list back in 2009 by Hansen et. al. showed that the majority of

nanomaterial-containing products are found in the health and fitness industry. Those

products can be broken down into different categories based on how the NP is used

within the product, such as whether the nanomaterial is in the bulk (i.e. nanocrystalline

copper), on the surface (as in films), or as particles (bound, suspended, or airborne). The

majority of these commercial products are using nanomaterials that are found in particle

form, where 19% have NPs bound to the surfaces, 37% use them suspended in liquids,

and 13% are suspended in solids. Products with NPs suspended in liquids or airborne

were further classified as “expected to cause exposure” based on the fact they are in

products that require contact with and application to human skin (e.g., sunscreen or

cosmetics). Surface-bound nanoparticles in consumer products were placed into the lesser

category of “may cause exposure,” due to the fact that though contact would occur,

removing the NPs from the surface would require much more force.23

One of the large unknowns when it comes to NP-containing consumer products is

what happens to the NPs during and after use? Different longevities are expected

depending on the type of consumer products they are used in, the concentration of the NP

within the product and how it is incorporated, as well as the predicted lifetime of the

5

product, frequency of its use, and how well it degrades. Life cycle analysis of three

common nanomaterials (nano-Ag, nano-TiO2, and CNTs) was performed by Mueller and

Nowack24 to estimate the predicted environmental concentrations (PEC) of highly

commercialized NPs that will likely be released from NP-containing consumer products.

By ratioing these PEC values to what are considered predicted no effect concentrations

(PNEC) the authors calculated a risk quotient (RQ) for each NP. Their analysis revealed

that nano-TiO2 showed a very high RQ (between 1 and 16 depending on if the realistic or

high-exposure model was used) because of its prolific use in sunscreen, paints, and

cosmetics. Meanwhile, nano-Ag and CNTs were predicted to have very low risks of

release associated with them, about 100 to 1000 times less than TiO2, most likely due to

the amount of products currently containing these nanoparticles. However, it is believed

that their RQs will increase with cheaper production and increasing incorporation in the

coming years. Though these results were calculated using some estimations and did not

include emission from production sites, it still speaks to the need to be critical of how

products incorporate NPs.

1.1.4 Release and Transformation of Engineered Nanoparticles in the Environment

The prolific use of nanomaterials in consumer products has substantiated the

growing concerns about NP release into the environment. It is believed that the most

common entry points for NPs to our waterways will be from industrial point sources, run-

off from farms and sewers, or the decomposition of NP-containing products in landfills.

There is concern that the natural process of NP removal, which has already been achieved

by organisms in a given ecosystem, will be disrupted by the introduction of engineered

species. This becomes a large problem when there are no natural analogues to the

6

manufactured/engineered NPs that are being released, or if they are being released in

quantities that are beyond the organism’s capability to recover from. For example,

quantum dots contain specific heavy metals (e.g., cadmium) that are not abundant in

nature, or some of the surface coatings created to increase hydrophilicity or

biocompatibility may not have a natural analogue.25 Currently, expectations are that low

NP concentrations (on the ng/L to pg/L scale) are the concentrations we would have to be

concerned about actually reaching an aquatic environment. Also, when mixing small

quantities of engineered NPs with those already existing in an ecosystem, it is believed

that heteroaggregation of these two types of colloids should be the dominant mechanism

of removal from aquatic systems.26

However, once released, NPs will have the potential to undergo transformations

within their environment including interactions with concomitant chemicals,

biotransformations within cells, mechanical force alterations, and photochemical induced

changes. A NP’s behavior can then be influenced by these transformations in some way,

whether it is their ultimate fate, how they are able to transport through the environment,

or how toxic they become. Lowry et. al’s review categorizes transformations into four

common types including chemical, physical, biological, or interactions with

macromolecules.27

Chemical transformations include, but are not limited to, the degradation of

surface coatings, dissolution by oxidation, or the adsorption/desorption of chemical

species to and from the surface. Studies on Ag28, 29, Cu30, and Fe31 NPs have seen these

chemical transformations which in turn affect their aggregation states, toxicity, and even

caused the formation of new particles from dissolved ions. Physical transformations are

7

most commonly associated with homo- and heteroaggregation of nanoparticles, usually

brought on by changes to the surface chemistry. Photolysis has also been shown to

change the surface chemistry of various particles, thereby inducing aggregation32 or

reduction33-35 of particle size.

Transformations within biological organisms are often caused by reduction or

oxidation mediated processes. These processes allow cells to break down biodegradable

polymers and proteins, or induce the dissolution of quantum dots and metal oxide NPs

depending on the compartment and cellular conditions when the NP is engulfed.36 Studies

using fungi show the foreign material, such as fullerols37 and quantum dots,38 are

degraded by these organisms through oxidation processes. When cerium oxide NPs were

introduced to cucumber plants, the NPs were reduced by biogenic reducing substances

and organic acids.39 Upon entering different environments NPs can become coated with

different proteins40, 41 and macromolecules42, 43 that exist within that particular

environment. These coatings change the inherent chemical and physical properties of the

NP, and thereby alter the way the particle transports and interacts with surfaces of

varying compositions.

1.1.5 Toxicity of Engineered Nanoparticles

The transformation processes discussed above can impose drastic changes in NP

toxicity to various cells and organisms. Key issues being investigated include: (i) is the

nature of a given material important, (ii) how does the NP change once it enters the body,

(iii) can the NP penetrate different mucosal barriers, (iv) can the NP be transported within

the body, (v) how does the NP change upon transport from organ to organ, and (vi) what

is the cellular response to these particles? These issues are all interrelated, which

8

complicates effective studies and results. A review of the toxicity literature by Elder et.

al. states that the two most common routes of exposure are estimated to be through the

skin or the respiratory tract. However, it is well documented that upon entering the body

nanomaterials will acquire a protein coating from the adsorption of biomolecules. This

protein corona can greatly influence the fate of a particle that is internalized. However,

this coating is not fixed, but will change as the particle moves to different areas of a cell

or the body, equilibrating with the surrounding media.44 For example, one study on silica

microspheres found that particle size and surface functionalization were key factors in

determining what proteins are adsorbed, though there were many proteins found to bind

to the particle surface in all circumstances.40

Size and composition are important factors when determining a NP’s toxicity.

This is because in ecological systems a variety of trophic levels exist, each corresponding

to a different feeding mechanism of the organisms there. The ability of an organism to

ingest or absorb the particles from their surroundings will be determined by the size and

shape of a given NP. Studies by Jeng et. al. and Karlsson et. al. showed that ZnO and

CuO were found to be much more toxic as a nanoparticle than a microparticle, where the

opposite was seen for TiO2. Meanwhile, iron oxides (Fe2O3, Fe3O4) showed very low

toxicity regardless of size.45, 46 One thing to notice is that metal oxides of both Ti and Fe

are found naturally, which may influence how toxic they are to different organisms.

Metal NPs are a slightly different story from metal oxides. A large concern is if

metallic NPs are able to dissolve into ions, which may turn out to be potentially toxic

themselves. When the toxicity of metallic NPs were examined for three different aquatic

species, Griffitt et. al.47 found that of Ag, Cu, aluminum (Al), nickel (Ni), and cobalt (Co)

9

NPs, Ag and Cu were the most toxic. As a comparison, no death was observed when the

same organisms were exposed to TiO2 NPs. However, review of the literature has shown

that Ag and Fe-Pt NPs are the most cytotoxic because they induce DNA damage in a

variety of bacterial species.48 In plant species (algae, fungi, seed-bearing plants), cell

porosity is disrupted by Ag, allowing more engineered NPs to pass through to the cytosol.

Indirect toxicity occurs when the amount of light that the leaf absorbs, and thus

photosynthesis, is inhibited by NP accumulation on plant surfaces. Toxicity was found to

be enhanced upon exposure to sunlight, potentially by generating reactive oxygen species

(ROS) and causing metal dissolution into ions.49

Johnston et. al.50 published a critical review of the available literature on the

toxicity of different carbonaceous NPs, examining studies conducted both in vivo and in

vitro. The largest percentage of these studies examined pulmonary cells due to the

likelihood that CNT powders, or suspensions aspirated into the air, may be inhaled.

Though the literature covers a broad range of topics, the authors propose that the CNT

toxicity is most likely determined by their length, metal content, tendency to aggregate,

and their surface chemistry. With the large aspect ratios of CNTs, a reasonable theory for

toxicity would be that a macrophage would not be able to engulf the CNT particle,

thereby inducing a persistent inflammatory response including the release of ROS.

Genotoxic and cytotoxic consequences are derived from this oxidative response. Of

course there are discrepancies between studies based on lack of characterization of

materials, differences in experimental set up, and appropriate controls studied, but overall

a decrease in toxicity appeared to correspond with multiwalled CNTs (MWCNTs),

functionalized CNTs, and lower metal impurity content.

10

1.1.6 Methods to Characterize and Study Nanoparticles

Operating on the assumption that exposure and release of nanomaterials into

different natural environments, accidental or otherwise, is a possibility then adequate risk

assessment will need to be designed around particle detection in complex matrices. One

of the biggest problems is that unreasonably high concentrations of NPs are used in

toxicity and release studies. Now using these concentrations can be important to define

limits of exposure or to enable more accurate quantification, but in fact the challenge is to

examine NPs at environmentally relevant concentrations (ng/L – pg/L). Unfortunately,

sufficiently low and environmentally relevant concentrations of NPs cannot adequately

be measured with the analytical techniques used by the majority of researchers due to

their limits of detection. The second challenge is that high backgrounds are often found in

natural environmental samples due to the presence of natural colloidal particles and other

chemicals.

The best strategy for coping with these challenges, according to the review by

Hassellöv et. al., may be to combine existing analytical techniques with new

methodology that simultaneously allows for screening capability and a highly selective

detection.51 Different methods give us complementary pieces of the puzzle: microscopy

enables researchers to image particles with atomic resolution; light scattering can give us

insights into the aggregation state and concentration of colloidal suspensions;

ultrafiltration, chromatography, and mass spectrometry allows for tunable size or charge

selection and separation; and a multitude of spectroscopy techniques, which take

advantage of the electromagnetic spectrum, can be used to evaluate everything from

chemical composition to electronic character. However, with all of this information, very

11

few of these techniques can actually reach down to the ng/L range of detection. With that

in mind, suggestions have been provided to improve environmental sampling for NP

species which include: careful extraction and pre-fractionation, knowing the aggregation

state of the sample, and using a combination of techniques that allows for good

separation with high resolution down to the single particle level (e.g., field flow

fractionation (FFF); inductively coupled plasma with mass spectrometry (ICP-MS) or

atomic emission spectroscopy (ICP-AES); and atomic force, scanning electron, or

transmission electron microscopy (AFM, SEM, TEM).

1.2 Significance and Objective

The probability of engineered nanomaterials entering the environment increases

every year as production of NPs intensifies with every new use for different areas of our

lives. It is for this reason that many researchers are focused on the environmental, health,

and safety concerns of this growing body of nanotechnology. As mentioned above, many

studies lack plausible scenarios to study NPs in the environment, for reasons such as the

concentration or dosage that is used is too high or very clean laboratory samples are used

instead of complex samples. Oftentimes, these are necessary steps to take to gain a

fundamental understanding of how colloids interact with their surroundings. Eventually,

better technology will be utilized to isolate and analyze samples on the nanogram or

picogram scale to give us a much more reliable look at the type of reality we face when

nanoparticles are released.

The objective of this dissertation is to form a bridge between different science and

engineering fields of study to provide the best interpretation of these studies’ results.

12

Merging the fields of analytical, physical, and surface chemistries with environmental

engineering and traditional colloid science, I hope to bring new insights into the nature of

colloidal and surface interactions. Two specific approaches were taken: (i) analytical

approaches are combined with mechanistic insights from the organic photochemistry

community to study the effects of ultraviolet (UV) light on the surface chemistry of

oxidized multiwalled carbon nanotubes (O-MWCNTs); and (ii) state-of-the-art

microscopy and colloidal theory are applied to examine individual anisotropic particles

diffusing across environmentally relevant surfaces. Though there has been significant

literature on both the transformations of carbon nanomaterials and the photochemistry of

small molecules, the two have remained exclusive from one another. Similarly,

traditional colloidal theory has been applied to environmental engineering problems in a

plethora of studies; however, these studies tend to examine the bulk diffusion of high

concentrations of particles and do not attempt to extract detailed information from an

individual particle. The work presented in this thesis takes a comprehensive analytical

approach to the study of colloidal particles in the environment while striving for common

language among different fields of study.

1.3 Summary and Dissertation Outline

This dissertation is organized to frame a discussion regarding various colloidal

particles and their interactions with the surrounding aquatic environment. Part I describes

a more qualitative understanding of colloidal particles in solution. Using O-MWCNT

suspensions, numerous spectroscopic and microscopic techniques are employed to

explain how UV light affects changes to the colloidal stability of these NPs. Chapter 2

13

outlines the experimental details for sample preparation and characterization, as well as

background information and parameters for each analytical technique used. Chapter 3

then discusses the results from studies examining the effects of high energy UV light (λ =

254nm) on the colloidal stability of O-MWCNT suspensions. Changes seen in the surface

chemistry of the CNTs before and after irradiation, as monitored by x-ray photoelectron

spectroscopy, Raman spectroscopy, and transmission electron microscopy, are arguably

similar to the photodecarboxylation process seen so prevalently in the organic

photochemistry literature on small molecules.

Part II takes a more quantitative approach to the analysis of colloidal suspensions

by delving into the colloidal theory of particle-surface interactions. Chapter 4 discusses

in depth the theoretical aspects of (i) the colloid and surface forces that exist between

spherical particles diffusing across planar surfaces, (ii) the colloid and surface forces that

exist between anisotropic particles diffusing across planar surfaces, and (iii) the colloid

and surface forces that dictate spherical particles diffusing across planar surfaces with

spherical asperities. The theories regarding anisotropic particles and diffusion of spheres

around asperities are then used to provide estimates of anisotropic particles navigating

asperity-covered planar surfaces. The complete experimental details for the studies in

Part II can be found in Chapter 5, including procedures for sample preparation,

background information and parameters for the various forms of microscopy used to

examine the colloidal particles, and details regarding data analysis.

The results for the studies investigating colloidal diffusion are found in Chapters

6 – 8. Chapter 6 takes advantage of newer and more sensitive technology to tackle an old

problem. Deviations of experimental results involving silica microspheres over silicate

14

surfaces from theoretical potentials predicted using traditional electrostatic repulsion and

van der Waals attraction (DLVO theory) are investigated. Sensitive measurements of

particle heights using total internal reflection microscopy (TIRM) enabled calculation of

gel-layer mediated interactions that introduced steric repulsion to the system. Two new

models were developed, both fitting experimental data very well, allowing for estimation

of the gel layer thickness at various solution conditions to be calculated and an

accompanying discussion of gel layer properties. Chapters 7 & 8 investigate the diffusion

of anisotropic colloidal particles and how they are influenced by solution chemistry and

surfaces. In Chapter 7, new algorithms were developed to track the diffusion of gold rods

over smooth silicate surfaces. The geometry of the sample cell and increasing ionic

strength conditions were found to profoundly impact the translational diffusion of these

particles, whereas rotational diffusion was almost insensitive to these factors. It was

found that the particle-wall surface separations decreased with increasing ionic strength

and the addition of a second wall. Chapter 8 discusses increasing the complexity of the

system by introducing a confined geometry filled with spherical asperities that the rods

must diffuse around.

1.4 References

1. Handy, R.; Owen, R.; Valsami-Jones, E., The ecotoxicology of nanoparticles and nanomaterials: current status, knowledge gaps, challenges, and future needs. Ecotoxicology 2008, 17, (5), 315-325.

2. Baalousha, M.; Lead, J. R.; von der Kammer, F.; Hofmann, T., Natural Colloids and Nanoparticles in Aquatic and Terrestrial Environments. In Environmental and Human Health Impacts of Nanotechnology, John Wiley & Sons, Ltd: 2009; pp 109-161.

15

3. Eda, G.; Fanchini, G.; Chhowalla, M., Large-area ultrathin films of reduced graphene oxide as a transparent and flexible electronic material. Nature Nanotechnology 2008, 3, (5), 270-274.

4. Park, J. H.; Lim, Y. T.; Park, O. O.; Kim, J. K.; Yu, J.-W.; Kim, Y. C., Polymer/Gold Nanoparticle Nanocomposite Light-Emitting Diodes:â€‰ Enhancement of Electroluminescence Stability and Quantum Efficiency of Blue-Light-Emitting Polymers. Chemistry of Materials 2004, 16, (4), 688-692.

5. Cao, Q.; Rogers, J. A., Ultrathin Films of Single-Walled Carbon Nanotubes for Electronics and Sensors: A Review of Fundamental and Applied Aspects. Advanced Materials 2009, 21, (1), 29-53.

6. Godovsky, D. Y., Device Applications of Polymer-Nanocomposites. In Biopolymers Â· PVA Hydrogels, Anionic Polymerisation Nanocomposites, Springer Berlin Heidelberg: 2000; Vol. 153, pp 163-205.

7. Kutvonen, A.; Rossi, G.; Puisto, S. R.; Rostedt, N. K. J.; Ala-Nissila, T., Influence of nanoparticle size, loading, and shape on the mechanical properties of polymer nanocomposites. The Journal of Chemical Physics 2012, 137, (21), -.

8. Acharya, S.; Kundu, S.; Hill, J. P.; Richards, G. J.; Ariga, K., Nanorod-Driven Orientational Control of Liquid Crystal for Polarization-Tailored Electro-Optic Devices. Advanced Materials 2009, 21, (9), 989-993.

9. Ray, P. C., Size and Shape Dependent Second Order Nonlinear Optical Properties of Nanomaterials and Their Application in Biological and Chemical Sensing. Chemical Reviews 2010, 110, (9), 5332-5365.

10. Selvan, R. K.; Perelshtein, I.; Perkas, N.; Gedanken, A., Synthesis of Hexagonal-Shaped SnO2 Nanocrystals and SnO2@C Nanocomposites for Electrochemical Redox Supercapacitors. The Journal of Physical Chemistry C 2008, 112, (6), 1825-1830.

11. Nair, A. S.; Peining, Z.; Babu, V. J.; Shengyuan, Y.; Ramakrishna, S., Anisotropic TiO2 nanomaterials in dye-sensitized solar cells. Physical Chemistry Chemical Physics 2011, 13, (48), 21248-21261.

12. Meng, H.; Wang, C.; Shen, P. K.; Wu, G., Palladium thorn clusters as catalysts for electrooxidation of formic acid. Energy & Environmental Science 2011, 4, (4), 1522-1526.

13. Thavasi, V.; Singh, G.; Ramakrishna, S., Electrospun nanofibers in energy and environmental applications. Energy & Environmental Science 2008, 1, (2), 205-221.

14. Roohani-Esfahani, S.-I.; Nouri-Khorasani, S.; Lu, Z.; Appleyard, R.; Zreiqat, H., The influence hydroxyapatite nanoparticle shape and size on the properties of

16

biphasic calcium phosphate scaffolds coated with hydroxyapatiteâ€“PCL composites. Biomaterials 2010, 31, (21), 5498-5509.

15. Safadi, B.; Andrews, R.; Grulke, E. A., Multiwalled carbon nanotube polymer composites: Synthesis and characterization of thin films. Journal of Applied Polymer Science 2002, 84, (14), 2660-2669.

16. Gojny, F. H.; Wichmann, M. H. G.; Fiedler, B.; Schulte, K., Influence of different carbon nanotubes on the mechanical properties of epoxy matrix composites â€“ A comparative study. Composites Science and Technology 2005, 65, (15â€“16), 2300-2313.

17. Bal, S.; Samal, S. S., Carbon nanotube reinforced polymer compositesâ€”A state of the art. Bulletin of Materials Science 2007, 30, (4), 379-386.

18. Avouris, P.; Chen, Z.; Perebeinos, V., Carbon-based electronics. Nature Nanotechnology 2007, 2, (10), 605-615.

19. Bondavalli, P.; Legagneux, P.; Pribat, D., Carbon nanotubes based transistors as gas sensors: State of the art and critical review. Sensors and Actuators B: Chemical 2009, 140, (1), 304-318.

20. Javey, A.; Guo, J.; Wang, Q.; Lundstrom, M.; Dai, H. J., Ballistic carbon nanotube field-effect transistors. Nature 2003, 424, (6949), 654-657.

21. Dekker, C., Carbon nanotubes as molecular quantum wires. Physics Today 1999, 52, (5), 22-28.

22. Center, W. W. The Project on Emerging Nanotechnologies: Consumer Products Inventory. http://www.nanotechproject.org/cpi/ (Feb 17),

23. Hansen, S. F.; Baun, A.; Michelson, E. S.; Kamper, A.; Borling, P.; Stuer-Lauridsen, F., Nanomaterials in Consumer Products. In Nanomaterials: Risks and Benefits, Linkov, I.; Steevens, J., Eds. Springer Netherlands: 2009; pp 359-367.

24. Mueller, N. C.; Nowack, B., Exposure Modeling of Engineered Nanoparticles in the Environment. Environmental Science & Technology 2008, 42, (12), 4447-4453.

25. Lowry, G. V.; Casman, E. A., Nanomaterial Transport, Transformation, and Fate in the Environment. In Nanomaterials: Risks and Benefits, Linkov, I.; Steevens, J., Eds. Springer Netherlands: 2009; pp 125-137.

26. Quik, J. T. K.; Stuart, M. C.; Wouterse, M.; Peijnenburg, W.; Hendriks, A. J.; van de Meent, D., Natural colloids are the dominant factor in the sedimentation of nanoparticles. Environmental Toxicology and Chemistry 2012, 31, (5), 1019-1022.

17

27. Lowry, G. V.; Gregory, K. B.; Apte, S. C.; Lead, J. R., Transformations of Nanomaterials in the Environment. Environmental Science & Technology 2012, 46, (13), 6893-6899.

28. Manoharan, V.; Ravindran, A.; Anjali, C. H., Mechanistic Insights into Interaction of Humic Acid with Silver Nanoparticles. Cell Biochemistry and Biophysics 2013, 1-5.

29. Hou, W.-C.; Kong, L.; Wepasnick, K. A.; Zepp, R. G.; Fairbrother, D. H.; Jafvert, C. T., Photochemistry of Aqueous C60 Clusters: Wavelength Dependency and Product Characterization. Environmental Science & Technology 2010, 44, (21), 8121-8127.

30. Mudunkotuwa, I. A.; Pettibone, J. M.; Grassian, V. H., Environmental Implications of Nanoparticle Aging in the Processing and Fate of Copper-Based Nanomaterials. Environmental Science & Technology 2012, 46, (13), 7001-7010.

31. Auffan, M. l.; Achouak, W.; Rose, J. r. m.; Roncato, M.-A.; ChaneÌ ac, C.; Waite, D. T.; Masion, A.; Woicik, J. C.; Wiesner, M. R.; Bottero, J.-Y., Relation between the Redox State of Iron-Based Nanoparticles and Their Cytotoxicity toward Escherichia coli. Environmental Science & Technology 2008, 42, (17), 6730-6735.

32. Hotze, E. M.; Labille, J.; Alvarez, P.; Wiesner, M. R., Mechanisms of Photochemistry and Reactive Oxygen Production by Fullerene Suspensions in Water. Environmental Science & Technology 2008, 42, (11), 4175-4180.

33. Hou, W.-C.; Jafvert, C. T., Photochemical Transformation of Aqueous C60 Clusters in Sunlight. Environmental Science & Technology 2009, 43, (2), 362-367.

34. Gorham, J.; MacCuspie, R.; Klein, K.; Fairbrother, D. H.; Holbrook, R. D., UV-induced photochemical transformations of citrate-capped silver nanoparticle suspensions. Journal of Nanoparticle Research C7 - 1139 2012, 14, (10), 1-16.

35. Lee, J.; Cho, M.; Fortner, J. D.; Hughes, J. B.; Kim, J.-H., Transformation of Aggregated C60 in the Aqueous Phase by UV Irradiation. Environmental Science & Technology 2009, 43, (13), 4878-4883.

36. Zhu, M.; Perrett, S.; Nie, G., Understanding the Particokinetics of Engineered Nanomaterials for Safe and Effective Therapeutic Applications. Small 2013, 9, (9-10), 1619-1634.

37. Schreiner, K. M.; Filley, T. R.; Blanchette, R. A.; Bowen, B. B.; Bolskar, R. D.; Hockaday, W. C.; Masiello, C. A.; Raebiger, J. W., White-Rot Basidiomycete-Mediated Decomposition of C60 Fullerol. Environmental Science & Technology 2009, 43, (9), 3162-3168.

18

38. Metz, K. M.; Mangham, A. N.; Bierman, M. J.; Jin, S.; Hamers, R. J.; Pedersen, J. A., Engineered Nanomaterial Transformation under Oxidative Environmental Conditions: Development of an in vitro Biomimetic Assay. Environmental Science & Technology 2009, 43, (5), 1598-1604.

39. Zhang, P.; Ma, Y.; Zhang, Z.; He, X.; Zhang, J.; Guo, Z.; Tai, R.; Zhao, Y.; Chai, Z., Biotransformation of Ceria Nanoparticles in Cucumber Plants. ACS Nano 2012, 6, (11), 9943-9950.

40. Lundqvist, M.; Stigler, J.; Elia, G.; Lynch, I.; Cedervall, T.; Dawson, K. A., Nanoparticle size and surface properties determine the protein corona with possible implications for biological impacts. Proceedings of the National Academy of Sciences 2008, 105, (38), 14265-14270.

41. Ravindran, A.; Singh, A.; Raichur, A. M.; Chandrasekaran, N.; Mukherjee, A., Studies on interaction of colloidal Ag nanoparticles with Bovine Serum Albumin (BSA). Colloids and Surfaces B: Biointerfaces 2010, 76, (1), 32-37.

42. Wilkinson, K. J.; Negre, J. C.; Buffle, J., Coagulation of colloidal material in surface waters: the role of natural organic matter. Journal of Contaminant Hydrology 1997, 26, (1â€“4), 229-243.

43. Smith, B.; Yang, J.; Bitter, J. L.; Ball, W. P.; Fairbrother, D. H., Influence of Surface Oxygen on the Interactions of Carbon Nanotubes with Natural Organic Matter. Environmental Science & Technology 2012, 46, (23), 12839-12847.

44. Elder, A.; Lynch, I.; Grieger, K.; Chan-Remillard, S.; Gatti, A.; Gnewuch, H.; Kenawy, E.; Korenstein, R.; Kuhlbusch, T.; Linker, F.; Matias, S.; Monteiro-Riviere, N.; Pinto, V. R. S.; Rudnitsky, R.; Savolainen, K.; Shvedova, A., Human Health Risks of Engineered Nanomaterials. In Nanomaterials: Risks and Benefits, Linkov, I.; Steevens, J., Eds. Springer Netherlands: 2009; pp 3-29.

45. Jeng, H. A.; Swanson, J., Toxicity of Metal Oxide Nanoparticles in Mammalian Cells. Journal of Environmental Science and Health, Part A 2006, 41, (12), 2699-2711.

46. Karlsson, H. L.; Gustafsson, J.; Cronholm, P.; Möller, L., Size-dependent toxicity of metal oxide particles: A comparison between nano- and micrometer size. Toxicology Letters 2009, 188, (2), 112-118.

47. Griffitt, R. J.; Luo, J.; Gao, J.; Bonzongo, J.-C.; Barber, D. S., Effects of particle composition and species on toxicity of metallic nanomaterials in aquatic organisms. Environmental Toxicology and Chemistry 2008, 27, (9), 1972-1978.

48. Kim, Y.; Platt, U.; Gu, M.; Iwahashi, H.; Niazi, J., Toxicity of Metallic Nanoparticles in Microorganisms- a Review. In Atmospheric and Biological Environmental Monitoring, Springer Netherlands: 2009; pp 193-206.

19

49. Navarro, E.; Baun, A.; Behra, R.; Hartmann, N.; Filser, J.; Miao, A.-J.; Quigg, A.; Santschi, P.; Sigg, L., Environmental behavior and ecotoxicity of engineered nanoparticles to algae, plants, and fungi. Ecotoxicology 2008, 17, (5), 372-386.

50. Johnston, H. J.; Hutchinson, G. R.; Christensen, F. M.; Peters, S.; Hankin, S.; Aschberger, K.; Stone, V., A critical review of the biological mechanisms underlying the in vivo and in vitro toxicity of carbon nanotubes: The contribution of physico-chemical characteristics. Nanotoxicology 2010, 4, (2), 207-246.

51. Hassellöv, M.; Readman, J.; Ranville, J.; Tiede, K., Nanoparticle analysis and characterization methodologies in environmental risk assessment of engineered nanoparticles. Ecotoxicology 2008, 17, (5), 344-361.

20

PART I: INTERACTIONS OF COLLOIDAL PARTICLES IN AQUEOUS SUSPENSIONS

21

Chapter 2:

Experimental Set-Up and Parameters for UV Studies

2.1 Chemicals and Materials

2.1.1 Chemicals

Sodium salts were all purchased from Fisher Scientific (Pittsburgh, PA, USA) and

used without purification. Sodium dihydrogen phosphate and sodium hydrogen phosphate

were used to prepare buffer solutions that stabilized carbon nanotube (CNT) suspensions

at various pH values during UV irradiation. Phosphate buffers were chosen because the

concentrations of dissolved ion species were not affected by changes in the dissolved

oxygen or dissolved carbon dioxide levels. Sodium chloride and sodium hydroxide were

used to increase the ionic strength and pH conditions of the CNT suspensions for

different UV irradiation experiments. Sodium hydroxide was also used in the cleaning

procedures for both CNTs and experimental glassware. Sodium deoxycholate was

purchased from Sigma-Aldrich (St. Louis, MO, USA) and used without further

purification to aid in the suspension of single-walled carbon nanotubes.

Nitric acid (70% w/w) and hydrochloric acid (9M) were purchased from Sigma-

Aldrich and used without further purification. Nitric acid was used solely for oxidizing

CNTs, whereas hydrochloric acid had a variety of purposes including setting low pH

values of CNT suspensions, expediting the aggregation of CNT suspensions, and

cleaning CNT powders and experimental glassware.

HPLC-grade ultrapure water purchased from VWR International (Philadelphia,

PA, USA) was used as received to prepare CNT stock suspensions and to dilute stocks to

22

experimental concentrations, as well as prepare solutions for actinometry chemical

measurements. This type of water was chosen because it helped reduce variability

between water samples when compared to lower purity samples obtained using

traditional filtered water systems. Ultra-high purity oxygen and nitrogen gas were

purchased from Air Gas (Salem, NH, USA). These gases were used to purge CNT

suspensions to change the levels of dissolved oxygen and dissolved carbon dioxide in an

experimental system. Standards to calibrate the meter reading for pH and conductivity

were purchased from Ricca Chemical Company (Arlington, TX, USA) and Oakton

(Vernon Hills, IL, USA), respectively.

Three fluorinated derivatizing agents were used in conjunction with x-ray

photoelectron spectroscopy to measure the level of oxidation imparted to CNT samples

(see section 2.4.5). 2,2,2-trifluoroacetic anhydride (TFAA), 2,2,2-trifluoroethanol (TFE)

with N,N’-di-tert-butyl-carbodiimide (DTBC) and pyridine, and 2,2,2-

trifluoroethylhydrazine (TFH) were all purchased from Sigma-Aldrich and used without

further purification.

Potassium ferrioxalate (K3[Fe(C2O4)3]), o-phenanthroline, sulfuric acid (18M),

and sodium acetate were the chemicals used for the calibration of light intensity. They

were purchased from Alfa Aesar (Ward Hill, MA, USA) and Fisher Scientific and used

without further purification to perform actinometric chemical measurements on the UV

lamps in the Rayonet photochemical reactor used in these studies. Actinometry

measurements will be discussed in section 2.3.2.

2.1.2 Materials

Pyrex glassware from Corning (Corning, NY, USA) was used to prepare and store

23

CNT stock suspensions. Each vessel underwent a rigorous cleaning procedure involving

sonication for 30min in 4M NaOH, then sonication in 4M HCl for 30min, and lastly

sonication in deionized (DI) water for 30min. Between each sonication the glassware was

rinsed thoroughly with DI water. Special quartz glassware was purchased for UV

irradiation experiments. Two quartz beakers (600mL and 1000mL volumes) and 25

quartz test tubes (15mL volume) were acquired from Technical Glass Products

(Painesville Twp., OH, USA). A rigorous but less harsh cleaning procedure was carried

out on the quartz vessels, using a diluted base and acid without sonication to avoid

scratching or etching the delicate quartz sidewalls.

2.2 Carbon Nanotubes

2.2.1 Oxidized Multiwalled Carbon Nanotubes (O-MWCNTs)

Two different commercially available O-MWCNTs were used. The first

CNT powder, which was used in the majority of the experiments discussed in the

following chapter, was purchased from NanoLab, Inc. (Newton, MA, USA) and had been

oxidized by the manufacturer using a 3:1 sulfuric:nitric mixture (PD15L5-20-COOH,

Outer diameter: 15+/-5nm, Length: 5-20μm, Purity: >95%). The second O-MWCNTs

were purchased from Cheaptubes (Brattleboro, VT, USA) and had also been oxidized

using a 3:1 sulfuric:nitric mixture (MWCNT-OH, Outer diameter: 10-20nm, Length: 10-

30μm, Ash: <1.5%, Purity: >95%). Two similar CNTs were purchased from different

manufacturers principally to compare the behavior of two common commercially

available O-MWCNTs. The O-MWCNTs from NanoLab, Inc. were subject to a rigorous

cleaning procedure outlined in Smith et. al.1 and others2, 3 to remove any amorphous

24

carbon or residual metals leftover from the manufacturer’s oxidation process. These O-

MWCNTs first underwent repeated rinsing/centrifugation cycles (4000rpm, 10min) with

deionized water, then 4M NaOH was used to remove any amorphous or graphitic carbon

byproducts remaining from oxidation that were adsorbed to the CNT surface through pi-

pi interactions. A 4M HCl solution was added next to neutralize the solution and help

dissolve any residual metal nanoparticles, and finally repeated rinses with DI water were

performed to significantly lower the electrolyte concentration in the supernatant. Samples

were considered clean when a low enough electrolyte concentration was reached. This

was measured by testing the resistivity of the supernatant, reaching >0.5MΩ upon

completion. Post cleaning, the O-MWCNTs were pipetted onto a cleaned glass

microscope slide and dried overnight at 70°C. Once dry, the powder was scraped off

using a razorblade and ball-milled for homogeneity. Cleaned and dried O-MWCNTs

were stored in small clean screw-cap vials prior to use. The oxidized CNTs purchased

from Cheaptubes did not undergo the same cleaning procedure, and were used as

received to compare how the cleaning and removal of amorphous carbon impacted the

overall effects of UVC irradiation. Characterization of these powders prior to use in

irradiation experiments was performed using a variety of methods discussed below.

2.2.2 Oxidized Single Walled Carbon Nanotubes (O-SWCNTs)

Commercially available pristine SWCNTs (SG-65, outer diameter: 0.93nm,

specific surface are: 800m2/g, carbon content: >90%) from Southwest Nanotechnologies

(Norman, OK, USA). were acquired from collaborators at Duke University (Durham,

NC, USA) and oxidized in our lab using various weight to weight (w/w) percentages of

nitric acid. The purpose of using different concentrations of nitric acid solution was to

25

increase the total surface concentration of oxygen grafted onto the CNT surface. Previous

studies in our lab have shown that this method can impart increasing levels of oxygen

with increasing nitric acid concentration.4 The analytical balance, a spatula, the sample

vial, and aluminum foil sheets were deionized before weighing out SWCNT powders.

Aluminum foil sheets were used to minimize loss of powders to traditional weigh paper.

100mg of pristine SWCNTs were dispersed by sonication in 250mL of either 10%, 40%,

or 70% w/w nitric acid solution for one hour. The mixture was then refluxed at 140°C for

1.5hrs with stirring. After cooling to room temperature, the acid-CNT mixture was

decanted into small test tubes and then repeatedly centrifuged to collect the newly

oxidized SWCNTs. The same cleaning procedure described above for the NanoLab, Inc.

O-MWCNTs was used to remove excess amorphous carbon and metal nanoparticles from

the O-SWCNTs. Once clean, the O-SWCNTs were dried, ball-milled, and stored before

use.

2.3 UV Irradiation Apparatus

2.3.1 Rayonet Photochemical Reaction Chamber

The quartz vessels described above were used to contain suspensions of oxidized

CNTs during irradiation in the Rayonet photochemical reaction chamber. An RPR-100

photochemical reactor unit equipped with 16, 35W (RPR-2537A) low pressure – low

intensity mercury lamps (90% 254nm emission, intensity = 12.5mW/cm2 two inches from

lamps) was purchased from Southern New England Ultraviolet (Branford, CT, USA).

The inside of the Rayonet chamber contained a smooth mirrored surface to reflect all of

the light output by the lamps. A cooling fan inside the bottom of the chamber was always

26

turned on while the lamps were on, which kept the photochemical reactor at 35°C.

Oftentimes an additional box fan was used externally to help cool the reactor when the

lamps were turned on for long periods of time (in excess of 12 hours). How the

irradiation experiments were performed is discussed in section 2.6.

2.3.2 Calibration of the Light Intensity

Knowing the intensity of light being produced by the experimental system is

necessary if one is ever to compare results between studies. Calibration can occur in one

of two ways: using an electronic photometer or by chemical actinometric measurements.

Chemical actinometry uses a compound that absorbs some quanta of radiation, promoting

it to an excited state. Once excited, the compound decomposes, yielding a colored

product whose concentration can be measured by UV-Visible spectroscopy. With more

intense radiation incident upon the solution, more molecules are excited and subsequently

decompose. Thus, a more intense color resulting from a higher concentration of the final

product is generated with longer exposures. In this case, the intensity of the radiation

corresponds to how many photons are being absorbed by the actinometric compound. For

the purpose of this investigation, actinometric measurements were performed at various

light intensities by changing the number of UV lamps that were turned on inside the

chamber. Actinometry experiments were performed in the dark because these chemical

compounds can decompose under ambient lighting conditions.

For this study, actinometry was performed using the chemical indicator,

potassium ferrioxalate. Irradiation with UV light causes the decomposition of the

ferrioxalate anion to form free Fe2+ ions. The addition of o-phenanthroline, a bidentate

ligand which strongly binds to most metal ions, to this solution causes the coordination of

27

Fe2+ ions following irradiation and forms a red colored complex with an absorption

maximum at λ = 510nm. The absorbance of this complex is measured as a function of the

length of time that the ferrioxalate has been exposed to the radiation. From these

absorbance measurements, it is possible to determine the concentration of Fe2+ ions and

thereby derive the quanta of light absorbed.

As described in the method by Hatchard and Parker,5 147.5mg of potassium

ferrioxalate were dissolved in 50mL of acidic water (45mL water + 5mL 1N H2SO4) to

form a 0.006M ferrioxalate solution. 3mL of this solution were put into the 15mL quartz

test tubes and irradiated for different prescribed periods of time (t). The test tubes were

irradiated while the carousel was turned on and rotating to achieve uniform exposure.

Once removed from the reaction chamber, each test tube was slid into an aluminum foil

sleeve to prevent extraneous light from continuing to decompose the ferrioxalate. After

all of the irradiation points had been completed, 2mL of the irradiated solution were

removed and placed into a clean glass screw cap vial that was also covered by an

aluminum foil sleeve. Here, 2mL of a 0.0055M o-phenanthroline solution and 1mL of a

pH 3.5 acetate solution were added to the vial to complex the Fe2+ ions and quench the

reaction. This 5mL volume was then diluted to a 20mL total volume and stored for at

least 1hr in the dark to ensure that all of the Fe2+ produced by the UVC light had

complexed with the o-phenanthroline. Afterwards, the optical density (D) of the solution

for each time point was measured at 510nm by UV-Vis to determine the concentration of

the colored complex. A plot of D vs. t was constructed and the slope of the linear region

was found by fitting the data points using a simple linear regression. Until the supply of

ferrioxalate has been consumed, D varies linearly with t. In this regime, Eq. 1 can be used

28

to determine the quantum flux (QF);5

3

1 3

2

10DVV NQF

t dV

(2.1)

where V1 is the volume of ferrioxalate solution irradiated (3mL), V3 is the total end

volume of solution (20mL), N is Avogadro’s number, Φ is the quantum yield at the

irradiation wavelength (for 254nm, Φ = 1.25), ε is the molar extinction coefficient of the

ferrioxalate-phenanthroline complex at 510nm (1.11 x 104 L/mol·cm), d is the path length

of the cuvette (1cm), and V2 is the volume of irradiated solution used for complexation

(2mL). It is important to note that once all available ferrioxalate has been consumed, the

plot of D vs. t will plateau. This plateaued region is not useful in the calculation of the

quantum yield, so time points were carefully chosen to ensure that six to eight data points

fell in the linear regime.

Actinometry was performed for all lamp arrangements used for various

experiments of light intensity. These experiments were conducted with different numbers

of symmetrically distributed UVC lamps in the Rayonet reaction chamber. An

experiment was also performed where the ferrioxalate solution was placed inside the

rotating carousel, but the lamps were not turned on. Time points were taken and analyzed

in the same fashion as mentioned above to account for any residual radiation that may be

produced by the lamps despite them being turned off. Results from these experiments can

be found in the following chapter.

29

2.4 Instrumentation

2.4.1 UV-Visible Spectroscopy

UV-Visible spectroscopy measures the electronic transitions of molecules

containing π-electrons or non-bonding (n) electrons (σ-electrons can be measured if the

spectrophotometer used can measure below 200nm). This includes transition metal ions,

highly conjugated organic compounds, and biological macromolecules. When a molecule

absorbs electromagnetic radiation it promotes these electrons to higher anti-bonding

states. The wavelength at which these molecules absorb corresponds to the size of the

HOMO-LUMO gap, which is associated with the types of functional groups present in

the molecule of interest. The resulting absorption spectrum can therefore aid in structure

determination. The Beer-Lambert law is the most common method used for the

measurement of either the concentration (c) or molar absorptivity (ε) of a molecule when

one or the other is known, where the absorbance (A) of a sample is related to the intensity

of light transmitted through the sample (I) by,

0log( / )A I I bc (2.2)

where I0 is the intensity of incoming radiation and b is the path length of the cuvette.6

This law maintains a linear relationship from A = 0 – 1.0 absorbance units.

For the purpose of these studies, the molar absorptivity of O-CNT samples was

estimated by preparing suspensions of varying concentrations and measuring the resulting

absorbance. However, since CNTs are not small, well-defined molecules with uniform

molecular weights, ε did vary with the O-CNT used. For that reason, a selected value of

optical density (A) was used as the main metric to establish O-CNT density rather than

30

attempting to use concentration or molarity. A Varian Cary 50 UV-Visible

spectrophotometer (Agilent Technologies) was used to set the absorbance of all O-CNT

suspensions in a 1cm path length quartz cuvette. Measurements consisted of five scans

taken over the range λ = 200 – 900nm, and the value at 350nm averaged and recorded.

The wavelength λ = 350nm was chosen because it allowed for strong signal without

interference from salts, acids, or bases. Further discussion is included in the supplemental

information section in Chapter 3.

2.4.2 Dynamic Light Scattering (DLS)

DLS is a technique used to determine the size of particles in solution by

measuring the intensity and fluctuations of scattered light. A laser is used to probe a

suspension of particles as they undergo Brownian motion. The intensity fluctuations are

measured as a function of time, which can then reveal information about the size of the

particles (i.e., smaller particles will move more quickly, thereby resulting in a shorter

amount of time between interference of the incoming light). An autocorrelation function

is used to calculate the size of the particles in the suspension. This correlates how quickly

a particle diffuses, which is directly related to a particle’s size by the Stokes-Einstein

relationship,

6

Bk TD

a (2.3)

where D is the particle diffusion coefficient, kB is Boltzmann’s constant, T is the

temperature, η is the solution viscosity, and a is the particle radius.7

O-CNTs are not spherical particles so the Stokes-Einstein equation is not

completely accurate, but it can be used to provide an estimation of their effective

31

hydrodynamic diameter, D(h). Particle size measurements were carried out in disposable

plastic cuvettes at 24°C using a Malvern ZetaSizer Nano-ZS. This instrument uses a He-

Ne laser (633nm) operating at 5mW as the optical probe, and non-invasive backscatter

measurements were taken at 173° from the incident beam. Five separate measurements

were taken per sample, each measurement consisting of 10 – 15 scans, with the average

and standard deviation calculated and recorded.

2.4.3 Zeta Potential

Zeta potential is a quantity used to describe the electric potential at the shear plane

of a particle’s electric double layer (EDL), and is often used as a measure of how stable a

suspension is. Particles with measured zeta potentials of greater than ±40mV are

considered to have good colloidal stability. Zeta potential (ζ) is related to the

experimentally determined electrophoretic mobility (μe) given by Smoluchowski’s model,

0re

(2.4)

where εr is the dielectric constant of medium, ε0 is the permittivity of free space, and η is

the viscosity of the medium.8 Electrophoretic mobility (EM) is an electrophoresis

method, so it measures the migration and velocity of charged particles under an electric

field. This is a dynamic method where the ions in the EDL are constantly moving with

and around the particle as it flows through the medium towards the charged endplate.

Therefore EM can only be an estimation of the electric potential and not an absolute

result.

The same Malvern ZetaSizer Nano-ZS used for DLS is equipped with zeta

potential capabilities. O-CNT suspensions were injected via syringe into a clear

32

disposable folded capillary zeta cell. Laser Doppler Micro-electrophoresis was used to

measure the zeta potential of a given particle suspension. An interferometric technique

measured the velocity of a particle, enabling the calculation of EM and then the zeta

potential using the Smoluchowski model described above. Five separate measurements

were taken per sample, each measurement consisting of 10 – 15 scans, with the average

and standard deviation calculated and recorded.

