Development of Icp-Ms Based Nanometrology Techniques for Characterization of Silver

DEVELOPMENT OF ICP-MS BASED NANOMETROLOGY TECHNIQUES FOR CHARACTERIZATION OF SILVER

NANOPARTICLES IN ENVIRONMENTAL SYSTEMS

by Denise Marie Mitrano A thesis submitted to the Faculty and the Board of Trustees of the Colorado School of Mines in partial fulﬁllment of the requirements for the degree of Philosophy of Science (Geochemistry).

Golden, Colorado Date

Signed: Denise Marie Mitrano

Signed: Dr. James F. Ranville Thesis Advisor

Golden, Colorado Date

Signed: Dr. David T. W. Wu Professor and Head Department of Chemistry and Geochemistry

ii ABSTRACT

The ubiquitous use of goods containing nanoparticles (NPs) will lead inevitably to environmental release and interaction with biota. Methods to detect, quantify, and characterize NPs in environmental matrices are highlighted as one of the areas of highest priority research in understanding potential environmental and health risks. Specifically, techniques are needed to determine the size and concentration of NPs in complex matrices. Particu- lar analytical challenges include distinguishing NPs from other constituents of the matrix (i.e. natural particles, humic substances, and debris), method detection limits are often higher than exposure concentrations, and differentiating dissolved metal and NPs. This work focuses on the development and optimization of two methods that address a number of challenges for nanometrology: single particle (sp)ICP-MS and asymmetrical flow field flow fractionation (AF4)-ICP-MS. Advancements in the spICP-MS method included systematic studies on distinction between ionic and NP fractions, resolution of polydisperse NP samples, and defining the techniques’ dynamic range (in terms of both particle size and concentration). Upon application of the technique, silver (Ag) NPs were discovered in raw wastewater treatment plant influent and effluent. Furthermore, methodical Ag NP stability studies determined the influence of particle capping agents and water chemistry parameters in a variety of synthetic, natural and processed waters. Method development for AF4-ICP-MS revolved around optimizing run conditions (i.e. operational flows, carrier fluid, membrane choice) to study detection limits, sample recovery, and resolution of polydisperse samples. Practical studies included sizing Ag NP in a sediment-dwelling, freshwater oligochaete (Lumbriculus variegatus) and the kinetics of accumulation of protein bound Ag+.Indirectcomparison, spICP-MS was found to be more versatile with less sample preparation and lower total analyte detection limit (ng/L vs. g/L), though AF4-ICP-MS could detect smaller particle sizes (2 nm vs. 25 nm) and elucidate NP/matrix interactions for changes in particle hydrodynamic

iii diameter. Additionally, spICP-MS afforded us the opportunity to determine the kinetic rate of Ag NP dissolution rate kinetics at environmentally relevent concentrations, the first study of its kind. We found significantly variable dissolution rates for differently capped NPs in addition to water chemistries. Tannic acid capping agent was least resistant to dissolution compared to citrate and PVP, while high concentrations of natural organic matter seemed to stabilize the particles over time in comparison to DI water. The residual chlorine in tap water increased the dissolution rates of all particles dramatically, which we hypothesize to be due to residual chlorine. Herein is described method development protocol and results of aforementioned studies comparing sp and AF4 ICP-MS and supporting their use as choice nanometrology techniques for quantitative environmental and toxicological studies.