2.4.4 X-ray Photoelectron Spectroscopy (XPS)

XPS is an analytical surface sensitive technique used to determine a material’s

chemical composition. Under ultra-high vacuum, x-rays are directed at a sample and

penetrate to a depth of about 10nm. These x-rays eject core level electrons from surface

atoms, and these electrons carry kinetic energies corresponding to the energy levels they

were ejected from. A hemispherical analyzer is set to only allow electrons of specific

kinetic energies to pass through, where all others are deflected, resulting in a destructive

collision with the side of the analyzer. The electrons are counted as they pass through to

the detector, and their binding energy is calculated using the equation,

KE BEh (2.5)

where hν is the energy of the x-ray, KE and BE are the kinetic energy and binding energy

respectively, and ϕ is the work function, which is the minimum energy required to

remove an electron from a solid.9 The binding energy of an electron is specific for a

particular element, as well as the energy level from which it was ejected. This allows for

determination and quantification of the elements (except hydrogen) composing the

surface. Quantification is accomplished by taking the area under each curve ratioed by a

sensitivity factor for each element. In addition to composition, XPS can also give

33

information on the oxidation states or chemical environment of an atom (e.g., C-F vs C-

O, NO3 vs NH2) based on detailed analysis of the specific binding energy of ejected

photoelectron spectral envelopes.

This technique was used to analyze O-CNT samples prior to and after irradiation

with UVC light to measure any chemical changes that may have been imparted to the O-

CNT particles by irradiation. XPS analysis of O-CNT samples was performed on a PHI

5400 XPS system using Mg Kα x-ray (1253.6eV) radiation. A small (~5mg) sample of

powdered O-CNTs was pressed onto a 1cm x 1cm piece of double sided copper tape that

was adhered to an XPS sample stub. Enough powdered sample was used to ensure no

copper tape was visible. The high energy electron analyzer was operated at a constant

pass energy to measure the ejected photoelectrons. A survey scan of the sample was

performed using AugerScan software at a pass energy of 178.95eV at a scan rate of

0.25eV/step for binding energies ranging from 1200eV to 0eV. Elemental quantification

was then performed on regions of interest using a pass energy of 58.7eV for general

purposes, and at 5.85eV for more detailed information on the envelope features. All O-

CNT samples were analyzed for total oxygen content in commercial software (CasaXPS)

to establish the carbon:oxygen ratio before and after irradiation with 254nm UVC light.

Oxygen functional group distributions were determined using wet chemical derivatization

techniques previously used in our lab to study CNTs.10, 11

2.4.5 Chemical Derivatization (CD)

Chemical derivatization is one of two common ways to determine the oxidation

state of an atom. Peak fitting of the XPS elemental spectral envelopes is the other method

and is more frequently used. However, the peak fitting method is often fraught with

34

ambiguity when attempting to deconvolute the spectral envelope. Deconvolution involves

assigning a peak position for a particular functional group within the overall peak of the

elemental region. Ambiguity arises due to discrepancies in the literature regarding peak

positions, which may encompass a small range rather than a specific value, and multiple

options for the full-width-half-max (FWHM) values of different chemical species.

Instead of arbitrarily assigning peaks to the spectral envelope, CD uses a chemical

tag for a particular functional group of interest. The method used in this study takes

advantage of fluorine-containing reagents to selectively and quantitatively tag specific

functional groups (hydroxyl, carboxyl, and carbonyl). Each O-CNT sample of interest

was derivatized for all three functional groups that can be quantified. Three small

separate aliquots (~5mg) of each O-CNT sample were added to three individual tiny

Pyrex cups. These cups were then placed inside three separate vacuum flasks, which

contain a cup holder suspended above the bottom of the flask. At the very bottom of each

flask beneath the cup holder, a particular derivatizing reagent was added (2,2,2-

trifluoroacetic anhydride (TFAA) for tagging hydroxyl groups; 2,2,2-trifluoroethanol

(TFE) for tagging carboxyl groups; and 2,2,2-trifluoroethylhydrazine (TFH) for tagging

carbonyl groups). Each flask was then attached to a vacuum line and inserted halfway

into liquid nitrogen so as to cover the bottom of the flask and freeze the liquid reagent.

Once frozen, the flask was opened to the vacuum line and allowed to pump away the

atmosphere inside. Upon closing the flask to the vacuum line, it was allowed to thaw

back to room temperature and the gas-phase reactions were allowed to proceed in the

vacuum flasks overnight. While thawing, the derivatizing reagent would create a gaseous

atmosphere that the functional group of interest would react with, labeling it with a

35

fluorine molecule. Figure 2.1 shows the derivatization schemes for each reaction to tag

the three functional groups of interest.

Figure 2.1 – Chemical derivatization reactions using fluorinated reagents to tag carbonyl, hydroxyl, and carboxyl functional groups.

Hydroxyl:

OHO

+

O

OH

F3C O

O

CF3

O+

O

F3C OH

O

CF3

O

Carbonyl:

+ +CF3HN

H2N

N

H2O

HN

CF3

Carboxyl:

+

CF3HO

C NN

N

OOCF3

+NH

NH

O

NO O

HN

Primary Product

Secondary Product

36

XPS analysis was performed on derivatized O-CNT samples from dark and

exposed UVC samples to monitor changes in the oxygen functional groups densities that

occurred as a result of exposure to UV light. Commercial software (CasaXPS) was used

to quantify the elemental regions of interest by integrating the area under the C(1s),

O(1s), N(1s), and F(1s) peaks.

Using the XPS results for the changes in oxygen-containing functional groups,

one can find an estimate of the amount of a given functional group by converting the

atomic percentages to weight percentages for the carbon and oxygen before and after

irradiation. This is accomplished by multiplying the atomic signal generated by the XPS

by the molecular weight of carbon or oxygen,

% %

% %

16

12wt atom

wt atom

O O

C C

(2.6)

Totaling the carbon and oxygen separately for before and after irradiation, one can then

divide the amount of carbon from the desired functional group to find the weight

percentage of carbon using the equation,

( ) %

( ) %% %

12X atom

X wtwt wt

CC

C O

(2.7)

where, X is the functional group of interest. Multiplying this percentage by the

concentration and volume of CNTs used in the study, one can estimate the mass of

carbon expected to be contained within each functional group before and after irradiation.

This can be useful when attempting to determine the amount of a given functional group

lost or gained during a particular experiment (see Section 2.4.8).

37

2.4.6 Transmission Electron Microscopy (TEM)

Electron microscopy takes advantage of an electron’s very small de Broglie

wavelength, which allows for imaging in much higher resolution than with traditional

optical microscopes. TEM is a technique that uses a focused beam of electrons passing

through a thin sample to image at the nanometer level. The electrons generated from a

high voltage source are focused by a series of special lenses and mirrors and are then sent

through the sample. The electron waves transmitted through the sample can be used to

create an image at very high magnifications, capable of observing individual atoms (and

in the case of larger atoms, resolving them).

TEM was not used to image individual atoms of O-CNTs, as increased

magnification requires increased energy, which using runs the risk of damaging the

sample from the high intensity electron beam. Instead, TEM was used to check for

structural changes on the CNTs imparted by UV light exposure, as high resolution

imaging can provide a view of the number of sidewalls present in a MWCNT, as well as

any amorphous carbon or defects in the sidewalls that may be present. Samples were

prepared for TEM analysis by dipping a holey-carbon TEM grid into suspensions of pre

and post exposure O-CNTs and allowing them to dry in air. The grids were then imaged

using a Philips CM 300 field emission gun TEM operating at 297kV. A CCD camera

mounted on a GIF 200 electron energy loss spectrometer was used for image collection.

2.4.7 Raman Spectroscopy

Raman spectroscopy is a light scattering technique often used to examine

complementary vibrational modes of molecules that are not IR-active. Effects are only

observed for molecules that can be polarized, and the extent of polarizability is directly

38

reflected in how intense the Raman scattering signal is; a stronger signal is generated for

molecules that are more polarizable. A laser is used to excite the molecule to a virtual

energy state, which is a state of higher energy that does not actually correspond to a

discrete energy level. Upon relaxation, the molecule will emit a photon of energy and

return to a different rotational or vibrational state from the one it originally began. This

difference between the starting and ending energy states of the molecule is what is

measured, appearing as a shift in frequency from the original excitation wavelength.

There are two different types of Raman effects: a Stokes shift is when the molecule

returns to a higher energy state than it began, and an anti-Stokes shift is when the final

energy state is lower than the initial state.

Carbon allotropes are often characterized with Raman spectroscopy because it is a

technique that is very sensitive to “highly symmetric covalent bonds with little or no

natural dipole moment.”12 This technique was used in this study to measure the amount

of damage imparted to the physical structure of the O-CNTs as a result of exposure to UV

irradiation in oxic and anoxic aquatic environments. Dried O-CNT samples from before

and after exposure were sent to Purdue University (West Lafayette, IN, USA) for

analysis. The powder was mounted onto a clean glass microscopic slide and a 100x

objective was used to focus on a specific sample area, then that area was analyzed over

10 second exposure times with three spectral averages using an XploRA ONE Raman

system from Horiba Scientific (Edison, NJ, USA). A solid state Nd:YAG laser (50mW)

was used as the excitation source, emitting a wavelength of 532nm. Instrument

parameters include: slit = 200 µm, hold= 300, grating = 1800, filter = 10%. Samples were

then scanned over the range of 500 – 3000 cm-1 to acquire the D and G band intensities of

39

each sample, then allowing for comparison of the ID:IG band ratio as a measure of

structural integrity.

2.4.8 Total Inorganic Carbon (TIC)

Measurements of the total inorganic carbon of an aquatic system are used to

determine the level of dissolved carbon dioxide and carbonate salts. This is often used as

an important metric to measure the health of an aquatic system, where increased levels of

CO2 in the atmosphere allows more CO2 to dissolve in water, forming carbonic acid

species that decrease the pH. To measure TIC a sample is acidified, which forces all

species towards the formation of CO2, and passed through an infrared/mass spectrum

analyzer. For the purpose of this study, TIC measurements were performed to test the

evolution of CO2 believed to arise from O-CNTs through photodecarboxylation as a

function of UV irradiation time. They were compared to dark controls (test tubes

wrapped in aluminum foil to prevent the suspension from being exposed to the UV light)

to eliminate the amount of CO2 already present in the system.

The concentration of CO2 expected to be produced by the irradiation of O-CNTs

is small, even if we assume that all of the carboxylic acid functional groups are removed,

so O-CNTs were prepared as a more concentrated stock for these experiments (25mg/L)

to achieve a measurable amount of CO2. Suspensions were set to pH 7 with the addition

of 3mM phosphate buffer and sent to Purdue University for analysis. Quartz test tubes

were filled with 8mL of O-CNT suspension, leaving 9mL of headspace. A stainless steel

syringe needle was inserted through a rubber septa capped onto each test tube and used to

sparge the CNT suspension with nitrogen gas for 30 minutes at a flow rate of 2.5mL/sec.

Half of the sample tubes were irradiated using a Rayonet RPR-100 photochemical reactor

40

(Southern New England Ultraviolet (SNEU), Branford, CT) with 16 RPR-2537A lamps

(24W). In the reactor, sample tubes were placed within a carousel at the center of the

reactor and rotated at 5 rpm to ensure uniform exposure.

After a specified period of irradiation, test tubes were removed and acidified with

85% phosphoric acid to pH < 3. Test tubes were shaken vigorously for 5min and allowed

to equilibrate overnight, after which 3mL aliquots from the headspace of each tube were

withdrawn for CO2 analysis. Dark controls were acidified and treated the same way for

analysis. The gas phase CO2 concentration was measured with a PDZ-Europa Elemental

Analyzer interfaced to a Sercon 20-20 Isotope Ratio Mass Spectrometer (Crewe,

England). Once this concentration had been determined, the amount of CO2 in the

aqueous phase could be calculated using the equation,

2

2

[ ]

[ ]g

Haq

COk

CO (2.8)

where kH is the Henry’s constant for CO2 dissolved in water, which is dimensionalized by

multiplying the constant by RT.13 The TIC can then be calculated from these two

concentrations by,

312 10g g aq aqTIC C V C V (2.9)

to calculate the total inorganic carbon content in milligrams.

2.4.9 Near-Infrared Fluorescence Spectroscopy (NIRF)

The near-infrared region of the electromagnetic spectrum (800 – 2500nm)

measures vibrational overtones (vibrational modes from the ground state that are excited

to the second or higher vibrational state) and combination bands (when two or more

41

fundamental vibrations (0 → 1) are excited simultaneously). A sample is probed by

irradiating with either a broadband visible-NIR source or specific wavelength laser

source. The resulting fluorescence spectra is often complex and requires deconvolution

for evaluation.

SWCNTs have been found to exhibit distinct fluorescence bands in the NIR

region corresponding to specific chiralities, the intensity of which is directly related to the

concentration of that chirality of CNT as seen in Figures 2.2 – 2.5. However,

fluorescence of SWCNTs can be quickly quenched by bundling or aggregation, so to

achieve individual tubes a surfactant is often added to the suspension and then

ultrasonicated. MWCNTs suffer from quenching by the multiple graphene sidewalls that

make up their structure, so they cannot be successfully quantified using this technique.

For this study, suspensions of pristine SWCNTs, O-SWCNTs at various oxidation

levels, and UV irradiated samples were sent to Duke University (Durham, NC, USA) to

be analyzed for NIRF. A Nanospectralyzer NS1 from Applied Nanofluorescence

(Houston, TX, USA) was used to evaluate suspensions made in a neutral 2% w/v SDC

solution. In some cases, ultracentrifugation was necessary to either concentrate the

volume of sample or remove bundled SWCNTs that would quench the fluorescence

signal. The supernatant was removed after centrifugation for 5.5hr at 60,000rpm and

22°C on a Beckman SW 60 rotor. The pelleted SWCNTs were resuspended by dissolving

a mass of sodium deoxycholate (SDC) into a volume of O-SWCNTs to produce a final

2%w/v suspension of SDC. Samples were then sonicated for 10min using a sonic horn

probe at 50% amplitude while immersed in a salt water ice bath. Three excitation

wavelengths were used: 638 nm at 32 mW (Figure 2.2), 691 nm at 31 mW (Figure 2.3),

an

6

pr

h

th

b

Fiw

nd 782 nm

000 – 11,50

ristine SG6

exagon com

hat the domi

atch of CNT

igure 2.2 – Fluwith excitation w

at 74 mW (

00cm-1.14 Th

5 SWCNTs

mpletely filled

inant SWCN

Ts its name.

uorescence emiwavelength 63

(Figure 2.4)

e diagram in

discussed

d in with blu

NT compone

ission spectra f8nm.

42

and fluores

ndicating th

in Section

ue correspon

nt in this ba

for pristine SW

scence emis

he different r

2.2.2 can b

nds to the ro

atch exhibits

WCNTs from So

ssion scanne

roll-up vecto

be found in

oll-up vector

s that vector.

outh West Nan

ed over the r

ors present i

Figure 2.5.

r (6,5), indic

. It also give

no Technologie

range

in the

. The

cating

es the

es

Fiw

Fiw







43

for pristine SW

for pristine SW

WCNTs from So

WCNTs from So

outh West Nan

outh West Nan

no Technologie

no Technologie

es

es

2

2

6m

so

ce

th

am

tr

Ficom

.5 Oxidize

.5.1 Prepar

mg) of the d

onicated (B

entrifuged a

he Pyrex flas

morphous c

ransferred to

igure 2.5 – Diaonfigurations th

more abundant t

d CNT (O-C

ration of Sto

Stock susp

desired CNT

Branson 151

at 1000rpm f

sk, and any C

carbon, very

o clean 200m

agram of differhat exist (armcthat species, th

CNT) Suspe

ock O-CNT S

pensions of O

Ts to 200mL

10, 70W)

for 5min to

CNTs that w

y lowly ox

mL Pyrex co

rent roll up vecchair vs. zigzagherefore the (6,

44

ensions

Suspensions

O-CNTs wer

of HPLC-g

for ~20hrs

remove any

were not take

xidized CN

ontainers for

ctors/chiralitiesg). The more b5) species is th

re prepared b

grade ultrapu

. After son

y glass that w

en up into su

Ts). This s

r storage. Th

s (m,n) for SWlue that is fille

he most abunda

by adding a

ure water. Sa

nication, su

was etched f

uspension (i.

stock suspe

hese stock s

WCNTs describied in each hexaant and is comp

known mass

amples were

uspensions

from the wa

.e., CNT bun

ension was

suspensions

ing the differenagon equates topletely blue.

s (4 –

e then

were

alls of

ndles,

then

were

nt o a

45

characterized by UV-Vis absorbance measurements, dynamic light scattering (DLS), and

zeta potential for consistency before being diluted for UV irradiation experiments.

2.5.2 Preparation of Experimental O-CNT Suspensions

Stock suspensions of O-CNTs were diluted to the same optical density of A = 0.34

at λ = 350nm (this corresponds to approximately 5mg/L for oxidized NanoLab CNTs, ε ~

0.065) using the ultrapure HPLC water. This ensured uniformity in experimental

suspensions regardless of type of manufacturer. Phosphate buffer was added to achieve a

final concentration of 3mM, and a small aliquot of HCl or NaOH was then added to

adjust the pH to ±0.1 unit with an Accumet Excel XL20 pH and conductivity meter from

Fisher Scientific. Depending on the conditions being examined, additional aliquots of 4M

NaCl were added to the suspension to bring the concentration of Na+ between 4.5 –

12mM. Once the suspension was appropriately set, a final scan consisting of five runs

was taken of the suspension from λ = 200 – 900nm, and then averaged to determine the

starting experimental concentration. 350nm was chosen as the wavelength for analysis

because it had a high absorbance without interference from buffer or salt.

For initial particle size measurements, a 0.5mL sample was removed from the

experimental suspension and diluted to a concentration of 1.25mg/L with ultrapure water

that had been buffered and pH balanced to identical solution conditions. Five separate

DLS measurements were taken per sample, each measurement consisting of 10-15 scans,

with the average and standard deviation calculated and recorded.

Initial zeta potential measurements were carried out on the samples used for DLS

analysis, having been diluted 1:10. Five separate measurements were taken per sample,

each measurement consisting of 10-15 scans, with the average and standard deviation

46

calculated and recorded.

Once initial absorbance and particle size measurements confirmed that a stable

and appropriately concentrated experimental suspension of O-CNTs had been prepared,

samples were transferred to the appropriate quartz vessel and purged with ultra-high

purity oxygen or nitrogen before being sealed first with Parafilm and then with aluminum

foil. The outside surface of the quartz glassware was wiped down with acetone to remove

dust and oils before being put into the Rayonet chamber.

2.6 UV Irradiation of O-CNT Suspensions

For the majority of UV studies, irradiation was performed with all 16 35W

RPR-2537A lamps in the chamber. Other experiments intended to test the effects of light

intensity were performed with 8, 4, or 2 lamps. To accomplish this, some of the lamps

were either symmetrically removed from the chamber or wrapped with aluminum foil to

ensure uniform light distribution around the entire chamber at experimental conditions

that tested lower intensities of UVC light. Experiments were performed by exposing

colloidal suspensions of O-CNTs to the effects of UVC irradiation in one of two

configurations: large batches or small sample volumes.

2.6.1 Large Batch Volumes

The 600mL or 1000mL quartz beaker was used to contain large volumes of O-

CNTs. The purpose of these experiments was to produce enough residual powdered

CNTs after irradiation for surface characterization and mass loss measurements. These

experiments were performed exclusively with 16 lamps and without the addition of extra

47

NaCl. The CNT sample was prepared as described above, and then the beaker was set on

a stand inside the chamber so that the sides of the glassware sat at a height equal to the

midpoint of the lamps so as to achieve the most intense exposure. O-CNT suspensions

were then irradiated in a static configuration until the UVC light was able to induce

aggregation and settling of O-CNT particles. The reactor was then shut down and the

beaker was removed.

At this point a small amount of 9M HCl was added to the beaker and the contents

stirred. This addition of strong acid was to help remove any small particles from

suspension that were too small to visually observe. The aggregated suspension of O-

CNTs was then centrifuged (4000rpm for 10min) to collect O-CNTs for analysis. The

supernatant was decanted and collected in a clean 1L Pyrex bottle after each

centrifugation cycle. Once all of the acidic supernatant had been removed, the O-CNTs

underwent a thorough rinsing cycle where DI water was added, the O-CNTs mixed, then

centrifuged and decanted until the resistivity of the supernatant was >0.5MΩ. The

remaining sample was dried on a cleaned glass microscope slide overnight at 70°C, then

scraped off and weighed on aluminum foil that had been deionized to remove static

charge.

UVC irradiated O-CNT samples were compared to a separate set of O-CNTs from

the same batch. These control studies were prepared in the same manner as those for

irradiation. However, the control suspensions were prepared and then destabilized by

stirring in a sufficient amount of 9M HCl to cause particle destabilization. Control and

UVC irradiated O-CNTs were used to examine any chemical and physical changes that

may have been imparted due to UVC irradiation.

48

2.6.2 Small Sample Volumes

A dozen 15mL quartz test tubes were used for small sample volume experiments.

The purpose of these experiments was to measure the changes in particle concentration

(absorbance) and size as a function of irradiation time and light intensity leading up to

visible aggregation. The light intensity was varied by removing or shielding lamps,

always decreasing the light intensity in half. Measurements were taken at 16, 8, 4, and 2

lamps turned on the chamber. In these experiments the test tubes were placed into a

rotating 12-slot carousel set to a height within the photochemical reaction chamber at the

lamp midpoint to ensure that each test tube received uniform exposure.

A time zero measurement was recorded where the sample had been prepared but

not yet exposed to UV irradiation. Quartz test tubes were then removed at discrete time

intervals over the course of irradiation. These samples were monitored for changes in the

solution chemistry using the pH and conductivity meter at each time point. For specific

experiments testing the effects of dissolved oxygen (DO) concentration, a Thermo

Electron Corporation Orion 5 Star DO meter (Thermo Fisher Scientific) was also used at

this stage. Each time a test tube was removed a replacement (containing already sampled

O-CNTs) was inserted into the vacated spot to maintain uniform light intensity. A 0.5mL

aliquot was taken first from the top of each test tube removed from the carousel (taking

care not to shake or disturb the contents) and placed into a clean 5mL screw cap vial and

set aside for particle size analysis. The rest of the suspension was gently centrifuged

(1000rpm) for 3 minutes to remove any large aggregates that would compromise

absorbance measurements by excessive scattering.

For absorbance measurements, an aliquot of the centrifuged O-MWCNT

49

suspension was pipetted into a 1cm path length quartz cuvette and the UV-Vis spectra

measured from 200 – 900nm. Absorbance readings were taken five times and the average

signal at 350nm was recorded and used as a measure of the particle concentration. These

concentrations were plotted as a function of irradiation time to track the progress of

UVC-induced aggregation. For particle size measurements, the 0.5mL sample placed in

the screw cap vial was diluted to 1.25mg/L with ultrapure buffered water and analyzed.

2.7 References


2. Salzmann, C. G.; Llewellyn, S. A.; Tobias, G.; Ward, M. A. H.; Huh, Y.; Green, M. L. H., The Role of Carboxylated Carbonaceous Fragments in the Functionalization and Spectroscopy of a Single-Walled Carbon-Nanotube Material. Advanced Materials 2007, 19, (6), 883-887.

3. Fogden, S. n.; Verdejo, R.; Cottam, B.; Shaffer, M., Purification of single walled carbon nanotubes: The problem with oxidation debris. Chemical Physics Letters 2008, 460, (1-3), 162-167.

4. Smith, B.; Wepasnick, K.; Schrote, K. E.; Cho, H.-H.; Ball, W. P.; Fairbrother, D. H., Influence of Surface Oxides on the Colloidal Stability of Multi-Walled Carbon Nanotubes: A Structure-Property Relationship. Langmuir 2009, 25, (17), 9767-9776.

5. Hatchard, C. G.; Parker, C. A., A New Sensitive Chemical Actinometer. II. Potassium Ferrioxalate as a Standard Chemical Actinometer. Proc. R. Soc. Lond. A. 1956, 235, (1203), 518-536.

6. Skoog, D. A.; Holler, F. J.; Nieman, T. A., Principles of Instrumental Analysis. 5th ed.; Saunders College Publishing: Philadelphia, 1998.

7. Berne, B. J.; Pecora, R., Dynamic Light Scattering: With Applications to Chemistry, Biology, and Physics. Wiley: New York, 2000.

8. Smulochowski, M., Contribution to the theory of electric endosmosis and some correlative phenomena. In Bulletin International de L'Académie Des Sciences de

50

Cracovie, Classe Des Sciences Mathématiques Et Naturelles, Imprimerie de l'Université: Krakow, 1903; pp 182-199.

9. Attard, G.; Barnes, C., Surfaces. Oxford University Press: New York, 2007.

10. Langley, L. A.; Villanueva, D. E.; Fairbrother, D. H., Quantification of Surface Oxides on Carbonaceous Materials. Chemistry of Materials 2005, 18, (1), 169-178.

11. Wepasnick, K. A.; Smith, B. A.; Schrote, K. E.; Wilson, H. K.; Diegelmann, S. R.; Fairbrother, D. H., Surface and structural characterization of multi-walled carbon nanotubes following different oxidative treatments. Carbon 2011, 49, (1), 24-36.

12. Hodkiewicz, J., Characterizing Carbon Materials with Raman Spectroscopy. In Scientific, T., Ed. Madison, 2010; Vol. 51901.

13. Stumm, W.; Morgan, J. J., Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters. 3rd ed.; Wiley-Interscience: New York, 1996.

14. Schierz, A.; Parks, A. N.; Washburn, K. M.; Chandler, G. T.; Ferguson, P. L., Characterization and Quantitative Analysis of Single-Walled Carbon Nanotubes in the Aquatic Environment Using Near-Infrared Fluorescence Spectroscopy. Environmental Science & Technology 2012, 46, (22), 12262-12271.

51

Chapter 3:

Photochemical Transformations of Oxidized Carbon Nanotubes as a Result of Exposure to UVC Irradiation

Motivated by the use of UVC radiation in drinking and waste water treatment

plants to destroy harmful pathogens, we have investigated the effect of 254nm (UVC)

radiation on the physical and chemical properties of oxidized multiwalled carbon

nanotube (O-MWCNT) suspensions. Absorbance and particle size measurements were

employed to monitor suspension stability, while x-ray photoelectron spectroscopy (XPS),

transmission electron microscopy (TEM), and Raman spectroscopy were used to

characterize any chemical and structural transformations imparted to the O-MWCNTs as

a result of irradiation. After an initial period of irradiation where suspensions appeared to

remain stable, exposure to 254nm light caused O-MWCNT particles to aggregate. The

length of this initially stable period increased as a function of increasing pH and

decreasing ionic strength. Photo-induced aggregation was found to be the result of a loss

of negatively charged carboxylic acid functional groups. Experiments performed in

solutions containing different levels of dissolved oxygen, used to increase the amount of

reactive oxygen species (ROS) generated upon irradiation, suggest that although ROS can

react with the graphenic sidewalls they are not directly involved in the

photodecarboxylation process. However, our results are consistent with a single photon

excitation process, previously identified in organic photochemistry literature as the

mechanism for removal of CO2 from small organics molecules. The pronounced changes

in the surface chemistry of O-MWCNTs that accompany UVC exposure, however,

proceed in the absence of any significant mass loss or changes in the O-MWCNT

52

structure. UVC irradiation is therefore incapable of mineralizing O-MWCNTs, although

it can transform the physical state and surface chemical properties of O-MWCNTs.

Preliminary studies on oxidized single walled carbon nanotubes (O-SWCNTs) showed

different behavior under similar conditions, but can be ascribed to the same mechanism

as the O-MWCNTs.

3.1 Introduction

The rapidly expanding use of nanoparticles (NPs), such as carbon nanotubes

(CNTs), nano-silver, and nano-scale metal oxides in consumer products has prompted

research into the potential health and safety effects of these NPs when they are released

into the environment.1-9 NPs could enter the environment from various point sources

during industrial manufacturing, as a result of spills that occur during transport, or from

the degradation of NP-containing products (e.g. polymer nanocomposites) at various

stages of their life cycle.2 Once NPs enter the environment they can undergo physical

and/or chemical transformations. Possible chemical transformations include degradation

of surface coatings, oxidation/reduction as a result of exposure to concomitant chemicals

such as hydroxyl radicals, ozone, or sulfides, or biotransformations when NPs are

exposed to cellular enzymes and proteins. All of these transformation processes could

modify a NP’s environmental behavior (e.g. transport, aggregation state, toxicity) with

direct relevance to their environmental health and safety effects.10 Consequently, it is

important to understand the nature and kinetics of transformations that occur to NPs.

Photolysis of carbon based nanomaterials, such as fullerenes, graphene oxide

nanosheets, and carbon nanotubes, is one of the most widely studied processes for

53

transformation. Depending on the original state of the carbon surface, the overall effect of

photolysis appears to fall into one of two categories, oxidation or reduction. Previous

studies have shown that when the surface is pristine or unfunctionalized, photolysis often

results in oxidation of the surface through the introduction of oxygen containing

functional groups. For example, in the case of fullerene clusters (nC60) exposed to UV

light in water, oxidation of the surface under either anoxic 11 and oxic12 conditions

reduces aggregate size, which is postulated to be a result of the surfaces becoming more

hydrophilic.11, 13 This oxidative process also occurs in the presence of natural organic

matter, with UV-induced oxidation enhancing the colloidal stability of humic acid coated

nC60.14 Similarly, when unfunctionalized (pristine) CNTs are irradiated with UV light,

photo-induced oxidation has been observed. For example, Savage et.al. studied pristine

MWCNT films in air exposed to both ambient light and 240nm UV irradiation.15 Using

thermoelectric power (TEP) as their gauge, TEP values became more positive the longer

the films were photolyzed in an oxygen-containing environment. The authors argued that

MWCNTs were being oxidized under these conditions, and that oxidation was occurring

preferentially at defects in the CNT sidewalls. Alvarez et. al.16 and Lee et. al.17 exposed

pristine single walled CNTs (SWCNTs) to UV irradiation, and showed with X-ray

photoelectron spectroscopy (XPS) that an increase in the surface oxygen content after UV

exposure. These findings were supported with attenuated total reflection-Fourier

transform infrared spectroscopy (ATR-FTIR) data which suggested that the oxygen

introduced was predominantly in the form of hydroxyl groups.

In contrast, photoreduction occurs when the carbon surface is already oxidized

prior to photolysis. For example, suspensions of graphene oxide (GO) nanosheets

54

exposed to UVA-visible irradiation (310 – 450nm)18-23 have shown a decrease in epoxide

and hydroxyl groups, as measured by XPS and IR. Obvious differences in the physical

structure of irradiated and unexposed GO nanosheets were also observed by AFM and

TEM, with large defect areas present after irradiation. This effect has been attributed to

the photo-induced removal of oxygen-containing carbon, which Stroyuk et. al.

determined that UVA irradiation (λ = 350nm, 3.54eV) was of sufficient energy to remove

C-O from epoxide (2.1eV) and hydroxyl (0.7eV) functionalities to induce the

photoelimination of peripheral carbonyl and hydroxyl groups.23 This is consistent with

data from Smirnov et. al., who used mass spectrometry to detect the evolution of H2O,

CO, O2, and CO2 from the surface of GO films.19 Matsumoto et. al. proposed

mechanisms for the evolution of these gases involving the excitation of the π-π* band to

produce holes and electrons that would facilitate the removal of CO2 and H2O from the

GO nanosheets.18

Similar photochemical transformations have been observed when oxidized CNTs

are photolyzed. For example, Chen and Jafvert examined carboxylated SWCNTs in water

under both sunlight irradiation and 350nm lamp light.24, 25 The oxidized SWCNTs (O-

SWCNTs) aggregated after irradiation at a variety of pH conditions compared to dark

controls, in which O-SWCNTs are normally known to be stable. They also monitored the

ROS production of singlet oxygen, superoxide radicals, and hydroxyl radicals from as-

received carboxylated SWCNTs and compared those results to heat treated/acid washed

SWCNTs. The singlet oxygen (1O2) and hydroxyl radical (•OH) production decreased for

the treated CNTs compared to the as-received. The authors suggested that this could be a

function of: (1) the decrease in amorphous carbon content, (2) the decrease in metal

55

content, (3) a decrease in –COOH functionalization, and (4) the change in the

aggregation state, which has been shown to occur for the thermal/acid treated CNTs. In a

related study, Hwang et. al. used 300-400nm (UVA) irradiation to study carboxylated

MWCNT suspensions in water with mono and divalent salts.26 A loss of surface oxygen

content was observed by XPS when oxidized MWCNTs were irradiated with UVA light.

This loss in oxygen resulted in destabilization and aggregation of MWCNTs as measured

by time resolved dynamic light scattering (TR-DLS) and quartz crystal microbalance

(QCM-D), respectively. A related study by the same researchers, also conducted with

UVA light saw a loss of oxygen in the form of carboxylic acid functional groups and an

increase in the structural disorder, monitored by XPS and Raman, respectively. The

addition of hydrogen peroxide was observed to enhance the rate of CNT aggregation, an

effect that was attributed to reactions of reactive oxygen species (ROS) with carboxylic

acid groups on the CNT surfaces.27

To date, studies on the photolysis of CNTs have focused on the effect of visible

light and lower energy UVA light. In contrast, comparatively little is known about the

effects of the more aggressive UVC irradiation, used in drinking and waste water

treatment plants, on its ability to degrade nanoparticles. With its potential to directly

excite electronic transitions and thereby open up new transformation pathways, the

known UVC-induced transformations of nanomaterials are limited. This has motivated us

to study the effects of UVC light on oxidized multi-walled carbon nanotubes (O-

MWCNTs) under various solution conditions (e.g. pH, ionic strength, dissolved oxygen)

as a function of irradiation time and light intensity. The concentration and size of

suspended particles were measured with UV-Vis and DLS respectively as a function of

56

irradiation time, while chemical and physical transformations were evaluated using

complementary analytical techniques including mass loss, TEM, Raman, and XPS with

chemical derivatization. Questions we sought to address included, (1) what are the

fundamental physical and chemical transformations of oxidized MWCNTs when they are

exposed to UVC irradiation, (2) how do water quality parameters affect the rate of

transformation, (3) can UVC degrade or mineralize oxidized MWCNTs, and (4) can we

provide mechanistic insights into the fundamental transformation processes.

3.2 Experimental

3.2.1 O-CNTs

Commercially available O-MWCNTs and O-SWCNTs were used for this study.

MWCNTs were purchased from NanoLab, Inc. (Newton, MA) that had been oxidized by

the manufacturer using a 3:1 sulfuric:nitric mixture (PD15L5-20-COOH, outer diameter:

15+/-5nm, Length: 5-20μm, Purity: >95%). Oxidized MWCNTs (MWCNT-OH, outer

diameter: 10-20nm, Length: 10-30μm, Ash: <1.5%, Purity: >95%) were also purchased

from Cheaptubes (Brattleboro, VT) and used as received, principally as a point of

comparison to the behavior of the NanoLab O-MWCNTs.

Pristine SWCNTs (SG-65, outer diameter: 0.93nm, specific surface are: 800m2/g,

carbon content: >90%) from Southwest Nanotechnologies (Norman, OK, USA) were

oxidized in our own laboratory using 10% w/w, 40% w/w, and 70% w/w nitric acid to

impart a systematically increasing amount of oxygen to the CNT surface used as a

mechanistic comparison to the two multiwall CNTs. In brief, the mixture was refluxed in

57

nitric acid at 140°C for 1.5hrs with stirring. After cooling to room temperature, the acid-

CNT mixture was decanted into small test tubes and then repeatedly centrifuged to

collect the newly oxidized SWCNTs (O-SWCNTs).

O-CNTs were subject to a rigorous cleaning procedure outlined in Smith et. al.28-

30 Briefly, the O-CNTs first underwent repeated rinsing/centrifugation cycles with

deionized water to remove as much of the nitric acid as possible. Then 4M NaOH was

used to remove amorphous carbon and excess carbon products remaining from oxidation.

4M HCl was added next to neutralize the solution, and finally repeated rinses with DI

water were performed to remove the electrolytes from the supernatant. The sample was

judged free of electrolyte when the resistivity of the supernatant reached >0.5MΩ. Post

cleaning, the CNTs were dried on a microscope slide, scraped off, and ball-milled for

homogeneity. Cleaned and dried CNTs were stored in small clean screw cap vials prior to

preparing colloidal suspensions and subsequent photolysis experiments. This cleaning

procedure was used for the oxidized CNTs purchased from NanoLab, Inc. and Southwest

Nanotechnologies, but not for the CNTs purchased from Cheaptubes. The purpose of

cleaning one MWCNT and not the other was to compare the effects, if any, of residual

metal and amorphous carbon on the CNT colloidal stability during UV irradiation.

3.2.2 Chemicals

Sodium chloride, sodium hydroxide, sodium dihydrogen phosphate, and sodium

hydrogen phosphate were purchased from Fisher Scientific and used without purification.

Nitric acid, hydrochloric acid, and derivatizing agents 2,2,2-trifluoroacetic anhydride

(TFAA), 2,2,2-trifluoroethanol (TFE), N,N’-di-tert-butyl-carbodiimide (DTBC), and

2,2,2-trifluoroethylhydrazine (TFH) were all purchased from Sigma-Aldrich and used

58

without further purification. HPLC-grade ultrapure water was purchased from VWR

International and used as received. Ultra-high purity oxygen and nitrogen were purchased

from Air Gas (Malvern, PA).

3.2.3 Preparation of O-CNTs Suspensions

Stock suspensions of O-CNTs were prepared by adding a known mass (4 – 6mg)

of the desired CNTs to 200mL of HPLC-grade ultrapure water. Samples were then

sonicated for 20hrs (Branson 1510, 70W). After sonication, suspensions were centrifuged

at 1000rpms for 5min to remove any glass that was etched from the walls of the flask,

and any CNTs that were not taken up into suspension (i.e., CNT bundles, amorphous

carbon, very lowly oxidized CNTs). This stock suspension was then transferred to clean

200mL Pyrex containers for storage.

A Varian Cary 50 UV-Vis was used to determine the absorbance of stock O-CNT

suspensions, and to set the absorbance of the final experimental suspensions to 0.34 at

350nm (this corresponds to approximately 5mg/L for oxidized NanoLab MWCNTs). A

typical UV-Vis spectra of the O-MWCNT suspension is shown in Figure S3.1. The

desired ionic strength was obtained by adding appropriate volumes of a 4M NaCl stock

solution. Suspensions of O-CNTs were set to the desired pH (±0.1) by adding phosphate

buffer (3mM final concentration) and HCl or NaOH as needed. The triprotic phosphate

buffer was chosen because it did not impact the absorption profile in the region being

examined, allowed stability over a range of pH values to be studied (4 – 10), and was not

negatively impacted by the removal of carbon dioxide, unlike carbonate buffers. Separate

control studies conducted with phosphate and acetate buffers revealed that the nature of

the phosphate buffer also did not influence the effect or kinetics of UVC exposure

59

(Figure S2). In the absence of a buffer, the pH of the suspension would undergo a drastic

decrease during irradiation, artificially destabilizing the O-CNTs before the UV radiation

had a chance to induce aggregation. The use of a buffer was therefore necessary to

stabilize the pH throughout an experiment, enabling us to isolate the effect of different

variables (pH, ionic strength, light intensity) on particle stability.

For initial particle size measurements prior to UVC photolysis, a 0.5mL aliquot

was removed and diluted to 1.25mg/L with buffered and pH balanced ultrapure water.

Particle size measurements were carried out on this sample in disposable plastic cuvettes

at 24°C with a Malvern ZetaSizer Nano-ZS using a 5mW He-Ne laser (633nm) to probe

CNTs. A non-invasive backscattering configuration (detection angle 173° with respect to

the incident laser light) was used to collect measurements. Effective hydrodynamic

diameter results were modeled by the Stokes-Einstein relationship and represent the

average of five separate measurements, each measurement consisting of 10-15 scans.

Once initial absorbance and particle size measurements confirmed that a stable

and appropriately concentrated experimental O-CNT suspension had been prepared,

samples were transferred to a quartz vessel. These vessels were then purged with ultra-

high purity oxygen or nitrogen before the vessel was sealed first with Parafilm and then

with aluminum foil. The outside surface of the quartz glassware was wiped down with

acetone before being exposed to UVC irradiation.

3.2.4 UVC Irradiation

Irradiation was performed on colloidal suspensions of O-CNTs in an RPR-100

Rayonet UV photochemical reaction chamber equipped with 16, 35W low pressure

mercury lamps (90% 254nm emission, irradiance with all 16 lamps = 15.4 – 16.0

60

mW/cm2), purchased from Southern New England Ultraviolet (Branford, CT). Most

experiments were performed with 16 lamps (photon flux ~ 1.32 x 1017 quanta/sec as

measured by actinometry). For experiments performed with less than 16 lamps, some

lamps were symmetrically wrapped with aluminum foil to ensure uniform light

distribution around the entire chamber. Photon flux measurements for experiments

performed with less than 16 lamps can be found in Figure S3 and Table S1.

Experiments were performed by exposing colloidal suspensions of O-MWCNTs

to UVC irradiation in one of two configurations:

3.2.4.1 Large Batch Volumes

A 600mL or 1000mL quartz beaker was used to contain large volumes of

O-MWCNTs during irradiation to produce enough sample for surface

characterization and mass loss measurements following UVC irradiation. In these

experiments a beaker was set on a stand inside the chamber so that the sides of the

glassware sat at a height equal to the midpoint of the lamps for maximum

exposure. O-MWCNT suspensions were irradiated in this static configuration

until UVC-induced aggregation and settling had occurred. At this point a small

amount of concentrated HCl was added to the beaker to aid in the removal of any

remaining suspended CNTs. The contents of the beaker were then centrifuged

(4000rpms for 10min), and the supernatant was decanted and collected in a clean

1L Pyrex bottle. The O-MWCNTs collected in the test tubes were then rinsed

thoroughly with DI water until the resistivity of the supernatant was >0.5MΩ and

dried on a cleaned glass microscope slide overnight at 70°C. The mass of O-

MWCNTs remaining after irradiation was determined by scraping the dried O-

61

MWCNTs off the glass and weighing them on aluminum foil that had been

deionized to remove static charge. Mass loss was determined by comparing the

values obtained from UVC irradiated O-MWCNT samples to control studies,

where suspensions of O-MWCNTs had been prepared and then crashed out of

solution by adding sufficient acid to cause particle destabilization without any

UVC exposure. O-MWCNTs generated from control and irradiation studies were

also used for surface characterization. O-SWCNTs were also irradiated in this

manner.

3.2.4.2 Small Sample Volumes

A dozen 15mL quartz test tubes were used to measure changes in particle

concentration (absorbance) and size as a function of irradiation time. In these

experiments the test tubes were placed into a rotating 12-slot carousel set to a

height at the midpoint of the UVC lamps to ensure that each test tube received

uniform exposure. Quartz test tubes were removed at discrete time intervals

during UVC exposure for sampling, and monitored for changes in the solution

chemistry using a pH meter and conductivity probe at each time point. Each time

a test tube was removed a replacement (containing an already sampled suspension

of O-MWCNTs) was inserted into its place to maintain uniform light intensity

within the carousel. Variations in solution conditions and light intensity were

studied with this method, so the absolute light intensity was calibrated for these

experiments (see below).

For each O-MWCNT suspension removed from the carousel, a 0.5mL

aliquot was extracted from the top of the test tube (taking care not to shake or

62

disturb the contents) and placed into a clean 5mL screw cap vial and set aside for

particle size analysis. For particle size measurements, the 0.5mL sample placed in

the screw cap vial was diluted to 1.25mg/L with ultrapure buffered water and

analyzed using a Malvern ZetaSizer Nano-ZS The suspension remaining in the

quartz test tube was gently centrifuged (1000rpms) for 3 minutes to remove any

large aggregates that would compromise absorbance measurements by excessive

scattering. For absorbance measurements, an aliquot of the centrifuged O-

MWCNT suspension was pipetted into a 1cm path length quartz cuvette and the

UV-Vis spectra measured from 200 – 900nm in a Varian Cary 50 UV-Vis

spectrophotometer. Five separate absorbance readings were taken and the average

signal at 350nm was used as a measure of the O-MWCNT particle concentration.

These concentrations were plotted as a function of irradiation time to track the

progress of UVC-induced aggregation.

3.2.5 Calibration of UVC Light Intensity

To determine the intensity of UVC light experienced by the O-MWCNTs during

irradiation, actinometry was performed using potassium ferrioxalate as a colorimetric

indicator. UV light causes the decomposition of ferrioxalate to Fe2+ ions. The addition of

o-phenanthroline to this solution causes the Fe2+ ions to form a red colored complex with

an absorption maximum at 510nm. By measuring the absorbance of this complex it is

possible to determine the concentration of Fe2+ ions and thereby derive the quanta of light

absorbed for all light intensities studied.31 More comprehensive details regarding the

calibration of the photochemical reactor can be found in the Supporting Information.

63

3.2.6 Characterization of O-MWCNT Powders

O-CNTs were analyzed before and after UVC induced aggregation to determine

the effect of irradiation on both the chemical and structural characteristics of the NPs.

3.2.6.1 Chemical characterization

To determine surface composition, a small sample of powdered O-CNTs

was pressed onto a 1cm x 1cm piece of double sided copper tape that was adhered

to an XPS sample stub, ensuring no copper was visible. A PHI 5400 XPS system

using Mg Kα x-ray (1253.6eV) irradiation was used to analyze O-MWCNT

samples. A high energy electron analyzer operating at a constant pass energy was

used to measure the ejected photoelectrons. Elemental quantification was

performed using a pass energy of 178.95eV at a scan rate of 0.25eV/step. O-

MWCNT samples were analyzed for total oxygen content as well as oxygen

functional group distribution using wet chemical derivatization techniques

previously used to study CNTs in our lab.32, 33 In brief this method uses fluorine-

containing reagents (2,2,2-trifluoroacetic anhydride (TFAA), 2,2,2-

trifluoroethanol (TFE), and 2,2,2-trifluoroethylhydrazine (TFH)) to selectively

and quantitatively tag specific functional groups. Commercial software

(CasaXPS) was used to quantify elemental regions by integrating the area under

the C(1s), O(1s), N(1s), and F(1s) peaks. O-SWCNTs did not produce enough

residual sample to perform chemical derivatization and total oxygen analysis.

64

3.2.6.2 Structural characterization

(i) TEM: O-MWCNTs were prepared for TEM analysis by dipping a

holey-carbon TEM grid into O-MWCNT-containing suspensions and allowing

them to dry in air. The grids were then imaged using a Philips CM 300 field

emission gun TEM operating at 297kV. A CCD camera mounted on a GIF 200

electron energy loss spectrometer was used for image collection.

(ii) Raman: Dried O-MWCNT samples were mounted onto a microscopic

slide and analyzed by Raman spectroscopy over 10 second exposure times with 3

spectral averages using an XploRA ONE Raman system from Horiba Scientific

(Edison, NJ). A solid state Nd:YAG laser (50mW) emitting an excitation

wavelength of 532 nm was used with 100x objective. Samples were scanned over

the range of 500 – 3000 cm-1. Instrument parameters include: slit = 200 µm, hold=

300, grating = 1800, filter = 10%.

(iii) Near Infrared Fluorescence (NIRF): Dried powders or suspensions of

O-SWCNTs were analyzed by NIRF using a Nanospectralyzer NS1 from Applied

Nanofluorescence (Houston, TX, USA). Suspensions were created in a 2% w/v

sodium deoxycholate (SDC) solution and analyzed over the range 6000 –

11,500cm-1 using excitation wavelengths of 638nm, 691nm, and 782nm.

3

3

ca

F

co

F

th

ey

th

d

an

su

ch

Fanaf5m

.3 Results

.3.1 Visual

arbon nanot

igure 3.1. F

olor or appe

igure 3.1) t

hen eventual

ye. The qual

he experime

epended upo

nd the light

uspensions i

hange in abs

0 hrs

igure 3.1 – Vinoxic conditiofterwards aggrmg/L. Photos b

and Discuss

l Effect of UV

An examp

tubes (O-MW

For an initia

arance. How

the O-MWC

lly larger, se

litative effec

nts reported

on the wate

intensity. Se

in quartz ve

sorbance or p

16

isual effects ofons at pH 10 aregation and sby Miranda Ga

sion of O-M

VC Irradiati

ple of the vi

WCNTs) as

al period of

wever, beyon

CNTs begin

ettleable agg

ct of UVC irr

d in this stud

er quality pa

eparate irrad

essels wrapp

particle size,

6 hrs

f UVC irradiatand 3mM Na+

settling are oballagher.

65

MWCNTs

ion

sual change

s a function

irradiation

nd a certain i

to aggregat

gregates that

radiation sho

dy, although

arameters, c

diation studie

ped with alu

, revealing th

28 hrs

tion on oxidize+. No observabbserved. Startin

s that occur

n of UVC i

the suspens

irradiation ti

te, first form

leave the su

own in Figur

h the time sc

characteristic

es conducted

uminum foil

hat the trans

3

ed multiwalledble change is ng O-MWCN

rred to oxidi

irradiation ti

ion displays

ime (greater

ming visible

upernatant c

re 3.1 was o

cale for aggr

cs of the O-

d in the Ray

l (dark cont

sformations o

8 hrs

d CNTs from Napparent over

NT concentratio

ized multi-w

ime is show

s little chan

r than 16 hou

particulates

clear to the n

observed for

regation to o

-MWCNTs

yonet on coll

trols) showe

observed we

54 hr

NanoLab, Inc. r the first 18 on is approxim

walled

wn in

nge in

urs in

s and

naked

all of

occur

used,

loidal

ed no

ere

rs

underhours;mately

66

strictly a result of UVC exposure (see Figure S3.5). The CNT aggregation and settling

observed in this study are qualitatively consistent with previous findings that studied

carboxylated single walled CNTs,24 as well as the decreased stability seen when

carboxylated multiwalled CNTs26 were exposed to UVA irradiation.

Phototransformations induced by UVC irradiation were examined more

quantitatively by acquiring absorbance and particle size measurements as a function of

irradiation times using UV-Vis and DLS, respectively. The UV-Vis absorbance profile

for suspensions of O-MWCNTs (Figure S3.1) are similar to previous studies26 with a

broad peak at λ = 264nm, corresponding to the π→π* transition in the conjugated

sidewall ring structure.34, 35 Although the intensity of the profile decreases with increasing

wavelength, scattering from O-MWCNT particles prevents the baseline from reaching an

absorbance value equal to zero. The sharp increase seen at wavelengths approaching λ =

200nm is due to the absorption of the phosphate buffer and NaCl added to each O-

MWCNT suspension (see Figure S3.2).36 As the UVC irradiation time increases, Figure

S3.1 shows that the intensity of the profile decreases without any significant changes in

shape. An example of the data acquired on the particle concentration and particle size

measured as a function of UVC irradiation time is shown in Figure 3.2. During the initial

stages of irradiation (up to approximately 18 hours in this case) absorbance

measurements illustrate that the concentration of O-MWCNTs in suspension remains

relatively unchanged, and the average particle size only increases slightly. However, for

irradiation times in excess of 24 hours a rapid increase in particle size (> 400nm) is

observed and the colloidal O-MWCNT concentration drops sharply. It is during this time

67

(shaded region of Figure 3.2) that small particulates are observed, and large settleable

aggregates start to form.

3.3.2 Effects of Water Quality Parameters

The effects of water quality were investigated at a variety of solution conditions,

defined by a specific pH and ionic strength. Figure 3.3 and Figure S3.6 illustrate that the

resistance of the O-MWCNTs to UVC-induced aggregation increases at lower ionic

strength and higher pH conditions. Thus, for NaCl concentrations of 4 – 5mM, the O-

MWCNT absorbance and particle size remains roughly unchanged for irradiation times

less than 18hrs. In contrast, for NaCl concentrations of 12mM the particles reach their

critical size after less than 6hrs and settleable aggregates are observed. In terms of pH

Figure 3.2 – Change in absorbance (filled red circles) and particle size (open blue squares) plotted as afunction of UVC irradiation time for oxidized multiwalled CNTs under anoxic conditions at pH 7 and3mM Na+ under radiation with 8 UVC lamps. The shaded region indicates the time where visibleaggregation of CNTs was observed.

Irradiation Time (hrs)

0 6 12 18 24 30

Ab

sorb

ance

at

350n

m

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35P

article Size A

verage, D

(h) (n

m)

100

200

300

400

500

600

68

effects, when the ionic strength is held constant, at pH 4 (black triangles) the O-

MWCNTs are quickly destabilized by UV irradiation, having almost completely

aggregated in less than two hours. In contrast, at pH 10 (green circles) the same particles

resisted the effects of UV-induced aggregation for over 48 hours. It should be noted that

these trends in increased resistance to UVC irradiation correspond to the same conditions

(low ionic strength, high pH) that increase the stability of negatively charged colloids

towards aggregation. The stability of pH and conductivity readings maintained with the

use of the phosphate buffer indicated that the pH and ionic strength of the suspension

were not the cause of the observed aggregation and settling. These results suggested to us

that the cause of CNT instability was due to a loss of surface oxygen.37, 38

Figure 3.3 – Absorbance profiles for oxidized multiwalled CNTs under anoxic conditions as a function of ionic strength at constant pH 7 (A) and pH at constant ionic strength 3mM Na+ (B) plotted as a function of UV irradiation time.


0 6 12 18 24 30 36

Ab

sorb

ance

at

350n

m

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

4.5mM4.7mM 6mM12mM 12mM

A

[NaCl]


0 8 16 24 32 40 48 56

Ab

sorb

ance at 350n

m

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

4 7 10

pH

B

69

3.3.3 Chemical Transformations to O-MWCNTs caused by UVC Irradiation

XPS was used to examine any chemical changes that occurred to the surface of O-

MWCNTs as a result of UVC-induced aggregation. Figures 3.4A and 3.4B show a

comparison of the C(1s) and O(1s) spectral envelopes obtained from a suspension of O-

MWCNTs that were not exposed to UV light (solid red line) and a suspension of the

same O-MWCNTs after UVC-induced aggregation (dashed blue line). Although the

C(1s) spectral envelope remains virtually unchanged after UVC irradiation, the O(1s)

envelope exhibits a noticeable decrease in intensity (from 9%O to 5.9%O). A similar

decrease in oxygen was observed for a number of O-MWCNTs under different solution

conditions (see Table S3.2) as a result of UVC-induced aggregation. The decrease in the

oxygen signal intensity in the absence of any marked changed to the carbon region is a

consequence of the fact that all of the oxygen functional groups are located at the surface

Figure 3.4 – XPS results for oxidized multiwalled CNTs in absence of any irradiation (solid red line) andafter UVC-induced aggregation and settling had occurred (dashed blue line). The solution conditions inthese experiments were 3mM NaCl and pH 10. Figures 4A and 4B show the C(1s) and O(1s) envelopesbefore and after irradiation. The distribution of oxygen-containing functional groups (C) determined bychemical derivatization illustrates how UVC irradiation changes the concentration of different oxygen-containing functional groups.

Binding Energy (eV)

526528530532534536538540542

Co

un

ts

beforeafter

O(1s) B

Total%

O

%O(C

-OH)

%O(C

OOH)

%O(C

=O)

%O(o

ther

)

Per

cen

t O

xyg

en

0

1

2

3

4

5

6

7

8

9 before after

C

Binding Energy (eV)

280282284286288290292294296

Co

un

ts

beforeafter

C(1s) A

70

of the O-MWCNTs, while approximately 68% of the C(1s) signal comes from the pure

carbon core of the O-MWCNTs. This calculation is based on a NanoLab, Inc. O-

MWCNT which has a manufacturer specified average of 10 walls with an interlayer

spacing of 0.34nm.39

To provide more detailed information on the effect that UVC irradiation has on

the surface chemistry of O-MWCNTs we used chemical derivatization to quantify

changes in hydroxyl, carboxyl, and carbonyl group densities.32, 33 Figure 3.4C shows an

example of the results from this analysis applied to O-MWCNTs before and after UVC-

induced aggregation. The data reveals that the carboxyl groups, as well as residual

(%Oother) oxygen functionalities such as ethers or esters that could also be present on the

O-MWCNT surface but cannot be tagged by chemical derivatization, exhibited a

measurable decrease in concentration. The loss of carboxylic acid groups was observed in

all of the UVC irradiated O-MWCNTs where chemical derivatization was performed,

indicative of a photodecarboxylation (PDC) process. Although there is some variability in

the results, chemical derivatization analysis showed that UVC-induced aggregation

produced a slight increase in the concentration of hydroxyl and carbonyl functionalities,

in contrast to the decrease in carboxyl group density.

The results shown in Figure 3.4 and Table S3.2 provide compelling evidence that

the UVC-induced aggregation of O-MWCNTs is caused by a PDC process, since the

negatively charged carboxylic acid groups are principally responsible for the colloidal

stability of O-MWCNTs. Previous studies have revealed that O-MWCNT stability is

strongly dependent on the concentration of surface bound carboxylic acid groups.38 As

UVC irradiation proceeds, the concentration of carboxylic acid groups decreases

71

systematically until a sufficient number of these groups, and thus the negative charge on

the O-MWCNTs, have been removed. This process results in a point where the O-

MWCNTs are no longer colloidally stable and consequently begin to aggregate (see

Figure 3.1). The extent of decarboxylation necessary to induce aggregation will depend

on the solution conditions. This provides a rationale for the data shown in Figure 3.3,

where the stability of O-MWCNTs towards UVC irradiation increases under solution

conditions that favor the stability of negatively charged colloids (i.e. low ionic strength

and high pH). As a consequence, the length of time necessary to induce aggregation also

increases under these same conditions. The increasing loss of carboxylic acid groups that

occurs as the irradiation time increases also rationalizes the changes in particle

concentration and size observed in Figure 3.2.

Separate experiments were conducted in an attempt to probe the evolution of CO2

during UVC irradiation. Results from these experiments (Figure S3.7) show slightly

higher levels of CO2 evolution for UVC irradiated samples, as compared to the

background levels measured in dark controls. Although the amount of CO2 evolved

during irradiation is comparable to the background CO2 levels from the dark controls, the

data shown in Figure S3.7 is qualitatively consistent with the idea of small (μg) quantities

of CO2 being evolved during the irradiation process. Based on the mass of O-MWCNTs

used in these experiments and the average percentage of carboxylic acid groups present

on these O-MWCNTs, the loss of all of the surface carboxylic acid groups would

correspond to a mass of CO2 evolved on the order of 1 – 2μg, which is qualitatively

consistent with our observations. More experimental detail on these studies can be found

in the Supplemental Information.

72

To better understand the mechanism of PDC, a kinetic study of UVC-induced

aggregation was performed as a function of light intensity by varying the number of

lamps in the Rayonet photoreactor. Figure 5 shows that the irradiation time needed to

affect UVC-induced aggregation increases systematically as the number of lamps

decreases, approximately doubling as the number of lamps is decreased by a factor of two

(16 to 8, 8 to 4, 4 to 2), except for a smaller relative change in going from 4 to 2 lamps.

Although the measurements of particle concentration by UV-Vis or particle size by DLS

provide only an indirect measure of the extent of reaction, we can take advantage of the

fact that if the fundamental nature of the transformation process remains invariant to the

light intensity, then the rate of the reaction will be directly proportional to the effective

rate constant (keff). We ensure this invariance occurs by keeping the initial particle

concentration and solution conditions constant between different light intensity

experiments, as is the case in Figure 3.5.

Under these conditions, the time taken to reach a common point in the

decarboxylation process will be inversely proportional to keff. In Figure 3.5 we choose the

irradiation time taken for the O-MWCNT suspension to decrease to half of its initial

absorbance value (A.U. = 0.17; Figure 3.5C) and for the particle size to increase in

effective hydrodynamic diameter to 630nm (Figure 3.5D) as examples of common points

in the reaction profile that can be measured with reasonable degrees of accuracy. More

importantly, since keff In, where I is the intensity of light (which is directly proportional

to the number of UV lamps) and n is the number of photons involved in the rate

determining step associated with PDC we can write;

73

0.17 630

1 1 or nm n

eff

t tk I

(3.1)

0.17 630log or log log log(# UV bulbs)nmt t n I n (3.2)

Figure 3.5 – Absorbance (A) and particle size (B) profiles for oxidized multiwalled CNTs at pH 7 and12mM NaCl exposed to different light intensities, plotted as a function of irradiation time. The dashed linesindicate t1/2 (A) where the absorbance reaches half of its initial value, and t630nm (B) where the particle sizereached ~630nm. Kinetic data is plotted as a log-log function of t1/2 (C) or t630nm (D) versus light intensity(I).


0 8 16 24 32 40 48 56

Ab

sorb

ance

at

350n

m

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

16 bulbs8 bulbs4 bulbs 2 bulbs

A


0 4 8 12 16 20 24 28 32 36

Par

ticl

e S

ize

Ave

rag

e, D

(h)

(nm

)

200

400

600

800

1000

1200

1400

1600

180016 bulbs8 bulbs 4 bulbs 2 bulbs

B

log(I)0.2 0.4 0.6 0.8 1.0 1.2 1.4

log

(t63

0nm

)

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

D

log (I)0.2 0.4 0.6 0.8 1.0 1.2 1.4

log

(t 1

/2)

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

C

74

where, t0.17 and t630nm represent the irradiation time required for the particle concentration

to decrease to 0.17 absorbance units or the particle size to increase to 630nm,

respectively. Thus, a plot of log (t0.17) or log (t630nm) as a function of log (# UV lamps)

should be linear with a gradient of –n. Analysis of Figures 3.5C and 3.5D shows these

two log-log relationships are close to linear (R2 = 0.9475 for 3.5C and 0.9623 for 3.5D)

with slopes of -0.89 and -0.92 respectively. Since photon induced processes require an

integral number of photons our results indicate that UVC photoinduced decarboxylation

is initiated by a single photon process.

Additional mechanistic insights into the PDC process were obtained by examining

the effect of dissolved oxygen on the stability of O-MWCNT suspensions towards UVC-

induced aggregation. Figure 3.6 shows the absorbance and particle size measurements

taken for experiments performed at the same pH and ionic strength, but under oxic (open

squares) and anoxic (filled circles) conditions. A comparison of the effect of UVC

irradiation on O-MWCNTs in nitrogen purged and oxygen saturated solutions revealed

essentially identical behaviors at the pH and ionic strength condition examined. Both

systems exhibited a decrease in colloidal O-MWCNT concentration (Fig. 3.6A) with a

corresponding increase in particle size (Fig. 3.6B) after approximately 5hrs of UVC

irradiation. Separate measurements (Figure S3.8) revealed that the concentration of DO

in the oxygen saturated solutions was ~3.5 times greater than the level in the nitrogen

purged solutions. These results combined indicate that the rate of photo-induced

aggregation is independent of the dissolved oxygen content. This in turn suggests that the

PDC process does not involve intermediate radicals of reactive oxygen species (such as

•OH), which should increase as the dissolved oxygen concentration increases.40-42 This is

75

validated by very short lifetimes of •OH in water43 and that decreasing concentrations of

•OH are generated from O-CNTs over the course of irradiation.24

Collectively, results from the light intensity and dissolved oxygen experiments

suggest a PDC mechanism that involves a direct single photon excitation of the

carboxylic acid group that triggers the expulsion of CO2. This mechanism is the same as

that which has been proposed and experimentally verified using laser flash photolysis,44

UV irradiation in a Rayonet photochemical reactor,45 and DFT calculations46 for PDC of

aliphatic or aromatic acids and esters such as ketoprofen, acetyl phenyl acetic acid, and

their derivatives. Photoexcitation promotes the molecule to an excited singlet state, which

expels CO2 as it undergoes a rapid intersystem crossing to a triplet state. Before relaxing


0 2 4 6 8 10 12 14

Ab

so

rban

ce a

t 35

0nm

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

N2 purged

O2 purged

A


0 1 2 3 4 5 6 7

Par

tic

le S

ize

Ave

rag

e, D

(h)

(nm

)

200

400

600

800

1000 N2 purged

O2 purgedB

Figure 3.6 – Absorbance (A) and particle size (B) profiles for oxidized multiwalled CNTs at pH 7 and12mM NaCl. plotted as a function of UVC irradiation time, conducted under anoxic (nitrogen purged) oroxic (oxygen purged) conditions.

76

back to a ground state configuration, the available radical or ionic species will extract a

hydrogen from a neighboring carbon atom or the surrounding medium.

Scheme 3.1 shows an example from the organic photochemistry literature of PDC

from 5H-dibenzo[a,d]cyclohepten-5-carboxylic acid,47 a reasonable representation of the

local chemical environment for carboxylic acids attached to CNT sidewalls. In this study,

the authors concluded that benzannelated acetic acids undergo PDC rapidly from a singlet

excited state, proceeding through a carbanion intermediate. They confirmed it was a

singlet state by probing the triplet excited state, which gave no PDC reaction, and the

carbanion intermediate with EPR measurements. Experiments performed in D2O proved

that the hydrogen captured upon returning to the ground state was from the surrounding

medium because the carbanion incorporated deuterium at the position vacated by CO2. A

review of the organic photochemistry literature on other small acid molecules shows that

PDC can occur when the acid is protonated or deprotonated, as well as in both aqueous

and organic solvents. Depending on the conditions (pH, oxic vs. anoxic) and solvent type,

PDC proceeds through either ionic (heterolytic cleavage)44, 45 or radical (mesolytic

cleavage)44, 46 intermediates, which are monitored by analytical techniques such as

Scheme 3.1 – Mechanism for the photodecarboxylation of 5H-dibenzo[a,d]cyclohepten-5-carboxylic acidin water proceeding through a carbanion intermediate upon irradiation with 254nm light based on work byMcAuley et. al.47

77

transient absorption, fluorescence, mass spectrometry, and NMR. A more detailed

explanation of carboxylic acid and ester PDC can be found in the review article by Budac

and Wan.48 Models from the organic photochemistry literature led us to propose the

following process (Scheme 3.2) for PDC and subsequent aggregation of O-MWCNTs.

The diversity of chemical environments under which PDC can occur suggests that

reactive oxygen species are not responsible for the photoreduction process exhibited by

O-MWCNTs. The idea that strongly oxidizing ROS, such as hydroxyl radicals, are

responsible for a net reduction process such as PDC is hard to rationalize from an

intuitive chemical perspective. Moreover, such a mechanism is inconsistent not only with

our experimental observations at different concentrations of dissolved oxygen, but also

the extensive body of already existing but seemingly overlooked organic photochemistry

literature which has identified the mechanism, including the fact that PDC can occur in

organic solvents where ROS should not exist.

In more general terms it should be noted, however, that although the changes in

particle size and absorbance values can be ascribed to PDC, the effects of UV irradiation

are not restricted to the carboxylic acid groups. For example, photo-induced

Scheme 3.2 – Proposed pathway for photodecarboxylation and subsequent aggregation of O-MWCNTsthrough the removal of carboxylic acid functional groups.

Aggregation

78

decarboxylation of ester groups48 and the photoreduction of ethers23 (the later

demonstrated in the photochemistry of graphene oxide) would be consistent with the

significant decrease in the concentration of “other” oxides (not hydroxyl, carbonyl or

carboxyl) that chemical derivatization cannot assay (see Figure 3.4). In this respect, a

potential role for reactive oxygen species (such as hydroxyl radicals) that are produced as

a result of UV exposure to oxidized CNTs24, 25 could be the cause of the increase seen in

hydroxyl group density, and also the slight increase in structural disorder observed by

Raman (Figure 3.8). Thus, it seems that the UVC-induced chemical transformations to O-

MWCNTs are likely a result of more than one process, though PDC is the dominant

process that directly influences particle stability and governs the fate and transport

properties of the O-MWCNTs.

3.3.4 Structural Transformations to O-MWCNTs

To determine if there were any structural transformations imparted to the O-

MWCNTs as a consequence of UVC irradiation, TEM and Raman were both performed

on O-MWCNTs before and after UVC-induced aggregation. TEM images (Figure 3.7)

revealed an absence of any obvious structural changes to the O-MWCNTs after UVC

exposure, with the shape, size, and number of sidewalls remaining constant based on

visual observations of different O-MWCNTs. Some amorphous carbon can be seen

attached to the outer walls of the individual CNTs in both sets of micrographs, with

perhaps slightly more on the outer walls of the tubes after UVC irradiation. TEM analysis

was performed on a number of O-MWCNTs before and after UVC irradiation under

different solution conditions (pH, dissolved gas) with the same qualitative finding. The

lack of physical transformations to the surface of O-MWCNTs in our irradiation

79

experiments is similar to the observations of Hwang et. al. when they irradiated

carboxylated MWCNTs for seven days under UVA light.26

Raman spectroscopy was performed on O-MWCNTs that had not been exposed to

UVC irradiation (control), as well as suspensions that had been irradiated under oxic and

anoxic conditions for sufficient times to induce aggregation. In Raman experiments, O-

MWCNTs from two different manufacturers, NanoLab, Inc. and Cheaptubes (Figure 3.8),

were studied. Prior to irradiation, O-MWCNTs from both manufacturers exhibited

significant D-band intensities indicative of a large amount of disordered (sp2) carbon

atoms, presumably as a result of the aggressive oxidizing conditions needed to introduce

surface oxides into the graphenic sidewalls.49 The NanoLab, Inc. CNTs exhibited

Figure 3.7 – Low (top row) and high (bottom row) resolution TEM micrographs of O-MWCNTs beforeand after UVC irradiation at pH 7 under anoxic conditions.

80

identical ID:IG band ratios of 0.94 for the control samples and for O-MWCNTs which had

aggregated under the influence of UVC irradiation in a nitrogen purged suspension. In

contrast, the ID:IG band ratio increased slightly to 1.01 when O-MWCNTs were subjected

to UVC irradiation in a solution saturated with dissolved oxygen. In comparison, the

Cheaptubes O-MWCNTs ID:IG band ratio increased slightly from 1.17 in the control, to

1.22 after UVC-induced aggregation in either oxic or anoxic conditions. These Raman

results indicate that UVC induced aggregation proceeds in the absence of any significant

structural changes, consistent with the TEM data shown in Figure 7. The very modest

increases in ID:IG band ratio observed upon UVC irradiation could potentially be the

result of reactions between small quantities of ROS generated from irradiation of the

suspension and the graphenic portions of CNT sidewalls, or from the removal of surface

functional groups and damage imparted to the sidewalls by the high energy UVC light.

However, if ROS, such as hydroxyl radicals, played any appreciable role in the

photochemical transformation process we would expect to see a significantly larger

Cheaptubes

Raman Shift (cm-1)

500 1000 1500 2000 2500 3000

Control Irradiated, N2Irradiated, O2

NanoLab, Inc.

Raman Shift (cm-1)

500 1000 1500 2000 2500 3000

No

rmal

ized

Inte

nsi

ty (

a.u

.)

ControlIrradiated, N2Irradiated, O2

Figure 3.8 – Raman spectroscopy showing the effects of UVC irradiation on O-MWCNTs purchased fromNanoLab, Inc. and Cheaptubes. Results are shown for O-MWCNTs before irradiation and after UVC-induced aggregation under oxic and anoxic conditions at pH 10.

81

increase in the degree of structural disorder to have occurred.42, 50 Thus, results from

TEM and Raman indicate that the significant changes in the O-MWCNT surface

chemistry caused by UVC irradiation are not accompanied by any measurable changes to

the physical structure of O-MWCNTs.

This absence of any significant structural changes to the O-MWCNTs is also

consistent with the lack of mineralization. In large batch studies greater than 90% of the

initial mass was recovered after aggregation, regardless of the solution conditions, time of

irradiation, source of O-MWCNTs, or whether the CNTs were cleaned before use (Table

1). The small amount of mass loss could potentially be ascribed to the CO2 evolution that

accompanies PDC and the concomitant release of a small amount of dissolved organic

carbon that arises from destruction of the O-MWCNT sidewalls. It should be noted that

upon comparing the recovery of O-MWCNTs from experiments performed with CNTs

that were base cleaned to remove excess amorphous carbon (NanoLab, Inc.) and those

that were not (Cheaptubes), the results are similar and therefore we can say that this

cleaning step did not influence the fundamental PDC process. Recoveries also suggest

that irradiation causes almost negligible destruction, and that UVC irradiation is unable to

achieve any appreciable mineralization of the O-MWCNT particle structure.

Table 3.1 – Mass loss experienced by O-MWCNTs from NanoLab Inc. and Cheaptubes as a result of UVCinduced aggregation under different solution conditions.

CNT Manufacturer pH Purged With Percent Remaining NanoLab, Inc. 10 Nitrogen 91% NanoLab, Inc. 10 Nitrogen 92% NanoLab, Inc. 10 Oxygen 94% NanoLab, Inc. 7 Nitrogen 97%

Cheaptubes 10 Oxygen 93%

82

3.3.5 Phototransformations of Oxidized Carbon Based Nanomaterials

Our results showed similar chemical changes to studies from Hwang et. al.26 and

Qu et. al.,27 who irradiated the same O-MWCNTs used in this study with UVA light.

Specifically, the three studies saw a photoreduction process that lacked change to the

physical structure. A significant difference, however, comes from the mechanism used to

explain the PDC process. Qu et. al. postulate that hydroxyl radicals were responsible for

the decarboxylation via a reaction where a carboxylic acid combines with a hydroxyl

radical to produce a molecule of CO2 and water. Meanwhile, results from our study and

from existing organic photochemistry studies indicate that PDC occurs in both oxic and

anoxic conditions, as well as in organic solvents where hydroxyl radicals could not exist.

Together, these results are inconsistent with a PDC mechanism that involves ROS

species.44-48

In addition to O-MWCNTs, a variety of complementary UV irradiation studies

have been performed on fullerols, graphene oxide (GO) nanosheets, and O-SWCNTs.

Collectively, these results point to a common phototransformation process that occurs

across a range of different excitation wavelengths. For O-MWCNTs, we have seen a

process dominated by transformations that occur exclusively at the surface, meaning that

the bulk of the carbon atoms in the central core of the O-MWCNT remain unaffected by

the UV radiation. However, if we were to consider that process in the context of O-

SWCNTs or fullerols then we would predict that irradiation would have a much greater

relative effect since all of the atoms are at the surface of these nanomaterials. Consistent

with this assertion, previous studies have reported decreases in the particle size of O-

SWCNTs and fullerols by DLS and TEM as a result of UV irradiation,51, 52 with

83

measurable amounts of CO2 released from fullerols that indicate almost 50%

mineralization of the fullerol molecule.52

Therefore, we can distinguish two subgroups of oxidized carbon based

nanomaterials: 1) when all of the carbon atoms are at the surface of the nanomaterial (O-

SWCNTs, fullerols, graphene oxide nanosheets), and 2) when a surface layer exists but

the majority of carbon atoms are contained below the surface (O-MWCNTs). We can

examine the different effects seen for each subgroup as a result of photolysis by

comparing the data obtained for O-MWCNTs in the present investigation to studies on

GO nanosheets. In terms of the chemical transformations, data from studies using GO

nanosheet suspensions18, 20-22 show a similar reduction in surface oxygen concentration to

O-MWCNTs. However, the type and amount of oxygen-containing functional groups are

different between these two carbon allotropes, where GO possesses a much higher

percentage of epoxides due to the oxidation method used.53, 54 Peak fitting of the XPS

carbon envelopes by Matsumoto, Guardia, and Koinuma showed almost complete loss of

the C-O peak, whereas the carbon envelope showed little change for O-MWCNTs. In

contrast to the qualitatively similar chemical changes imparted by photolysis,

comparisons of the physical transformations exhibit a much more measurable difference

between O-MWCNTs and GO nanosheets. Minimal structural damage was observed for

O-MWCNTs in all three above-mentioned studies, but AFM micrographs and TEM

images of GO nanosheets show distinct holes created by the removal of oxygen. Raman

results from the GO nanosheets actually saw a decrease in the ID:IG band ratio, indicating

a less disordered surface, whereas Raman showed an increase in the disorder with O-

MWCNTs.

84

Therefore, the main difference between these two subgroups lies in the physical

transformations photolysis produces, and the resulting structural integrity of the

nanomaterial. Photoreduction and the subsequent loss of volatile carbon atoms create

noticeable defect areas in nanomaterials consisting of a single sheet of graphene.

Removal of portions of the sheet simultaneously degrades the material but lowers the

overall defect density by creating a smooth graphene-like structure with large defect

areas.18 This eventually leads to large scale disruption of the nanomaterial, including

mineralization and fragmentation.17, 51 When the same size portion of graphene is

removed from the outermost wall of an O-MWCNT, the remaining walls beneath this

hole show a defect site in the carbon layers, but do not impact the overall integrity of the

nanomaterial. Removal of oxygen functional groups simply causes the particle to become

unstable in aqueous solution. The magnitude of the van der Waals attraction between the

carbon cores causes aggregation once enough carboxylic acid groups are removed. Thus,

UVC light is unable to effect mineralization of O-MWCNTs.

3.3.6 Environmental Implications

Results from our study have shown that UVC irradiation can destabilize a

colloidal suspension of O-MWCNTs regardless of solution conditions. Many facilities

across the country have implemented UVC irradiation to disinfect their water resources.

However, these municipal drinking and wastewater treatment plants operate at a

multitude of volumes, ranging from 1 – 450 million gallons per day.55 Though the EPA

has issued standard dosages for water treatment plants to enact specific log kills of

various waterborne bacteria and viruses,56 there are not yet standards for nanoparticles.

To complicate matters, any number of variables including the concentration of

85

nanoparticles in the water being treated, the dissolved organic matter content, the output

wattage of lamps used, the water contact time, reactor depth, and flow rate of a given

plant differ between facilities. Even if the UV light intensity in a commercial water

treatment plant is an order of magnitude higher than the experiments described in this

study (photon flux ~ 1.32 x 1017 quanta/sec with all lamps operational), if we consider the

volumes of water disinfected each day in one of these plants it seems unlikely that UVC

treatment will transform O-MWCNTs to an appreciable extent given the timescales

(hours) required for PDC.

3.4 Results and Discussion of O-SWCNTs

3.4.1 Visual Effect of UVC Irradiation

Large batch experiments, like those performed on the O-MWCNTs, were used to

study colloidal suspensions of 40% w/w nitric acid treated O-SWCNTs to see if the

effects from UV radiation were similar. Upon irradiation with 254nm light, a drastically

different mechanism governing the MW and SW CNT suspensions was observed. Where

the O-MWCNTs took just over two days to aggregate and settle out of suspension, the O-

SWCNTs took over two weeks before the supernatant became noticeably clear to the eye.

However, the most significant difference was in the aggregation state of the CNTs. The

O-MWCNT suspension exhibited an initial period of stability where there were no

observable changes to the suspension, after which aggregates began to form and settle to

the bottom of the vessel. This was not the case for the O-SWCNTs. The color slowly

began to fade from the suspension of O-SWCNTs after the first few time points, and over

the course of irradiation no visible aggregates could be seen floating in the supernatant.

86

Eventually, fine black flecks began to appear at the bottom of the quartz beaker, and the

supernatant cleared completely. Figure 3.9 shows the progression of O-SWCNTs under

irradiation.

Absorbance and particle size measurements of the O-SWCNT suspensions

confirmed that a different process was occurring, validating the visual difference

observed between the two types of oxidized CNTs. Absorbance measurements in Figure

3.10 showed a profile that almost instantaneously begins to decline at the onset of

irradiation, and throughout there were regions of leveling where the absorbance value did

not change between tested samples. This was almost directly opposite to the profile

generated by the O-MWCNTs, where absorbance remained steady at the onset and then

declined sharply after a critical particle size was reached. Corresponding particle size

measurements for the O-SWCNT suspension showed very steady hydrodynamic diameter

readings for the first four days while the absorbance was rapidly decreasing. In

0 days 6 days 10 days 12 days 17 days

Figure 3.9 – Visual effects of UVC irradiation on single walled CNTs from Southwest Nanotechnologiesoxidized with 40% nitric acid. Irradiation was performed under ambient conditions at pH 7. No observableaggregation is observed for the first 10 days, as only the color of the suspension lightens over the course of this time. Starting O-SWCNT concentration is approximately 13.6mg/L. Photos by Miranda Gallagher.

87

the regions where the absorbance measurements reached a brief plateau a rise in particle

size was seen. As the decrease in absorbance became more linear, the particle size

continued to grow, eventually reaching far larger measurable particle sizes than the O-

MWCNTs. Once the threshold of 1000nm was reached, the absorbance began rapidly

decreasing. This decline eventually slowed until an absorbance value was eventually

reached corresponding to an estimated concentration of the supernatant that was less than

5% of the initial starting concentration. These changes occurred within ±0.1 of pH 7 and

relatively steady conductivity measurements (conductivity measurements showed much

higher recorded values for O-SWCNTs than O-MWCNTs).

3.4.2 Chemical Transformations to O-SWCNTs caused by UVC Irradiation

The large volume samples used for the absorbance and particle size measurements

above were cleaned and dried in the same fashion as the O-MWCNTs from Section 3.2.1.

Irradiation Time (days)0 3 6 9 12 15 18

Ab

sorb

ance

at

350n

m

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Particle S

ize Ave

rage, D

(h) (n

m)

0

500

1000

1500

2000

2500

3000

Figure 3.10 – Change in absorbance (filled red circles) and particle size (open blue squares) plotted as afunction of UVC irradiation time for oxidized single walled CNTs under ambient conditions at pH 7 underradiation with 16 UVC lamps.

88

However, mass recovery proved difficult as there was little residual powdered sample

remaining after irradiation. Weighing the remaining sample on the analytical balance

indicated that greater than 90% of the starting mass was lost during the course of UVC

irradiation, as is illustrated for two examples in Table 3.2. XPS analysis on the remaining

powder indicated a loss of surface oxygen, indicating a decrease from 16% on the starting

material to 10 – 12% on the powder remaining post irradiation. Chemical derivatization

for oxygen-containing functional groups was unable to be performed on O-SWCNT

samples exposed to UVC light due to the small quantity recovered. However, it is

believed that a similar photodecarboxylation (PDC) mechanism is responsible for the loss

of oxygen, as well as the mineralization observed.

3.4.3 Structural Transformations to O-SWCNTs

Fullerenes are similar analogues to SWCNTs, consisting of only a single layer of

carbon. Studies examining the effects of UV radiation on fullerenes and SWCNTs have

shown decreases in particle size, as well as slow loss of supernatant intensity over the

course of irradiation.13, 51 Hou and Jafvert suggested that the fullerenes were in fact being

either mineralized to CO2 or converted to another product like volatile organic

compounds.13 One needs to take into consideration that oxidation, especially more

Mass to Start (mg) Mass Recovered (mg) Percent Remaining 13.6 0.63 4.6% 13.6 0.55 4%

Table 3.2 – Mass loss experienced by O-SWCNTs from Southwest Nanotechnologies as a result of UVCinduced aggregation at pH 7.

89

aggressive methods of oxidation (i.e., nitric acid, 3:1 sulfuric:nitric acids) already imparts

large structural defects to the CNT surface. When these treatment methods are performed

on SWCNTs, it destroys the structural integrity of the material by punching holes into the

sidewall and breaking longer tubes into smaller pieces. Having the sidewall already

compromised by introducing 16% surface oxygen, it seems likely that the energy

absorbed by the O-SWCNTs during irradiation is enough to not only remove and alter

surface functionalization through UV-induced PDC, but also to fragment the tube into

smaller and smaller pieces.

Suspensions were created with small samples of powdered SWCNTs to

compare pristine SG65s to O-SWCNTs with different levels of surface oxygen present.

These varying surface oxygen levels were created by refluxing the pristine CNTs in

different w/w concentrations of nitric acid. These powders were suspended in a 2%w/v

SDC solution before being analyzed by near infrared fluorescence spectroscopy (NIRF).

It was discovered that the more aggressive treatments with nitric acid (i.e., 40% and 70%

HNO3) destroyed the fluorescence signal of the SWCNTs even before any exposure to

irradiation (Figure 3.11). Those O-SWCNTs that still showed some fluorescence

character after oxidation (10% HNO3) displayed selective removal of the main (6,5)

diameter tube peak at about 10,200cm-1. This result suggests that the oxidation process

may selectively oxidize smaller tubes first, destroying the sidewall structure and reducing

the fluorescence signature. Figure 3.11 illustrates this selective reduction by comparing a

sample from pristine SG65 SWCNTs versus the 10%, 40%, and 70% nitric acid treated

sample.

One suspension of O-SWCNTs that still showed some fluorescence after

90

oxidation was selected to test the effects of UV radiation. Several small volume samples

of the O-SWCNT suspensions were exposed to UV radiation and subsequently analyzed

by NIRF to mark any changes in the fluorescence signal as a result of UV light exposure.

Figure 3.12 shows a comparison of the control which was kept wrapped in aluminum foil,

to samples that were exposed to UVC radiation for 10, 30, and 60min. Though oxidation

removed the majority of the signal contribution from the (6,5) SWCNTs, as was seen in

Figure 3.11, UVC radiation seemed to destroy the electronic signature of the larger

diameter tubes ((7,5) and (8,4) tubes at approximately 9650cm-1 and 8950cm-1,

respectively) sooner as the time progression shows. The exposed samples lost all

fluorescence signal after just 60min of UV irradiation.

Wavenumber (cm-1)

7000 8000 9000 10000 11000

Flu

ore

scen

ce E

mis

sio

n I

nte

nsi

ty (

nW

/cm

-1)

0

2e-6

4e-6

6e-6

8e-6

1e-5

Pristine10% HNO340% HNO370% HNO3

A

Wavenumber (cm-1)7000 8000 9000 10000 11000

Flu

ore

sce

nce

Em

issi

on

In

ten

sity

(n

W/c

m-1

)

0

1e-6

2e-6

3e-6

4e-6

5e-6

6e-6


B

Wavenumber (cm-1)

7000 8000 9000 10000 11000

Flu

ore

scen

ce E

mis

sio

n I

nte

nsi

ty (

nW

/cm

-1)

0

5e-6

1e-5

1e-5

2e-5


C

Figure 3.11 – NIRF signal for pristine SWCNTS compared to differently oxidized SWCNT under excitation wavelengths (A) 638nm, (B) 691nm, and (C) 782nm. Suspension concentration was slightly varied as a result of centrifugation to remove bundling..

91

3.5 Conclusions

UVC irradiation performed using a Rayonet photochemical reaction chamber

resulted in transformations of O-MWCNTs principally through a photodecarboxylation

process. This process was found to be mediated by a one photon, direct excitation

mechanism consistent with existing literature in the organic photochemistry community,

and inconsistent with a process mediated by reactive oxygen species. During the initial

stages of irradiation the extent of carboxylic acid group removal from the surface of the

O-MWCNTs leads to a slow increase in particle size, but is not sufficient to cause

settleable aggregates to form. During this time, the particle concentration remains

constant and the suspension remains visibly unchanged to the naked eye. However, once

a sufficient number of carboxylic acid groups have been removed, a critical point is

reached where the electrostatic repulsion between O-MWCNTs is no longer sufficient to

prevent the CNTs from rapidly aggregating. UVC-induced aggregation occurs at all light

intensities studied and under both oxic and anoxic conditions where the resistance of

Wavenumber (cm-1)

7000 8000 9000 10000 11000

Flu

ore

scen

ce E

mis

sio

n I

nte

nsi

ty (

nW

/cm

-1)

0

1e-6

2e-6

3e-6

4e-6

5e-6

6e-6

Control10min Exposure 30min Exposure60min Exposure

C

Wavenumber (cm-1)

7000 8000 9000 10000 11000

Flu

ore

scen

ce E

mis

sio

n I

nte

nsi

ty (

nW

/cm

-1)

-1e-6

0

1e-6

2e-6

3e-6

4e-6

5e-6

Control10min Exposure30min Exposure60min Exposure

B

Wavenumber (cm-1)

7000 8000 9000 10000 11000

Flu

ore

scen

ce E

mis

sio

n I

nte

nsi

ty (

nW

/cm

-1)

0

2e-6

4e-6

6e-6

8e-6Control10min Exposure30min Exposure60min Exposure

A

Figure 3.12 – NIRF results for lightly oxidized SWCNT control suspension versus exposures to 8 UVC lamps for 10, 30, and 60min under excitation wavelengths (A) 638nm, (B) 691nm, and (C) 782nm. Suspension concentration was 10mg/L at pH 10.

92

O-MWCNTs towards photo-induced aggregation is enhanced in solution conditions that

favor the stabilization of negatively charged colloids (high pH and low ionic strength).

XPS, Raman, and TEM show significant changes in surface chemistry and aggregation

state induced by UVC exposure, but in the absence of any discernible structural

transformations or mineralization.

3.6 Acknowledgements

We acknowledge financial support by the Environmental Protection Agency

(R834858). The authors would also like to thank Dr. Ken Livi of the Earth and Planetary

Sciences Department at JHU for the TEM imaging, and the Materials Science

Department at JHU for use of the surface analysis laboratory. JLB would also like to

thank Miranda Gallagher and Ronald Lankone for their help with cleaning and processing

CNT samples.

3.7 Supplemental Information

3.7.1 UV-Visible Spectroscopy of O-MWCNT Suspensions

Absorbance of O-MWCNT suspensions was an easy way to monitor the colloidal

stability of CNTs during UVC irradiation. Figure S3.1 shows a representative example of

a typical UV exposure experiment, where the absorbance profile did not change

significantly over the course of the experiment. The dashed line indicates the wavelength

of UVC light being emitted from the lamps in the Rayonet reaction chamber, which

coincides with where the O-MWCNTs absorb most strongly (λ = 264nm). This peak is

93

due to the π→ π* transition in the conjugated ring structures. The sharp rise in

absorbance close to 200nm is due to the presence of ions in the suspension from the

added buffer and salt added to the O-MWCNT suspension. Figure S3.2 illustrates the

contribution of the various suspension components to the UV-Vis absorbance spectra.

Wavelength (nm)

200 300 400

Ab

sorb

ance

(a

.u.)

0.0

0.2

0.4

0.6

0hrs 2hrs 3hrs 4hrs 5hrs 6hrs 7hrs 8hrs 9hrs 10hrs 11hrs 12hrs

200 300 400 500 600 700 800 9000.0

0.1

0.2

0.3

0.4

0.5

Figure S3.1 – UV-Vis absorbance spectra from 200 – 450nm of oxidized multiwalled CNTs at pH 7 and 12mM NaCl, purged with nitrogen and measured as a function of irradiation time. Absorption maximum (λ = 264nm) corresponds to the π→ π* transition in the conjugated sidewall ring structure. The dashed line indicates the irradiation wavelength of 254nm, and the solid line indicates the wavelength at which measurements were taken (350nm). Inset shows the full spectra ranging from 200 – 900nm.

94

3.7.2 Choosing a Suitable Buffer

Choosing an appropriate buffer that will not degrade or influence the O-

MWCNTs while under irradiation with 254nm light was key in having reproducible and

reliable results. We avoided the use of buffers that contained conjugated ring structures

because those would be degraded by the UV light, and we avoided the use of buffers that

could potentially produce electrons or reactive oxygen species that could influence the

transformation of O-MWCNTs in suspension. It was decided that the buffer needed to be

polyprotic so it could be used over a range of pH values we wished to study. Carbonate is

a typical buffer used in previous experiments with O-MWCNTs,28, 37, 38 but upon purging

with nitrogen the carbonate is not stable due to the reduction of oxygen and carbon

Wavelength (nm)200 400 600 800

Ab

sorb

ance

(A

.U.)

0.0

0.1

0.2

0.3

0.4

0.5

HPLC water O-MWCNT suspension

Wavelength (nm)200 205 210 215 220

Ab

so

rban

ce

(A.U

.)

0.0

0.1

0.2

0.3

0.4HPLC blank 3mM phosphate buffer12mM NaClbuffer + NaCl

Figure S3.2 – UV-Vis absorbance spectra from 200 – 900nm for the individual constituents that make up an O-MWCNT suspension. The inset shows the region from 200 – 220nm to show the increase displayed in the absorbance profiles of the 3mM phosphate buffered water and the 12mM NaCl solution. These contributions can be seen in the profiles of the experimental O-MWCNT suspension from Figure S1.

95

dioxide dissolved in the medium. Therefore, we chose phosphate salts as the buffer of

choice for our experiments because it allowed the flexibility necessary with the pH range,

was unaffected by the UV light, and did not influence transformation.

Figure S3.3 shows two common buffers used for experiments performed at pH 4.

Suspensions were prepared as in the experimental section, buffered using either acetate or

phosphate, and purged with nitrogen. The two quartz vessels were put into the Rayonet

chamber and run simultaneously, monitoring the absorbance and pH as a function of

irradiation time. Results indicate that the phosphate behaved similarly to the acetate,

confirming our choice of using phosphate salts as a stable buffer for the remainder of our

experiments.


0.0 0.5 1.0 1.5

Ab

sorb

ance

at

350n

m

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Acetate Phosphate

Figure S3.3 – Absorbance measurements for O-MWCNTs from NanoLab, Inc. under anoxic conditions using two common buffers to keep the suspension stable at pH 4.

96

3.7.3 Calibration of Light Intensity via Actinometry

To determine the quantum flux we used a colorimetric actinometry experiment to

measure the radiation of our UV lamps. As described in the experimental section of this

paper, potassium ferrioxalate decomposes to form Fe2+ ions under irradiation, which we

can then complex with o-phenanthroline to form a red solution. The intensity of this red

solution can be measured as a function of irradiation time to gauge the conversion to

Fe2+, thus measuring the intensity of light being emitted from our lamps.

Actinometry experiments are performed in the dark to eliminate light from the

windows and overhead lights contributing to the decomposition of the ferrioxalate.

Experimentally, 3mL of 0.006M ferrioxalate solution were put into the same 15mL

quartz test tubes used to irradiate O-MWCNTs and exposed to irradiation for different

prescribed periods of time (t). After each exposure period, 2mL of the irradiated solution

was removed and mixed with 2mL of 0.0055M o-phenanthroline and 1mL of a pH 3.5

acetate solution to quench the reaction. This 5mL volume was then diluted to a 20mL

total volume and stored for at least 1hr in the dark. This ensured that all of the Fe2+

produced by the UVC light had complexed with the o-phenanthroline. The optical density

(D) of the solution was then measured at 510nm by UV-Vis to determine the

concentration of the Fe2+:o-phenanthroline complex and a plot of D vs. t was constructed.

Until the supply of ferrioxalate has been consumed, D varies linearly with t, and in this

regime Eq. S3.1 can be used to determine the quantum flux (QF);31

3

1 3

2

10DVV NQF

t dV

Eq. S3.1

97

where V1 is the volume of ferrioxalate solution irradiated, V3 is the total end volume of

solution, N is Avogadro’s number, Φ is the quantum yield at the irradiation wavelength

(for 254nm, Φ = 1.25), ε is the molar extinction coefficient of the ferrioxalate-

phenanthroline complex at 510nm (1.11 x 104 L/mol·cm), d is the thickness of the

cuvette, and V2 is the volume of irradiated solution used for complexation.

Part A of Figure S3.4 illustrates the linear portion of our actinometry curves for

experiments performed with 16, 8, 6, 4, 2, and 0 lamps. A minimum of six data points

were collected and fit using a linear regression to determine the absorbance versus time

dependence for each set of exposures. What the results suggest is that when the lamps are

off, there is no residual light or photons being emitted from the lamps to induce any

degradation of the ferrioxalate compound. As the number of lamps in the Rayonet

chamber increases, it was found that intensity varied linearly as a function of the number

Ferrioxalate Exposure Time (sec)

0 20 40 60 80 100 120

Ab

sorb

ance

at

510n

m

0.0

0.5

1.0

1.5

2.0

1686420

A

Number of Lamps

0 2 4 6 8 10 12 14 16

Qu

an

tum

Flu

x (

qu

an

ta/s

ec

)

0.0

2.0e+16

4.0e+16

6.0e+16

8.0e+16

1.0e+17

1.2e+17

1.4e+17

B

Figure S3.4 – Calibration curves (A) and the calculated quantum flux (B) for various lamp intensities measured with the ferrioxalate actinometry experiments.

98

of lamps, except for experiments conducted with 16 lamps. By the time the maximum

number of lamps is reached, there is no real difference in the slope from 8 to 16 lamps.

This effect is most likely due to the light intensity at both 8 and 16 lamps being

sufficiently high to cause the ferrioxalate solution (0.006M, A254nm = 3.6) to convert

completely to Fe2+. Part B of Figure S3.4 shows the quantum flux calculated for each set

of exposures using Eq. S3.1. A linear regression was used to extrapolate what the likely

quantum flux for 16 lamps would be. In this instance the flux would be expected to reach

approximately 1.32 x 1017 quanta/sec, as opposed to the measured flux of 7.73 x 1016

quanta/sec. The values for all lamp intensities measured with actinometry can be found in

Table S3.1.

For experiments performed with O-MWCNTs, the optical density of the O-

MWCNT suspensions (5mg/L, A254nm = 0.34) is an order of magnitude lower than that of

the ferrioxalate. Consequently, the absorbance of the suspension in the quartz test tubes is

expected to follow a Beer-Lambert law regardless of the number of lamps used. Under

Number of Lamps Quantum Flux (quanta/sec)

0 3.07 x 1011 2 1.94 x 1016

4 2.59 x 1016 6 5.82 x 1016 8 6.30 x 1016

16 (measured) 7.73 x 1016 16 (extrapolated) 1.32 x 1017

Table S3.1 – Calculation of the quantum flux for 16, 8, 6, 4, 2, and 0 lamps. Flux is determined by actinometric measurements performed with potassium ferrioxalate. The quantum flux listed for 16 lamps appears twice to indicate that which was actually measured during the actinometry experiment, and what the likely flux is based on extrapolation via linear regression of the data from 0 – 8 lamps.

99

these conditions the effective light intensity within each test tube will be directly

proportional to the number of lamps.

3.7.4 Water Quality Parameters

An important factor in examining the effect of UVC radiation on suspensions of

O-MWCNTs is that they remain stable while not under the influence of light. To test this,

suspensions were run simultaneously, where half were being exposed to the UVC lamps

and the other half were wrapped in aluminum foil to prevent exposure. Figure S3.5 shows

that the O-MWCNT suspensions were in fact stable over the course of a given

experiment and showed no change in absorbance from start to finish.

Irradiance Time (hrs)0 3 6 9 12 15

Ab

sorb

an

ce a

t 35

0nm

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

ControlIrradiated

Figure S3.5 – Comparison of absorbance measurements for a control/dark sample and a sample exposed to UVC radiation at pH 7 and 3mM Na+.

100

As was discussed in Section 3.3.2, absorbance and particle size was measured as a

function of time for various solution conditions. Figure S3.6 illustrates the particle size

measurements corresponding to the absorbance readings from Figure 3.3.

3.7.5 Chemical Characterization of Large Volume Irradiation Experiments

XPS and chemical derivatization was performed on a series of large volume

experiments performed with O-MWCNT suspensions at various solution conditions.

Large volumes were used to ensure enough sample remained after exposure to text for

total percent surface oxygen and carboxylic acid functional groups. Table S3.2 lists the

measured atomic percentages of oxygen from control and irradiated samples. All samples

showed a decrease in carboxylic acid group density, even if there was no real change in

the total percent oxygen.

Irradiation Time (hrs)0 8 16 24 32 40 48

Particle S

ize, D(h

) (nm

)

200

300

400

500

600

B

pH4 7 10


0 6 12 18 24 30 36

Par

ticl

e S

ize,

D(h

) (n

m)

200

300

400

500

600

4.5mM4.7mM6mM12mM 12mM

[NaCl]

A

Figure S3.6. – Particle size measurement profiles for oxidized multiwalled CNTs under anoxic conditions as a function of ionic strength (A) and pH (B) plotted as a function of UV irradiation time.

101

3.7.6 Measurement of Total Inorganic Carbon (TIC) and Calculation of CO2 Evolved

O-MWCNTs were prepared as a 25mg/L stock suspension and set to pH 7 with

3mM phosphate buffer. This stock suspension was sent to Purdue University for total

inorganic carbon analysis to determine the amount of CO2 evolved over the course of

irradiation. Quartz test tubes were filled with 8mL CNT suspension and contained 9mL

headspace. A stainless steel syringe needle was inserted through a rubber septa capped

onto each test tube and used to sparge the CNT suspension with nitrogen gas for 30

minutes at 2.5mL/sec flow rate. Sample tubes were irradiated using Rayonet RPR-100

merry-go-round photochemical reactor (Southern New England Ultraviolet (SNEU),

Branford, CT) with 16 RPR-2537A lamps (24 watts). In the reactor, sample tubes are

placed within the merry-go-round at the center of the reactor and rotated at 5 rpm to

ensure uniform exposure.

Table S3.2 – XPS measurements performed on various O-MWCNTs before and after irradiation with 254nm UVC light for various O-MWCNTs under oxic or anoxic conditions at pH 10. Only the total oxygen percent is shown, the percent carbon is neglected, but the %C + %O = 100%. For example, if the %O = 7.5%, the carbon peak result was 92.5%. The numbers in parentheses show the percentage of carboxylic acid groups that were measured before and after irradiation. The asterisk (*) indicates that the experiment was performed at pH 7 instead of pH 10.

Manufacturer Gas Purge %Oxygen Before

Irradiation %Oxygen After UV-Induced

Aggregation NanoLabs, Inc. Nitrogen 7.5 (1.7) 5.1 (0.6) NanoLabs, Inc. Oxygen 7.5 (1.7) 6.1 (1.3)

Cheaptubes Nitrogen 9.0 (0.7) 7.2 (0.5) Cheaptubes Oxygen 9.0 (0.7) 5.1 (0.4)

NanoLabs, Inc. Nitrogen 9.0 (3.2) 5.9 (2.2) Cheaptubes Oxygen 6.9 (0.9) 6.8 (0.3)

NanoLabs, Inc. Nitrogen 9.5 (1.7) 6.5 (0.8) NanoLabs, Inc.* Nitrogen 7.9 (1.8) 6.3 (0.4)

102

After a specified period of irradiation, test tubes were removed and acidified with

85% phosphoric acid to pH < 3. Test tubes were shaken vigorously for 5min and allowed

to equilibrate overnight, after which 3mL aliquots from the headspace of each tube were

withdrawn for CO2 analysis. Controls, test tubes wrapped in aluminum foil to prevent the

suspension from being exposed to the UV light while in the Rayonet chamber, were

acidified and treated the same way for analysis. The gas phase CO2 concentration was

measured with a PDZ-Europa Elemental Analyzer interfaced to a Sercon 20-20 Isotope

Ratio Mass Spectrometer (Crewe, England). The amount of CO2 in the aqueous phase

can be calculated using the equation,

2

2

[ ]

[ ]g

Haq

COk

CO Eq. S2

where kH is the Henry’s constant for CO2 dissolved in water, which is dimensionalized by

multiplying the constant by RT.57 The TIC can then be calculated from these two

concentrations by,

312 10g g aq aqTIC C V C V Eq. S3

to get the total carbon in milligrams. Results from the TIC analysis are shown in Figure

S3.7.

The XPS results for the changes in oxygen-containing functional groups can be

used to estimate of the amount of carboxylic acid groups removed during

photodecarboxylation. This is accomplished by converting the atomic percentage of

carboxylic acids determined by XPS to the expected weight percentage before and after

irradiation. First, the total carbon and oxygen signal before and after irradiation is

103

determined by multiplying the atomic signal by the molecular weight of carbon or

oxygen,

% %

% %

16

12wt atom

wt atom

O O

C C

Eq. S4

The total carbon and oxygen results can be combined, then the amount of carbon from the

carboxylic acid groups is divided by this total to find the weight percentage of carbon

using the equation,

( ) %

( ) %% %

12COOH atom

COOH wtwt wt

CC

C O

Eq. S5

Multiplying this percentage by the concentration and volume of CNTs used in the study,

we can estimate the mass of CO2 expected to be evolved as a result of irradiation. For

example, if all 10 atom% of the surface oxygen detected on the control sample by XPS

Irradiation Time (hrs)0 3 6 9 12 15 18

To

tal I

no

rgan

ic C

arb

on

(g

)

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

dark controlirradiated

Figure S3.7 – CO2 measurements from irradiated and dark control samples performed at pH 7.

104

were carboxylic acid groups, that would correspond to 12.9 wt% of carbon from

carboxylic acids. If all of the carboxylic acid groups were lost due to

photodecarboxylation during irradiation, in the above experiment performed with 8mL of

a 5mg/L concentration O-MWCNT suspension that would translate into production of

5.2μg of CO2.

3.7.7 Oxic versus Anoxic Conditions

The amount of dissolved oxygen (DO) water can hold varies by temperature,

where oxygen saturated water is usually between 8-9mg/L between 20-25°C (EPA).Our

experiments were performed using nitrogen or oxygen as the primary dissolved gas

species. Besides absorbance and particle size measurements shown in Figure 3.4 of this

manuscript, the DO of the suspensions was monitored when comparing two experiments

at the same solution conditions. Plotted in Figure S3.8 we can see how at time zero, the

DO level is at its lowest point for nitrogen purged samples, and its highest point for

oxygen purged samples. As irradiation occurs the DO level changes, presumably due to

the generation of hydrogen ions and oxygen species as they are cleaved from the CNT

surface. The fits to both curves in Figure S3.8 illustrate that the DO level increases or

decreases to a certain point, approximately half the irradiation time it takes to reach the

final aggregation state (around 6hrs). From that point on, the DO level then remains

around an equilibrium point, ending around 3.3mg/L and 6.2mg/L for the nitrogen and

oxygen purged experiments respectively.

105

3.8 References

1. Hristozov, D.; Malsch, I., Hazards and Risks of Engineered Nanoparticles for the Environment and Human Health. Sustainability 2009, 1, (4), 1161-1194.

2. Lowry, G. V.; Casman, E. A., Nanomaterial Transport, Transformation, and Fate in the Environment. In Nanomaterials: Risks and Benefits, Linkov, I.; Steevens, J., Eds. Springer Netherlands: 2009; pp 125-137.

3. Metz, K. M.; Mangham, A. N.; Bierman, M. J.; Jin, S.; Hamers, R. J.; Pedersen, J. A., Engineered Nanomaterial Transformation under Oxidative Environmental Conditions: Development of an in vitro Biomimetic Assay. Environmental Science & Technology 2009, 43, (5), 1598-1604.

4. Wiesner, M. R.; Lowry, G. V.; Jones, K. L.; Hochella, J. M. F.; Di Giulio, R. T.; Casman, E.; Bernhardt, E. S., Decreasing Uncertainties in Assessing Environmental Exposure, Risk, and Ecological Implications of Nanomaterials. Environmental Science & Technology 2009, 43, (17), 6458-6462.

5. Nowack, B.; Ranville, J. F.; Diamond, S.; Gallego-Urrea, J. A.; Metcalfe, C.; Rose, J.; Horne, N.; Koelmans, A. A.; Klaine, S. J., Potential scenarios for nanomaterial release and subsequent alteration in the environment. Environmental Toxicology and Chemistry 2011, 31, (1), 50-59.


0 2 4 6 8 10 12 14

Dis

solv

ed O

xyg

en C

on

cen

trat

ion

(m

g/L

)

0

5

10

15

20

25

N2 purged

O2 purged

Figure S3.8 – The change in total dissolved oxygen plotted as a function of irradiation time measured from the same experiment as shown in Figure 6.

106

6. Petersen, E. J.; Zhang, L.; Mattison, N. T.; O'Carroll, D. M.; Whelton, A. J.; Uddin, N.; Nguyen, T.; Huang, Q.; Henry, T. B.; Holbrook, R. D.; Chen, K. L., Potential Release Pathways, Environmental Fate, And Ecological Risks of Carbon Nanotubes. Environmental Science & Technology 2011, 45, (23), 9837-9856.

7. Lowry, G. V.; Gregory, K. B.; Apte, S. C.; Lead, J. R., Transformations of Nanomaterials in the Environment. Environmental Science & Technology 2012, 46, (13), 6893-6899.

8. Batley, G. E.; Kirby, J. K.; McLaughlin, M. J., Fate and Risks of Nanomaterials in Aquatic and Terrestrial Environments. Accounts of Chemical Research 2013, 46, (3), 854-862.

9. Maurer-Jones, M. A.; Gunsolus, I. L.; Murphy, C. J.; Haynes, C. L., Toxicity of Engineered Nanoparticles in the Environment. Analytical Chemistry 2013, 85, (6), 3036-3049.

10. Mudunkotuwa, I. A.; Pettibone, J. M.; Grassian, V. H., Environmental Implications of Nanoparticle Aging in the Processing and Fate of Copper-Based Nanomaterials. Environmental Science & Technology 2012, 46, (13), 7001-7010.

11. Lee, J.; Cho, M.; Fortner, J. D.; Hughes, J. B.; Kim, J.-H., Transformation of Aggregated C60 in the Aqueous Phase by UV Irradiation. Environmental Science & Technology 2009, 43, (13), 4878-4883.

12. Hwang, Y. S.; Li, Q., Characterizing Photochemical Transformation of Aqueous nC60 under Environmentally Relevant Conditions. Environmental Science & Technology 2010, 44, (8), 3008-3013.

13. Hou, W.-C.; Jafvert, C. T., Photochemical Transformation of Aqueous C60 Clusters in Sunlight. Environmental Science & Technology 2009, 43, (2), 362-367.

14. Qu, X.; Hwang, Y. S.; Alvarez, P. J. J.; Bouchard, D.; Li, Q., UV Irradiation and Humic Acid Mediate Aggregation of Aqueous Fullerene (nC60) Nanoparticles. Environmental Science & Technology 2010, 44, (20), 7821-7826.

15. Savage, T.; Bhattacharya, S.; Sadanadan, B.; Gaillard, J.; Tritt, T. M.; Sun, Y.-P.; Wu, Y.; Nayak, S.; Car, R.; Marzari, N.; Ajayan, P. M.; Rao, A. M., Photoinduced oxidation of carbon nanotubes. Journal of Physics: Condensed Matter 2003, 15, (35), 5915.

16. Alvarez, N. T.; Kittrell, C.; Schmidt, H. K.; Hauge, R. H.; Engel, P. S.; Tour, J. M., Selective Photochemical Functionalization of Surfactant-Dispersed Single Wall Carbon Nanotubes in Water. Journal of the American Chemical Society 2008, 130, (43), 14227-14233.

107

17. Lee, S. H.; Jung, Y. C.; Kim, Y. A.; Muramatsu, H.; Teshima, K.; Oishi, S.; Endo, M., Optical spectroscopic studies of photochemically oxidized single-walled carbon nanotubes. Nanotechnology 2009, 20, (10), 105708.

18. Matsumoto, Y.; Koinuma, M.; Ida, S.; Hayami, S.; Taniguchi, T.; Hatakeyama, K.; Tateishi, H.; Watanabe, Y.; Amano, S., Photoreaction of Graphene Oxide Nanosheets in Water. The Journal of Physical Chemistry C 2011, 115, (39), 19280-19286.

19. Smirnov, V. A.; Arbuzov, A. A.; Shul'ga, Y. M.; Baskakov, S. A.; Martynenko, V. M.; Muradyan, V. E.; Kresova, E. I., Photoreduction of graphite oxide. High Energy Chemistry 2011, 45, (1), 57-61.

20. Yeh, T.-F.; Chan, F.-F.; Hsieh, C.-T.; Teng, H., Graphite Oxide with Different Oxygenated Levels for Hydrogen and Oxygen Production from Water under Illumination: The Band Positions of Graphite Oxide. The Journal of Physical Chemistry C 2011, 115, (45), 22587-22597.

21. Guardia, L.; Villar-Rodil, S.; Paredes, J. I.; Rozada, R.; Martínez-Alonso, A.; Tascón, J. M. D., UV light exposure of aqueous graphene oxide suspensions to promote their direct reduction, formation of graphene-metal nanoparticle hybrids and dye degradation. Carbon 2012, 50, (3), 1014-1024.

22. Koinuma, M.; Ogata, C.; Kamei, Y.; Hatakeyama, K.; Tateishi, H.; Watanabe, Y.; Taniguchi, T.; Gezuhara, K.; Hayami, S.; Funatsu, A.; Sakata, M.; Kuwahara, Y.; Kurihara, S.; Matsumoto, Y., Photochemical Engineering of Graphene Oxide Nanosheets. The Journal of Physical Chemistry C 2012, 116, (37), 19822-19827.

23. Stroyuk, A. L.; Andryushina, N. S.; Shcherban', N. D.; Il'in, V. G.; Efanov, V. S.; Yanchuk, I. B.; Kuchmii, S. Y.; Pokhodenko, V. D., Photochemical reduction of graphene oxide in colloidal solution. Theoretical and Experimental Chemistry 2012, 48, (1), 2-13.

24. Chen, C.-Y.; Jafvert, C. T., Photoreactivity of Carboxylated Single-Walled Carbon Nanotubes in Sunlight: Reactive Oxygen Species Production in Water. Environmental Science & Technology 2010, 44, (17), 6674-6679.

25. Chen, C.-Y.; Jafvert, C. T., The role of surface functionalization in the solar light-induced production of reactive oxygen species by single-walled carbon nanotubes in water. Carbon 2011, 49, (15), 5099-5106.

26. Hwang, Y. S.; Qu, X.; Li, Q., The role of photochemical transformations in the aggregation and deposition of carboxylated multiwall carbon nanotubes suspended in water. Carbon 2013, 55, 81-89.

27. Qu, X.; Alvarez, P. J. J.; Li, Q., Photochemical Transformation of Carboxylated Multiwalled Carbon Nanotubes: Role of Reactive Oxygen Species. Environmental Science & Technology 2013, 47, (24), 14080-14088.

108


29. Fogden, S. n.; Verdejo, R.; Cottam, B.; Shaffer, M., Purification of single walled carbon nanotubes: The problem with oxidation debris. Chemical Physics Letters 2008, 460, (1-3), 162-167.

30. Salzmann, C. G.; Llewellyn, S. A.; Tobias, G.; Ward, M. A. H.; Huh, Y.; Green, M. L. H., The Role of Carboxylated Carbonaceous Fragments in the Functionalization and Spectroscopy of a Single-Walled Carbon-Nanotube Material. Advanced Materials 2007, 19, (6), 883-887.

31. Hatchard, C. G.; Parker, C. A., A New Sensitive Chemical Actinometer. II. Potassium Ferrioxalate as a Standard Chemical Actinometer. Proc. R. Soc. Lond. A. 1956, 235, (1203), 518-536.

32. Langley, L. A.; Villanueva, D. E.; Fairbrother, D. H., Quantification of Surface Oxides on Carbonaceous Materials. Chemistry of Materials 2005, 18, (1), 169-178.

33. Wepasnick, K. A.; Smith, B. A.; Schrote, K. E.; Wilson, H. K.; Diegelmann, S. R.; Fairbrother, D. H., Surface and structural characterization of multi-walled carbon nanotubes following different oxidative treatments. Carbon 2011, 49, (1), 24-36.

34. Wang, X. S.; Liu, P.; Zheng, H. T.; Hu, H.; Zheng, W. J.; Suye, S. I., Preparation of Nicotinamide Adenine Dinucleotide Functionalized Multi-Walled Carbon Nanotube and its Application to Dehydrogenase Biosensor. Advanced Matieral Research 2011, 298, 121-127.

35. Anslyn, E. V.; Dougherty, D. A., Modern Physical Organic Chemistry. University Science Books: Sausalito, CA, 2006.

36. Noto, V. D.; Mecozzi, M., Determination of Seawater Salinity by Ultraviolet Spectroscopic Measurements. Applied Spectroscopy 1997, 51, (9), 1294-1302.

37. Smith, B.; Wepasnick, K.; Schrote, K. E.; Bertele, A. R.; Ball, W. P.; O'Melia, C.; Fairbrother, D. H., Colloidal Properties of Aqueous Suspensions of Acid-Treated, Multi-Walled Carbon Nanotubes. Environmental Science & Technology 2009, 43, (3), 819-825.

38. Smith, B.; Wepasnick, K.; Schrote, K. E.; Cho, H.-H.; Ball, W. P.; Fairbrother, D. H., Influence of Surface Oxides on the Colloidal Stability of Multi-Walled Carbon Nanotubes: A Structure-Property Relationship. Langmuir 2009, 25, (17), 9767-9776.

39. Cho, H.-H.; Wepasnick, K.; Smith, B. A.; Bangash, F. K.; Fairbrother, D. H.; Ball, W. P., Sorption of Aqueous Zn[II] and Cd[II] by Multiwall Carbon

109

Nanotubes: The Relative Roles of Oxygen-Containing Functional Groups and Graphenic Carbon. Langmuir 2009, 26, (2), 967-981.

40. Cho, M.; Chung, H.; Choi, W.; Yoon, J., Linear correlation between inactivation of E. coli and OH radical concentration in TiO2 photocatalytic disinfection. Water Research 2004, 38, (4), 1069-1077.

41. Almquist, C. B.; Biswas, P., A mechanistic approach to modeling the effect of dissolved oxygen in photo-oxidation reactions on titanium dioxide in aqueous systems. Chemical Engineering Science 2001, 56, (11), 3421-3430.

42. Haag, W. R.; Hoigné, J., Photo-sensitized oxidation in natural water via ·OH radicals. Chemosphere 1985, 14, (11-12), 1659-1671.

43. Song, W.; Ravindran, V.; Pirbazari, M., Process optimization using a kinetic model for the ultraviolet radiation-hydrogen peroxide decomposition of natural and synthetic organic compounds in groundwater. Chemical Engineering Science 2008, 63, (12), 3249-3270.

44. Martínez, L. J.; Scaiano, J. C., Transient Intermediates in the Laser Flash Photolysis of Ketoprofen in Aqueous Solutions: Unusual Photochemistry for the Benzophenone Chromophore. Journal of the American Chemical Society 1997, 119, (45), 11066-11070.

45. Xu, M.; Wan, P., Efficient photodecarboxylation of aroyl-substituted phenylacetic acids in aqueous solution: a general photochemical reaction. Chemical Communications 2000, (21), 2147-2148.

46. Ding, L.; Fang, W.-H., Exploring Photoinduced Decarboxylation Mechanism of o-Acetylphenylacetic Acid from the Combined CASSCF and DFT Studies. The Journal of Organic Chemistry 2010, 75, (5), 1630-1636.

47. McAuley, I.; Krogh, E.; Wan, P., Carbanion intermediates in the photodecarboxylation of benzannelated acetic acids in aqueous solution. Journal of the American Chemical Society 1988, 110, (2), 600-602.

48. Budac, D.; Wan, P., Photodecarboxylation: mechanism and synthetic utility. Journal of Photochemistry and Photobiology A: Chemistry 1992, 67, (2), 135-166.

49. Wepasnick, K. A.; Smith, B. A.; Bitter, J. L.; Fairbrother, D. H., Chemical and structural characterization of carbon nanotube surfaces. Analytical and Bioanalytical Chemistry 2010, 396, (3), 1003-1014.

50. Wang, X.; Huang, X.; Zuo, C.; Hu, H., Kinetics of quinoline degradation by O3/UV in aqueous phase. Chemosphere 2004, 55, (5), 733-741.

51. Chae, S.-R.; Watanabe, Y.; Wiesner, M. R., Comparative photochemical reactivity of spherical and tubular fullerene nanoparticles in water under ultraviolet (UV) irradiation. Water Research 2011, 45, (1), 308-314.

110

52. Kong, L.; Tedrow, O. N.; Chan, Y. F.; Zepp, R. G., Light-Initiated Transformations of Fullerenol in Aqueous Media. Environmental Science & Technology 2009, 43, (24), 9155-9160.

53. Hummers, W. S.; Offeman, R. E., Preparation of Graphitic Oxide. Journal of the American Chemical Society 1958, 80, (6), 1339-1339.

54. Dreyer, D. R.; Park, S.; Bielawski, C. W.; Ruoff, R. S., The chemistry of graphene oxide. Chemical Society Reviews 2010, 39, (1), 228-240.

55. TrojanUV UV Resources. http://www.trojanuv.com/uvresources (December 5),

56. Malcolm Pirnie, I.; Carollo Engineers, P. C.; The Cadmus Group, I.; Linden, K. G.; Malley, J. P., ULTRAVIOLET DISINFECTION GUIDANCE MANUAL FOR THE FINAL LONG TERM 2 ENHANCED SURFACE WATER TREATMENT RULE. In Agency, U. S. E. P., Ed. Washington, D.C., 2006.

57. Stumm, W.; Morgan, J. J., Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters. 3rd ed.; Wiley-Interscience: New York, 1996.

111

PART II: INTERACTIONS OF PARTICLES AND SURFACES IN AQUATIC ENVRIRONMENTS

112

Chapter 4:

Theory

4.1 Solution and Surface Chemistry

4.1.1 Solution Chemistry

Accurately determining the pH and ionic strength of the solution for each

experiment is necessary to properly understand how colloidal forces of the target particle

are affected by the medium. Knowing the concentration of each ion present in each

experimental solution allows for the calculation of a more accurate ionic strength, and

therefore a more correct estimate of the magnitude of forces acting on a particle. The

concentration and pH of the solution are directly related to the inverse Debye length, κ-1,

and surface potentials, ψ. CO2 saturated water has an equilibrium pH value of 5.8 at

25°C. One can determine the concentration of [H+] using the definition of pH as,

pHH 10 (4.1)

This can then be used to find [OH-] based on the dissociation equilibrium for H2O as,

OH HWk (4.2)

From here, the partial pressure of CO2, the pKa values for the different carbonic acid

species, and Henry’s law can be employed to determine the concentrations of all carbonic

acid species in water given by,

2

2

2

COCO

CO

PC

kh (4.3)

113

2 3 2 2H CO CO COC Kh C (4.4)

2 3

3

1a H CO

HCOH

K CC

C

(4.5)

3

23

2a HCO

COH

K CC

C

(4.6)

22 3 3 3

T H CO HCO COA C C C (4.7)

where PCO2 and khCO2 are the partial pressure and Henry’s law constant of CO2 at 25°C

respectively, KhCO2 is the hydration equilibrium constant for carbonic acid at 25°C, Ka1

and Ka2 are the dissociation constants for the diprotic carbonic acid species, and Cx are

the concentrations of the various ions in water.

Knowing the amount of salt, acid, or base added (in mol/L) and the total working

volume allows then for the calculation of the concentrations of K+, Na+, and Cl- in

solution by measuring the pH and conductivity of the newly made solution. One can then

establish the total concentration of the ions in solution by measuring the conductivity,

KM, and then using a salt dependent equation for conductivity given as,

Mall

T

KC

(4.8)

23 3

2 30 0

0

( , )

( , )

H HCO CO K OHT H CO KOH KOH

T T

Na ClNaCl NaCl

T

C C C C CC

C C

C CC

C

(4.9)

0.50 0 0( , ) ( )i iC A B C (4.10)

114

0 c c a az z (4.11)

where c and a stand for cation and anion respectively, Λ0 is the molar ionic conductivity

at infinite dilution for a given salt, z is the valence charge of the ion, λ is the molar ionic

conductivity for a specific ion, Λ(Ci, Λ0) is the concentration dependent molar ionic

conductivity of a particular species, A and B are constants determined by the ratio of

cations to anions, ΛT is the total calculated molar ionic conductivity for the given

solution, Ci indicates the concentrations of ionic species present in the solution, and Call is

the actual ionic strength of the solution determined from the measured value of KM. From

the total concentration of dissolved ions in solution we can determine the Debye

screening length (κ) for all experiments using,

0.5220 ( )

1...7

Ai i

m B

e Nz C

k T

i

(4.12)

where e0 is the elementary charge, NA is Avogadro’s number, εm is the dielectric

permittivity of the medium, kB is Boltzmann's constant, T is absolute temperature, zi is the

valence of a given ion, and Ci are the calculated concentrations of those ions.

4.1.2. Surface Chemistry

To accurately predict the net potentials for a given experiment, knowing the

solution chemistry is important, but one also needs to know how the prepared solution

affects the surface chemistry of the particles and surfaces of interest. Using a streaming

potential apparatus or measuring the zeta potential of a surface or particle, respectively, is

a traditional method for determining its surface chemistry. However, those have been

found to only be reliable as an estimate of the Stern layer potential. To model pH and

115

ionic strength dependent SiO2 surface potentials, literature measurements on quartz and

an associated amphoteric site binding model were found to accurately capture surface

potential values (ψ). By fitting these results with a convenient expression as,

( , ) ( ) ( ) exp[ ( ) ]pH C I C x C y C pH (4.13)

where the units on all constants are either mV, pH, or M where appropriate. The

measured pH and conductivity of the experimental solutions allows the derivation of a

pH and concentration dependent surface potential for silica particles and surfaces.

4.2 Colloidal and Surface Interactions of Spherical Particles

4.2.1 Net Potential Energy Interactions

The net particle-surface interaction potential for spherical colloidal particles of

diameter 2a is equal to the sum of gravitational body forces, colloidal, and surface forces

in aqueous solutions as given by,

( , ) ( , ) ( ) ( , ) ( )NETpw Epw G VDWpw Spwu h C u h C u h u h C u h (4.14)

where h is the surface-to-surface separation between a given particle and the wall and C

is the total ionic concentration of the solution. The subscripts E, G, VDW, and S stand for

electrostatic repulsion, gravitational body force, van der Waals attraction, and steric

repulsion, respectfully. This net potential is obtained by measuring a statistically

significant number of height excursions and creating a normalized equilibrium height

histogram, p(h). This histogram can be related to the height-dependent net particle-

surface interaction potential, uNETpw(h), via Boltzmann’s equation as,

116

lnNETpw Bu h k T p h (4.15)

where kB is Boltzmann's constant and T is absolute temperature.

The net potential, as well as the diffusivity landscape, D(h), can also be obtained

by measuring the time-dependent probability of heights, p(h,t), and using these to solve

the Smoluchowski equation given by,1

, , ,p h t p h t p h t dU h

D ht h h kT dh

(4.16)

This equation reduces to Boltzmann’s equation at steady state conditions. The diffusion

of spherical particles will be discussed in later sections.

4.2.2 Gravitational Body Force

Gravitational body effects are a result of a given particle’s buoyant weight, G,

multiplied by its relative height above the wall given by

( )Gu h Gh mgh (4.17)

34( )

3 p mG a g (4.18)

where a is the particle radius, ρp and ρm are the densities of the particle and medium

respectively, and g is acceleration due to gravity.

4.2.3 Electrostatic Repulsion

Electrostatic interactions result from the overlapping of electric double layers

(EDLs), which are defined as charged regions of ion density at or near a surface. In this

work, these double layers are present when a system contains ions in solution, consisting

117

of the intrinsic charge on a surface and a diffuse layer of counterions in the medium that

interact with the ions present on the surface. The resulting EDL is a direct result of a

particle or wall’s surface charge (σ) and surface potential (ψ), and they induce

electrostatic repulsion as the two surfaces approach one another. This repulsive

electrostatic interaction between the EDLs for a constant potential (ψ) can be well

modeled as an exponential function, and is expressed as

( , ) exp[ ( ) ]Epw pwu h C B C h (4.19)

2

0 00

( , )64 tanh tanh

4 4p wB

pw mB B

pH Ck TB a ze ze

ze k T k T

(4.20)

where κ is the Debye screening length, and B takes into account the pH and ionic strength

dependent surface potentials (ψp and ψw), which were discussed in Section 4.1.1.

The two surface potentials are considered to be equal in the case of individual

silica spheres interacting with silica surfaces as discussed in Chapter 6, but may in fact be

different when the particle and the surface are not composed of the same material (as in

Chapters 7 and 8).

The expression for electrostatic repulsion between a spherical particle and a flat

surface in Eq. 4.19 is actually based off of the work by Russian scientist Boris Derjaguin

published in 1934.2 The actual expression listed above is the result from the linearization

of a more complicated differential equation, but it is well suited for situations where the

particle-surface separation is small compared to the radius of the sphere.3 However, under

different circumstances, the linearized Poisson-Boltzmann theory (or linear superposition

approximation, LSA) is better suited. Scenarios detailing when LSA would be more

118

appropriate can be found in the Section 4.4.

4.2.4 The Derjaguin Approximation

Derjaguin’s work2 related the interaction of finite sized spherical bodies to the

overlapping double layers on semi-infinite planar surfaces. This geometric calculation

simplified the curved surface of a sphere into a series of ring-shaped flat discs. This series

of “stepped” surfaces with planar geometries can then be integrated over the series to

determine the potential over the entire curved surface area of interaction. This

approximation is useful for determining the interaction of two spherical bodies, as well as

a spherical body interacting with a flat surface. However, the Derjaguin approximation is

limited to the regime where the radius of the particle is far larger than the particle-surface

separation. In the case of interest here, a sphere greater than 1μm interacts with a flat

plate, where the plate is modeled as a sphere with a diameter much greater than the first

sphere (on the order of 106μm) so that it appears infinitely large in comparison.

The Derjaguin approximation states that the interaction potential between two

spherical particles of equal radii, or between a particle and surface (wall), can be obtained

from the energy per unit area of interaction between two flat plates, EXww(l), at a function

of separation, l. For our purpose we will examine only the particle-wall interaction,

uXpw(z), which is given by4

2pw wwX X

z

u z a E l dl

(4.21)

The electrostatic potential for the interaction of two charged surfaces is related to

the force between a charged spherical particle and a charged flat wall as,

119

2pw wwE EF z aE z (4.22)

where FEpw is the electrostatic force experienced between the particle and the wall,

respectively, and EEww is the electrostatic potential between two planar walls. EE

ww is

given by,

1 1 264 tanh tanh exp4 4

wwE B b

ze zeE l k Tn l

kT kT

(4.23)

where is given as,

1 22 22 b m Be Z n k T (4.24)

Force and potential energy are related by,

pw pwE E

dF z u z

dz (4.25)

where uEpw is the electrostatic interaction between a charged particle and wall surfaces,

given by,

2 exppwEu z B z a (4.26)

where B is a constant that takes into account the particle radius (a) and the surface

potentials (ψ) on the particle and wall. From here, we can write

exppw pwE E

du z B z a u z

dz (4.27)

This now allows us to relate our original equation for the electrostatic potential defining

the interaction of two charged surfaces with our definition for the force between a particle

and a wall in terms of potentials as,

120

2pw wwE E

du z aE z

dz (4.28)

We can insert Equations 4.23 and 4.27 into the equation above since we have the

definitions for each side. This yields,

2 1 2128 tanh tanh exp4 4

pwE B b

B B

ze zeu z ak Tn z a

k T k T

(4.29)

If we were to now substitute the equation for in and simplify, we obtain,

2

1 264 tanh tanh exp4 4

pw BE m

B B

k T ze zeu z a z a

ez k T k T

(4.30)

Using our expression for uEpw, we can pull out the definition of B as,

2

1 232 tanh tanh4 4

Bm

B B

k T ze zeB a

ez k T k T

(4.31)

The Derjaguin equation is used to describe the electrostatic repulsion in the

system discussed in Chapter 6 because it follows the parameters for applicability: the

range of interaction (surface to surface separation) is far less than the radii of the sphere.

The silica particles used were 2.1μm in diameter, whereas the separation distance was on

the order of zero to 0.4μm.

4.2.5 van der Waals Attraction

The attractive forces resulting from van der Waals forces arise from quantum

electrodynamics. Every molecule carries either a dipole moment or has the ability to

spontaneously form an instantaneous dipole, which results from the position of electrons

in an atom or molecule at a given moment in time. Two dipoles can induce an attractive

121

force between the two atoms or molecules. These attractive forces between atoms can be

extrapolated to explain the long-range interactions between a particle and a wall through

Lifshitz theory.

Lifshitz theory ignores the effects of individual atoms composing a material and

instead examines the material as a continuum. This takes into consideration how the

intrinsic bulk properties of a material and the medium comprising the system, such as the

dielectric constant, affect the overall charge or potential of a given surface, thereby

making the interaction attractive or repulsive. The distance that separates two objects in a

medium governs how strong or weak this force will be. For a spherical particle

interacting with a planar surface we use Lifshitz theory with a Derjaguin geometric

correction, as was discussed in Section 4.2.4. Application of Lifshitz theory enables us to

take into account retardation (reduction of attractive energy from slowed propagation of

radiation through a material or medium)5 and screening (effect of ions that emerge with

increasing ionic strengths) on the Hamaker constant.

The Hamaker constant is a proportionality constant that is related to the number

of atoms in the interacting bodies. This constant represents the strength of van der Waals

interactions. However, when we are talking about macroscopic surfaces and media, the

constant becomes defined by an analytical expression for the Hamaker function, Avdw, for

two half spaces with a distance h in separation. This function can be broken into two

terms,

inf 0( , ) ( ) ( , )VDW salt nA h C A h A h C (4.32)

122

The first term is the Hamaker function at infinite salt concentration, Ainfsalt that is

calculated as,6

inf1

3( ) [1 ] ln[1 ]

2 n

x xpm wmsalt B r pm wm

n

A h k T x e e dx

(4.33)

j k k jjk

j k k j

s s

s s

(4.34)

j k k jjk

j k k j

s s

s s

(4.35)

2 2 22

2( ) ( )n

k k

ls x

c

(4.36)

22 nn

lr

c

(4.37)

( )k k ni (4.38)

n

nkT

(4.38)

where jk is a function of the dielectric properties of the materials composing the particle

(p), wall (w), and medium (m). The analytical solution for the Hamaker function at

infinite salt concentration can then be fit with a rational expression. The second term that

accounts for screening effects is the zero frequency term of the Hamaker function, An=0,

which is calculated as,

0 0( , ) 0.5 [1 2 ( ) ]exp[ 2 ( ) ]nA h C A C h C h (4.39)

123

which uses the concentration dependent Debye screening length, , and Hamaker

constant, A0. The particle-wall van der Waals interaction potential then becomes a

function of separation and ionic strength that is calculated as,

610

2

( , )( , )

6

mp w VDWBVDWpw h

p w

a a A l Ck Tu h C dl

a a l

(4.40)

where ap and aw are the radii of the particle and wall respectively, and aw>>ap.

4.2.6 Steric Repulsion

Steric potentials are oftentimes used to describe repulsive forces caused by

polymers, polyelectrolytes, or macromolecules that are adhered or absorbed to a surface.

These polymers often form what is called a brush architecture, where at high enough

concentrations, the attractive portion of the polymer pack around the surface and

repulsive portions are forced to extend out into the surrounding medium, forming a

brush-like structure. Steric repulsive forces are very short-ranged, and arise only as the

two macromolecular coated surfaces approach one another and the outer edges of the

layers begin to overlap. The interaction of these surfaces depends on the change in free

energy of the two macromolecular layers under compression and penetration. The brush

architecture of a polymer is defined by two important variables: its thickness and free

energy. These variables will change in value depending on the type and chemical

structure of a polymer. For a polymer layer with a brush architecture on a planar surface

of an uncompressed thickness, δ0, and free energy per area, f0, the compressed free

energy per area, f(δ), for 1/2<δ/δ0<1 can be captured accurately by,7

0 0( ) 1 expB Bf f (4.41)

124

where ΓB and γB are dimensionless constants specific to the brush architecture of a

particular polymeric material. Using this expression in the Derjaguin approximation (see

Section 4.2.4), the steric potential between two symmetric macromolecular brush layers

as can be obtained as,

, 0 0 0( ) 16 exp 2S B B Bu h af h (4.42)

This equation is versatile because it can be used to obtain the steric repulsive

force for a variety of materials that exhibit different interfacial macromolecular

architectures with different decaying density profiles at their periphery, by applying

different values of Γ and γ in Eq. 4.41. Eq. 4.42 can be used to accurately model

macromolecular layer repulsion for a broad range of δ0, f0, Γ, and γ using a general

repulsive steric potential of the form,

( ) expSpwu h h (4.43)

This equation should be broadly applicable by lumping unknown constants together. This

expression is particularly useful when the uncompressed layer properties and architecture

being studied are not well characterized.

4.3 Diffusion Modes of Spherical Particles

4.3.1 Diffusion of Spheres near a Flat Surface

The Stokes-Einstein equation is used to explain the Brownian diffusion of an

unbounded spherical particle in a bulk viscous fluid which extends out infinitely in all

directions. The diffusion (D) is the inverse of the drag coefficient given by,

125

0 6Bk T

Da

(4.44)

where is the fluid medium viscosity. A sphere would be modeled simply with Stokes-

Einstein in ideal circumstances. However, when a spherical particle diffuses close to a

wall the diffusion slows as it approaches the surface. This occurs because as a particle

approaches a wall, the liquid separating the two objects must be removed from between

them for the surfaces to come into contact. This takes energy and induces a drag force on

the particle. Now the particle diffusion becomes a function of the sphere’s height above a

planar surface as,

0( ) ( )D h D f h

(4.45)

where f(h) is a factor used to account for the particle-wall hydrodynamic interactions as

described by Brenner.8 His expression describes how the motion of the sphere is slowed

as it approaches a surface, which can be accurately captured by the more simple

expression,

2

2 2

6 2( )

6 9 2

h ahf h

h ah a

(4.46)

from Bevan and Prieve.9

The translational diffusion of a sphere through the bulk medium can be captured

by the mean squared displacement (MSD), X2. For a particle moving in one dimension (x

or y directions) the MSD is given by,

2 22X D t x t

(4.47)

where Δx is simply the change in the sphere’s position during an increment of time as,

126

2

02

0

i

i

x xx t

t t

(4.48)

For spheres that may sample a variety of heights while diffusing, D(h) can be

used in the equation for MSD by obtaining an average height that a given particle

samples over a course of time. This average can be calculated by integrating over a height

range such as,

0

0

( ) ( )

( )

( )

n

avg n

D h p h dh

D h

p h dh

(4.49)

where p(h) is the probability distribution of heights sampled and is obtained from the net

potential energy defined as,

( )

( ) exp NETpw

B

u hp h

k T

(4.50)

which is simply the inverse of Eq. 4.15.

4.3.2 Diffusion of Spheres through Obstacles

Calculation of the diffusion of spheres over a flat surface hindered by spherical

obstacles can take advantage of the translational diffusion equations discussed in Section

4.3.1. However, to be accurate they must include factors that account for the size and

number of asperities in the field that the diffusing sphere may come into contact with.

Saxton10 suggested that the MSD for the lateral diffusion of a sphere through obstacles in

two dimensions follows the form,

127

2 2

0

4 2 ( , )x t Dt xdxC x t x

(4.51)

for a given point in time, t, where D is a function of the concentration of obstacles at a

particular position and time, C(x,t). These two variables are related by,

21( , ) exp

4 4

xC x t

Dt Dt

(4.52)

Kim and Torquato11 took a similar approach using an “exclusion” probability

function, Eυ(r), and a “void” nearest-neighbor probability density, Hυ(r). These two

variables are defined as the fraction of space available for exploration in a field of

spherical obstacles of density, ρ, by a particle, and the probability that an arbitrary point

in the system lies near a spherical obstacle between r and r + dr, respectively where r is

the radius of a spherical empty space in the grid. Eυ(r) and Hυ(r) are used in the

calculation of the volume fraction, ϕ, and specific surface, s, respectively. From here they

suggest a calculation for the effective diffusion coefficient, De, for a Brownian particle of

radius b in a system with spherical obstacles of radius a as De[ϕ(ρ,a);b]. They relate the

De in fluid saturated porous media to the MSD by,

2

26 ( )e

XD

X

(4.53)

where X2 is the MSD and τ(X2) is the average time for a Brownian particle to hit a surface

for the first time.

128

4.4 Colloidal and Surface Interactions of Rod-Shaped Particles

4.4.1 Net Potential Energy Interactions

At relatively low ionic strength conditions where the particle sits at a height above

the surface sufficiently far away so that van der Waals attraction is not contributing to the

net potential (>10nm), the net particle-surface interaction potential for a rod-shaped

particle over a planar surface, u(h), is the sum of the gravitational attractive forces and

electrostatic repulsion. Theoretical models of this potential energy can be calculated by

the addition of the above mentioned contributing potentials as,

( ) ( ) ( )NETpw G Epwu h u h u h (4.54)

where the subscripts G and E refer to the gravitational and electrostatic interactions

respectively, and h is the particle surface to planar surface separation distance. The

gravitational potential is associated with a body force governed by the size and shape of

the particle, whereas the electrostatic potential is associated with colloidal surface forces.

However, both forces are intimately dependent on the length of the rod. This differs from

spherical particles where all colloidal interactions of rod-shaped particles are derived per

unit length.

4.4.2 Gravitational Body Forces

The gravitational potential energy of each rod depends on its separation from the

wall, h, multiplied by its buoyant weight, G, as given by,

( )Gu h Gh mgh (4.55)

2 ( )p fG a L g (4.56)

129

where m is buoyant mass, g is acceleration due to gravity, a is the particle radius, L is the

particle length, and ρp and ρf are particle and fluid densities respectively. The geometric

parameter (πa2L) corresponds to a rectangular cylinder.

4.4.3 Electrostatic Repulsion derived from the Derjaguin Approximation

As mentioned above in Section 4.2.3, the Derjaguin approximation is most

applicable when the radius of the particle is far greater than the particle-surface

separation. It is also used to describe the electrostatic potential when double layers on a

rod and planar surface are sufficiently thin (κa>>1), corresponding to higher ionic

strength conditions. Therefore, the Derjaguin approximation can be used in conjunction

with the superposition, non-linear Poisson-Boltzmann equation for a 1:1 monovalent

electrolyte3 to give the rod-wall potential as,4

exp ( )2E

au h LB h

(4.57)

2

64 tanh tanh4 4

p wBm

B B

e ek TB

e k T k T

(4.58)

1 22

2( )Ai i

im B

e Nz C

k T

(4.59)

where κ is the Debye screening length, εm is the solvent dielectric constant which is the

product of permittivity in a vacuum (ε0) and the relative permittivity of water (εw), e is the

elemental charge, ψp and ψw are the surface potentials of the particle and the wall,

respectively, NA is Avogadro's number, Ci is the electrolyte molarity of species i, and zi is

the ion valence.

130

The height of a rod-shaped particle above a planar surface can be determined by

taking the derivative of the net potential and solving for h, as

1 lnABL

hG

(4.60)

where L is the length, G is the prefactor from the equation for the gravitational potential,

B is the prefactor from electrostatic repulsion, and A is a constant that contains geometric

corrections. This would predict the most ideal height of the particle. However, if the

particle samples many heights, an average height can be derived by integrating over the

probability distribution of heights sampled from Equation 4.50, similar to the average

diffusion coefficient,

0

0

( )

( )

n

avg n

hp h dh

h

p h dh

(4.61)

4.4.4 Electrostatic Repulsion from the Linear Superposition Approximation

For thick double layers (κa~1) the above expression will generally over-estimate

the interaction since the Derjaguin approximation no longer holds. We instead model the

rod as a rigid chain of touching spheres and approximate the electrostatic interaction

between the rod and wall by summing up all sphere-wall interactions based on the linear

superposition approximation (LSA) for thick double layers,

3,

1

4(z) ( )

3

p

G sphere s f ii

U a gz

(4.62)

and the electrostatic repulsion contribution given as,

131

,

1

(z) exp( ( ))p

E sphere ii

U B z a

(4.63)

where zi is the mass center for ith particle composing the rod, and B is the prefactor for the

electrostatic repulsive interactions based on the linear superposition approximation

(LSA), which can be used for thick double layers, given as,

2 22

42 4

kT a eB

e h a kT

(4.64)

4.5 Diffusion Modes of Rod-Shaped Particles

4.5.1 Diffusion of Rods in the Bulk

The same string of beads model that was used to determine the electrostatic

repulsion from LSA was used to calculate theoretical equations for the free diffusion of a

rod-shaped particle in the bulk, DB(p). These diffusion equations were calculated for rods

of varying aspect ratios (p), which is the length (L) divided by the radius (a), by adding

beads of radius, a, to the model. These equations consist of a Stokes-Einstein type

diffusion coefficient, D0, multiplied by a factor dependent on the aspect ratio of the

particle, f(p), where the bulk translational diffusion coefficient parallel to the long axis is

given by,

|| 0|| ||( ) ( )BD p D f p (4.65)

0|| 2Bk T

Dpd

(4.66)

2

|| 2

0.4536 1.772 41.5( ) ln( )

34.38 18.96

p pf p p

p p

(4.67)

132

and the translational bulk diffusion coefficient perpendicular to the long axis is given by,

0( ) ( )BD p D f p (4.68)

0 4Bk T

Dpd (4.69)

2

2

0.3604 28.36 72.63( ) ln( )

36.29 34.9

p pf p p

p p

(4.70)

where η is the fluid medium viscosity, and p is the particle aspect ratio where L is the

length and d the diameter.

A 2-dimensional (2-D) translational diffusion coefficient for the motion of a rod

in bulk medium can be established by combining the parallel and perpendicular modes to

achieve,

|| ( ) ( )( )

2B B

T

D p D pD p

(4.71)

A set of theoretical equations can be similarly derived for the rotational motion of

a rod-shaped particle freely diffusing in the bulk, DB(p), as given by,

0( ) ( )BR R RD p D f p (4.72)

0 3

3

( )B

R

k TD

pd (4.73)

3 2

3 2

1.373 19.39 148.1 265.2( ) ln( )

56.43 54.35 268.4R

p p pf p p

p p p

(4.74)

133

4.5.2 Diffusion of Rods near a Flat Surface

When a rod diffuses in the bulk it is only dependent on the rod length determined

by p. However, when a rod diffuses near a surface, the surface can influence the rod

diffusion depending on how close the two objects are to one another. This in turn makes

the diffusion coefficient for a rod diffusing near a surface dependent on the distance or

height between the rod and the surface, where the particle would be expected to slow as it

approaches the surface, similar to Brenner’s theory from Section 4.2.7.

Translational diffusion coefficient can be obtained for the motion of a rod of

aspect ratio, p, as a function of its particle surface-wall separation, h, above a planar

surface by multiplying the bulk diffusion coefficient, DB(p), by a correction factor

dependent on h and another correction factor dependent on p. The motion parallel to its

long axis takes the form,

|| || || ||( , ) ( ) ( )BD p h D f h g p (4.75)

3 2

|| 3 2

0.9909 0.3907 0.1832 0.001815( )

2.03 0.3874 0.07533

z a z a z af h

z a z a z a

(4.76)

|| ( ) 1.1669 0.0091g p p (4.77)

and the translational diffusion coefficient for the motion perpendicular to its long axis is,

( , ) ( ) ( )BD p h D f h g p (4.78)

3 2

3 2

0.9888 0.788 0.207 0.004766( )

3.195 0.09612 0.1523

z a z a z af h

z a z a z a

(4.79)

134

|| ( ) 1.2239 0.0120g p p (4.80)

Similar to the translational motion in the bulk, a 2-dimensional translational

diffusion coefficient for the motion of a rod near a flat surface can be established by

combining the parallel and perpendicular modes to achieve,

|| ( , ) ( , )( , )

2T

D p h D p hD p h

(4.81)

Subsequently, the rotational diffusion coefficient for a given rod-shaped particle

as function of its aspect ratio and height above the planar surface can be obtained the

same way and is given by,

( , ) ( ) ( )R BR R RD p h D f h g p (4.82)

3 2

3 2

0.998 131.1 21.25 0.01275( )

128.7 121.1 2.897R

z a z a z af h

z a z a z a

(4.83)

( ) 1.154 0.0096Rg p p (4.84)

4.5.3 Diffusion of Rods between Two Parallel Surfaces

A second set of translational diffusion coefficients can be obtained for a scenario

where a rod of aspect ratio, p, is diffusing between two parallel plates with separation, Δ.

An approximate solution for these diffusion coefficients can be calculated from the linear

superposition approximation12 in the form,

11 1

|| || || || ||( , ) ( ) ( ) (( ) 2 ) 1BD p h D f h g p f a z a

(4.85)

11 1

( , ) ( ) ( ) (( ) 2 ) 1BD p h D f h g p f a z a

(4.86)

135

for the parallel and perpendicular translational diffusion coefficients, respectively. This

can also be applied to the rotational diffusion a rod of between two parallel plates given

as,

11 1

( , ) ( ) ( ) (( ) 2 ) 1R BR R R RD p h D f h g p f a z a (4.87)

This approximation takes into account the hydrodynamic effects from the top and bottom

walls.

Just as was discussed in the second half of Section 4.3.1, the mean squared

displacement (MSD) is a useful tool for tracking the translational diffusion of rods.

Equations 4.47 and 4.48 from Section 4.2.7 can be applied to track the center of mass

position of a rod to help understand the translational motion as a function of time.

Another useful metric is the mean squared angular displacement (MSθ), which tracks the

total rotation of a particle as a function of time. This can be calculated by monitoring the

position of the end points of the rod in relation to the center of mass. The angular position

of each end point on the rod with respect to the center of mass, i, can be calculated as,

arctan( )cm ii

cm i

y y

x x

(4.88)

where xcm/ycm is the x or y center of mass position, and xi/yi is the position of one end of

the rod at some time point i. Values of θ can be implemented into Eq. 4.47 in place of

steps in the x or y direction to obtain the MSθ as,

2MS 2D t t

(4.89)

136

4.5.4 Diffusion of Rods through Obstacles

As discussed in Section 4.3.2, the theoretical expressions for the diffusion of

spheres through spherical obstacles can be extrapolated to the movement of rod-shaped

particles through a field of obstructions. Similar variables will be necessary to describe

the field including a volume or area fraction of spherical obstacles, ϕ, which can be used

as a dimensionalized variable in place of a concentration in Saxton’s equation. A factor

that would replace the spherical geometry with a cylindrical shape would be the first step

in the calculation. The next step would be to know the center of mass and rod end point

coordinates at a given point in time to allow for the calculation of the MSD or MSθ,

which can then be directly related to an effective diffusion coefficient. This De can also

be compared to the diffusion coefficient calculated for rod-shaped particles under the

same conditions without the presence of physical obstacles that may induce

hydrodynamic hindrances. With these pieces, one could produce a reliable estimation of

the diffusion of rods through obstacles of varying concentration densities.

However, like Saxton discusses, the volume or area fraction of the obstacles is

important for explaining the MSD. Above a given ϕ, called the percolation threshold, a

particle can become trapped in a small space between clusters of obstacles. In this

instance the diffusion of the particle may appear faster than a particle that is diffusing

more freely; this can be the result of the trapped particle’s diffusion appearing more one

dimensional (1-D) in nature as opposed to 2-D. In that case, analyzing the particle’s

trajectory as 2-D motion would falsely give the particle a faster displacement. In these

cases, one must consider that as the area fraction of obstacles increases, there is the

likelihood that a particle’s motion is not strictly 1-D or 2-D, but a fractal dimension

137

between these two. Thus, a parameter indicating the dimension describing the particle

motion may also be employed to accurately capture the diffusion of rods through spheres.

4.6 References

1. Murphy, T. J.; Aguirre, J. L., Brownian Motion of N Interacting Particles.1. Extension of Einstein Diffusion Relation to N-Particle Case. J. Chem. Phys. 1972, 57, (5), 2098.

2. Derjaguin, B., Untersuchungen über die Reibung und Adhäsion, IV. Kolloid-Zeitschrift 1934, 69, (2), 155-164.

3. Russel, W. B.; Saville, D. A.; Schowalter, W. R., Colloidal Dispersions. Cambridge University Press: New York, 1989.

4. Israelachvilli, J., Intermolecular and Surface Forces. 3rd ed.; Academic Press: New York, 2011.

5. Bevan, M. A.; Prieve, D. C., Direct Measurement of Retarded van der Waals Attraction. Langmuir 1999, 15, (23), 7925-7936.

6. Prieve, D. C.; Russel, W. B., Simplified predictions of Hamaker constants from Lifshitz theory. Journal of Colloid and Interface Science 1988, 125, (1), 1-13.

7. Eichmann, S. L.; Meric, G.; Swavola, J. C.; Bevan, M. A., Diffusing Colloidal Probes of Protein-Carbohydrate Interactions. Langmuir 2013.

8. Brenner, H., The Slow Motion of a Sphere Through a Viscous Fluid Towards a Plane Surface. Chem. Eng. Sci. 1961, 16, (3-4), 242-251.

9. Bevan, M. A.; Prieve, D. C., Hindered diffusion of colloidal particles very near to a wall: Revisited. The Journal of Chemical Physics 2000, 113, (3), 1228-1236.

10. Saxton, M. J., Lateral diffusion in an archipelago. Single-particle diffusion. Biophysical Journal 1993, 64, (6), 1766-1780.

11. Kim, I. C.; Torquato, S., Diffusion of finite-sized Brownian particles in porous media. The Journal of Chemical Physics 1992, 96, (2), 1498-1503.

12. Happel, J., Low Reynolds number hydrodynamics. In Brenner, H., Ed. Prentice-Hall: Englewood Cliffs, N.J., 1965.

138

Chapter 5:

Experimental Set-Up and Parameters for Microscopy Studies

5.1 Chemicals and Materials

5.1.1 Chemicals

Acetone and isopropanol were purchased from Fisher Scientific (Pittsburgh, PA,

USA) and used to clean microscope slides. Sulfuric acid and hydrochloric acid were

purchased from Sigma-Aldrich (St. Louis, MO, USA). Sulfuric acid was mixed with

Nochromix powder (Godax Labs, Takoma Park, MD, USA) to create an oxidizing

solution used to remove organic impurities from the surfaces of all microscope slides and

coverslips. Hydrochloric acid was used to treat the slides as well as adjust the pH of the

solutions. All chemicals were used as received and without purification.

Potassium hydroxide and sodium chloride were also purchased from Fisher

Scientific and used without purification. Potassium hydroxide was used to treat

microscope surfaces prior to experiments and adjust the pH of solutions. Sodium chloride

solutions were used to set the ionic strength of a given experiment to the desired

conditions.

5.1.2 Materials

Plain glass microscope slides from Fisher Scientific had a manufacturer reported

density of 2.48 grams/cm3 and a soda lime composition of approximately 72% silicon

dioxide, 14% sodium oxide, 6% calcium oxide, 4% magnesium oxide, 1% aluminum

oxide, 1% potassium oxide, <1% other trace elements. The microscope slides were first

wiped clean with lens paper and then sonicated in a Branson Ultrasonics Corporation

139

1510 ultrasonicator (Danbury, CT, USA) for 30min in acetone and 30min in isopropanol.

The slides were then rinsed thoroughly with deionized (DI) water (18.3MΩ) from a Milli-

Q Academic ultrafiltration system (Milli-Pore, Danvers, MA, USA) using a Millipak 20

Express filter (0.22μm). To remove impurities from the glass surface, the slides were

soaked in Nochromix overnight. In the morning, slides were rinsed thoroughly with DI

water, then soaked in 0.1M KOH for 30min prior to use. Slides were again rinsed with

deionized water and dried with nitrogen (Airgas, Salem, NH, USA) before use.

Glass coverslips (long: 24cm x 60cm; small: 18cm x 18cm) were purchased from

Corning (Corning, NY, USA). The long coverslips were cleaned in the same manner as

the microscope slides for all experiments where rod-wall interactions were being

measured (Chapter 7). For porous media experiments (Chapter 8) these slides were

soaked in acidic (pH = 1) water after soaking in Nochromix overnight. Small coverslips

were wiped with lens paper purchased from Fisher Scientific and place directly into the

Nochromix. Upon removal they were rinsed with DI water and soaked in 0.1M KOH

solution for 30min, rinsed again, and then dried with high purity nitrogen from Airgas

(Salem, NH, USA).

Viton O-rings (5mm ID) were purchased from McMaster Carr (Robbinsville, NJ,

USA) and used to contain colloidal samples in one-wall systems. Vacuum grease

purchased from Dow Corning (Midland, MI, USA) and Loctite professional heavy duty

epoxy from Henkel Consumer Adhesives (Avon, OH, USA) were used to seal sample

cells, to simultaneously contain the colloidal suspension being examined, and to prevent

convection and dust from entering the sample cell. Lens paper was used to wipe away

excess dust from slides and coverslips before cleaning, and also to wick away liquid from

140

the edges of the small coverslip that created the top of a confined cell unit.

Refractive index matching oil (n=1.515) from Cargille (Cedar Grove, NJ, USA)

was used to couple the microscope slide/coverslip cells to a 68º dovetail prism from Red

Optronics (Mountain View, CA, USA). The prism was used to create the evanescent

wave for use in total internal reflection microscopy (TIRM) experiments (Chapter 6).

5.2 Colloids

5.2.1 Silica Microspheres

Colloidal SiO2 (nominal diameter of 2.34 microns) were purchased from

Bangs Laboratories (Fishers, IN, USA) and used without further purification. The non‐

porous amorphous SiO2 colloids are synthesized by a precipitation method using pure

tetraethyl orthosilicate and reagents with minimal trace elements. Colloidal SiO2

dispersions used for TIRM experiments were prepared by diluting 0.7μL of the

manufacturer stock dispersion (10% solids in ethanol) into 1mL of a solution which had

been set to the desired pH and ionic strength. These solutions were made by diluting

concentrated stocks of NaCl (4M), KOH (1M), and HCl (1M). Each day before use, the

stock solutions were measured using an Accument AR20 pH and conductivity probe from

Fisher Scientific to obtain an accurate concentration of each salt. From here, a calculated

volume of each salt, based on the stock concentration, was dispensed and diluted

accordingly to create the solutions needed for experiments. After the final addition of the

silica, this suspension was sonicated for 15min, then diluted 100 with the appropriate

solution and sonicated again before use.

141

For experiments where the silica microspheres were used as spacer particles a

diluted stock was prepared. For experiments where silica was sparsely implemented to

keep confining walls at a set distance (Chapter 7), the dispersion consisted of 0.5μL of

silica stock diluted in 4mL of DI water. For porous media experiments (Chapter 8), a

silica dispersion was created by adding 100μL of the stock to a 1mL total volume of

0.1mM NaCl solution.

5.2.2 Gold Rods

Gold rods were synthesized at Pennsylvania State University (State College, PA,

USA) using an electrochemical deposition process. This process was adapted from

several deposition methods.1, 2 An anodic aluminum oxide (AAO) membrane from

Whatman Inc. (purchased from Sigma Aldrich) was coated on one side with a thin film of

silver using a Kurt Lesker (Jefferson Hills, PA, USA) Lab-18 electron beam evaporator

before deposition to increase conductivity. A two-electrode system was used for

deposition, the membrane serving as the working electrode and platinum as the reference.

A sacrificial layer of silver was first deposited into the pores of the membrane, then gold

was added on top by electrochemically growing the rods to a prescribed length using a

current density of -1.24mA/cm2. The alumia template was first rinsed with DI water and

dried. Then the entire membrane was soaked in 1:1 HNO3 to dissolve the silver and free

the rods, followed by soaking in 0.5M NaOH to dissolve the AAO membrane. The rods

were rinsed with DI water until neutral. This process led to the creation of rods on the

order of 1 x 109 per mL.2 The rods used in these experiments have a diameter of

approximately 300nm as determined by scanning electron microscopy.

142

5.2.2.1 Estimation of Surface Potential

Zeta potential measurements were used as an estimation of the surface

potential (ψ) for gold rods at four different ionic strength conditions. Though the

zeta potential exists further away from the surface, encompassing both the

adsorbed and diffuse layers of ions, it is often used as an approximation when

direct measurements of are not possible. Using a Malvern ZetaSizer Nano-ZS,

three sets of five separate measurements were taken per sample, each

measurement consisting of 10 – 15 scans, to obtain the average and standard

deviation. The Smoluchowski model was used to determine the zeta potential

from electrophoretic mobility measurements. A more in depth explanation of zeta

potential can be found in Section 2.4.3.

5.3 Preparation of Samples

5.3.1 One-Wall Cells

Experiments where only the interaction of the particle with a single planar wall

was of interest were performed using what from this point forward is referred to as a one-

wall cell. These cells are the most commonly used, in all forms of microscopy discussed

above, and rely on particles that are of significant weight such that they will levitate at a

height relatively close to the bottom wall. These cells are created by taking a clean glass

microscope slide and bonding a 5mm (ID) O-ring to the slide with vacuum grease that

envelops the O-ring and covers up to but not over its outer edge. A 120μL aliquot of the

colloidal dispersion of interest was added to the O-ring and allowed to sediment for one

143

minute before a small glass coverslip was used to top the O-ring. The coverslip was then

pressed down into the excess vacuum grease around the O-ring’s edge to seal the small

glass coverslip to the O-ring, preventing unwanted air currents that would cause

convection, or dust particles from falling into the sample during experiments.

5.3.2 Confined Cells

Confined cells, also called two-wall cells, were used to trap particles within a

small gap. This method was used for very small particles (diameters usually on the order

of <1μm) because gravity is not great enough to contain these particles at focal distances

comparable to the objectives being used so that they appear blurry in bright field

microscopy, or because they levitate at distances outside the evanescent wave in TIRM.

The sample used for these cells is a mixture of colloidal rods and “spacer” particles,

which in this case are the same 2.1μm silica particles used above. It must be a more

concentrated suspension since a much smaller volume is used. To create the sample,

aliquots of salt solution, gold rods, and spacer particles were mixed in a 34:15:1 ratio.

The cells are assembled by taking a cleaned long glass coverslip and applying a 10μL

aliquot of the prepared gold rod-silica spacer mixture to the center of the coverslip. A

clean small glass coverslip was then placed on top of the liquid mixture.

Using a piece of lens paper, excess liquid was then wicked away from the edge of

the coverslip by placing the lens paper along the edge of the top coverslip. Using a

fingertip to hold the lens paper in place at one end, the other hand gently pressed down on

the surface and wiped along the edge. This was repeated for all four sides of the small

coverslip. Then the lens paper was placed on top of the sample cell so that it covered the

small coverslip completely, and very gently a fingertip was run across all four sides. This

144

was performed twice, and then the sample cell was checked for diffraction patterns and

small air pockets at the edges of the glass, which indicate that enough liquid had been

wicked away. If too much liquid had been removed, large dry areas would appear

between the two coverslips. Once enough liquid was removed, the epoxy resin and

hardener were mixed and immediately used to seal around all four sides.

5.3.3 Porous Media

To create slides for 2-dimensional (2-D) porous media experiments, a

concentrated silica stock in 0.1mM NaCl was applied to the surfaces of long glass

coverslips using a Laurell Technologies Corporation WS-400BZ-6NPP/LITE spin coater

(North Wales, PA, USA). Placing the coverslip onto the spin coater, a 100μL aliquot of

the silica stock was added to the surface. Closing the lid, the spin coater was then run at

1000rpms for 40s to evenly distribute the silica microspheres. Depending on the desired

density of porous media, this process was performed one to five times, where for more

dense concentrations the same coverslip was left on the spin coater and a second, third,

etc. 100μL aliquot of silica was added in succession. Afterwards, the coverslips were

moved to a hotplate and tented with aluminum foil, then left to dry overnight at 50°C.

The next day, the coverslip was removed from the hotplate and rinsed gently with

DI water from a squirt bottle, then placed back on the hotplate to dry for ten minutes.

This rinsing step was performed to remove any excess salt crystals that may have formed

upon the 0.1mM NaCl solution drying. This procedure was repeated three to five times

depending on the density of porous media on the coverslip. After the last drying step, a

10μL aliquot of the gold rod stock was deposited onto the center of the coverslip. A clean

small glass coverslip was then placed on top of the liquid mixture, and the sample cell

145

was removed of excess liquid and sealed with epoxy in a manner identical to that for

confined cells.

5.4 Microscopy Techniques

5.4.1 Bright Field

Bright field microscopy is the most traditional form of optical microscopy, which

uses a compound microscope to image small objects. These microscopes consist of a light

source, a condenser lens to focus the light onto the sample stage, and oculars to view the

sample.3 Various illumination sources and objectives can be used to enhance the

resolution and performance in bright field microscopy to image objects on the order of

~250nm across. The resolution of the objective being used is defined as the ability to

distinguish small details of an object in an image. Good optical microscopes today use

Köhler illumination, which implements a tungsten lamp as the illumination source and a

series of lenses and condensers to split the light into two different light paths. The

separate illuminating and imaging paths help produce even illumination of the sample,

resulting in high resolution magnified images. The resolution of an image is also a

product of the ocular magnification, the objective magnification, focal length, and

numerical aperture (NA).

The total magnification of the system is the product of the magnification of the

objective and the magnification of the eyepiece. Eyepieces typically have a magnification

of 10, but the magnification of objective lenses can range from 4 to 100 which

correspond to focal lengths equaling 40mm and 2mm respectively. Magnification is

146

related to the focal length (f) by,

oo

dM

f (5.1)

where d is the distance from the lens to the object being examined and the subscript o

stands for objective. The focal length measures the distance at which the rays of light are

brought to a single focal point.

The objective and oculars create the ability to magnify a sample, but the resolving

power of the objective is determined by its NA. The NA is a dimensionless number that

describes the range of angles at which the light will be accepted to pass through the

objective, defined by,

NA sinn (5.2)

where n is the refractive index of the medium in which the objective is working (nair =

1.00) and θ is the half-angle of the maximum accepted cone of light. The NA is changed

by adjusting the band around the objective, allowing more or less light in, which helps

resolve the image. The finest detail that can be resolved by an objective is proportional to

0.61λ/NA, where λ is the wavelength of light being used or emitted, if the NA of the

objective and condenser are the same.

For experiments performed in bright field in the following chapters, a Zeiss

Axioplan 2 upright optical microscope and a Zeiss Axio Observer A1 inverted optical

microscope (Oberkocken, Germany) were used in conjunction with either a 40 objective

(air NA = 0.65) or a 63 objective (air NA = 0.6), and 10 eyepieces. Particle tracking

was captured on video with a Hamamatsu Photonics ORCA-ER 12bit CCD camera

147

(Hamamatsu, Japan) operated in 4-binning mode at ~27.6fps over 30,000 frames. The

exposure time was set between 2 – 5ms for imaging silica particles, and 0.5ms for gold

rods. The software program Streampix 3.2.1, by Norpix (Montreal, Quebec, Canada), was

used to track the lateral diffusion and rotation of particles.

5.4.2 Dark Field

Dark field microscopy is a technique used in optical microscopy to enhance the

contrast of a sample without the use of a stain. A special condenser attachment is used to

block the central beam light path, allowing only light from a thin ring around the edge to

illuminate the sample. The small amount of light is focused and passed through the

sample, but only light that is scattered off the sample at oblique angles is collected by the

objective. The rest of the light is directed away from the objective opening and is not

collected, giving the resulting image a dark background with bright features that is

characteristic of the technique.4 A schematic of the dark field set condenser and its

relation to the light source and objective can be seen in Figure 5.1.5

For experiments performed in dark field in the following chapters, a Zeiss Axio

Observer A1 inverted optical microscope was used in conjunction with a 63 objective

(air NA = 0.6), a Zeiss dry dark field condenser (NA = 0.8/0.95), and 10 eyepieces.

Particle tracking was captured on video with a Hamamatsu Photonics ORCA-ER 12bit

CCD camera (Hamamatsu, Japan) operated in 4-binning mode at 10fps over 30,000

frames. The exposure time was set between 25 – 35ms for imaging gold rods and silica-

based porous media. Streampix 3.2.1 was used to capture the lateral diffusion and

rotation of particles.

148

5.4.3 Total Internal Reflection

Total internal reflection microscopy (TIRM) is a technique developed in the

1990s that uses the non-intrusiveness of light scattering to measure the colloidal

interactions of a single particle with a flat plate using an optical microscope6

Measurements of particles are made possible by tracking the light scattered by the

particle and recording the light intensity as a function of time. This allows for the

separation distance between a spherical colloidal particle and a planar surface to be

determined, while achieving greater sensitivity in measurements of force and energy

Figure 5.1 – Schematic representation of dark field set-up. Adapted from Hu et. al.5

Objective

CCD Camera

149

compared to other surface force techniques (e.g., surface force apparatus, SFA; atomic

force microscopy, AFM). This is possible because instead of using a cantilever to

measure the separation distance, which relies on the limitations of a spring constant to

determine the force, a laser is used. More recently, measurements of ensembles of

particles was made possible by developing a technique that combined TIRM with video

microscopy, which allowed for the scattering intensities of multiple individual particles to

be tracked simultaneously.7

The sample of interest is placed on top of a specially cut dove-tail prism, which is

then situated onto the translating stage of the optical microscope. A laser is focused onto

one side of the prism at an angle so that the laser is completely reflected back at the

interface of the prism and the sample so that it exits on the opposite side. This internal

reflection is predicted by Snell’s law as,

1 2sin sini rn n (5.3)

where θr > θi and the incident laser passes through a material of a higher refractive index

n2

n1

θr

θi

Figure 5.2 – Internal reflection of a laser as predicted by Snell’s Law.

150

(n) than that of the material or medium on the other side of the interface. Figure 5.2

illustrates this relationship where the black arrows depict the path of the internally

reflected beam when θr is sufficient to cause internal reflection, and the red arrow shows

when θr is not large enough.

This internally reflected light generates an evanescent wave that propagates into

the medium above it and decays exponentially with distance from the surface. In our

case, this would be the sample cell sitting atop the prism. As a particle of interest

undergoes Brownian motion within the sample cell it will translate in the x-y plane, but it

will also experience a variety of height excursions in the z-direction. As the particle

samples different heights (h) the intensity of the scattered light will change, where the

particle scatters more light as it gets closer to the surface, and therefore, farther into the

evanescent wave. The intensity of the scattered light also decays exponentially and is

given by,

0( ) exp( )I h I h (5.4)

where I0 is the scattering intensity of a particle in direct contact with the planar surface (h

= 0) and β is the penetration depth of the laser, which can be calculated as,

2 21 2

4sin in n

(5.5)

This exponential decay allows for high resolution measurements in the z-axis (down to

1nm). A schematic representation of the TIRM set-up is illustrated in Figure 5.3.7

The intensity of a given particle at any time can be used to calculate the height of

that particle, and then create a distribution of heights over the duration of the experiment.

151

The height a particle samples the most frequently is considered the most probable height

(hm). This hm corresponds to the lowest potential energy experienced by the particle. The

number of times all heights are sampled can be translated into a potential energy via

Boltzmann’s equation,

( ) ( ) ( )

ln( )

m m

B

u h u h p h

k T p h

(5.6)

where kB is Boltzmann’s constant and T is the temperature.

TIRM experiments were performed using the Zeiss Axioplan 2 upright optical

microscope, to which the dovetail prism was coupled. A 15mW HeNe laser (λ =

632.8nm) from Melles Griot (Carlsbad, CA, USA) was focused onto one end of the

prism, striking it at an incident angle of 68° to create an evanescent wave with a decay

CCD, PC h

I(h)

Figure 5.3 – Schematic representation of TIRM set-up, inset shows exponential decay of evanescent wavewith a spherical particle scattering light. Adapted from Wu and Bevan.7

152

length (β-1) of 113.7nm. Sample cells were indexed matched with oil to the prism.

Streampix 3.2.1 was used to capture the lateral diffusion and variations in scattering

intensity of particles.

5.5 Image and Data Analysis

Image analysis algorithms8 coded in FORTRAN were used to analyze videos

taken in the above-mentioned experiments. Compaq Visual Fortran 6.6 (Houston, TX,

USA) was used to generate specific codes to separately track the scattering intensity,

lateral trajectories, and rotational diffusion of spherical and anisotropic particles in bright

field, dark field, or total internal reflection microscopy. These codes provided the ability

to translate scattering intensities of ensembles of particles into potential energies using

Equations 5.4 – 5.6, calculate the mean squared translational and angular displacements,

calculate lengths of colloidal rods, and determine the heights sampled by particles using

the theories described in Chapter 4. These codes also created tagged image files (*.tif)

that were then utilized as a tool for tracing particle behavior (e.g., irreversibly bound,

diffused out of frame, collided with another particle).

Directly from Streampix, small video clips were extracted to make sizeable files

for presentation material in either audio video interleave (*.avi) or *.tif formats.

Videomach from Gromada, Wright Cell Imaging Facility (WCIF) ImageJ (Toronto, ON,

Canada), and Scion Image from the Scion Corporation (Frederick, MD, USA) were used

to edit these video clips. SigmaPlot from Systat Software Incorporated (San Jose, CA,

USA), MathCad from the Microsoft Corporation (Redmond, WA, USA), and MATLAB

2011 from Mathworks (Natick, MA,USA) were used to process all data files resulting

153

from image analysis including plotting and theoretical fits.

5.6 References

1. Yu, J.-S.; Kim, J. Y.; Lee, S.; Mbindyo, J. K. N.; Martin, B. R.; Mallouk, T. E., Template synthesis of polymer-insulated colloidal gold nanowires with reactive ends. Chemical Communications 2000, (24), 2445-2446.

2. Wang, W.; Castro, L. A.; Hoyos, M.; Mallouk, T. E., Autonomous Motion of Metallic Microrods Propelled by Ultrasound. ACS Nano 2012, 6, (7), 6122-6132.

3. Saferstein, R., Forensic science handbook. Prentice Hall: 2001.

4. Davidson, M. W. Darkfield Illumination. http://micro.magnet.fsu.edu/primer/techniques/darkfield.html (January 24, 2013),

5. Hu, M.; Novo, C.; Funston, A.; Wang, H.; Staleva, H.; Zou, S.; Mulvaney, P.; Xia, Y.; Hartland, G. V., Dark-field microscopy studies of single metal nanoparticles: understanding the factors that influence the linewidth of the localized surface plasmon resonance. Journal of Materials Chemistry 2008, 18, (17), 1949-1960.

6. Prieve, D. C., Measurement of colloidal forces with TIRM. Advances in Colloid and Interface Science 1999, 82, (1-3), 93-125.

7. Wu, H.-J.; Bevan, M. A., Direct Measurement of Single and Ensemble Average Particle-Surface Potential Energy Profiles. Langmuir 2005, 21, (4), 1244-1254.

8. Crocker, J. C.; Grier, D. G., Methods of Digital Video Microscopy for Colloidal Studies. J. Colloid. Interface Sci. 1996, 179, 298-310.

154

Chapter 6:

Anomalous Silica Colloid Stability and Gel Layer Mediated Interactions

Adapted from: J.L. Bitter, G.A. Duncan, D.J. Beltran-Villegas, D.H. Fairbrother, M.A. Bevan, “Anomalous Silica Colloid Stability and Gel Layer Mediated Interactions”.

Langmuir, 29 (28): 8835-8844, 2013. DOI: 10.1021/la401607z

Total internal reflection microscopy (TIRM) is used to measure SiO2 colloid

ensembles over a glass microscope slide to simultaneously obtain interactions and

stability as a function of pH (4-10) and NaCl concentration (0-100mM). Analysis of SiO2

colloid Brownian height excursions yields kT-scale potential energy vs. separation

profiles, U(h), diffusivity vs. separation profiles, D(h), and whether particles are levitated

or irreversibly deposited (i.e., stable). By including an impermeable SiO2 “gel layer”

when fitting van der Waals, electrostatic, and steric potentials to measured net potentials,

gel layers are estimated to be ~10nm thick and display an ionic strength collapse. The

D(h) results indicate consistent surface separation scales for potential energy profiles and

hydrodynamic interactions. Our measurements and model indicate how SiO2 gel layers

influence van der Waals (e.g., dielectric properties), electrostatics (e.g., shear plane), and

steric (e.g., layer thickness) potentials to understand the anomalous high ionic strength

and high pH stability of SiO2 colloids.

6.1 Introduction

Silica is ubiquitous. It makes up 60% of the earth’s crust and silicates make up

90% of all minerals. It is present in amorphous and crystalline forms important for

155

everyday use, optics, and microelectronics. It is present in food, pharmaceuticals, and

organisms. As such, understanding the chemical and physical properties of bulk silica,

silica surfaces, and colloidal silica is crucial to numerous applications. A great deal of

silica chemistry is well understood and catalogued in the comprehensive book by Iler.1

However, the stability of colloidal silica against aggregation at high ionic strengths and

high pHs is often referred to as “anomalous”2 because it is not well described by the

Derjaguin-Landau-Verwey-Overbeck (DLVO) theory.3, 4

Historical reviews of possible stabilizing mechanisms that might account for

anomalous silica colloid stability are contained within representative papers on the

topic.2, 5-8 Direct measurements of force vs. distance curves with the surface forces

apparatus and the atomic force microscope indicate a short-range repulsion.5-8 While this

measured repulsion appears sufficient to account for anomalous colloidal stability, its

physical origin remains an open question. Two mechanisms suggested in the literature

include structural forces due to interfacial water9 or steric interactions between silica gel

layers.10 The water structuring mechanism does not appear that it would be unique to

silica. The presence of silica gel layers is supported by measurements of adhesion,

friction, contact angle,5 and surface density profiles via scattering/spectroscopic

methods.11-13 Despite some evidence in favor of silica gel layers, direct measurements5-8

do not conclusively support either mechanism or a quantitative potential model. It is also

not clear that the role of silica gel layers has been treated self-consistently in terms of

their effects on all interactions including van der Waals, electrostatic, and steric

interactions.

In this work, we simultaneously measure the interactions and stability of silica

156

colloids over a glass microscope slide as a function of pH (4-10) and ionic strength (0-

100mM NaCl). TIRM is used to non-intrusively measure weak kT-scale interactions

between a glass microscope slide and an ensemble of silica colloids14 by analyzing their

Brownian height excursions, which also reveals whether they are irreversibly deposited

or levitated (i.e., stable). This has the advantage that separation dependent interactions are

obtained simultaneously with measurements of stability, so the two can be

unambiguously linked in the same material system. We also simultaneously obtain

potential energy vs. separation, U(h), and diffusivity vs. separation, D(h), profiles by

fitting the Smoluchowski equation coefficients to the measured particle dynamic

trajectories. The D(h) trajectories yield additional information about particle-wall

separation and fluid mechanics important to interpretation of electrostatic and steric

interactions. As such, the present study provides new measurements and models of silica

gel layer mediated interactions that lead to anomalous silica colloid stability.

6.2 Theory

6.2.1 Potential Energy Profiles

By measuring a statistically significant number of height excursions, h, of a

spherical particle above a planar wall surface, a normalized equilibrium height histogram,

p(h), can be related to net separation dependent interaction potential, U(h), via

Boltzmann’s equation as,

expp h U h kT (6.1)

where k is Boltzmann's constant and T is absolute temperature. Equation 6.1 can be

157

inverted to obtain a measurement of U(h) from measured p(h) data as,

lnU h kT p h (6.2)

Theoretical models of U(h) can be computed from superposition of contributing

potentials as,

( ) ( ) ( ) ( ) ( )G E V SU h U h U h U h U h (6.3)

where the subscripts refer to the gravitational (G), electrostatic (E), van der Waal (V), and

steric (S) interactions. The gravitational potential is associated with a body force, whereas

the other potentials are associated with surface forces. Electrostatic and van der Waals

potentials were considered in the original DLVO theory.3, 4

The gravitational potential energy of each particle depends on its height, h, of the

particle above the wall, multiplied by its buoyant weight, G, as given by,

3( ) 4 3 ( )G p fU h Gh mgh a gh (6.4)

where m is buoyant mass, g is acceleration due to gravity, and p and f are particle and

fluid densities.

The interaction between electrostatic double layers on two plates (from

superposition, non-linear Poisson-Boltzmann equation, 1:1 monovalent electrolyte)15 can

be used in conjunction with the Derjaguin approximation to give the particle-wall

potential as,16

expEU h B h (6.5)

158

2

1 264 tanh tanh4 4

kT e eB a

e kT kT

(6.6)

1 22

2( )Ai i

i

e Nz C

kT

(6.7)

where κ is the inverse Debye length, ε is the solvent dielectric constant, e is the elemental

charge, ψ1 and ψ2 are surface potentials, NA is Avogadro's number, Ci is electrolyte

molarity, and zi is ion valence. The inverse Debye length and surface potentials of the

particle and wall is essential to the fit of the electrostatic repulsion portion of the net

potential and directly related to the solution chemistry. See sections 6.8.1 and 6.8.2 in the

Supplemental Information for greater detail on accurately determining these quantities.

van der Waals attraction between two plates as predicted from the Lifshitz

theory17 (that includes retardation and screening effects) can be used in conjunction with

the Derjaguin approximation to give the particle-wall potential as,18

26V

h

U h a A l l dl

(6.8)

where A(l) is the Hamaker function given by,19, 20

13 23 13 230

3( ) ' [1 ] ln[1 ]

2rn

x x

n

A l kT x e e dx

(6.9)

where the Δ terms include the frequency dependent dielectric properties of the particle

(1), wall (2), and medium (3), and the remainder of the terms are defined in previous

papers.18, 19 The prime (') next to the summation indicates that the first term (n = 0) is

multiplied by ½(1+2l)exp(-2l) to account for screening of the zero-frequency

159

contribution.

The interaction between surfaces coated with macromolecular layers depends on

the free energy change of layers under compression.21-23 For a layer with a brush

architecture on a planar surface with an uncompressed thickness, δ0, and free energy per

area, f0, the compressed free energy per area, f(δ), for ½ < δ/δ0 <1 can be captured

accurately by,24

0 0( ) 1 expB Bf f (6.10)

where ΓB and γB are dimensionless constants specific to the brush architecture.24 Using

this expression in the Derjaguin approximation, the potential between two symmetric

macromolecular brush layers as can be obtained as,24

, 0 0 0( ) 16 exp 2S B B BU h af h (6.11)

For different interfacial macromolecular architectures with different decaying density

profiles at their periphery, different values of Γ and γ can be used in Equation 6.10.

Because Equation 6.11 can be used to accurately model adsorbed macromolecular layer

repulsion for a broad range of δ0, f0, Γ and γ, a general repulsive steric potential of the

form,

( ) expSU h h (6.12)

is broadly applicable (by lumping unknown constants together), particularly when the

uncompressed layer properties and architecture are not well characterized (which has also

been shown for asymmetric interactions between layers of different properties24).

160

6.2.2 Diffusivity Profiles

In contrast to measuring the equilibrium probability p(h) to obtain U(h) via

Equation 6.2, measurements of the time-dependent probability, p(h,t), can be used to

obtain both U(h) and the separation dependent diffusivity, D(h), as described by the

Smoluchowski equation,25

, , ,p h t p h t p h t dU h

D ht h h kT dh

(6.13)

which reduces to Boltzmann’s equation in the long-time limit as equilibrium is

approached. In previous work, we have reported non-equilibrium analysis of colloidal

trajectories to obtain U(h) and D(h). Measured D(h) are modeled using,

0( ) ( )D h D f h (6.14)

where D0 is the Stokes-Einstein coefficient of an unbounded spherical particle given by,

0 6

kTD

a (6.15)

where η is the fluid medium viscosity, and f(h) accounts for particle-wall hydrodynamic

interactions from Brenner,26 which is accurately captured by the simple expression,27

Table 6.1 – Constants used in theoretical fits.

Variable (units) Value Equation ρp (g/cm3) 1.96 (6.4)

ρf (g/cm3) 1.00 (6.4)

εw 78 (6.7)

T (K) 295 (6.7)

η (Pa·s) 1.002 x 10-3 (6.15)

161

2

2 2

6 2( )

6 9 2

h ahf h

h ah a

(6.16)

6.3 Materials and Methods

6.3.1 Colloids and Surfaces

Hydrochloric acid, potassium hydroxide, sodium chloride (all from Fisher

Scientific), and colloidal SiO2 (nominal diameter of 2.34 microns, Bangs Laboratories)

were used as purchased and without further purification. The non‐porous amorphous SiO2

colloids are were synthesized by a precipitation method using pure reagents with minimal

trace elements. Glass microscope slides (Fisherbrand Plain Microscope Slides) had a

manufacturer reported density of 2.48 grams/cm3 and a soda lime composition

(approximately 72% silicon dioxide, 14% sodium oxide, 6% calcium oxide, 4%

magnesium oxide, 1% aluminum oxide, 1% potassium oxide, <1% other trace elements).

The microscope slides were sonicated for 30min in acetone, 30min in isopropanol, rinsed

with deionized water, and soaked in Nochromix overnight. Slides were rinsed thoroughly

before soaking in 0.1M KOH for 30min prior to use. Slides were again rinsed with

deionized water and dried with nitrogen before use. Colloidal SiO2 dispersions were

prepared by diluting 0.7μL of the manufacturer stock dispersion into 1mL of the desired

pH and ionic strength solution, which was sonicated for 15min before diluting 100X

before introduction into the measurement cell.

6.3.2 Ensemble Total Internal Reflection Microscopy

All experiments were performed in cells consisting of a 5mm ID Viton O-ring

(McMaster Carr) sealed with vacuum grease (Corning) to cleaned microscope slides.

162

100μL of colloidal SiO2 dispersions were added to the O-ring and covered with a

coverslip. Experiments were performed using a Zeiss Axioplan 2 optical microscope with

a 40x objective. Particle scattering was recorded with a 12bit CCD camera (Hamamatsu

ORCA-ER) operated in 4-binning mode at ~27.6fps for 30,000 frames. The evanescent

wave was generated by a 15mW 632.8nm HeNe laser (Melles Griot) focused onto a

dovetail prism (Red Optronics) at an incident angle of 68° to create a decay length of

113.7nm. Image analysis algorithms28 coded in FORTRAN were used to track the lateral

trajectories and scattering intensity of each particle.

6.3.3 Diffusivity Landscape Analysis

Fitting measured particle dynamics with the Smoluchowski equation (Equation

6.13) to obtain the coefficients (i.e., U(h), D(h)) is described in previous29 and recent30

papers from our group. Our previous analysis of local dynamics at each elevation

provides a more intuitive explanation of how D(h) can be extracted simultaneously with

U(h).29, 31 In this work, we employ a less obvious but numerically more robust scheme

based on a global analysis of excursions between all elevations to find an optimal fit to

the Smoluchowski equation. The development of the global analysis algorithm is

described elsewhere,32 and specific details relevant to colloidal interactions are provided

in our recent paper.30 In brief, a FORTRAN program was used to construct a matrix

enumerating the number of times all particles jumped from each initial height to all other

heights on a given time scale. By using a Monte Carlo sampling scheme, values of U(h)

and D(h) are optimized to fit the measured data. Convergence is determined when U(h)

and D(h) fluctuate about a solution that shows a minimum difference with the measured

dynamics. The magnitude of the fluctuations about the solution provides an estimate of

163

error bars on the solution.

6.4 Results and Discussion

6.4.1 Example Deviation from DLVO Theory

DLVO theory is used to understand the stability of charged colloids in aqueous

media in terms of pair potentials that are the superposition of electrostatic repulsion and

van der Waals attraction. However, to demonstrate how DLVO theory can be an

oversimplification, even in apparently model systems, Figure 6.1 shows a measured

potential for ~2.1μm SiO2 colloids interacting with a glass microscope slide in 20mM

NaCl at pH = 10. As noted in more detail in the materials sections, the colloids are non‐

porous, amorphous, pure SiO2 and the microscope slide has a soda lime glass

composition. We chose this system to be representative of typical commercially available

silica materials with compositions that might also be encountered in environmental

applications. The theoretical prediction using only electrostatic and van der Waals

potentials (Equations 6.5, 6.8) with independently measured parameters displays a deeper

secondary minimum than the experimental data and does not match the potential shape.

The predicted potential also indicates a lower energy barrier and particle stability against

deposition on the wall. This finding is consistent the anomalous SiO2 colloid stability

reported in the past.2-8

By simply adding an additional repulsive potential, it is possible to more

accurately capture the attractive well depth, shape, and range. As already reviewed in the

introduction of this paper, a solvated gel layer, which is also possibly a polyelectrolyte, is

164

thought to provide a stabilization mechanism through a repulsive steric interaction.

However, simply introducing a short range repulsion cannot provide stability in the

presence of a long range van der Waals attraction; the mechanism must be more complex

than this simple picture and it must be physically realistic. In the following, we provide

more measurements and potential fits like the one illustrated in Figure 6.1 to understand

the mechanism of silica colloid stability beyond the standard DLVO theory.

6.4.2 Interaction Potentials vs. Ionic Strength (at fixed pH = 10)

Figure 6.2A shows potential energy profiles between 2.1μm SiO2 colloids

interacting with a glass microscope slide at pH=10 for [NaCl] = 0 – 85mM. The

gravitational potential energy, which corresponds to a body force, has been subtracted

Figure 6.1 – Example of disagreement between ensemble TIRM measured particle-wall potential energy profile (points) and DLVO theory (red solid line) for 2μm SiO2 in [NaCl]=20mM at pH=10. Addition of a steric potential to the DLVO potentials produces a net potential prediction (blue dashed line) in better agreement with the depth of the secondary minimum and produces an energy barrier consistent with the particles’ observed stability.

(h-hm)/nm

-50 0 50 100 150

U(h

)/kT

-4

-2

0

2

4

165

from the measured potential energy profiles to leave only the contributions due to

colloidal/surface forces. The remaining net potentials display the correct qualitative trend

based on expectations that the range of electrostatic repulsion decreases with increasing

ionic strength to reveal a shorter range attractive interaction. However, quantitative curve

fits to the net interaction potentials, which are discussed in detail in the following

sections, are not accurately captured for all conditions by DLVO potentials alone.

To analyze the measured interactions reported in Figure 6.2A, the net potentials

were fit with either DLVO potentials only (Equation 6.3 with UE+UV) shown by solid

lines or DLVO plus a steric contribution (Equation 6.3 with UE+UV+US) shown by

dashed lines. The DLVO potentials were fit to the measured profiles for [NaCl] = 0 –

5mM, and the DLVO plus steric potential was fit to measured profiles for [NaCl] = 10 –

85mM. The solid lines representing purely DLVO interactions were obtained without any

adjustable parameters by using independent measurements of the solution conductivity

and pH to predict from Equation 6.7 and using a literature model summarized in the

Supporting Information (SI).33, 34 Although the SiO2 values in our experiments could

differ from this literature model based on different compositions or cleaning procedures,

we proceed with this model to minimize adjustable parameters and ultimately show it

captures our measured DLVO potentials with no adjustable parameters. The van der

Waals attraction was modeled using literature dielectric properties for water and SiO2

described in our previous work.18, 35 The agreement of the measured potentials at low

ionic strengths with DLVO theory is consistent with previous TIRM measurements of

interactions between different colloidal materials including SiO2 colloids and glass

surfaces.14, 36 We return to a discussion of the non-DLVO potential fits after first

166

Figure 6.2 – Ensemble TIRM measurements of (A) potential energy profiles, U(h), and (B) diffusivity profiles, D(h), for 2.1μm SiO2 at pH = 10 with [NaCl] = 0.1 – 100mM. The color scheme for lines and points indicates [NaCl] given in the legend in (A). In (A), the points are measured data from an equilibrium analysis of particle trajectories using Equation 6.1, solid lines indicate DLVO potentials only (Equation 6.3 with UE+UV), and dashed lines indicate DLVO plus a short range steric contribution (Equation 6.3 with UE+UV+US). In (B), the points are measured data from a non-equilibrium analysis of particle trajectories using Equation 6.13, solid lines are fits to theoretical predictions from Equation 6.14, and error bars are explained in the Methods section.

h/nm

0 100 200 300 400

U(h

)/k

T

-4.0

-1.5

1.0

3.5

6.0

0.1mM2mM5mM10mM15mM20mM40mM60mM80mM

A

h/nm

0 100 200 300 400

D(h

)/(n

m2/m

s)

0

20

40

60

80

B

167

presenting measurements of separation dependent diffusivity profiles, D(h), to provide

additional information on the SiO2 particle-wall interaction.

6.4.3 Hydrodynamic Interactions vs. Ionic Strength (at fixed pH = 10)

As part of verifying both the DLVO and non-DLVO potential fits in Figure 6.2A,

it is useful to have an independent measurement of absolute separation between the

particle and wall. In Figure 6.2B, we report measurements of particle diffusivity profiles,

D(h), obtained from a dynamic analysis based on Equation 6.13 (see the Methods

section). This analysis also yields potential energy profiles, U(h), essentially identical to

those obtained with the standard Boltzmann inversion in Equation 6.2. This confirms the

dynamic analysis successfully recovers the potential energy due to conservative forces,

which provides confidence in the D(h) data obtained simultaneously in this analysis.

Figure 6.2B shows fit theoretical D(h) curves using the literature value for the

viscosity of water and the particle radius obtained from the gravitational potential energy

fit (Equation 6.4, subtracted from data in Figure 2A). By fitting the measured D(h) curves

to the theoretical prediction in Equations 6.14 – 6.16, we obtain an estimate of the

absolute separation scale by setting h = 0 as the location where D(h) = 0. By measuring

and fitting the complete functional form of D(h), we obtain a more accurate estimate of

separation than previous measurements of spatially averaged diffusivities.27 The D(h)

data become scattered at larger separations due to increasing signal noise and lower

statistical sampling (at corresponding higher energies in the U(h) data), but the curve fits

display good agreement with the less noisy data at short separations.

For the [NaCl] = 0 – 5mM data fit with only DLVO potentials (UE+UV) in Figure

6.2A, the particle-wall absolute separation scales from the U(h) and D(h) fits are in good

168

agreement with no adjustable parameters (see Table 6.1 and Figure S6.5). Specifically,

the location of the most probable separation, hm, at the potential energy minimum, and

where the sum of the forces equal zero, is similar in both the DLVO potential fits to the

U(h) data using Equation 6.3 and the hydrodynamic interaction model fits to the D(h)

data using Equation 6.14. As already noted, this finding is consistent with numerous

previous TIRM studies that report excellent agreement with DLVO theory at low ionic

strengths,14, 18, 27, 36, 37 as well as, measurements that have confirmed the accuracy of the

theoretical expression for D(h).27, 38, 39 When fitting the DLVO + steric potentials to the

data for [NaCl] > 10mM in Figure 6.2A, we also use D(h) data to confirm the validity of

the separation scale inferred from the conservative forces and to confirm some

assumptions about the fluid flow in the presence of gel layers. Before discussing the

potential fits that include steric contributions in Figure 6.2A, we first discuss several

conceptual issues related to including a gel layer in a manner that consistently treats

electrostatic and van der Waals potentials.

6.4.4 Role of “Gel” Layer in van der Waals, Electrostatic, and Steric Potentials

The primary conceptual issue to address for computing net potentials in the

presence of silica gel layers is the reference separation for each potential. Four illustrative

cases are depicted in Figure 6.3 showing: (1) no gel layer with UE and UV on a scale, h,

between the H2O/SiO2 interfaces, (2) a gel layer composed of nearly pure SiO2 with UE

and UV on the h scale, and a steric interaction, US, on a separation scale, L, between the

SiO2 gel/bulk interfaces, (3) a gel layer with mostly water properties that is permeable to

fluid flow and has charge on the SiO2 gel/bulk interfaces; this suggests UE, UV, and US

should all be on the L scale, and (4) a gel layer with mostly water properties that is

169

impermeable to fluid flow and has a no-slip surface/potential originating on the H2O/SiO2

gel interfaces; this suggests UV and US should all be on the L scale and UE should be on

the h scale. This list is not exhaustive, but these are the four physically meaningful cases

that bound other cases including gel layer dielectric properties intermediate to pure

H2O/pure SiO2 and/or charge distributed (and a shear plane) between the H2O/SiO2 and

SiO2 gel/bulk interfaces.

These cases can be compared and contrasted to decide on an appropriate model

for the net potentials in Figure 6.2A. Case 1 is the standard model for the DLVO theory.

Case 2 illustrates a gel layer that has no stabilizing effect. In particular, if the gel layer is

composed almost entirely of SiO2, the van der Waals attraction will still be as strong on

the h-scale as in Case 1.40-42 However, since steric repulsion between the gel layers is not

generated until h = 0, the strong van der Waals attraction for h > 0 would cause

irreversible surface adhesion. In fact, Case 1 and Case 2 have the same van der Waals and

electrostatic interactions, but the repulsion at contact is weaker (soft steric repulsion

instead of hard wall repulsion). Clearly a different mechanism is required to produce

stabilization by a silica gel layer.

Cases 3 and 4 illustrate gel layers with a predominantly water composition and

therefore water dielectric properties. This effectively weakens van der Waals on the h-

scale by moving UV to the L-scale in the limit of a purely water layer. The difference

between Cases 3 and 4 is whether the electrostatic potential originates at the SiO2

gel/bulk interfaces (Case 3) or the H2O/SiO2 gel interfaces (Case 4). These cases are

limits of intermediate cases that depend on whether the layer is a polyelectrolyte, how

charge is spatially distributed within the layer, and whether ions are mobile within the

170

Figure 6.3 – Schematics and predicted potentials (for pH = 10, [NaCl] = 80mM in Figure 6.2A) based on various cases for including SiO2 gel layers. In the schematics and predictions, h is the separation between the outer edges of the SiO2 gel layers (i.e. the H2O/SiO2 gel interfaces), and L is the separation between the inner edges of the SiO2 gel layers (i.e. the SiO2 gel/bulk interface). See text for detailed explanation of each case, but in brief: (top-to-bottom) (1) the typical configuration with no gel layers considered in the DLVO theory, (2) gel layers of mostly SiO2 composition, (3) gel layers of mostly H2O composition that are permeable to fluid flow, (4) gel layers of mostly H2O composition that are impermeable to fluid flow. The potentials are color coded as: electrostatics (red), van der Waals (blue), steric (yellow), and net (green).

h

L

silicasilica watergel gel

h/nm-5 0 5 10 15

u(h

)/kT

-20

-10

0

10

20

L/nm

15 20 25 30 35

h

silica silicawater

h/nm0 5 10 15 20

u(h

)/k

T

-20

-10

0

10

20

h

L

water silicasilica gel gel

h/nm0 5 10 15 20

u(h

)/kT

-20

-10

0

10

20

L/nm10 15 20 25 30

h

L

water silicasilica gel gel

h/nm-20 -15 -10 -5 0

u(h

)/kT

-20

-10

0

10

20

L/nm0 5 10 15 20

1

2

3

4

171

layer.10, 43 Both of these cases can produce net potentials that correspond to stable

particles and can be fit to the data in Figure 6.2A. In short, Case 4 has more electrostatic

repulsion and as a result has a smaller steric contribution, whereas Case 3 has less

electrostatic repulsion and therefore requires a greater steric contribution. The key effect

in Cases 3 and 4 compared to Case 2 is that van der Waals attraction is weaker due to the

gel layer.

6.4.5 Fitting DLVO and Steric Interactions in the presence of a “Gel” Layer

Based on the discussion of the different cases in Figure 6.3, the measured profiles

for [NaCl] > 10mM in Figure 6.2A are fit using models based on Cases 3 and 4. As in the

purely DLVO fits for [NaCl] < 5mM, the same ionic strength and pH dependent κ from

Equation 6.12 and the ψ model in the SI27, 28 are used in the DLVO potentials for [NaCl]

> 10mM. For the steric potential in Equation 6.12, a prefactor of Γ = 100kT is assumed,

and the steric inverse decay length, γ, becomes the sole adjustable parameter in the net

potential. While Γ = 100kT is somewhat arbitrary, it is of the correct order of magnitude

based on the few cases where the steric prefactor has been estimated22 and based on what

is necessary to generate the observed stability. We decided to fix Γ = 100kT for several

reasons including (1) a greater sensitivity of the fit to the decay length than the prefactor

when both parameters are varied, (2) some uncertainty in the strong, short-range

repulsion due to noise, which affects estimates of the intercept, and (3) the prefactor in

steric interactions cannot generally be predicted a priori based on independent

parameters.22 On the basis on this prefactor, the gel layer thickness, Δ, can be estimated

as 2Δ = L – h = 5γ-1, which corresponds to a decay from 100kT to ~0.5kT based on the

properties of the exponential function. As a result, the difference between Cases 3 and 4

172

is simply whether all potentials are on the same separation scale (Case 3) or whether the

electrostatic potential is shifted outward by 5γ-1 (Case 4). We return to a discussion of the

validity of the assumed prefactor after reporting and discussing the fit steric decay

lengths.

Figure 6.4 reports the inferred decay lengths, γ-1, and gel thicknesses, Δ= 0.5(L –

h) = 2.5γ-1 as a function of ionic strength from the fits to the pH=10 profiles in Figure 6.2.

Data corresponding to fits based on Case 3 indicate decay lengths of γ-1 ≈ 4 – 7nm and

thicknesses of Δ ≈ 10 – 17nm, whereas fits based on Case 4 give γ-1 ≈ 2 – 4nm and Δ≈ 5

– 10nm. It should be noted that these fit parameters are obtained by generating a

repulsion that results in the correct attractive well depth, which then determines the

potential energy minimum location at the most probable separation, hm. This approach is

necessary since the actual repulsive decay corresponds to a strong force approaching the

Figure 6.4 – Steric decay length, γ-1, (left) and gel layer thickness, Δ, (right) vs. [NaCl]/mM at pH=10 from fits in Figure 6.2A based on models for Cases 3 (red triangles) and 4 (blue circles) in Figure 6.3.

[NaCl]/mM

0 25 50 75 100

n

m

1.5

3.0

4.5

6.0

7.5

/n

m

4

8

12

16

173

noise limit of the TIRM method,44, 45 which limits the accuracy of the measured repulsive

decay length. In short, noise softens strong forces (i.e., large energy changes over small

distances) at short separations and high ionic strengths, so that matching the potential

energy minimum well depth and location is a better measure of the net repulsion than the

decay length.

When considering the absolute values of the layer dimensions inferred in Figure

6.4, it is important to recall that Cases 3 and 4 bound some limiting physical models. For

example, distributing charge anywhere in between the SiO2 gel/bulk interfaces (Case 3)

or the H2O/SiO2 gel interfaces (Case 4) would produce gel layer estimates in between the

two curves shown in Figure 6.4. If the layers contain a higher concentration of SiO2 than

the nearly pure H2O layers in Cases 3 and 4, then the two cases in Figure 6.4 represent

lower bounds where the layer thickness would diverge to infinity as the layers approach

pure SiO2 (as in case 2 where the gel layer is not capable of generating a stabilizing

repulsion beyond the range of van der Waals attraction). One way to overcome this

problem, is to also include surface roughness in addition to the gel layer, which will

weaken van der Waals18 and still allow gel layers without pure water properties.46, 47

6.4.6 Do Inferred Gel Layer Properties Make Sense?

The inferred gel layers in Figure 6.4 from the measured potentials in Figure 6.2

display the expected trend by showing a decreasing thickness vs. increasing ionic

strength. This behavior is consistent with the gel behaving as a polyelectrolyte where

screening of electrostatic repulsion within the layers allows for a dimensional collapse.48,

49 It is also expected that this collapse will not occur until high ionic strengths when the

Debye length is on the order of the separation of charges within the gel layer. This

174

dimensional collapse will expel water from the gel layer and enrich it in pure SiO2

properties, which is reminiscent of solvent quality mediated collapse of adsorbed polymer

layers.42, 46, 50, 51 Such a collapse will reduce stability both by decreasing steric repulsion

and increasing the van der Waals attraction due to changing layer dielectric properties.42

The observed stability behavior also has the character of a solvent quality

mediated collapse of a repulsive steric interaction. Specifically, an attractive energy

minimum evolves beyond the range of an infinitely repulsive steric barrier, which

progressively increases bond lifetimes with an exponential dependence on well-depth.24,

43 When the attractive well is deep enough to produce most probable bond lifetimes

longer than the observation time, then the particle appears to be irreversibly deposited.

This type of destabilization mechanism contrasts the typical mechanism for only DLVO

interactions, where the height of an energy barrier determines the probability of forming

an irreversible bound state involving strong van der Waals attraction at contact. Our

results appear to display stability behavior consistent with a decreasing range of steric

repulsion due to collapsing impenetrable gel layers

The values of the inferred thicknesses are larger than the ~2nm estimates from

mechanical force measurements (e.g. SFA,5 AFM6) but are comparable to gel thicknesses

obtained from surface spectroscopic/scattering methods (e.g. nuclear resonance

profiling,11 neutron, x-ray reflectivity12, 13). One way to reconcile the differences between

layer thicknesses inferred from SFA and AFM force profiles and the TIRM potential

energy profiles in Figure 6.2A is the much higher sensitivity to small energies (and

forces) with TIRM.37 In particular, very weak steric interactions between silica gel layers

could go undetected with mechanical methods until they generate ≥10pN of repulsion. In

175

contrast, TIRM is capable of detecting the very onset of silica gel layer compression on

the kT energy scale and fN force scale (e.g., 1kT/100nm ≈ 10fN). This sensitivity can be

expected to produce thicker layer estimates, and because TIRM is not mechanically

limited, it might produce estimates closer to non-intrusive spectroscopic/scattering

methods. In short, the 5 – 17nm SiO2 gel layers are reasonable based on literature neutron

and x-ray measurements.12, 13

It is also useful to consider the D(h) data and fits in Figure 6.2B. The D(h) fits

serve the purpose of confirming that steric potentials occur on separation scales

consistent with hydrodynamically measured surface separation. In particular, estimates of

hm from non-DLVO fits to U(h) data in Figure 6.2A for both Cases 3 and 4 agree with hm

estimates from the D(h) fits in Figure 6.2B within the uncertainty of the measurements

(see Figure S6.1). The pH = 10, [NaCl] = 80mM potentials in Figure 6.3 can be used to

illustrate this point; the net potential in Figure 6.3C for Case 3 has Lm = 13nm, whereas

the net potential in Figure 6.3D for Case 4 has hm = 8nm, which are both within the

uncertainty of hm = 7nm from the D(h) fit in Figure 6.2B (it should be noted the Lm and

hm are used in this example since L is the hydrodynamic separation scale in Case 3 and h

is the hydrodynamic separation scale in Case 4). As a result, steric potentials with Γ =

100kT and layer thicknesses of Δ= 2.5γ-1 produce net U(h) profiles on the same

separation scale as the D(h) data and fits. However, the D(h) fits do not resolve difference

between Cases 3 and 4.

Of Cases 3 and 4 presented in Figure 6.3, Case 4 is more likely for several

reasons. Because the electrostatic repulsion is longer range in Case 4, the silica gel layer

thicknesses in Case 4 are thinner and closer to the estimates from spectroscopic/scattering

176

methods. Based on previous measurements of adsorbed polymers,46 it is more likely the

gel layers are impermeable to flow (Case 4). The electrostatic potential appears more

likely to originate from a surface potential at the outer edge of the silica gel layer (Case

4).

6.4.7 Potentials and Stability vs. Ionic Strength and pH

To understand how the silica gel layer influences stability as a function of both

pH and ionic strength, Figure 6.5 reports results that summarize both stability and

potential energy profile measurements. In addition to the pH = 10 results already

discussed in Figures 6.2 – 6.4, ionic strength dependent results are shown for pHs of 7,

5.5, and 4. The points show several states indicating whether: (1) all particles were

Figure 6.5 – Summary of whether DLVO theory fit measured potentials and the degree of particle stability vs. solution pH and [NaCl]. Points indicate: (1) robust levitation, accurately modeled by DLVO theory (green circles), (2) robust levitation, modeled by DLVO + steric repulsion (green triangles), (3) slow deposition of particles, levitated particles are modeled by DLVO + steric repulsion (yellow inverted triangles), and (4) irreversible deposition (red squares).

[NaCl]/mM

0.01 0.1 1 10 100

pH

4.0

5.5

7.0

8.5

10.0

177

robustly levitated throughout the entire observation time and had potentials captured by

DLVO theory (green circles), (2) all particles were robustly levitated and had potentials

that required DVLO theory + a steric repulsion (green triangles), (3) some particles

became irreversibly deposited during the observation time and the particles that remained

levitated had potentials that required DVLO theory + a steric repulsion (yellow inverted

triangles), or (4) all particles were irreversibly deposited during the observation time (red

squares). The measured ionic strength dependent potentials for each pH are included in

the SI in Figures S6.4 – S6.6 with fit parameters reported in Table 1. The agreement

between measured potentials and theoretical fits for other pHs in Figures S6.4 – S6.6 are

similar to the agreement observed for the pH = 10 data in Figure 6.2A.

At each pH, there is a clear progression with increasing ionic strength through

states 1 – 4 described in the previous paragraph. The ionic strength dependence is

different at each pH showing a more compressed transition through states 1 – 4 at lower

pHs. For example, at pH = 10 the transition from stable particles described by DLVO

theory to irreversibly deposited particles occurs between [NaCl] = 5 – 100mM whereas

the same transition occurs at pH=4 between [NaCl] = 0.1 – 10mM. At each pH, the

potentials and stability are well described by DLVO theory at low ionic strengths, but an

additional repulsion, presumably due to steric gel layer interactions, is required to fit the

measured potentials and capture the stability at high ionic strengths.

The agreement between measured low ionic strength potentials with DLVO

theory is easy to understand at all pHs; the long range electrostatic repulsion does not

allow particle-wall separations to become small enough to observe a steric repulsion

between silica gel layers in contact. However, as the ionic strength is increased at each

178

pH, a steric repulsion is required to capture the observed repulsion, lack of attraction, and

stability at higher ionic strengths. The fits to measured potentials at all pHs and ionic

strengths (Figures 6.2A, S6.4 – S6.6) have no adjustable parameters in the electrostatic

and van der Waals contributions by using κ from Equation 6.7 and the ψ model in the

SI.27, 28

The different trends at each pH in Figure 6.5 are accounted for by the ionic

strength and pH dependence of the van der Waals and electrostatic potentials, which

suggests the silica gel layer repulsion that is relatively insensitive to pH. This is perhaps

most readily illustrated by noting the steric decay length vs. solution ionic strength is

essentially the same for each pH within the limits of uncertainty of the fit points (see γ-1

data in Table 6.2). Because intramolecular electrostatic repulsion within a polyelectrolyte

brush determines its degree of swelling,48 decreasing such interactions either by screening

at elevated ionic strengths or reducing the total charge via pH dependent weak acid

groups could produce a dimensional collapse of the layers and an associated decreasing

steric repulsion. The steric interaction indeed appears to weaken as the Debye length

changes from ~5nm down to <1nm, which is consistent with the screening length

becoming comparable to spatial dimensions of intramolecular charge separation within

the gel layer to cause its dimensional collapse.48 By analogy, it might be expected that

decreasing charge density with decreasing pH might also influence the average charge

separation and intramolecular electrostatic repulsion within the gel layers to also cause a

dimensional collapse. Although it is non-trivial to demonstrate quantitatively, we

speculate that the average charge separation (i.e., inverse of charge density) is smaller

than the characteristic range of intramolecular electrostatic repulsion for pH > 4 so that

179

γ-1 does not change in this pH range (but might be expected to decrease as pH is

lowered further). Ultimately, the results in Fig. 6.5 in addition to the results in Figures 6.2

pH 2a

(μm) NaCl (mM)

κ-1 (nm)

pH -ψ

(mV)

-1

(#3) (nm)

hm-U (nm)

-1 (#4) (nm)

hm-U (nm)

hm-D (nm)

10

2.17 0.11 30.0 10.02 120 - 320.5 - 320.5 300 2.15 2.1 6.6 10.01 100 - 76.4 - 76.4 80 2.15 5.2 4.2 10.00 83 - 48.1 - 48.1 50 2.16 9.9 3.0 9.93 61 7.2 39.2 4.2 33.4 45 2.13 16.1 2.5 10.01 44 6.3 35.1 3.4 25.9 30 2.17 21.1 2.1 9.91 36 5.9 29.7 3.5 21.5 25 2.13 40.6 1.5 9.88 26 4.7 17.6 2.33 11.6 15 2.15 60.8 1.2 9.92 25 4.3 14.1 2.1 9.1 7 2.15 84.2 1.1 9.77 24 4.1 13.4 2.0 7.5 7

x 111 0.96 9.94 24 x x x x x

7

2.13 0.05 44.2 7.01 94 - 443.6 - 443.6 2.22 2.2 6.8 7.10 85 - 76.7 - 76.7 2.20 5.0 4.3 6.99 70 7.0 47.6 3.1 47.8 2.15 11.3 3.0 7.05 49 6.7 37.7 2.6 30.6 2.13 17.2 2.5 7.00 36 6.4 35.4 3.4 23.8 2.08 22.8 2.1 7.01 30 5.5 25.1 2.7 18.4

x 34.1 1.8 7.08 25 x x x x

5.5

2.22 0.04 51.2 5.52 70 - 480.0 - 480.0 2.10 2.1 6.8 5.48 64 - 73.6 - 73.6 2.08 5.1 4.3 5.50 54 8.2 53.0 2.7 47 2.07 10.3 3.0 5.51 40 8.5 47.8 7.0 30.3 2.02 16.3 2.5 5.55 30 6.8 34.8 4.8 21.6 2.08 20.9 2.1 5.44 24 7.6 43.3 6.0 20.3

x 31.8 1.8 5.46 19 x x x x

4

2.18 0.13 29.9 3.98 30 - 245.7 - 245.7 2.13 2.0 6.6 4.00 29 10.0 67.7 3.0 62.6 2.03 4.5 4.2 3.97 25 9.3 57.6 4.0 39.7

x 11.1 3.0 3.98 15 x x x x

Table 6.2 – Experimental parameters for each pH and ionic strength condition examined. The column labeled as “-1 (#3)” is the steric decay length from a net potential fit based on Case 3 in Fig. 3, and “-1 (#4)” is the steric decay length from a net potential fit based on Case 4. The columns labeled as hm-U are most probable particle-wall separations obtained from potential energy profile fits, and hm-D is the most probable height from diffusivity profile fits. Dashes indicate cases without a steric contribution, and “x”s indicate irreversibly deposited particles where potential energy and diffusivity profiles could not be measured.

180

– 6.4show potentials and stability that are well described by a silica gel layer that reduces

van der Waals attraction, preserves electrostatic repulsion, and contributes an additional

steric repulsion.

6.5 Conclusions

TIRM was used to measure SiO2 colloid ensembles on a glass microscope slide to

simultaneously obtain particle-wall potential energy profiles, diffusivity profiles, and

stability as a function of ionic strength and pH. To interpret the measured potentials and

explain anomalous high ionic strength stability, a model was developed based on

electrostatic and van der Waals potentials from the DLVO theory plus a steric repulsion

attributed to silica gel layers on the particle and wall. For such a model to successfully

quantify the measured potentials, the van der Waals attraction must be weakened by a

layer that has some solvent composition rather than pure silica properties, although

surface roughness could account for some weakening. By including an impermeable gel

layer when fitting van der Waals, electrostatic, and steric potentials to measured net

potentials, gel layer thicknesses of 5 – 10nm were obtained from the model, consistent

with literature scattering measurements. Such gel layers also indicate consistent surface

separation scales for both potential energy profiles and diffusivity profiles based on

theoretical models. The net potential model reported here accurately captures measured

potentials and stability for [NaCl] = 0 – 100mM and pH = 4 – 10 by including a gel layer

that collapses at high ionic strengths but is relatively insensitive to pH. Our findings

indicate a model of silica gel layers that captures measured van der Waals, electrostatic,

steric, and hydrodynamic interactions and their role in the anomalous high ionic strength

stability of silica colloids.

181

6.6 Appendix

Here we describe a curve fit to A(l) computed by the Lifshitz theory (Equation

6.9). The form of the expression provides an accurate representation of both retardation

and screening effects captured by the Lifshitz theory. This expression is convenient for

computing the van der Waals potential via the Derjaguin Approximation (Equation 6.8)

while avoiding the complexity and computational expense of re-computing A(l) from

Equation 6.9 each time. The form we choose is,

0 11 2 [1 2 ]exp[ 2 ]A l l l A A l (6.17)

where A0 is obtained from Equation 6.9 for n = 0 (without the prefactor of ½(1+2l)exp(-

2l) indicated by the prime symbol), and A1∞(l) is obtained from Equation 6.9 for n = 1 –

∞. This form accounts for the fact that the zero frequency term (n = 0) is screened but not

retarded and all higher frequency terms (n > 0) are not screened but are retarded. The

function of A1∞(l) is accurately fit by,

21 1f f f fA l a b l c l d l (6.18)

where the constants A0, af, bf, cf, and df are reported in Table 6.1. Figure 6.6 shows the

expression in Equation 6.17 accurately captures of A(l) curves computed using the

Lifshitz theory (Equation 6.9).

182


We acknowledge financial support by the National Science Foundation (CHE-

1112335, CBET-1066254).

Table 6.3 – Constants used in to fit the Hamaker function from Lifshitz theory.

Variable (units) Value Equation A0 (kT) 1.501 (6.17) af (kT) 1.962 (6.18)

bf (kT·nm-1) 0.0281 (6.18)

cf (nm-1) 0.0593 (6.18)

df (nm-2) 0.0033 (6.18)

Figure 6.6 – Hamaker functions for two silica half spaces vs. separation and medium ionic strength. Points were computed from the Lifshitz theory in Eq. 6.9 for salt concentrations of 0.01 mM (blue), 0.1 mM (pink), 1 mM (green), 10 mM (red), 100 mM (black), as well as an infinite salt case computed by neglecting the n=0 term in Eq. 6.9 (black triangles). The infinite salt case was fit by Eq. 6.18 (solid black line), which was used in Eq. 6.17 to capture all other salt concentrations (dashed lines).

l/nm

0.1 1 10 100 1000

A(l

)/kT

0.00

0.75

1.50

2.25

3.00

183


6.8.1 Solution Chemistry

Understanding how the surface forces of the particle or surface are affected by the

medium is directly related to the solution chemistry of a sample. To accurately determine

the inverse Debye length, , and surface potentials, , used in the calculation of the

electrostatic potential it is necessary to know the solution ionic strength and pH. For the

aqueous media consisting of CO2 saturated H2O, with added NaCl, KOH, and HCl, it is

necessary to determine the following concentrations: [H+], [OH-], [K+], [Na+], [Cl-],

[CO2], [H2CO3], [HCO3-], and [CO3

-2]. By measuring pH, the concentration of [H+] is

determined using the definition of pH as,

pHH 10 (S6.1)

which can be used to determine [OH-] based on the dissociation equilibrium for H2O as,

OH HWk (S6.2)

(where kw and all other constants in this section are reported in Table S1). The

equilibrium value of [CO2] can be determined from Henry’s law as,

2 22 CO COCO P kh (S6.3)

where PCO2 is the partial pressure of CO2 in the atmosphere and khCO2 is Henry’s constant

at 25°C. The values of [H+] and [CO2] can then be used to determine the equilibrium

concentrations of dissolved CO2 species as,

22 3 2H CO CO COKh (S6.4)

184

2 3

3 1

H COHCO

HaK

(S6.5)

32

3 2

HCOCO

HaK

(S6.6)

where KhCO2 is the hydration equilibrium constant for carbonic acid at 25°C, Ka1 and Ka2

are the dissociation constants for the diprotic carbonic acid species. While [K+], [Na+],

and [Cl-] are known from careful solution preparation and dilutions, these values can be

confirmed by measuring the solution conductivity, KM, which can be compared with the

predicted conductivity based on the concentrations of all charged species using,

M

0 0 0 1 2

0

( , ) ( )

i ii

i i i i

c c a ai i i i i

K C

C A B C

z z

(S6.7)

where Λi0 is the molar ionic conductivity at infinite dilution for a given salt, z is the

valence charge of the ion, λ is the molar ionic conductivity for a specific ion, Λi(C, Λi0) is

Variable (units) Value Equation kW (M2) 1 x 10-14 (S6.2)

PCO2 (Pa) 40.53 (S6.3) khCO2 (Pa·L/mol) 3.015 x 106 (S6.3)

KhCO2 1.7 x 10-3 (S6.4) Ka1 (mol/L) 4.266 x 10-7 (S6.5) Ka2 (mol/L) 4.677 x 10-11 (S6.6)

A [(S·cm2/mol)·(L/mol)1/2] 60.2 (S6.7) B (L/mol)1/2) 0.232 (S6.7)

Table S6.1 – Constants used in theoretical fits.

185

the concentration dependent molar ionic conductivity of a particular species, A and B are

constants determined by the ratio of cations to anions. From the total concentration of

dissolved ions in solution we can accurately determine the Debye screening length (κ-1)

for all experiments.

6.8.2 pH and Ionic Strength Dependent Surface Potentials

To model pH and ionic strength dependent SiO2 surface potentials, we found

literature measurements on quartz33 and an associated amphoteric site binding model34

that accurately capture values inferred from our U(h) data. We fit these results with a

convenient expression as,

Figure S6.1 – Empirical fit to literature data33 and model34 for quartz surface potential vs. pH and ionic strengths of 1mM (green squares), 10mM (red triangles), and 100mM (black circles). The lines are fits to the data given by Eq. (S6.8).

pH

3.0 4.5 6.0 7.5 9.0

/m

V

-100

-75

-50

-25

0

186

( , ) ( ) ( ) exp[ ( ) ]

136 111 1 exp 105

( ) 209 2.06 0.0129

( ) 0.313 0.358 1 exp 54.9

C pH a C b C b C pH

a C C

b C C

c C C

(S6.8)

where the units on all constants are either mV, pH, or M where appropriate. Figure S6.1

shows the expression in Equation S6.8 accurately captures literature data33 and model34

for quartz.

6.8.3 Potential Energy Profiles at Various Solution Conditions

As seen in section 6.4.7, experiments were conducted at a variety of pH and ionic

strength conditions to determine if there was a dependence of the gel layer on the solution

conditions. Figures S6.2 – S6.4 show the experimental data for pH values 7, 5.5, and 4,

and the net potential curves derived from fitting the only adjustable parameter, the steric

inverse decay length, .

Figures S6.2 – Ensemble TIRM measurements of potential energy profiles, U(h), at pH 7 with same format as pH = 10 data and fits in Fig. 6.2A. The ionic strengths range from [NaCl] = 0.1 – 20mM with exact values reported in Table 6.2.

h/nm

0 200 400 600 800

U(h

)/kT

-1.5

0.0

1.5

3.0

4.5

187

h/nm

0 200 400 600 800

U(h

)/kT

-1.5

0.0

1.5

3.0

4.5

Figures S6.3 – Ensemble TIRM measurements of potential energy profiles, U(h), at pH 5.5 with same format as pH = 10 data and fits in Fig. 6.2A. The ionic strengths range from [NaCl] = 0.1 – 20mM with exact values reported in Table 6.2

h/nm

0 125 250 375 500

U(h

)/kT

-1.5

0.0

1.5

3.0

4.5

Figures S6.4 – Ensemble TIRM measurements of potential energy profiles, U(h), at pH 4 with same format as pH = 10 data and fits in Fig. 6.2A. The ionic strengths range from [NaCl] = 0.1 – 5mM with exact values reported in Table 6.2.

188

6.8.4 Comparison of hm Derived from Different Gel Layer Scenarios

The most probable height, hm, can be estimated in one of two ways: through

fitting the net potential curve to experimental data or by fitting diffusivity data. Both

approaches were undertaken in this study in an effort to provide the most rigorous

estimate under all conditions examined. Using Equation 6.3 we were able to fit data from

experiments performed at pH 10 with different gel layer scenarios (see Section 6.4.5). By

testing the extreme cases of where the charge on the silica surface would lie, analysis

produced two different estimations of gel layer thicknesses and surface to surface

separations, where thinner layers were produced when the charge laid at the edge of the

gel layer (hm – Case 4), and thicker gel layers when the charge was at the interface

Figure S6.5 – The most probable separation at the potential energy minimum, and where the sum of the forces equal zero, for the potential energy profiles at pH = 10 in Fig. 6.2A. The x-axis shows estimates of hm from DLVO (closed symbols) and non-DLVO (open symbols) fits of Eq. 6.3 to the U(h) data in Fig. 6.2A for both Cases 3 (red triangles) and 4 (blue circles), and the y-axis shows estimates of hm from fits of Eq. 6.14 to the D(h) data in Fig. 6.2B. The x-axis is labeled as Lm and hm since L is the hydrodynamic separation scale in Case 3 and h is the hydrodynamic separation scale in Case 4. A 1:1 line shows when the two measurements are equivalent.

Lm - U(L)/nm , hm - U(h)/nm

10 100

hm

- D

(h)/

nm

10

100

189

between the gel layer and bulk silica (Lm – Case 3). The results were then compared to

heights determined by fitting diffusivity profiles with Equation 6.13 and are plotted in

Figure S6.5. A linear correlation indicated good agreement between the two methods of

finding the surface to surface separation, which supported not only the method of analysis

but also the degree of accuracy of our results.

6.9 References

1. Iler, R. K., The Chemistry of Silica. Wiley: New York, 1979; p 866.

2. Healy, T. W., Stability of Aqueous Silica Sols. In The Colloid Chemistry of Silica, American Chemical Society: 1994; Vol. 234, pp 147-159.

3. Derjaguin, B. V.; Landau, L., Theory of the stability of strongly charged lyophobic sols and of the adhesion of strongly charged particles in solutions of electrolytes. Acta Physicochim. URSS 1941, 14, 633-662.

4. Verwey, E. J. W.; Overbeek, J. T. G., Theory of the stability of lyophobic colloids. Elsevier: Amsterdam, 1948.

5. Vigil, G.; Xu, Z. H.; Steinberg, S.; Israelachvili, J., Interactions of Silica Surfaces. Journal of Colloid and Interface Science 1994, 165, (2), 367-385.

6. Adler, J. J.; Rabinovich, Y. I.; Moudgil, B. M., Origins of the Non-DLVO Force between Glass Surfaces in Aqueous Solution. Journal of Colloid and Interface Science 2001, 237, (2), 249-258.

7. Valle-Delgado, J. J.; Molina-Bolivar, J. A.; Galisteo-Gonzalez, F.; Galvez-Ruiz, M. J.; Feiler, A.; Rutland, M. W., Hydration forces between silica surfaces: Experimental data and predictions from different theories. The Journal of Chemical Physics 2005, 123, (3), 034708-12.

8. Grabbe, A.; Horn, R. G., Double-Layer and Hydration Forces Measured between Silica Sheets Subjected to Various Surface Treatments. Journal of Colloid and Interface Science 1993, 157, (2), 375-383.

9. Derjaguin, B. V.; Churaev, N. V., Structural component of disjoining pressure. Journal of Colloid and Interface Science 1974, 49, (2), 249-255.

10. Tadros, T. F.; Lyklema, J., Adsorption of potential-determining ions at the silica-

190

aqueous electrolyte interface and the role of some cations. Journal of Electroanalytical Chemistry and Interfacial Electrochemistry 1968, 17, (3–4), 267-275.

11. Lanford, W. A.; Davis, K.; Lamarche, P.; Laursen, T.; Groleau, R.; Doremus, R. H., Hydration of soda-lime glass. Journal of Non-Crystalline Solids 1979, 33, (2), 249-266.

12. Richardson, R. M.; Dalgliesh, R. M.; Brennan, T.; Lovell, M. R.; Barnes, A. C., A neutron reflection study of the effect of water on the surface of float glass. Journal of Non-Crystalline Solids 2001, 292, (1-3), 93-107.

13. Brennan, T.; Dalgliesh, R. M.; Lovell, M. R.; Richardson, R. M.; Barnes, A. C.; Sergeant, S. A., An X-ray and Neutron Reflection Study of Water Penetration into Fluorocarbon Doped Silica Gel Filmsâ€ Langmuir 2003, 19, (19), 7761-7767.

14. Wu, H. J.; Bevan, M. A., Direct Measurement of Single and Ensemble Average Particle-Surface Potential Energy Profiles. Langmuir 2005, 21, (4), 1244-1254.


16. Bike, S. G.; Prieve, D. C., Measurements of Double-Layer Repulsion for Slightly Overlapping Counterion Clouds. Int. J. Multiphase Flow 1990, 16, (4), 727-740.

17. Dzyaloshinskii, I. E.; Lifshitz, E. M.; Pitaevskii, L. P., The general theory of van der Waals forces. Adv. Phys. 1961, 10, 165-209.

18. Bevan, M. A.; Prieve, D. C., Direct measurement of retarded van der Waals attraction. Langmuir 1999, 15, (23), 7925-7936.

19. Prieve, D. C.; Russel, W. B., Simplified Predictions of Hamaker Constants from Lifshitz Theory. J. Colloid Interface Sci. 1988, 125, 1.

20. Parsegian, V. A., Van der Waals Forces. Cambridge University Press: Cambridge, 2005.

21. Milner, S. T., Compressing Polymer Brushes - a Quantitative Comparison of Theory and Experiment. Europhysics Letters 1988, 7, (8), 695-699.

22. Milner, S. T., Polymer Brushes. Science 1991, 251, (4996), 905-914.

23. Milner, T.; Witten, T. A.; Cates, M. E., Theory of the grafted polymer brush. Macromolecules 1988, 21, (8), 2610-2619.

24. Eichmann, S. L.; Meric, G.; Swavola, J. C.; Bevan, M. A., Diffusing Colloidal Probes of Protein-Carbohydrate Interactions. Langmuir 2013.

191

25. Murphy, T. J.; Aguirre, J. L., Brownian Motion of N Interacting Particles.1. Extension of Einstein Diffusion Relation to N-Particle Case. J. Chem. Phys. 1972, 57, (5), 2098.

26. Brenner, H., The Slow Motion of a Sphere Through a Viscous Fluid Towards a Plane Surface. Chem. Eng. Sci. 1961, 16, (3-4), 242-251.

27. Bevan, M. A.; Prieve, D. C., Hindered Diffusion of Colloidal Particles Very Near to A Wall: Revisited. J. Chem. Phys. 2000, 113, (3), 1228-1236.

28. Crocker, J. C.; Grier, D. G., Methods of Digital Video Microscopy for Colloidal Studies. J. Colloid. Interface Sci. 1996, 179, 298-310.

29. Beltran-Villegas, D. J.; Sehgal, R. M.; Maroudas, D.; Ford, D. M.; Bevan, M. A., Fokker–Planck Analysis of Separation Dependent Potentials and Diffusion Coefficients in Simulated Microscopy Experiments. J. Chem. Phys. 2010, 132, 044707.

30. Beltran-Villegas, D. J.; Edwards, T. D.; Bevan, M. A., Self-Consistent Colloidal Energy and Diffusivity Landscapes in Macromolecular Solutions. submitted 2013.

31. Kopelevich, D. I.; Panagiotopoulos, A. Z.; Kevrekidis, I. G., Coarse-grained kinetic computations for rare events: Application to micelle formation. J. Chem. Phys. 2005, 122, 044908.

32. Hummer, G., Position-dependent diffusion coefficients and free energies from Bayesian analysis of equilibrium and replica molecular dynamics simulations. New J. Phys. 2005, 7, 34.

33. Li, H. C.; De Bruyn, P. L., Electrokinetic and adsorption studies on quartz. Surface Science 1966, 5, (2), 203-220.

34. James, R. O., Characterization of Colloids in Aqueous Systems. Advances in Ceramics 1987, 21, 349-410.

35. Parsegian, V. A.; Weiss, G. H., Spectroscopic Parameters for Computation of van der Waals Forces. J. Colloid Interface Sci. 1981, 81, 285-289.

36. Wu, H.-J.; Pangburn, T. O.; Beckham, R. E.; Bevan, M. A., Measurement and Interpretation of Particle−Particle and Particle−Wall Interactions in Levitated Colloidal Ensembles. Langmuir 2005, 21, (22), 9879-9888.

37. Prieve, D. C., Measurement of Colloidal Forces with TIRM. Adv. Colloid Interface Sci. 1999, 82, (1-3), 93-125.

38. Oetama, R. J.; Walz, J. Y., Simultaneous investigation of sedimentation and diffusion of a single colloidal particle near an interface. J. Chem. Phys. 2006, 124, (16), 164713.

192

39. Pagac, E. S.; Tilton, R. D.; Prieve, D. C., Hindered Mobility Of A Rigid Sphere Near A Wall. Chem. Eng. Comm. 1996, 148, 105.

40. Vold, M. J., The Effect of Adsoprtion on the van der Waals Interaction of Spherical Colloidal Particles. Journal of Colloid Science 1961, 16, 1-12.

41. Parsegian, V. A., Model for van der Waals Attraction between Sphereical Particles with Nonuniform Adsorbed Polymer. J. Colloid Interface Sci. 1975, 51, (3), 543-546.

42. Bevan, M. A.; Petris, S. N.; Chan, D. Y. C., Solvent quality dependent continuum van der Waals attraction and phase behavior for colloids bearing nonuniform adsorbed polymer layers. Langmuir 2002, 18, (21), 7845-7852.

43. Napper, D. H., Polymeric Stabilization of Colloidal Dispersions. Academic Press: New York, 1983.

44. Odiachi, P. C.; Prieve, D. C., Total internal reflection microscopy: Distortion caused by additive noise. Industrial & Engineering Chemistry Research 2002, 41, (3), 478-485.

45. Odiachi, P. C.; Prieve, D. C., Removing the effects of additive noise from TIRM measurements. Journal of Colloid and Interface Science 2004, 270, (1), 113-122.

46. Bevan, M. A.; Prieve, D. C., Forces and hydrodynamic interactions between polystyrene surfaces with adsorbed PEO-PPO-PEO. Langmuir 2000, 16, (24), 9274-9281.

47. Dagastine, R. R.; Bevan, M. A.; White, L. R.; Prieve, D. C., Calculation of van der Waals forces with diffuse coatings: Applications to roughness and adsorbed polymers. J. Adhesion 2004, 80, (5), 365-394.

48. Zhulina, E. B.; Klein Wolterink, J.; Borisov, O. V., Screening Effects in a Polyelectrolyte Brush: Self-Consistent-Field Theory. Macromolecules 2000, 33, (13), 4945-4953.

49. Fleer, G. J.; Stuart, M. A. C.; Scheutjens, J. M. H. M.; Cosgrove, T.; Vincent, B., Polymers at Interfaces. Chapman & Hall: New York, 1993.

50. Israelachvili, J. N., Intermolecular and Surface Forces. 2nd ed.; Academic Press: New York, 1992; p 450.

51. Zhulina, E. B.; Borisov, O. V.; Pryamitsyn, V. A.; Birshtein, T. M., Coil-Globule Type Transitions in Polymers. 1. Collapse of Layers of Grafted Polymer Chains. Macromolecules 1991, 24, 140.

193

Chapter 7:

Diffusion of Micron-Sized Gold Rods across Silicate Surfaces and through Slit Pores

Optical microscopy was used to track the translational and rotational diffusion of

micron sized gold rods over silicate surfaces in both two wall (confined) and one wall

(open) systems. Experiments were performed as a function of NaCl concentration (0-

5mM) to determine the effects on the absolute separation and rate of diffusion between

particles and walls. Brownian excursions were analyzed using novel tracking and analysis

algorithms developed in our lab. Calculations were performed using a string of beads

model to acquire equations for bulk and interfacial diffusion coefficients of cylindrically

shaped particles as a function of aspect ratio and absolute separation above a surface. The

equations derived from these simulations were used to fit experimental diffusion

coefficients from rods of various lengths by using a squared error function to fit the

height parameter of the translational diffusion data. The equations developed from our

simulation model were found to accurately capture the concentration dependent

separation of gold rods over silicate surfaces, for a one-wall system. This research has

allowed us to establish a simpler format to explain the translational and rotational

diffusion of cylindrical particles as a function of their length and height.

7.1 Introduction

Micro and nanomaterials of widely varying sizes, shapes, and compositions are

employed in a variety of industries including food and pharmaceuticals, optics,

194

electronics, manufacturing, and energy. With such widespread usage in everyday

consumer products one has to consider the environmental implications of such

technology. The fate of these engineered materials could be in our lakes and oceans if

released from point sources, or our bodies if used for medicinal or agricultural purposes.

Rod/cylinder-shaped, or anisotropic, particles are among those of interest for use in

various technologies. One especially important consideration is that these materials,

regardless or the environment they find themselves in will be interacting with any

number of surfaces. However, tracking the motion of rod-shaped micro and

nanomaterials as they interact with cellular or mineral surfaces can be difficult, and

obtaining ways of accurately measuring the diffusion and interaction energies of these

particles in a simple model that is easy to implement is an even more difficult task.

Many groups in the last century have used theoretical applications to determine

various coefficients and equations related to the diffusion of anisotropic particles. In

1915, G. B. Jeffery attempted to solve for the motion of a viscous fluid caused by the

rotation of axisymmetric particles (ellipsoid, circular disk, 2 non-concentric spheres, and

a sphere with a plate).1 Though thorough in his calculations, his approach attacked the

problem from the standpoint of the liquid motion and not the motion of the particle, but it

did take into consideration the extra resistance that would be imparted when a particle

was close to a surface. By the 1950s two different approaches were being undertaken:

determining the diffusion rates of a cylinder-shaped particle in the bulk medium2 and

between or next to a wall.3Both methods were now approaching the problem by

examining the drag on the cylinder to calculate the translational and rotational motion of

the particle. However, these equations were very long and complex, often involving any

195

assortment of derivatives, integrals, and matrices, and they also were different for each

particle shape and shear flow through the medium. Howard Brenner attempted to unify

the variety of equations by creating a general theory to explain the rheological properties

of rigid axisymmetric Brownian particles (spheroids, long slender bodies, dumbbells, and

circular disks).4

Using an experimental approach, Broersma5, 6 attempted to correct discrepancies

found in earlier experimental findings of lengths for tobacco mosaic viruses (TMV) that

were determined by solving for their rotational diffusion coefficients. The author

concluded that the discrepancy arose from a geometric problem that was not being

accounted for. By creating end corrections that would assume a cylinder with a flattened

top and bottom, and adding these to an updated version of existing theories, Broersma

was able to extract rotational and translational diffusion coefficients. In doing so, the

author was able to produce a table of values that came much closer to lengths for TMV

found using electron microscopy.

Broersma’s work was used as the standard calculations for rod-like particles until

Tirado and de la Torre7published their equations in 1979. They modeled rigid cylindrical

particles with concentric beads, and use their symmetry aspects to simplify the

calculation of a translational friction tensor. A cylindrical coordinate system was

employed to reduce the number of individual friction elements one must take into

account, using sets of symmetry elements instead, thereby allowing for a greater number

of total elements in their model and wider ranges of aspect ratios (p = L/d) to be

examined. Their results were similar to Broersma’s as p approached infinity, but as p

decreased their results diverged so much that Tirado stated that Broersma’s work “should

196

be disregarded as representative of the frictional behavior of finite cylinders.” Broersma

soon corrected his original work,8 and these two groups are frequently cited when

examining the diffusion behavior of cylindrically shaped particles in the bulk.

Computational simulations of cylindrical particles using various models9-11

greatly increased the knowledge of these hydrodynamic interactions. Aragon and

Flamik11 created tables of equations that only required the input of the correct

dimensional qualities of the desired component that were much easier to implement than

many previous papers. A study by Mukhija and Solomon12 combined experiment and

simulation, using confocal microscopy to study fluorescent poly(methyl methacrylate)

rods. By altering the viscosity of the medium, it allowed them to explore the 3-

dimensional degrees of freedom by computing the azimuthal and polar coordinates and

angles of the suspended rods. Good agreement was found between their experiments and

previously published literature on diffusion of rods in a bulk medium.

Work on cylinders next to or confined by walls was a question being explored

simultaneously. Experimental13, 14 and theoretical15, 16strides were made in determining

the influence of a wall on the drag and torque acting on a single rod-shaped particle

(using a string of beads model). Padding and Briels17 used molecular dynamics

simulations to examine the friction on a rod-shaped particle of p = 10 as a function of

distance and angle with a planar surface. Their results suggested that friction does not

become significant until the distance is on the same order as the rod. Their results were

used to capture experiments of carbon nanofibers tethered to the surface at one end to test

very small separation distances.18 Once tethered, all diffusion was purely rotational, and

197

the authors suggest that either diffusion is caused by the tethering action itself or a stick-

slip adherence to the surface.

All of the previously mentioned studies have in some way helped contribute to

our current understanding of the hydrodynamics for cylinder-shaped particles. This is

important for many biological and industrial purposes. In biology, studying the diffusion

behavior of model systems, such as the tobacco mosaic virus (TMV),19-22fd virus,23, 24 or

various bacteria25, 26 can bring insight on the hydrodynamic interactions of these particles

in a bulk fluid or near a surface. It has been suggested that cylindrical particles show

enhanced circulation through blood vessels,27 that coupled with model systems research

can open doors to explore how biological moieties such as DNA fragments,28kinesin

molecules,29 or actin filaments30 translate through the body. For industrial purposes,

understanding how polymer and micelles,31 cellulose whiskers,32 and carbon nanotubes33-

35 diffuse is important for building better composite materials and electronic devices. It is

also important to understand their diffusion in thin films and interfaces36-38 for the

advancement of devices such as semiconductors, antireflective coatings, and thin film

batteries.

This study examined the diffusion behavior of varying length micron-sized gold

rod particles as they interact with silicate surfaces under various solution conditions. New

tracking algorithms were developed to analyze videos of particles taken with an optical

microscope. Lengths, as well as translational and rotational diffusion coefficients were

experimentally determined for each individual rod in an ensemble. These coefficients

were plotted as a function of aspect ratio and fit using equations generated from

simulations with a string of bead model to determine the hydrodynamic interactions at the

198

bulk and interfacial separation distances. The distances determined for the particle height

above a surface was compared to theoretical potential energies to gather information

regarding surface potential heterogeneities on the gold particles.

7.2 Theory

7.2.1 Potential Energy Profiles

Theoretical models of the potential energy for a rod-plate interaction, U(z),

can be calculated by the addition of contributing potentials as,

( ) ( ) ( )G EU z U z U z (7.1)

where the subscripts refer to the gravitational (G) and electrostatic (E) interactions, z is

the center particle to surface distance seen in Figure 1 (z = h + a) where h is the particle

surface to planar surface separation distance and a is the radius of the rod. The

gravitational potential is associated with a body force, whereas the electrostatic potential

is associated with surface forces, but both are dependent on the length of the rod.

The gravitational potential energy of each rod depends on its length, L, and its separation

from the wall, z, multiplied by its buoyant weight, G, as given by,

2( ) ( ) ( ) ( )G p fU z Gz mg h a a L g h a

(7.2)

where m is buoyant mass, g is acceleration due to gravity, and rp and rf are particle and

fluid densities respectively.

The interaction between the electrostatic double layers on a rod and plate depend

on how thick they are. For thin double layers (κa>>1), the interaction can be described

199

from the superposition, non-linear Poisson-Boltzmann equation for a 1:1 monovalent

electrolyte,39 and can be used in conjunction with the Derjaguin approximation to give

the rod-wall potential as,40

exp ( )

2E

aU z LB h a

(7.3)

2

64 tanh tanh4 4

p we ekT

Be kT kT

(7.4)

1 222( )A

i ii

e Nz C

kT

(7.5)

where κ is the Debye screening length, ε is the solvent dielectric constant which is the

product of permittivity in a vacuum (ε0) and the relative permittivity of water (εw), e is the

elemental charge, ψp and ψw are the surface potentials of the particle and the wall

respectively, NA is Avogadro's number, Ci is electrolyte molarity, and zi is ion valence.

For thick double layers (κa~1) the above expression will generally over-estimate

the electrostatic interaction since the Derjaguin approximation no longer holds. We

instead model the rod as a rigid chain of touching spheres, and approximate the potential

energy for a rod levitated above wall by summing up for the potential energy of the

spheres composing the rod. This means that the net potential uses the geometry of a series

of spheres to determine the gravitational contribution,

3,

1

4(z) ( )

3

p

G sphere s f ii

U a gz

(7.6)

and the electrostatic repulsion contribution given as,

200

,

1

(z) exp( ( ))p

E sphere ii

U B z a

(7.7)

where zi is the mass center for ith particle composing the rod, and B is the prefactor for the

electrostatic repulsive interactions based on the linear superposition approximation

(LSA), which can be used for thick double layers, given as,

2 22

42 4

kT a eB

e h a kT

(7.8)

The electrostatic repulsion for a series of spheres from Equation 7.7 can be related to the

UE(z) for a rod through the aspect ratio, p,

, ,( ) ( )E rod E sphereU z pU z (7.9)

UE,sphere(z) is the result of an integration over all spheres in the rigid chain. More detail

regarding this calculation can be found in the Supplemental Information.

The most probably height (hm) of a rod-shaped particle above a planar surface can

be determined by taking the derivative of the net potential and solving for h. The value of

hm calculated from either approximation using a form similar to,

1 lnm

ABLh

G

(7.10)

where L is the length, G is the prefactor from the equation for the gravitational potential,

B is the prefactor from electrostatic repulsion, and A is a constant that contains geometric

corrections. However, it is likely that as a rod undergoes Brownian motion it will sample

many heights. We can determine this height average (havg) by integrating over a range of

heights that may be sampled as,

201

0

0

( )

( )

n

avg n

hp h dh

h

p h dh

(7.11)

where p(h) is the probability distribution of heights sampled obtained from the net

potential energy given by,

( )( ) exp NETu h

p hkT

(7.12)

7.2.2 Bulk Diffusion Modes

Theoretical equations for the parallel and perpendicular translational motion of a

rod-shaped particle freely diffusing in the bulk, DB(p), were derived for rods of varying

aspect ratios (p) using Brownian dynamics (BD) simulations. These equations consist of

Stokes-Einstein type diffusion coefficient, D0, multiplied by a factor dependent on the

aspect ratio of the particle, f(p), where the bulk parallel diffusion coefficient is given by,

|| 0|| ||( ) ( )BD p D f p (7.13)

0|| 2

kTD

pd

(7.14)

2

|| 2

0.4536 1.772 41.5( ) ln( )

34.38 18.96

p pf p p

p p

(7.15)

and the perpendicular bulk diffusion coefficient is given by,

0( ) ( )BD p D f p (7.16)

202

0 4

kTD

pd (7.17)

2

2

0.3604 28.36 72.63( ) ln( )

36.29 34.9

p pf p p

p p

(7.18)

where h is the fluid medium viscosity, and p is the particle aspect ratio where L is the

length and d the diameter.

7.2.2.1 Bulk Translational Diffusion Coefficients

A 2-dimensional translational diffusion coefficient for the motion of a rod

in bulk medium can be established by combining the parallel and perpendicular

modes to achieve,

|| ( ) ( )( )

2B B

T

D p D pD p

(7.19)

7.2.2.2 Bulk Rotational Diffusion Coefficients

The same set of theoretical equations can be derived for the rotational

motion of a rod-shaped particle freely diffusing in the bulk, DB(p), as given by,

0( ) ( )BR R RD p D f p (7.20)

0 3

3

( )R

kTD

pd

(7.21)

3 2

3 2

1.373 19.39 148.1 265.2( ) ln( )

56.43 54.35 268.4R

p p pf p p

p p p

(7.22)

203

7.2.3 Interfacial Diffusion Modes

A translational diffusion coefficient can be obtained for the parallel motion for a

rod of aspect ratio, p, as a function of its particle surface-wall separation, h, above a

planar surface by multiplying the bulk diffusion coefficient, DB(p), by a correction factor

dependent on h and another correction factor dependent on p in the forms,

|| || || ||( , ) ( ) ( )BD p h D f h g p (7.23)

3 2

|| 3 2

0.9909 0.3907 0.1832 0.001815( )

2.03 0.3874 0.07533

z a z a z af h

z a z a z a

(7.24)

|| ( ) 1.1669 0.0091g p p (7.25)

and the translational perpendicular diffusion coefficient is,

( , ) ( ) ( )BD p h D f h g p (7.26)

3 2

3 2

0.9888 0.788 0.207 0.004766( )

3.195 0.09612 0.1523

z a z a z af h

z a z a z a

(7.27)

|| ( ) 1.2239 0.0120g p p (7.28)

For a rod of aspect ratio, p, a second set of translational diffusion coefficients can

be obtained for a scenario where a rod is diffusing between two parallel plates of

separation, Δ, in the forms,

11 1

|| || || || ||( , ) ( ) ( ) (( ) 2 ) 1BD p h D f h g p f a z a (7.29)

204

11 1( , ) ( ) ( ) (( ) 2 ) 1BD p h D f h g p f a z a

(7.30)

7.2.3.1 Interfacial Translational Diffusion Coefficients

A 2-dimensional translational diffusion coefficient for the motion of a rod near a

flat surface can be established by combining the parallel and perpendicular modes

to achieve,

|| ( , ) ( , )( , )

2T

D p h D p hD p h

(7.31)

7.3.2.2 Interfacial Rotational Diffusion Coefficients

The rotational diffusion coefficient for a given rod-shaped particle as

function of its aspect ratio and height above the planar surface is given by,

( , ) ( ) ( )R BR R RD p h D f h g p (7.32)

3 2

3 2

0.998 131.1 21.25 0.01275( )

128.7 121.1 2.897R

z a z a z af h

z a z a z a

(7.33)

( ) 1.154 0.0096Rg p p (7.34)

and for a rod of diffusing between two parallel plates of separation, Δ,

11 1( , ) ( ) ( ) (( ) 2 ) 1R BR R R RD p h D f h g p f a z a

(7.35)

205


7.3.1 Colloids & Surfaces

Potassium hydroxide, sodium chloride (both from Fisher Scientific) and colloidal

SiO2 (nominal diameter of 2.34 microns, Bangs Laboratories) were used as received and

without further purification. Gold rods were synthesized by electrochemically growing

them to a prescribed length in the pores of an anodic aluminum oxide membrane. The

alumina template was then dissolved in base, and the rods were freed from a thin film of

gold using nitric acid.41 Zeta potential (ζ) was used as an estimation of the surface

potential for gold rods (ψp), and was measured at four ionic strength conditions using a

Malvern ZetaSizer Nano-ZS. Three sets of five separate measurements were taken per

sample, each measurement consisting of 10 – 15 scans, to create an average and standard

deviation. The Smoluchowski model is used to determine the zeta potential from

electrophoretic mobility measurements.

Table 7.1 – Constants used in theoretical fits.

Variable (units) Value Equation a (μm) 0.3 (7.2), (7.3), (7.6), (7.7), (7.8), (7.10)

ρp (g/cm3) 19.3 (7.2), (7.6)

ρf (g/cm3) 1.00 (7.2), (7.6)

ε0 (farad/m) 8.852 x 10-12 (7.4), (7.5), (7.8)

εw 78 (7.4), (7.5), (7.8)

T (K) 294 (7.4), (7.5), (7.8), (7.12), (7.14), (7.17), (7.21)

e (C) 1.602 x 10-19 (7.4), (7.5), (7.8)

η (Pa·s) 1.002 x 10-3 (7.14), (7.17), (7.21)

Δ (μm) 2.1 (7.29), (7.30), (7.35)

206

Long glass microscope coverslips (Corning, 24 x 60 mm) were wiped clean with

lens paper, then sonicated for 30min in acetone and30min in isopropanol before being

soaked in Nochromix overnight. Small glass coverslips (Corning, 18 x 18 mm) were

wiped with lens paper and placed directly into Nochromix. Coverslips were rinsed with

deionized (DI, 18.3MΩ) water and soaked in 0.1M KOH for 30min, then rinsed with DI

water again and dried with nitrogen before use.

Particle solutions were prepared by making sodium chloride solutions of desired

concentrations and measuring the pH and conductivity. A colloidal silica spacer solution

was created by diluting 0.5μL of the stock silica suspension in 4mL of DI water.

Colloidal gold rod dispersions were prepared by diluting 60μL of the electrochemically

prepared stock dispersion into 136μL of the desired pH and ionic strength solution, and

adding 4μL of silica spacer solution. This vial was then sonicated for 15min before use.

Confined sample cells were created by dropping 10μL of the gold/silica spacer

particle mixture onto the center of a clean and dried long coverslip. A small coverslip was

placed on top of the droplet to sandwich the particle solution between the two walls. Lens

paper was then used to wick away extraneous solution from between the walls until

enough solution has been removed to cause interference patterns to appear. Once this

occurred, the two coverslips were sealed together using Loctite Epoxy.

O-ring sample cells were constructed by using vacuum grease (Corning) to adhere

a 5mm ID Viton O-ring (McMaster Carr) to a cleaned long coverslip. An aliquot of the

sample used in the confined experiments was diluted using the appropriate NaCl stocks,

and then sonicated for 15min. 120μL of the newly prepared gold rod dispersion was

207

pipetted into the O-ring, and a clean small coverslip was placed on top and pressed into

the vacuum grease around the O-ring to seal the sample cell.

7.3.2 Ensemble Video Microscopy

Two wall (confined) experiments were performed using a Zeiss Axioplan 2

upright optical microscope with a 63x objective. One wall (O-ring) experiments were

performed on a Zeiss Axio Observer A1 inverted microscope with a 63x objective.

Particle diffusion was recorded with a 12bit CCD camera (Hamamatsu ORCA-ER)

operated in 4-binning mode at ~27.6fps for 30,000 frames. Image analysis algorithms

coded in FORTRAN were used to track the lateral and rotational trajectories of each

particle.

7.3.3 Image Analysis

A new image analysis algorithm was coded in FORTRAN to determine the

translational and rotational diffusion of rod-shaped particles from video microscopy

(VM) experiments. The rods block the incident light and appear as dark regions on the

otherwise gray image as shown in Figure 7.1A. The grayscale of each image is first

inverted to make the rods appear as white regions on a black background as shown in

Figure 7.1B. The image can then be thresholded, as shown in Figure 7.1C, to identify the

coordinates of each point on the backbone of the rods. To discriminate between multiple

rods in the image, the coordinates found from thresholding are grouped using a search

algorithm to identify neighboring coordinates. To find points neighboring a coordinate i,

a 25 square pixel box is centered on coordinate i. All coordinates within this box are

labeled as neighbors of coordinate i and therefore a part of the same rod. The search box

si

al

tr

id

ca

ize used wa

lgorithm wa

To de

racked over

dentified, th

alculated as,

Figure 7.1 – inverted versioon the invertepoints (D). Afunction of tim

as found em

s able to effi

etermine the

time. With

he coordinat

,

Cropped (120 on of the sameed version of tlso included a

me to illustrate

mpirically as

ficiently proc

translationa

h the coordi

tes of the c

pixel x 120 pe image (B), ththe image (C)are plots of thehow the tracki

208

individual

cess large set

al diffusivity

inates of ea

center of m

ixel) and scalehe result after t, and the orige center of maing algorithm w

rods were i

ts of images

y of each rod

ach point on

mass of each

ed 2x: the origthe thresholdin

ginal image noass position (Eworks.

identified ac

s.

d, the center

n individual

h rod, xcm

ginal experimeng algorithm how with markeE) and angular

ccurately an

r of each rod

l rods, xi an

and ycm, ca

ntal image (A)has been perfored centers and r rotation (F)

nd the

d was

nd yi,

an be

), an rmed

end as a

209

n

ii

cm

xx

n

(7.36)

n

ii

cm

yy

n

(7.37)

To determine the rotational diffusivity of each rod, the position of the end points

of the rod were tracked over time in polar coordinates. The root mean squared distance of

each point on the rod from the center of mass, ri, can be calculated as,

2 2( ) ( )i cm i cm ir x x y y

(7.38)

The center of mass of each rod is indicated in Fig 7.1D with a red cross. The angular

position of each point on the rod with respect to the center of mass, i, is calculated as

arctan( )cm i

icm i

y y

x x

(7.39)

To obtain unique polar angles for each point on the rod ranging from 0 to 2π, the

arctan2 function in FORTRAN is used to account for the quadrant each coordinate is

located and 2π is added when i < 0. The end points of the rod are then determined based

on the polar coordinates of each point on the backbone of the rod. The coordinate i in

quadrant 1 or 4, where 0 < i < π/2 or 3π/2 < i < 2π, with the largest ri is labeled as one

endpoint and the coordinate j in quadrant 2 or 3, where π/2 ≤ j ≤ 3π/2, with the largest rj

is labeled as the second endpoint. Including both a distance and orientation criteria in the

endpoint determination ensures that two unique endpoints are found on opposite ends of

210

the rod and are shown in Fig 7.1D as green crosses. Sample changes in radii and θ that

are tracked with this analysis method are plotted in Figures 7.1E and 7.1F.

7.3.4 Measuring System Noise

Inherent in any system is noise from fluctuations in light intensity or camera

speed and vibrations due to thermal changes. To measure any influence of noise on the

experimental data, stuck particles were imaged at binning 4 and binning 1 for a set

duration of time and analyzed using the above-mentioned tracking codes. Figure S7.1

shows the length distributions and angles of θ observed for colloidal rods that were

irreversibly bound to the bottom surface of the sample cell. The error measured in this

manner was accounted for by including error bars on the diffusivity measurements

equaling one pixel length in binning 4 (385nm) to the aspect ratio.

7.3.5 Calculations of Position-Dependent Diffusivities

The height-dependent diffusivities obtained in Equations 7.23 – 7.35 were

calculated using a rigid chain of spheres to model the rod-shaped particle. The various

resistance forces acting on each individual sphere in the chain were calculated for a rod

sitting parallel to a flat surface. First, translational and rotational diffusivities at an

infinite distance from the surface were calculated for chains of varying length (i.e., aspect

ratios, p). These diffusivities were then used as normalization factors for finite values of h

for rods of different p. Two correction factors were introduced for the chain of spheres

diffusing near a surface: f, which is a height-dependent factor that approaches 1 as h

approaches infinity and 0 at contact; and g, which corrects for the rod length for surface

to surface separations less than ten radii. Figure 7.2 illustrates the important variables

n

ca

ecessary for

alculations c

Figure 7.2 –pertinent scale

r these calc

can be found

– Comparison es and variable

culations in

d in the Supp

of the schems

211

the two sc

plemental Inf

matics used for

chematics. M

formation.

r experiments

More detail

and calculati

regarding

ions, including

these

g all

212


7.4.1 Measuring the Mean Squared Displacement for Translational and Rotational

Diffusion

The tracking algorithms we created produce output data files with the

translational displacement information in the x and y directions for the center of mass of

each rod in frame, as well as the rotational displacement for each end of the rod in the x-y

plane (θ). This data can be plotted as a function of time and fit with a linear regression to

obtain the translational and rotational diffusion coefficients for a rod of given length.

Histograms of the lengths sampled by each rod while remaining in the experimental

window were used to obtain the weighted average length, which was then used to

determine the aspect ratio, p. Translational and rotational displacements for systems

where the rods interact with a single wall (top) and in an arrangement where the rods are

confined between two parallel walls (bottom) in a solution of 0.1mM NaCl are plotted as

a function of time in Figure 7.3. The trend exhibited by both datasets shows shorter rods

translating and rotating more quickly than their longer counterparts. It can be noted that

this trend is followed more strictly by the rotational diffusion than the translational

diffusion. This arguably makes sense considering the energy penalty for rotating a body

at its center would increase dramatically as the aspect ratio increased, whereas only the

perpendicular translational motion would be greatly affected by a longer body.

Comparing the two experimental geometries, we can see that the colloidal rods in

the more open experimental system with only one wall exhibit faster diffusion rates in

their translational motion than their confined counterparts. The addition of a second wall

above the rods to form a closed system confines the rods to a window of space about 2μm

213

high. This confined arrangement introduces a second surface for the rods to interact with

and be electrostatically repelled from. The presence of electric double layers on both

walls and the surfaces of the rods in close proximity can help to explain the slower

diffusion rates observed for the two-wall geometries. However, it is noticed that this

MS

D,

m2

0.00

0.04

0.08

0.12

0.162.03um4.02um4.05um4.36um4.59um4.92um5.11um5.47um5.58um5.61um

A

Time, ms0 100 200 300 400

MS

D,

m2

0.00

0.02

0.04

0.06

0.08

0.10 3.49um3.90um4.65um4.84um4.96um5.53um6.26um

C

Time, ms0 100 200 300 400

MS rad

2

0.00

0.02

0.04

0.06

0.08

0.103.49um3.90um4.65um4.84um4.96um5.53um6.26um

D

MS , rad

2

0.00

0.02

0.04

0.06

0.084.02um4.05um4.36um4.59um4.92um5.11um5.47um5.58um5.61um

B

Figure 7.3 – Mean squared displacement data plotted as a function of time for the translational (A and C) and rotational (B and D) trajectories of varying length colloidal rods in an open system (top) and confined system (bottom). Closed and open symbols represent actual experimental data acquired from the tracking algorithms and solid lines represent the best fit linear regression to the first five data points in each set.

214

effect is much less pronounced for the rotational diffusion, where the two plots actually

show very similar rates of diffusion for both systems.

Occasionally, deviations were observed from the general trend where shorter rods

diffuse faster. These deviations could be caused by irregularities in the cylindrical shape

of the rods, where a slight curvature in the rod could influence the rotational diffusion.

When a perfectly straight cylinder rolls along the long axis we are unable to distinguish

that movement with our technique. However, if the rod is slightly bent or curved, the end

would look to our tracking codes as if it were rotating more quickly when the rod rolls

and the end being tracked jumps over a measurable distance. This rolling can and has

been observed in rods that showed a curved structure. Deviations could also arise from

heterogeneities in the surface potentials of the silicate surfaces or on the gold rods. A

more negative surface charge on one rod would initiate more electrostatic repulsion

between it and the walls, thereby allowing the rod to achieve a greater surface to surface

separation and potentially a faster translation than one may expect for a rod of that length.

Lastly, deviations may result from the tracking algorithms which assign the center and

end points of each rod (see Figure 7.1). Experiments were performed at binning 4 to

achieve resolution without sacrificing recording speed and creating immense file sizes.

Pixel resolution at binning 4 is 385nm/pixel, meaning that as a rod rotates and translates,

a single pixel may be added or removed from the ends of a rod in each frame depending

on its orientation. This could cause errors in the estimation of rod length, as well as the

position of each end point which are used to calculate the rotational diffusion. This is

why the length of each rod is calculated from a weighted average.

215

7.4.2 Effects of Ionic Strength on Diffusion Coefficients

Experiments in confined and open geometries were performed at multiple ionic

strength conditions to determine if there exists a dependence of ionic strength on either

the translational or rotational diffusion. The resulting diffusion coefficients from the

linear fits to mean square displacement data for translational and rotational data were

plotted as a function of p for both experimental configurations at all ionic strengths

examined. Diffusion coefficients were then fit with the theoretical lines generated by

Equations 17 – 29 for interfacial diffusion coefficients, where the only adjustable

parameter was the height of the gold rod above a surface (h). The highest (hmax, dash),

lowest (hmin, dot-dot-dash), and best fit (hm, solid) lines are plotted in Figure 7.4 with the

corresponding experimental data points for different ionic strength conditions. Best fit

lines were achieved by using a least squares error function to find the particle-surface

separation that would result in the lowest error. The measured surface to surface

separation values can be found in Tables 2 and 3 for one-wall and two-wall geometries,

respectively.

We could see from Figure 3 that the one-wall experiments very clearly show

faster translational diffusion coefficients at short aspect ratios. However, comparing the

resulting diffusion coefficients from the two experimental geometries over the same

range of p (Figure 4) one can see that the one-wall system does exhibit greater diffusivity

than that seen in the confined geometries. As was seen in the mean squared displacement

data for rotational motion in Figure 3, very similar rotational diffusion coefficients were

observed for both systems over the same range of p, though much faster rotation was

216

observed for smaller rods. The experimental parameters for each system configuration

can be found in Tables 2 and 3.

The general trend for both systems shows that with increasing ionic strength

(decreasing inverse Debye screening length, κ-1) rods of similar aspect ratios show a

DR , rad

2/ms

0.0

5.0e-5

1.0e-4

1.5e-4

2.0e-4

2.5e-4DI water0.1mM NaCl1mM NaCl5mM NaCl

B

Aspect Ratio10 12 14 16 18 20 22

DR , rad

2/ms

0.0

5.0e-5

1.0e-4

1.5e-4

2.0e-4


D

DT, m

2/m

s

0.0

5.0e-5

1.0e-4

1.5e-4

2.0e-4


A

Aspect Ratio10 12 14 16 18 20 22

DT, m

2/m

s

0.0

5.0e-5

1.0e-4

1.5e-4

2.0e-4


C

Figure 7.4 – Translational (A and C) and rotational (B and D) diffusion coefficients determined from the fits to mean squared displacement data for each ionic strength condition examined is plotted as a function of rod aspect ratio for colloidal rods in an open system (top) and confined system (bottom). The highest (hmax, dash), lowest (hmin, dot-dot-dash), and best fit (hm, solid) lines are plotted for each data set.

217

decrease in their translational diffusion coefficients. Fits to position dependent

diffusivities show that the rods exhibit smaller and smaller surface to surface separations

as the electrostatic repulsion decreases with increasing ionic strength. This is represented

by the value of h necessary to fit the theoretical line to the experimentally derived

diffusion coefficients. The decrease in separation is directly related to the increase in

ionic strength, which systematically depresses κ-1 through screening effects, and

simultaneously reduces the surface potential on both silicate and gold surfaces. Both of

these effects allow for the two surfaces to approach closer to one another before repulsion

is initiated.

Once again, the differences displayed by the rotational diffusion coefficients at

various ionic strengths are minimal. This assertion applies to both experimental

geometries examined. These results indicate that the drag imposed by a second wall, as

well as the height at which the rod sits above a surface, has a discernable effect on the

[NaCl] (mM) κ-1 (nm) hmin,t (nm) hmax,t (nm) hmin,r (nm) hmax,r (nm) hm (nm)0.03 (DI) 54.5 180 375 285 750 307

0.13 26.7 172.5 262.5 150 345 207 1.06 9.3 127.5 187.5 202.5 345 145.5 5.14 4.3 31.5 82.5 112.5 232.5 45

Table 7.2 – Measured values used in the fitting of one-wall experimental data.

Table 7.3 – Measured values used in the fitting of two-wall experimental data.

[NaCl] (mM) κ-1 (nm) hmin,t (nm) hmax,t (nm) hmin,r (nm) hmax,r (nm) hm (nm)0.03 (DI) 54.5 127.5 184.5 150 244.5 153

0.13 26.7 112.5 154.5 135 168 123 1.06 9.3 93 129 142.5 187.5 111 5.14 4.3 28.5 49.5 60 75 43.5

218

parallel and perpendicular translational motion of anisotropic particles. However, these

same parameters show very little effect on the rotational motion of the rods in the x-y

plane. These two observations have been seen with Stokesian dynamics simulations of

particle clusters,42 which is similar to the string of beads model used for our own

calculations. The collective diffusive modes simulating translation of the cluster slowed

with decreasing particle-wall separation, whereas the relative diffusive mode simulating

rotation showed a relatively uniform diffusion over a range of heights with very small

fluctuations.

7.4.3 Comparing Diffusion Coefficients from Experiment and Theory

To determine how close our theoretical calculations were to the measured values,

translational diffusion coefficients were calculated using the values of hm derived from

various wall surface potentials from both the Derjaguin and linear superposition

approximations. Figure 7.5 illustrates the measured data for the one-wall geometries

(points) and the calculated diffusion coefficients (lines) at three wall surface potentials

(ψw = -10mV, -50mV, -100mV). The resulting diffusion coefficients from each ionic

strength condition were ratioed to the bulk diffusion values. As the ionic strength

increases, the ratio should drop as the surface to surface separation becomes farther from

infinity. For lower ionic strength conditions the values obtained from LSA were used;

and at higher ionic strengths where κ-1 is less than 10nm, the Derjaguin approximation

was employed. We see that the theoretically calculated values show relatively good

agreement with our measured diffusion coefficients. Though the values of hm calculated

at each surface potential increased as the surface potential became more negative, the

values for -50mV and -100mV showed only small differences, especially as the ionic

219

strength increased. This indicates that the surface potentials become relatively insensitive

at higher ionic strengths and values greater than -50mV. The calculated values of hm and

havg (the integral average height from Equation 7.11 can be found in Table 7.4.

Aspect Ratio (p)0 5 10 15 20 25

Dif

fusi

on

Co

effc

ien

t R

atio

, DT(m

easu

red

)/D

T(b

ulk

)

0.0

0.2

0.4

0.6

0.8

1.0

DI water0.1mM NaCl1mM NaCl5mM NaCl

Figure 7.5 – Measured diffusion coefficients from one-wall experiments ratioed to the calculated diffusion coefficients from values of hm obtained from Derjaguin and linear superposition approximations at various ionic strength conditions.

[NaCl] (mM)

κ-1 (nm)

ζ (ψp) (mV)

hm (nm) ψw = -10mV

hm (nm)ψw = -50mV

hm (nm) ψw = -

100mV

havg (nm) ψw = -10mV

havg (nm) ψw = -50mV

havg (nm) ψw = -

100mV 0.03 (DI)

54.5 -24 412 / 385

496 / 360

523 / 390 453 / 320 538 / 402 564 / 429

0.13 26.7 -24 231 / 165

272 / 195

285 / 210 278 / 200 319 / 240 332 / 250

1.06 9.3 -20 93.2 / 75 107 / 75 112 / 90 153 / 102 168 / 114 168 / 140 5.14 4.3 -15 46.1 / 45 52.6 / 45 54.7 / 45 99.3 / 87 106 / 90 109 / 97

Table 7.4 – Calculated values of the most probable and average heights from the Derjaguin and linear superposition approximations (Derjaguin/LSA) used to find diffusion coefficients.

220

7.5 Conclusions

Video microscopy was used to track the translational and rotational diffusion of

colloidal gold rods as a function of ionic strength. Two systems were investigated: an

“open” geometry where the rods interacted with a single planar surface, and a “confined”

geometry where rods diffused between two parallel surfaces in a slit pore. Newly

developed video tracking algorithms allowed us to monitor rod diffusion and determine

the length of individual rods in the frame. However, pixel resolution was responsible for

a variety of lengths being calculated for each rod, as well as small defects such as

curvature affected the calculation of the rotational diffusion coefficients. Using a string-

of-beads model enabled the calculation of translational and rotational diffusion

coefficients of rod-shaped particles as a function of aspect ratio and surface to surface

separation in the bulk and near a planar surface. Fits to these curves showed that the

average surface to surface separation decreased with increasing ionic strength for both

system configurations, where the confined geometry slows the translational diffusion but

not rotational. In fact, it was found that fits to the translational diffusion were a much

more sensitive metric of the surface to surface separation than fits to the rotational

diffusion, providing evidence for similar results seen with Stokesian dynamics

simulations of particle clusters. Electrostatic potentials modeled by the Derjaguin

approximation (high ionic strength conditions) and LSA (low ionic strength conditions)

for a rod-planar surface interaction were used to calculate surface to surface separations

for one-wall geometries. Using zeta potential as an estimation of the particle surface

potential, diffusion coefficients calculated from hm were satisfactorily close to our

experimental results, thereby confirming the legitimacy of our theory.

221



1112335, CBET-1066254). We would also like to thank Wei Wang and Tom Mallouk

from Pennsylvania State University for giving us samples of the gold rods to use in these

studies.


7.7.1 Numerical Calculation of the Electrostatic Repulsion between a Rod in a Parallel

Configuration and a Wall

A rod can be modeled as a rigid linear chain of touching spheres, and the electrostatic

interaction between rod and wall can be approximated by summing up all of the sphere-

wall interactions based on the linear superposition approximation (LSA) for thick double

layers as,

,( ) ( )E E sU h pU h (S7.1)

where p is the aspect ratio. UE,s is the electrostatic repulsion potential between a single

sphere and a wall, which is obtained via integration of the force along the elevation as,

, ( )

h

E sU h Fdh

(S7.2)

where F is the electrostatic force between the sphere and the wall, which only has a z

component due to symmetry. F is given by the stress tensor, σ, integrating through a

222

surface Ω enclosing the sphere and the unit normal vector of the surface pointing outward

(n) as,

ds

F σ n (S7.3)

1Tr( )

2p σ EE I EE I

(S7.4)

and σ contains the electrostatic Maxwell stress and the osmotic pressure, p. I is the unit

tensor, Tr is the trace operation, and ε is the dielectric constant of the solution. The

integration surface, Ω, is chosen as the mid-plane between a sphere and the wall, where

the linear superposition will hold with the minimum amount of error. The electric field,

E, is given by E = −ψ (i.e. the negative gradient of the electrostatic potential field). The

electrostatic potential ψ at Ω is approximated by the sum of the electrostatic potential

generated separately by each sphere and the wall. Mathematically,

w p (S7.5)

,0 exp( )w w h (S7.6)

,0 exp( ( ))pp r a

r

(S7.7)

where ψw,0 and ψp,0 are the surface stern potential of the wall and sphere, h is the distance

to the wall, r is the distance to the mass center of the sphere.

The osmotic pressure from the concentration of NaCl in solution is obtained from the

ideal gas law and the Boltzmann distribution given by,

223

0(exp( ) exp( ))

B B

e ep kT

k T k T

(S7.8)

where ρ0 is the bulk concentration of the salt.

7.7.2 Stokesian Dynamics Simulations

To obtain the height dependent diffusivities of an individual rod above a planar, no-slip

wall, we modeled the rod as a rigid linear chain of spheres. We first employ the grand

resistance tensor (R) for the sphere-chain system above a wall, as given by43

( )

-1PW 2B W 2B, W,R = (M ) + R R R R

(S7.9)

This tensor includes the many-bodied, far-field resistance tensor above a no-slip plane

((M∞PW)-1), as well as the pair-wise lubrication interactions. The pair-wise lubrication

interaction is obtained by first adding the two-body particle-particle exact resistance

tensor R2B and the two-body particle-wall exact resistance tensor RW. The far-field two-

body resistance tensor (R2B∞ + RW∞) is subtracted subsequently the to avoid double

counting the far-field particle-particle and particle-wall interaction in ((M∞PW)-1 and R2B

+ RW. The elements in these tensors can be found in the literature.43-45

After obtaining R, we can calculate the translational and rotational diffusivities

for a rod oriented parallel and with a surface to surface separation distance h from the

wall by the method introduced by Carrasco.46 The diffusivities in the bulk (i.e. setting

h→∞) were first calculated as a function of aspect ratio p, given the diameter of rod, and

then used as a normalization factor for diffusivities with finite values of h.

224

The diffusivities of rods with aspect ratios ranging from 10 to 25 at various

elevations were calculated and normalized by the bulk value. The diffusivity of a rod

with aspect ratio p and at elevation h can be represented as

|| || ||( , ) ( )pBD p h D f h

(S7.10)

where f is the height dependent coefficient function, whose form generally depends on

the value of p. f will approach 1 as h approaches infinity, and approach 0 contact. Other

diffusive modes will take a similar form as Equation S7.10.

7.7.2.1 Approximate diffusivities for cylindrical rod above wall

Here we showed that the height dependent coefficients from the chain-

sphere rod model can be used to approximate the diffusivities of cylinder rod. In

the far-field limit, the flow field generated by the translating object, irrespective

of its shape, can be approximated by the Stokelet, or Oseen, tensor. This

represents the leading term of the multipole expansion of the flow field generated

by a force distribution. In the far-field limit, the hydrodynamic interaction

between a cylindrical object and a wall can therefore be approximated by the

interaction between sphere-chain rod and wall.47 The leading error in the

approximation is expected not to exceed O((a/(h+a))3). Therefore, at h/a >> 1, the

height dependent coefficients of the sphere-chain rod model are the approximate

and upper-bound of the height-dependent coefficients of the cylindrical rod.

At separations of h/a << 1, the height-dependent coefficient can be

approximated using the analytical results for infinite cylinders adjacent to a plane.

The analytical result is the upper-bound for the diffusivities of a finite long

225

cylinder due to the hydrodynamic screening effect (a cylinder surface of length 2L

will experience less drag than 2x that of a cylindrical surface of length L). When h

is about the radius of the sphere, it is no longer appropriate to approximate the

cylindrical particle by either method. The coefficient within this range is found by

fitting a curve that matches the near-field lubrication result and the far-field

Instead, a simpler function accounting for the variation of both p and h will be

needed in practical purpose such as,

|| || || ||( , ) ( ) ( )BD p h D f h g p (S7.11)

where g is the correcting factor for different length of rod, for h < 10a. For

situations where h > 10a, the height dependent coefficient function has a very

weak dependence on the length of rod in the form,

|| || ||( , ) ( )BD p h D f h (S7.12)

7.7.3 Evaluating Stuck Particle Behavior

As discussed in Section 7.3.4, particles that are stuck to the silicate surface can

give information on the noise that is inherent to the system. Figure S7.1 shows example

histograms of a stuck particle resulting from the image analysis algorithms.

226

7.8 References

1. Jeffery, G. B., On the Steady Rotation of a Solid of Revolution in a Viscous Fluid. Proceedings of the London Mathematical Society 1915, s2_14, (1), 327-338.

2. Prager, S., Interaction of Rotational and Translational Diffusion. The Journal of Chemical Physics 1955, 23, (12), 2404-2407.

3. Takaisi, Y., Note on the Drag on a Circular Cylinder moving with Low speeds in a Viscous Liquid between Two Parallel Walls. Journal of the Physical Society of Japan 1956, 11, (9), 1009-1013.

4. Brenner, H., Rheology of a dilute suspension of axisymmetric Brownian particles. International Journal of Multiphase Flow 1974, 1, (2), 195-341.

5. Broersma, S., Rotational Diffusion Constant of a Cylindrical Particle. The Journal of Chemical Physics 1960, 32, (6), 1626-1631.

6. Broersma, S., Viscous Force Constant for a Closed Cylinder. The Journal of Chemical Physics 1960, 32, (6), 1632-1635.

7. Tirado, M. M.; de la Torre, J. G., Translational friction coefficients of rigid, symmetric top macromolecules. Application to circular cylinders. The Journal of Chemical Physics 1979, 71, (6), 2581-2587.

8. Broersma, S., Viscous force and torque constants for a cylinder. The Journal of Chemical Physics 1981, 74, (12), 6989-6990.

Rod 2

Total Length (nm)

4400 4500 4600 4700 4800 4900 5000 5100

Co

un

t

0

1000

2000

3000

4000

5000

6000

Rod 2

Theta 1 (degrees)

162 164 166 168 170 172 174 176

Co

un

t

0

500

1000

1500

2000

2500

3000Rod 2

Theta 2 (degrees)

162 164 166 168 170 172 174 176

Co

un

t

0

500

1000

1500

2000

2500

3000

Figure S7.1 – Histograms of the various length (A) and theta (B and C) values tracked for a rod that was irreversibly bound to the surface. These measurements help to inform on the extent of noise originating from the system as well as the tracking algorithms used.

A B C

227

9. Ortega, A.; de la Torre, J. G., Hydrodynamic properties of rodlike and disklike particles in dilute solution. The Journal of Chemical Physics 2003, 119, (18), 9914-9919.

10. de la Torre, J. G.; del Rio Echenique, G.; Ortega, A., Improved Calculation of Rotational Diffusion and Intrinsic Viscosity of Bead Models for Macromolecules and Nanoparticles. The Journal of Physical Chemistry B 2007, 111, (5), 955-961.

11. Aragon, S. R.; Flamik, D., High Precision Transport Properties of Cylinders by the Boundary Element Method. Macromolecules 2009, 42, (16), 6290-6299.

12. Mukhija, D.; Solomon, M. J., Translational and rotational dynamics of colloidal rods by direct visualization with confocal microscopy. Journal of Colloid and Interface Science 2007, 314, (1), 98-106.

13. Stalnaker, J. F.; Hussey, R. G., Wall effects on cylinder drag at low Reynolds number. Physics of Fluids 1979, 22, (4), 603-613.

14. Marshall, B.; Davis, V.; Lee, D.; Korgel, B., Rotational and translational diffusivities of germanium nanowires. Rheologica Acta 2009, 48, (5), 589-596.

15. Jeffrey, D.; Onishi, Y., The slow motion of a cylinder next to a plane wall. The Quarterly Journal of Mechanics and Applied Mathematics 1981, 34, (2), 129-137.

16. Ristow, G. H., Wall correction factor for sinking cylinders in fluids. Physical Review E 1997, 55, (3), 2808-2813.

17. Padding, J. T.; Briels, W. J., Translational and rotational friction on a colloidal rod near a wall. The Journal of Chemical Physics 2010, 132, (5), 054511-8.

18. Neild, A.; Padding, J. T.; Yu, L.; Bhaduri, B.; Briels, W. J.; Ng, T. W., Translational and rotational coupling in Brownian rods near a solid surface. Physical Review E 2010, 82, (4), 041126.

19. Wilcoxon, J.; Schurr, J. M., Dynamic light scattering from thin rigid rods: Anisotropy of translational diffusion of tobacco mosaic virus. Biopolymers 1983, 22, (3), 849-867.

20. Schumacher, G. A.; van de Ven, T. G. M., Brownian motion of rod-shaped colloidal particles surrounded by electrical double layers. Journal of the Chemical Society, Faraday Transactions 1991, 87, (7), 971-976.

21. Cush, R. C.; Russo, P. S., Self-Diffusion of a Rodlike Virus in the Isotropic Phase Macromolecules 2002, 35, (23), 8659-8662.

228

22. Cush, R.; Dorman, D.; Russo, P. S., Rotational and Translational Diffusion of Tobacco Mosaic Virus in Extended and Globular Polymer Solutions. Macromolecules 2004, 37, (25), 9577-9584.

23. Newman, J.; Swinney, H. L.; Day, L. A., Hydrodynamic properties and structure of fd virus. Journal of Molecular Biology 1977, 116, (3), 593-603.

24. Lettinga, M. P.; Barry, E.; Dogic, Z., Self-diffusion of rod-like viruses in the nematic phase. EPL (Europhysics Letters) 2005, 71, (4), 692.

25. Klein, J. D.; Clapp, A. R.; Dickinson, R. B., Direct measurement of interaction forces between a single bacterium and a flat plate. Journal of Colloid and Interface Science 2003, 261, (2), 379-385.

26. Tavaddod, S.; Charsooghi, M. A.; Abdi, F.; Khalesifard, H. R.; Golestanian, R., Probing passive diffusion of flagellated and deflagellated Escherichia coli. The European Physical Journal E C7 - 16 2011, 34, (2), 1-7.

27. Daum, N.; Tscheka, C.; Neumeyer, A.; Schneider, M., Novel approaches for drug delivery systems in nanomedicine: effects of particle design and shape. Wiley Interdisciplinary Reviews: Nanomedicine and Nanobiotechnology 2012, 4, (1), 52-65.

28. Tirado, M. M.; Martinez, C. L.; de la Torre, J. G., Comparison of theories for the translational and rotational diffusion coefficients of rod-like macromolecules. Application to short DNA fragments. The Journal of Chemical Physics 1984, 81, (4), 2047-2052.

29. Hunt, A. J.; Gittes, F.; Howard, J., The force exerted by a single kinesin molecule against a viscous load. Biophysical journal 1994, 67, (2), 766-781.

30. Li, G. L.; Tang, J. X., Diffusion of actin filaments within a thin layer between two walls. Physical Review E 2004, 69, (6).

31. Lehner, D.; Lindner, H.; Glatter, O., Determination of the Translational and Rotational Diffusion Coefficients of Rodlike Particles Using Depolarized Dynamic Light Scattering. Langmuir 2000, 16, (4), 1689-1695.

32. De Souza Lima, M. M.; Wong, J. T.; Paillet, M.; Borsali, R.; Pecora, R., Translational and Rotational Dynamics of Rodlike Cellulose Whiskers. Langmuir 2002, 19, (1), 24-29.

33. Tsyboulski, D. A.; Bachilo, S. M.; Kolomeisky, A. B.; Weisman, R. B., Translational and Rotational Dynamics of Individual Single-Walled Carbon Nanotubes in Aqueous Suspension. ACS Nano 2008, 2, (9), 1770-1776.

229

34. Streit, J. K.; Bachilo, S. M.; Naumov, A. V.; Khripin, C.; Zheng, M.; Weisman, R. B., Measuring Single-Walled Carbon Nanotube Length Distributions from Diffusional Trajectories. ACS Nano 2012, 6, (9), 8424-8431.

35. Wu, L.; Gao, B.; Tian, Y.; Muñoz-Carpena, R.; Zigler, K. J., DLVO Interactions of Carbon Nanotubes with Isotropic Planar Surfaces. Langmuir 2013, 29, (12), 3976-3988.

36. Levine, A. J.; Liverpool, T. B.; MacKintosh, F. C., Mobility of extended bodies in viscous films and membranes. Physical Review E 2004, 69, (2), 021503.

37. Lee, M. H.; Lapointe, C. P.; Reich, D. H.; Stebe, K. J.; Leheny, R. L., Interfacial Hydrodynamic Drag on Nanowires Embedded in Thin Oil Films and Protein Layers. Langmuir 2009, 25, (14), 7976-7982.

38. Rovner, J. B.; Lapointe, C. P.; Reich, D. H.; Leheny, R. L., Anisotropic Stokes Drag and Dynamic Lift on Cylindrical Colloids in a Nematic Liquid Crystal. Physical Review Letters 2010, 105, (22), 228301.


40. Israelachvilli, J., Intermolecular and Surface Forces. 3rd ed.; Academic Press: New York, 2011.

41. Yu, J.-S.; Kim, J. Y.; Lee, S.; Mbindyo, J. K. N.; Martin, B. R.; Mallouk, T. E., Template synthesis of polymer-insulated colloidal gold nanowires with reactive ends. Chemical Communications 2000, (24), 2445-2446.

42. Lele, P. P.; Swan, J. W.; Brady, J. F.; Wagner, N. J.; Furst, E. M., Colloidal diffusion and hydrodynamic screening near boundaries. Soft Matter 2011, 7, (15), 6844-6852.

43. Swan, J. W.; Brady, J. F., Simulation of hydrodynamically interacting particles near a no-slip boundary. Physics of Fluids 2007, 19, (11), 113306.

44. Jeffrey, D. J.; Onishi, Y., Calculation of the resistance and mobility functions for two unequal rigid spheres in low-Reynolds-number flow. J Fluid Mech 1984, 139, 261-290.

45. Bossis, G.; Meunier, A.; Sherwood, J. D., Stokesian dynamics simulations of particle trajectories near a plane. Physics of Fluids A: Fluid Dynamics 1991, 3, (8), 1853.

230

46. Carrasco, B.; Garcıa de la Torre, J., Improved hydrodynamic interaction in macromolecular bead models. J Chem Phys 1999, 111, (10), 4817-4826.

47. Cox, R. G.; Brenner, H., Effect of finite boundaries on the Stokes resistance of an arbitrary particle Part 3. Translation and rotation. J Fluid Mech 1967, 28, (02), 391-411.

231

Chapter 8:

Colloidal Rod Diffusion through Model 2-Dimensional Porous Media

Porous media is not only an essential component in the natural filtration of

groundwater in aquifers, but it also an important instrument of study to understand the

diffusion of molecules and particles in the fields of chemistry, engineering, and medicine.

By creating a model 2-dimensional porous media out of 2.1μm silica microspheres, this

study employs the use of dark field microscopy to track individual colloidal gold rods as

they undergo Brownian diffusion. Preliminary results showed that the length of the

particle and the area fraction of silica spheres were key parameters in determining the

mechanism of diffusion for each rod. Increasing area fraction showed the rods diffusing

more quickly to areas with less obstruction. It was also found that for similarly sized rods

there was an increase in their rate of diffusion with increasing area fraction up to a point

where diffusion was limited by the packing arrangement of silica microspheres. To our

knowledge, this is one of the first reports of real-time tracking of individual rods through

porous media.

8.1 Introduction

Porous media is an important component in a range of different fields including

biology, industry, and the environment. A skeletal component makes up the framework

of the porous media, while the voids within the frame compose the pore space. There

exist two important and distinct frameworks when discussing porous media: i) a network

232

of capillaries which can describe blood flow, acoustics through caves, or the empty space

associated with air pockets in Styrofoam and molded plastics; and ii) an array of close-

packed solid particles which includes packed sand, capillary columns, or even bacterial

colonies. Understanding the diffusion and transport of particles through these types of

media to either retain or screen a particle of molecule of given weight, size, shape, or

chemistry is an important component in science and engineering technology. Especially

with the increasing use of 1D and 2D nanomaterials, knowing how their size and shape

affect their transport and interactions with the surrounding porous media framework is of

great interest.

In biology and medicine where the porous media framework is composed of cells

with varying shapes and surface chemistries, it is being realized that the shape of a drug

delivery particle is a significant variable in determining if the cargo will reach its

destination. Geng et. al. examined the importance of shape on the circulatory lifetime and

delivery capabilities of spheres versus filaments. They found that the anisotropic particles

were able to effectively deliver their cargo and persisted ten times longer in the

circulatory system of mice than spheres because of their ability to streamline through

blood vessels.1 Membranes for filtration or capillary columns in HPLC and GC used for

the separation of particles and molecules are important examples of porous media for

industrial purposes in biology and chemistry. Separation of DNA and proteins using

microfabricated nanochannels2 and lattices3 or ultrafiltration membranes4 occurs when

molecules diffuse through these pore spaces differently due to their size, shape, or surface

charge. Tuning the chemical identity of the porous media framework or creating

gradients of pore sizes is an effective way to capture the ideal molecules for a specific

233

purpose. Studies involving the pore structure of capillary columns5, 6 and microfluidic

devices7 helps industry better understand how separation is dependent on variables such

as flow, pore size, polarity, and particle size, and allows them to create more effective

separation mechanisms.

Transport of liquids and particles through porous media is one of the most widely

studied mechanisms by environmental engineers in efforts to understand how these

materials move through the groundwater system. The sand and clay beneath our feet act

as a natural filter for the water cycle, and by using columns packed with quartz sand8-11

and silica beads11-14 the effects of straining, particle shape, organic matter, and aquatic

solution conditions can be studied. Using engineered materials (latex colloids) to study

the effect of shape, researchers have shown that an increase in the aspect ratio increased

the retention,12 but if they were small enough along their minor-axis than they could flow

similarly to their spherical counterparts.9 Transport of bacteria through columns showed

similar findings, where longer and wider cells were often strained out by the porous

media, and those cells that passed through had aspect ratios greater than 0.5.8 It has also

been found that solution chemistry and the surface chemistry of the media and particles

plays a large role in whether anisotropic bacteria10 or particles are retained in the media.14

However, these types of batch study experiments do not allow us to track a single particle

through the media. To give us more insight into the mechanism of retention, simulations

of these pore structures allows for a more quantifiable examination of the convection-

dispersion theories that determine the flow of particles.15, 16

An important step towards understanding the diffusion of particles through a

porous media structure is examining how colloidal particles diffuse in concentrated

234

dispersions. Medina-Noyola17 postulated that short and long-time self-diffusion worked

by different mechanisms, where short-time diffusion (DS) was a purely hydrodynamic

interaction and long-time diffusion (DL) involved many body collisions. DL is therefore

smaller due to the applied friction from these collisions, and related to DS by the radial

distribution factor (g(r)). Brady18 expanded on this idea, stating that there exists a

temporal transition from short to long-time diffusion behavior that is controlled by the

thermodynamic equilibrium of the overall colloidal structure, which changes as particles

diffuse away from the center. This equilibrium is dependent on the volume fraction of

particles in the suspension, where diffusion ceases as the maximum volume fraction for

packing is reached. A concentrated colloidal dispersion can therefore be likened to a

porous media, where the volume fraction is going to play a significant role in the

diffusion of tracer particles. However, the rate of transition between short and long-time

diffusion will not be related to radial distribution of the particles making up the media

because they will be set in their equilibrium positions for the entirety of a given

experiment.

Within the last decade, direct imaging of spheres diffusing around obstacles has

been observed using different microscopy techniques. Kuznar and Elimelech19 used

fluorescence microscopy to directly image 1μm fluorescently tagged latex spheres as they

flowed through. They examined how solution conditions affected the transition from

secondary minimum deposition, to primary minimum deposition. Eral et. al.20 used

confocal microscopy to examine how the volume fraction of particles in solution affected

their settling and translational diffusion over smooth and rough surfaces. They found that

diffusion was slowed by the presence of sintered particles on the surface, and at certain

235

volume fractions the particles become trapped in layers on top of the sintered particles,

but was not so much as to form a glass. Differential dynamic microscopy was used by He

et. al.21 to track the diffusion of nanoparticles through microfabricated silica posts spaced

1.2 – 10μm apart. Results showed the translational diffusion slowed 25% between the

largest and smallest post spacing, where nanoparticles exhibited similar diffusion rates as

those seen without any posts present when diffusing through post spacing greater than

8μm. Simulation and modeling studies performed by Kim and Torquato22 and Saxton23

corroborate the experimental evidence from Eral and He. Their models suggest that

diffusion not only decreases due to exclusion-volume effects22 with increasing volume

fraction (ϕ) of diffusing particles or obstacles approaching the percolation threshold, but

trapping of a particle becomes more and more likely.23 These studies are useful

comparisons to this study regarding anisotropic particle diffusion because they discuss

sedimentation and Brownian diffusion as opposed to particle motion under flow

conditions24 or external fields.25

Experimentally measuring the diffusion of anisotropic particles in space26, 27 and

near flat surfaces28, 29 has become a subject of great interest in the last decade, building on

the work of Broersma30 and Tirado.31 Brownian dynamics simulations have greatly aided

in the understanding of anisotropic diffusion to help predict the various translational and

rotational diffusion modes. These results have shown that the addition of walls enhances

drag on a particle when the particle-wall separation distance is on the same order as the

particle diameter.29 Results from our previous study examining the influence of system

geometry and solution conditions indicated that the translational diffusion was more

sensitive to these parameters than the rotational diffusion. This has been seen before with

236

simulations performed on clusters of colloidal crystals.32 Our observed slowed diffusion

rates for confined geometries and increased ionic strength conditions correlated to a

measurable decrease in the particle to surface separation. We were able to establish those

heights using equations for the diffusion coefficients that included particle aspect ratio

and separation dependent factors.

Studies that combine the tracking of a single anisotropic particle while it

undergoes Brownian diffusion through a porous medium are few and far between. That is

why in this study we examine the diffusion behavior of anisotropic colloidal particles as

they navigate increasing area fractions of 2-dimensional porous media structures.

Previously developed and new tracking algorithms were used to analyze experiments

performed with dark field microscopy by taking advantage of the scattering efficiency of

two different materials (gold and silica). Lengths, as well as translational and rotational

diffusion coefficients were experimentally determined for individual rods, and were

compared relative to the area fraction of the porous media to determine the effective role

media concentration played in their diffusion behavior. In this chapter, preliminary

results are reported and briefly discussed.

8.2 Theory

To our knowledge, there is not currently a theory that accurately describes the

motion of individual rods through porous media. However, one can make some

qualitative guesses based on previously existing literature, such as the Brownian diffusion

of spherical particles and the diffusion of rods near surfaces.

237

8.2.1 Diffusion of Spheres through Obstacles

Calculation of the diffusion of spheres over a flat surface hindered by spherical

obstacles takes advantage of the widely used translational diffusion equations for spheres

(i.e., mean squared displacement (MSD)). However, to accurately describe the diffusion

of a sphere through 2-D porous media there must be factors that account for the size and

number of obstacles in the field that the diffusing sphere may come into contact with.

Saxton23 suggested that the MSD for the lateral diffusion of a sphere through obstacles in

two dimensions follows the form,

2 2

0

4 2 ( , )x t Dt xdxC x t x

(4.51)

for a given point in time, t, where D is a function of the concentration of obstacles at a

particular position and time, C(x,t). These two variables are related by,

21( , ) exp

4 4

xC x t

Dt Dt

(4.52)

His simulation results suggest that at moderate concentrations of random

obstacles the diffusion would decrease, but the probability of a diffusing sphere sampling

many points on the grid would not change. It would not be until the percolation threshold

is reached that a particle would no longer be able to experience long-range paths, but

instead may get trapped in a small area.

Kim and Torquato22 took a similar approach to Saxton by using an “exclusion”

probability function, Eυ(r), and a “void” nearest-neighbor probability density, Hυ(r).

These two variables are defined as the fraction of space available for exploration in a

field of spherical obstacles of density, ρ, by a particle, and the probability that an

238

arbitrary point in the system lies near a spherical obstacle between r and r + dr,

respectively, where r is the radius of a spherical empty space in the grid. Eυ(r) and Hυ(r)

are used in the calculation of the volume fraction, ϕ, and specific surface, s, respectively.

From here they suggest a calculation for the effective diffusion coefficient, De, for a

Brownian particle of radius b in a system with spherical obstacles of radius a as

De[ϕ(ρ,a);b]. They relate the De in fluid saturated porous media to the MSD by,

2

26 ( )e

XD

X

(4.53)

where X2 is the MSD and τ(X2) is the average time for a Brownian particle to hit a surface

for the first time.

8.2.2 Diffusion of Rod-Shaped Particles

As discussed at length in Section 7.2.3, we have shown that the theory we

calculated for the motion of rod-shaped particles near a flat surface agrees well with

experimental measurements. The equations generated by our rigid chain of sphere model

can be implemented into the equations for the diffusion of spheres through obstacles.

8.2.3 Diffusion of Rods through Obstacles

Similar variables will be necessary to describe the field in which the rod particles

are diffusing. These include a volume or area fraction of spherical obstacles, ϕ, which can

be used as a dimensionalized variable in place of a concentration in Saxton’s equation,

and a factor that would replace the spherical geometry of diffusing particles with a

cylindrical shape. The next step would be to know the center of mass and rod end point

coordinates at a given point in time to allow for the calculation of the MSD or MSθ,

239

which can then be directly related to an effective diffusion coefficient. This De can also

be compared to the diffusion coefficient calculated for rod-shaped particles under the

same conditions without the presence of physical obstacles that may induce

hydrodynamic hindrances. With these pieces, one could produce a reliable estimation of

the diffusion of rods through obstacles of varying concentration densities.

However, like Saxton discusses, the volume or area fraction of the obstacles is

important for explaining the MSD. Above a given ϕ, called the percolation threshold, a

particle can become trapped in a small space between clusters of obstacles. In this

instance the diffusion of the particle may appear faster than a particle that is diffusing

more freely; this can be the result of the trapped particle’s diffusion appearing more one

dimensional (1-D) in nature as opposed to 2-D. In that case, analyzing the particle’s

trajectory as 2-D motion would falsely give the particle a faster displacement. In these

cases, one must consider that as the area fraction of obstacles increases, there is the

likelihood that a particle’s motion is not strictly 1-D or 2-D, but a fractal dimension

between these two. Thus, a parameter indicating the dimension describing the particle

motion may also be employed to accurately capture the diffusion of rods through spheres.


8.3.1 Colloids and Surfaces

Hydrochloric acid, potassium hydroxide, sodium chloride (purchased from Fisher

Scientific) and colloidal SiO2 (nominal diameter of 2.34 microns, Bangs Laboratories)

were used as received and without further purification. Gold rods were synthesized at

240

Pennsylvania State University (State College, PA, USA) using an electrochemical

deposition process, and suspended in deionized water. Details regarding their synthesis

can be found in Section 7.3.1. Silica microspheres were made into a stock solution where

100μL of silica was removed from the bottle and diluted to 1mL in 0.1mM NaCl

solution.

Long glass microscope coverslips (Corning, 24 x 60 mm) were wiped clean with

lens paper, then sonicated for 30min in acetone and 30min in isopropanol before being

soaked in Nochromix overnight. Coverslips were rinsed with deionized (DI, 18.3MΩ)

water and soaked in 0.1M HCl for 30min, then rinsed with DI water again and dried with

nitrogen. Once dry, the long coverslips were placed on a WS-400BZ-6NPP/LITE spin

coater (Laurell Technologies Corporation, North Wales, PA), the silica stock was added

to the slide in 100μL aliquots, and then spun for 40sec at 1000rpms. These coverslips

were made with aliquots of silica ranging from one to five to increase the density of silica

present on the slide. After spin coating, the coverslip was placed on a hotplate and dried

overnight at 50°C. The next day silica coated coverslips were gently rinsed with DI water

from a squirt bottle and placed back on the hotplate to dry for 10min. Rinsing was

repeated three to five times to remove any crystalized salt from the silica stock before

use.

Small glass coverslips (Corning, 18 x 18 mm) were wiped with lens paper and

placed directly into Nochromix where they soaked overnight. The following day

coverslips were rinsed with DI water and soaked in 0.1M KOH for 30min. These

coverslips were rinsed and then dried with nitrogen before use.

241

Sample cells were created by dropping 12μL of the gold rod stock onto the center

of long coverslip coated with silica microspheres. A small coverslip was then placed on

top of the droplet to sandwich the particle solution between the two walls. Lens paper

was then used to wick away extraneous solution from between the walls until enough

solution has been removed to cause interference patterns to appear. Once this occurred,

the two coverslips were sealed together using Loctite Epoxy.

8.3.2 Dark Field Microscopy

Porous media experiments were performed using a Zeiss Axio Observer A1

inverted microscope with a Zeiss dry dark field condenser attachment (NA = 0.8/0.95).

All experiments were imaged using a 63x objective and particle diffusion was recorded

with a 12bit CCD camera (Hamamatsu ORCA-ER) operated in 4-binning mode at ~10fps

for approximately 30,000 frames.

8.3.3 Image Analysis

Image analysis algorithms coded in FORTRAN that were previously developed in

our lab were employed to determine the translational and rotational diffusion of rod-

shaped particles as the navigated the 2-D porous media. Additionally, a new algorithm

written in MATLAB allowed for the determination of the silica microspheres location

that make up the porous media. An image is exported from the video sequence where the

rods and the silica are not in contact with one another. This image is used by the program

to detect the silica using an 8-point connectivity thresholding process similar to that from

Chapter 7. Once an adequate threshold level has been reached that sufficiently captures

the silica microspheres, a solidity factor is entered. This factor corresponds to the percent

of a thresholded object that is filled (white pixels). Due to the different scattering

242

intensities in dark field, the silica scatters far less than the gold and appears dimmer. The

spheres appear as white halos around dark centers. This solidity factor uses those halos to

A B

C D

E F

Figure 8.1 – Different concentrations of porous media resulting in increasing area fractions of (A) 0.048, (B) 0.085, (C) 0.114, (D) 0.156, (E) 0.210, and (F) 0.245.

243

differentiate between a gold rod and a silica microsphere. From here, boundaries are

drawn around objects that meet the required solid percentage. Boundaries are not drawn

around objects above this limit (i.e., the gold rods). Once the boundaries are determined,

the area fraction (ϕ) of the image contained within those boundaries is calculated and

overlaid onto the video. This enables us to investigate the rods as a function of area

fraction of silica to determine the effects the increasing porous media concentration has

on the diffusion of rods. Figure 8.1 illustrates different area fractions determined by the

thresholded boundary conditions. Sometimes lower values of ϕ may produce more

interesting pores for the rods to navigate because the silica forms smaller clusters as in

Figure 8.1C and 8.1E, whereas the a higher ϕ shows larger, more spaced our crystals of

silica spheres (Figure 8.1D and 8.1F).


8.4.1 Tracking Rod-Shaped Particles through Porous Media

The same tracking algorithm from Chapter 7 was used to monitor the diffusion of

gold rods diffusing through silica spheres. Figure 8.2 shows the trajectories of three

different length particles through a moderately dispersed array of silica spheres. As can

be seen by the trajectories, the rods tend to move more quickly through areas with a

lower concentration of porous media, and often may remain in an open area or pocket

between clusters. All three rods exhibit some part of their trajectory where they remain in

space before diffusing elsewhere, the smallest rod being most evident (blue). The two

longer rods showed similar paths where they moved quickly through the open areas and

were able to weave between the clusters. However, we notice that when the rods do come

244

in close enough contact with the silica, they tend to remain near a cluster for a period of

time before moving on. This could be because the rod and sphere are now close enough

to feel attraction between the two surfaces. The hovering near the silica could be the

result of attractive and repulsive forces balancing each other as the rod samples different

areas of the silica cluster surface. The longest rod, shown in pink, exhibits this type of

sampling where it diffuses, hovers for a while, and then diffuses away.

8.4.2 Deciphering Calculated Mean Squared Displacements

Mean squared displacements (MSD) can give insight into what the trajectory of

the particle reflects about the porous media. Figure 8.3 shows the MSD for the three rods

at three different time intervals. Figure 8.3A examines the very short-time diffusion of

the rods. We would think that the fastest diffusing rod would be the shortest length rod at

this short window of time, but all three profiles show similar slopes. The shortest and

longest rods exhibit the same slope, and examining their position at time zero they are

Figure 8.2 – Plotted trajectories of three differently sized gold rods maneuvering through porous media with an area fraction 0.085. Green = 3.9μm, Blue = 3.4μm, and Pink = 4.5μm.

245

both further away from the silica than the rod that is 3.9μm. The increased slope could be

the result of electrostatic repulsion between the silica and the gold. At longer time

intervals (Figures 8.3B and 8.3C) we see the displacements start to deviate from one

another. At 60sec the slopes of all three rods are still similar, but we see the 3.9μm rod

slow down as it hovers in the open space between clusters. As time continues to progress,

the shortest rod takes advantage of its smaller size and the open pockets of space between

clusters to cover more space in the same time frame. The longest rod followed a similar

progression, but covered a longer lateral distance even if it was not the greatest overall

distance.

8.4.3 Comparing the Effect of Silica Area Fraction

By increasing the concentration of obstacles in the porous media, we hope to see

disruptions in the hydrodynamic interactions of the gold rods. Figure 8.4 shows how the

MSD of rods of similar length (4.5 – 4.9μm) is affected by changes in ϕ. At short

Time, ms

0 100 200 300 400 500 600

MS

D,

m2 /m

s

0.00

0.04

0.08

0.12

0.16

0.20

A

Time, sec0 10 20 30 40 50 60

MS

D, m

2/m

s

0

5

10

15

20

B

Time, sec0 100 200 300 400 500 600

MS

D,

m2/m

s

0

200

400

600

800

C

Figure 8.3 – Mean squared displacement calculations at (A) 600ms, (B) 60sec, and (C) 600sec for the three gold rods maneuvering through porous media with an area fraction 0.085. Green = 3.9μm, Blue = 3.4μm, and Pink = 4.5μm.

246

diffusion times the profiles exhibit very similar slopes, which we would expect for rods

of similar sizes. It does not take long to see changes in the diffusion profiles, and a wide

variety of slopes after 10min. There does not appear to be any real trend that the MSD

follows as a function of ϕ, but this could be in part to several factors. Since these are not

standard oriented media, the chance that a rod will sample the same area or same

trajectory as another rod is slim. Increasing the area fraction also does not necessarily

mean that the distribution of silica is uniform. The rods that appear to be translating faster

even under conditions of higher silica concentrations could be the result of the rod

finding a larger open pocket of plain silica surface to diffuse over, or in the opposite case

may be propelled by electrostatic repulsion between itself and the surrounding silica

spheres.

Time, ms0 100 200 300 400 500 600

MS

D,

m2 /m

s

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

A

Time, sec

0 10 20 30 40 50 60

MS

D, m

2/m

s

0

5

10

15

20

B

Time, sec

0 100 200 300 400 500 600

MS

D,

m2/m

s

0

50

100

150

200

250

300

C

Figure 8.4 – Mean squared displacement calculations at (A) 600ms, (B) 60sec, and (C) 600sec for the gold rods of similar sizes (4.5 – 4.9μm) maneuvering through porous media with area fractions 0.0 (black filled circles), 0.048 (red open squares), 0.085 (yellow filled triangles), 0.114 (green open upside down triangles), and 0.156 (blue filled diamonds).

247

8.5 Conclusions

Preliminary results of investigating the effect of area fraction of silica spheres on

the diffusion of gold rods of varying size show some positives and negatives.

Examination of the MSD plots at different time intervals allows us to decipher a bit about

how the porous media structure plays a role in diffusion. We saw faster translation when

rods had wide paths between clusters, but also that large open spaces essentially

“trapped” particles causing them to hover in the space for an extended time. Attractive

interactions between the gold and silica can be observed from the MSD, which shows the

rods hovering near the edges of silica clusters. Experiments examining area fraction

density are still inconclusive, as there are too many independent variables that need to be

isolated and tested. More in depth examination is needed to better understand how

exactly the porous media influences the hydrodynamic interactions of rod-shaped

particles diffusing through it.



1112335, CBET-1066254). We would also like to thank Wei Wang and Tom Mallouk

from Pennsylvania State University for giving us samples of the gold rods to use in these

studies.

248

8.7 References

1. Geng, Y.; Dalhaimer, P.; Cai, S.; Tsai, R.; Tewari, M.; Minko, T.; Discher, D. E., Shape effects of filaments versus spherical particles in flow and drug delivery. Nature Nanotechnology 2007, 2, (4), 249-255.

2. Balducci, A.; Mao, P.; Han, J.; Doyle, P. S., Double-Stranded DNA Diffusion in Slitlike Nanochannels. Macromolecules 2006, 39, (18), 6273-6281.

3. Chou, C.-F.; Bakajin, O.; Turner, S. W. P.; Duke, T. A. J.; Chan, S. S.; Cox, E. C.; Craighead, H. G.; Austin, R. H., Sorting by diffusion: An asymmetric obstacle course for continuous molecular separation. Proceedings of the National Academy of Sciences 1999, 96, (24), 13762-13765.

4. Opong, W. S.; Zydney, A. L., Diffusive and convective protein transport through asymmetric membranes. AIChE Journal 1991, 37, (10), 1497-1510.

5. Deen, W. M., Hindered transport of large molecules in liquid-filled pores. AIChE Journal 1987, 33, (9), 1409-1425.

6. Hlushkou, D.; Bruns, S.; Höltzel, A.; Tallarek, U., From Pore Scale to Column Scale Dispersion in Capillary Silica Monoliths. Analytical Chemistry 2010, 82, (17), 7150-7159.

7. Huang, L. R.; Cox, E. C.; Austin, R. H.; Sturm, J. C., Continuous Particle Separation Through Deterministic Lateral Displacement. Science 2004, 304, (5673), 987-990.

8. Weiss, T. H.; Mills, A. L.; Hornberger, G. M.; Herman, J. S., Effect of Bacterial Cell Shape on Transport of Bacteria in Porous Media. Environmental Science & Technology 1995, 29, (7), 1737-1740.

9. Xu, S.; Liao, Q.; Saiers, J. E., Straining of Nonspherical Colloids in Saturated Porous Media. Environmental Science & Technology 2008, 42, (3), 771-778.

10. Kim, H. N.; Bradford, S. A.; Walker, S. L., Escherichia coli O157:H7 Transport in Saturated Porous Media: Role of Solution Chemistry and Surface Macromolecules. Environmental Science & Technology 2009, 43, (12), 4340-4347.

11. Liu, X.; O'Carroll, D. M.; Petersen, E. J.; Huang, Q.; Anderson, C. L., Mobility of Multiwalled Carbon Nanotubes in Porous Media. Environmental Science & Technology 2009, 43, (21), 8153-8158.

249

12. Salerno, M. B.; Flamm, M.; Logan, B. E.; Velegol, D., Transport of Rodlike Colloids through Packed Beds. Environmental Science & Technology 2006, 40, (20), 6336-6340.

13. Bozorg, A.; Gates, I. D.; Sen, A., Real time monitoring of biofilm development under flow conditions in porous media. Biofouling 2012, 28, (9), 937-951.

14. Yang, J.; Bitter, J. L.; Smith, B. A.; Fairbrother, D. H.; Ball, W. P., Transport of Oxidized Multi-Walled Carbon Nanotubes through Silica Based Porous Media: Influences of Aquatic Chemistry, Surface Chemistry, and Natural Organic Matter. Environmental Science & Technology 2013, 47, (24), 14034-14043.

15. Adler, P. M.; Brenner, H., Multiphase Flow in Porous Media. Annual Review of Fluid Mechanics 1988, 20, (1), 35-59.

16. Adler, P. M.; Jacquin, C. G.; Quiblier, J. A., Flow in simulated porous media. International Journal of Multiphase Flow 1990, 16, (4), 691-712.

17. Medina-Noyola, M., Long-Time Self-Diffusion in Concentrated Colloidal Dispersions. Physical Review Letters 1988, 60, (26), 2705-2708.

18. Brady, J. F., The long-time self-diffusivity in concentrated colloidal dispersions. Journal of Fluid Mechanics 1994, 272, 109-134.

19. Kuznar, Z. A.; Elimelech, M., Direct microscopic observation of particle deposition in porous media: Role of the secondary energy minimum. Colloids and Surfaces A: Physicochemical and Engineering Aspects 2007, 294, (1â€“3), 156-162.

20. Eral, H. B.; Mugele, F.; Duits, M. H. G., Colloidal Dynamics Near a Particle-Covered Surface. Langmuir 2011, 27, (20), 12297-12303.

21. He, K.; Babaye Khorasani, F.; Retterer, S. T.; Thomas, D. K.; Conrad, J. C.; Krishnamoorti, R., Diffusive Dynamics of Nanoparticles in Arrays of Nanoposts. ACS Nano 2013, 7, (6), 5122-5130.

22. Kim, I. C.; Torquato, S., Diffusion of finite-sized Brownian particles in porous media. The Journal of Chemical Physics 1992, 96, (2), 1498-1503.

23. Saxton, M. J., Lateral diffusion in an archipelago. Single-particle diffusion. Biophysical Journal 1993, 64, (6), 1766-1780.

24. Kemps, J. A. L.; Bhattacharjee, S., Particle Tracking Model for Colloid Transport near Planar Surfaces Covered with Spherical Asperities. Langmuir 2009, 25, (12), 6887-6897.

250

25. Ghosh, P. K.; Hänggi, P.; Marchesoni, F.; Martens, S.; Nori, F.; Schimansky-Geier, L.; Schmid, G., Driven Brownian transport through arrays of symmetric obstacles. Physical Review E 2012, 85, (1), 011101.

26. Mukhija, D.; Solomon, M. J., Translational and rotational dynamics of colloidal rods by direct visualization with confocal microscopy. Journal of Colloid and Interface Science 2007, 314, (1), 98-106.

27. Aragon, S. R.; Flamik, D., High Precision Transport Properties of Cylinders by the Boundary Element Method. Macromolecules 2009, 42, (16), 6290-6299.

28. Marshall, B.; Davis, V.; Lee, D.; Korgel, B., Rotational and translational diffusivities of germanium nanowires. Rheologica Acta 2009, 48, (5), 589-596.

29. Padding, J. T.; Briels, W. J., Translational and rotational friction on a colloidal rod near a wall. The Journal of Chemical Physics 2010, 132, (5), 054511-8.

30. Broersma, S., Viscous force and torque constants for a cylinder. The Journal of Chemical Physics 1981, 74, (12), 6989-6990.

31. Tirado, M. M.; de la Torre, J. G., Translational friction coefficients of rigid, symmetric top macromolecules. Application to circular cylinders. The Journal of Chemical Physics 1979, 71, (6), 2581-2587.

32. Lele, P. P.; Swan, J. W.; Brady, J. F.; Wagner, N. J.; Furst, E. M., Colloidal diffusion and hydrodynamic screening near boundaries. Soft Matter 2011, 7, (15), 6844-6852.

251

JULIE L. BITTER Department of Chemistry, Johns Hopkins University

3400 N. Charles St. NCB 228 Baltimore, MD 21218 [email protected]

609-433-7583 EDUCATION: The Johns Hopkins University, Baltimore, MD, USA Ph.D. Chemistry (Mar 2014)

M.A. Chemistry (Jan 2011) University of Central Florida, Orlando, FL, USA B.S. Forensic Science (May 2008)

Minors: Chemistry & Criminal Justice LABORATORY SKILLS: Chromatography: GC-MS, LC-MS, Thin Layer (TLC) Microscopy: Atomic Force (AFM), Comparison, Confocal, Dark Field, Optical, Polarized

Light (PLM), Scanning Electron (SEM), Total Internal Reflection (TIRM) Spectroscopy: FTIR/ATR, Ion Mobility (IMS), Mass Spectrometry (ion trap, quadrupole),

UV-Visible, X-ray Photoelectron (XPS) Analytical Techniques: Conductivity, Contact Angle, pH, Pipetting, Refractometry,

Sedimentation, Surface Charge Titration, Total Organic Carbon, Wet Chemical Extraction

Light Scattering: Dynamic Light Scattering (DLS), Zeta Potential Computer: AugerScan; CASA XPS; ChemDraw; ImageJ; Mathcad; Microsoft Excel,

PowerPoint, Word; SigmaPlot; StreamPix; VideoMach RESEARCH EXPERIENCE: The Johns Hopkins University, Baltimore, MD, USA Graduate Research Assistant (Sep 2010 to Present) Supervisor: Dr. Michael Bevan Study potential energy interactions of silica microspheres with environmental

surfaces under various aquatic conditions in confined experiments to determine contributions of attractive repulsive forces using TIRM, as well as determine the importance of the role that surface chemistry plays in these interactions

Elucidate new information on the hydrodynamic interactions of rod-shaped particles with silica surfaces and with other particles to garner quantitative measurements of force, potential, and diffusivity.

Track colloidal rods as they navigate 2D porous media to gain understanding of the forces that regulate the transport of nanomaterials in the environment

252

Perform literature research, edit documents for coworkers and the principal investigator (manager), develop and edit written and visual proposal/grant content including figures, tables, and plots using Microsoft Office applications and SigmaPlot for my own research or for annual reports to funding agencies

Graduate Research Assistant (Sep 2008 to Present) Supervisor: Dr. Howard Fairbrother Characterize and analyze carbon nanotube structure, colloidal, and transport

properties using UV-Vis, AFM, DLS, XPS, and chemical derivatization techniques

Examine how the physical and chemical nature of CNTs exposed to UV light is altered, as well as examine the products formed by CNTs degraded by the UV light

Create samples for use in the laboratory as well as for others within and outside the university, maintain detailed records of all experimental procedures and results including tables and plots, standardize instruments and keep them in good working order

Perform literature research, edit documents for coworkers and the principal investigator (manager), develop and edit written and visual proposal/grant content including figures, tables, and plots using Microsoft Office applications and SigmaPlot for my own research or for annual reports to funding agencies

Provide XPS analytical services to seven independent research groups both affiliated and not affiliated with The Johns Hopkins University leading to two peer reviewed publications

Provide FTIR analytical services to undergraduate classes and independent research groups both affiliated and not affiliated with The Johns Hopkins University

Train graduate and undergraduate research students in applying knowledge to create and perform useful experiments, and teach them good laboratory and safety protocols

University of Central Florida/National Center for Forensic Science, Orlando, FL, USA Undergraduate Research Assistant (Jan 2006 to May 2008) Supervisor: Dr. Michael Sigman Examined the retention times of explosives (e.g., TNT, RDX, and PETN) on

different solid matrices by analyzing residues with GC-MS and IMS to determine detection limits

Examined pre and post blast material of TATP to compare signature features for trace detection using GC-MS, LC-MS, and IMS

Created methods for sample handling and preservation of explosive residues on evidence to be used in field tests and real life scenarios

253

New Jersey State Police Office of Forensic Sciences, Central Regional Lab, Hamilton, NJ, USA Summer Intern (Jun 2007 to Aug 2007) Supervisor: Edward Gainsborg, FSIII Prepared and mounted samples of canine hair requested from purebred breeders

all over the country for a microscopy collection Created tests for hair/fiber examiners using the canine hair samples to discern

differences among similar colors, lengths, and within breeds Shadowed examiners from physical evidence section to assist with casework by

applying knowledge of presumptive and confirmatory tests for blood, semen, saliva, and urine, and other spectroscopic and microscopic techniques for the analysis of trace evidence

TEACHING EXPERIENCE: Independent Tutor General Chemistry I and II to three undergraduate students A.P. Chemistry to one high school student

The Johns Hopkins University Graduate Teaching Assistant (Sep 2008 to June 2010) Teaching assistant for undergraduate physical chemistry lab for three semesters Teaching assistant for undergraduate introduction to chemistry lecture for one

semester Duties included teaching laboratory experiments, grading lab reports, running

recitation sessions, grading exams, and grading oral presentations University of Central Florida Undergraduate Teaching Assistant (Aug 2007 to Dec 2007) Teaching assistant for undergraduate microscopy course for one semester Duties include teaching laboratory experiments, running laboratory

demonstrations, grading quizzes PRESENTATIONS: Platform

1. Bitter, J.L.; Duncan, G.A.; Bevan, M.A.; Fairbrother D.H.; “Quantitative Analysis of Particle-Surface Interactions in Aqueous Environments using Total Internal Reflection Microscopy.” 245th American Chemical Society National Meeting, New Orleans, LA, USA. April 7-11, 2013.

2. Bitter, J.L.; Duncan, G.A.; Fairbrother D.H.; Bevan, M.A. “Using Total Internal Reflection Microscopy to Quantitatively Interrogate Particle Interactions at Silica-Water Interfaces.” 245th American Chemical Society National Meeting, New Orleans, LA, USA. April 7-11, 2013.

254

3. Bitter, J.; Duncan, G.; Fairbrother H.; Bevan, M. “Direct Measurements of Non-DLVO Forces between Silica Colloids and Surfaces.” 86th Colloids and Surfaces Symposium, Baltimore, MD, USA. June 10-13, 2012.

4. Bitter, J.; Duncan, G.; Eichmann, S.; Smith, B.; Bevan, M.; Fairbrother, H. “Using Real Time Imaging to Quantify Nanoparticle Dynamics and Surface Interactions in Aquatic Media.” 241st American Chemical Society National Meeting, Anaheim, CA, USA. March 27-31, 2011.

Poster

1. Bitter, J.; Yang, J.; Beigzadehmilani, S.; Jafvert, C.; Fairbrother, H. “Photochemical Transformations of Carbon Nanotubes.” 2nd Gordon Research Conference on Environmental Nanotechnology, Stowe, VT, USA. June 2-7, 2013.

2. Bitter, J.L.; Smith, B.A.; Wepasnick, K.A.; Fairbrother, H. “Influence of Oxygen Containing Functional Groups on the Transport Properties of Carbon Nanotubes in Saturated Porous Media.” 48th Eastern Analytical Symposium, Somerset, NJ, USA. November 16-19, 2009.

3. Bitter, J.; George, K.; Clark, D.; Sigman, M. “Explosive Recovery Off of Solid Matrices Over Time.” 33rd Federation of Analytical Chemistry and Spectroscopy Societies Meeting, Lake Buena Vista, FL, USA. September 24-28, 2006.

JOURNAL PUBLICATIONS:

1. J.L. Bitter, Y. Yang, G.A. Duncan, D.H. Fairbrother, M.A. Bevan, “Capturing the Diffusion of Micron-Sized Gold Rods Across Silicate Surfaces and through Slit Pores.” In Prep.

2. J.L. Bitter, J. Yang, S. Beigzadehmilani, C. Jafvert, D.H. Fairbrother, “Photochemical Transformations of Oxidized Multiwalled Carbon Nanotubes as a Result of Exposure to UVC Irradiation”. In Prep.

3. J. Yang, J. Bitter, B. Smith, H. Fairbrother, W. Ball, “Transport of Oxidized Multi-Walled Carbon Nanotubes through Silica Based Porous Media: Influences of Aquatic Chemistry, Surface Chemistry and Natural Organic Matter”. Environmental Science & Technology, 47: p. 14034-14043, 2013.

4. L. Tang, P. Yi, W. Gu, J. Bitter, H. Fairbrother, K.L. Chen, “Bacterial Anti-Adhesive Properties of Polysulfone Membranes Modified with Poly(allylamine hydrochloride) and Poly(acrylic acid) Multilayers”. Journal of Membrane Science, 446: p. 201-211, 2013.

5. J.L. Bitter, G.A. Duncan, D.J. Beltran-Villegas, D.H. Fairbrother, M.A. Bevan, “Anomalous Silica Colloid Stability and Gel Layer Mediated Interactions”. Langmuir, 29: p. 8835-8844, 2013.

6. B. Smith, J. Yang, J. Bitter, W.P. Ball, and D.H. Fairbrother, “Influence of Surface Oxygen on the Interactions of Carbon Nanotubes with Natural Organic Matter”. Environmental Science and Technology, 46: p. 12839−12847, 2012.

7. W. Boncher, E. Gorlich, K. Tomala, J. Bitter, S. Stoll, “Valence and Magnetic Investigations of Alkali-metal Doped Europium Sulfide”. Chemistry of Materials, 24: p. 4390–4396, 2012.

255

8. K. Wepasnick, B. Smith, J. Bitter, D.H. Fairbrother, “Chemical and structural characterization of carbon nanotube surfaces”. Analytical and Bioanalytical Chemistry, 396: p. 1003-1014, 2010.

9. M. Sigman, C.D. Clark, K. Painter, C. Milton, E. Simatos, J. Frisch, M. McCormick, J. Bitter, “Analysis of oligomeric peroxides in synthetic triacetone triperoxide samples by tandem mass spectrometry.” Rapid Communications in Mass Spectrometry, 23: p. 349-356, 2009.

AWARDS: Top 3 Posters – 2013 Gordon Conference on Environmental Nanotechnology University Honors: 2008 Student-Mentor Academic Research Team (SMART) Grant: 2006 Dean’s List – College of Sciences: 2004 to 2008 PROFESSIONAL AFFILIATIONS: American Chemical Society: 2009 to present American Academy of Forensic Sciences: 2006 to 2008

March 2014

fate and tracking of engineered nanomaterials in aqueous environments › bitstream › handle ›...

Documents