iv TABLE OF CONTENTS

ABSTRACT ...... iii

LISTOFFIGURES ...... xii

LISTOFTABLES ...... xiv

LISTOFABBREVIATIONS ...... xv

ACKNOWLEDGMENTS ...... xvii

CHAPTER1 INTRODUCTION ...... 1

1.1 There’s plenty of room at the bottom: the rise of the nanotechnology industry . 2

1.2 Silver nanotechnology and the environment: bactericidal eﬀects and consequences of nanosilver ...... 5

1.2.1 Ionic versus particle silver behavior ...... 5

1.2.2 Toxicityof(nano)silver: acellularapproach ...... 7

1.2.3 Nanosilver, bacteria, and wastewater treatment plants ...... 9

1.3 Potential transformations of silver nanoparticles in natural and treated watersystems ...... 13

1.3.1 Potential scenarios for nanoparticle release and subsequent transformation in the environment ...... 14

1.3.2 Alterations of silver nanoparticles in complex media ...... 16

1.3.3 Occurance of silver nanoparticles in wastewater treatment plants and subsequent transformation ...... 20

1.4 Nanometrology: analytical techniques for nanoparticle detection ...... 22

1.5 Purpose and Signiﬁcance ...... 25

1.6 Organization of This Work ...... 27

v CHAPTER 2 INVESTIGATIONS FOR THE FEASIBILLITY OF DETECTING NANOPARTICULATE SILVER USING SINGLE PARTICLE INDUCTIVELY COUPLED PLASMA-MASS SPECTROMETRY . . . 28

2.1 Abstract...... 28

2.2 Introduction ...... 29

2.3 Materials and Methods ...... 32

2.3.1 AccompanyingStudies ...... 39

2.4 ResultsandDiscussion ...... 39

2.4.1 ICP-MS Recovery ...... 39

2.4.2 Filtration ...... 40

2.4.3 Analysis of Commercial Colloidal Silver Suppliment (ASAP) via Sp-ICP-MS ...... 42

2.4.4 Wastewater Inﬂuent and Outﬂuent Samples ...... 44

CHAPTER 3 SILVER NANOPARTICLE CHARACTERIZATION USING SINGLE PARTICLE (SP)-ICP-MS AND ASYMMETRICAL FLOW FIELD FLOW FRACTIONATION (AF4)-ICP-MS ...... 48

3.1 Abstract...... 48

3.2 Introduction ...... 49

3.3 Materials and Methods ...... 52

3.3.1 Materials ...... 52

3.3.2 Instrumentation-sp-ICP-MS ...... 53

3.3.3 Instrumentation-AF4-ICP-MS ...... 53

3.3.4 Data collection, conversion to particle size, and quality of analysis - spICP-MS ...... 54

3.3.5 Data collection, converstion to particle size, and quality of analysis - AF4-ICP-MS ...... 57

vi 3.3.6 Size, detection limit, and resolution experimental parameters ...... 58

3.3.7 Multi-form analysis ...... 59

3.4 ResultsandDiscussion ...... 60

3.4.1 Optimization for sp-ICP-MS ...... 60

3.4.2 MethodOptimizationforAF4-ICP-MS ...... 63

3.4.3 spICP-MS and AF4-ICP-MS comparison: Detection limit, NP size . . . 64

3.4.4 spICP-MS and AF4-ICP-MS comparison: Dynamic range, NP concentration ...... 66

3.4.5 spICP-MS and AF4-ICP-MS comparison: Resolution ...... 70

3.4.6 Multi-form analysis: Dissolved versus NP constituents...... 74

3.4.7 Multi-formanalysis: NPcomplexes ...... 76

3.4.8 Multi-form analysis: Multiple metals analysis ...... 76

3.5 Conclusions ...... 77

3.5.1 Advantages and limitations of using spICP-MS in NP characterization . 78

3.5.2 Advantages and limitations of using AF4-ICP-MS in NP characterization...... 79

3.5.3 Future development in the areas of spICP-MS and AF4-ICP-MS . . . . 79

CHAPTER 4 TRACKING TRANSFORMATIONS OF SILVER NANOPARTICLES IN SYNTHETIC, NATURAL, AND PROCESSED WATERS USING SINGLE PARTICLE (SP)ICP-MS . . . 81

4.1 Introduction ...... 82

4.2 Materials ...... 85

4.3 Methods ...... 86

4.3.1 Instrumentation...... 88

4.3.2 Data Collection, Conversion to Particle Size, and Quality Assurance . . 88

vii 4.3.3 Characterization of Particle Stability and Silver Release ...... 89

4.3.4 Dissolution Rate Kinetics ...... 90

4.3.5 Preliminary Control Studies ...... 90

4.4 ResultsandDiscussion ...... 91

4.4.1 Preliminary Control Studies ...... 91

4.4.2 Eﬀects of Water Chemistry on Ag NP Stability - Mechanistic Studies . 94

4.4.3 Environmental Systems ...... 99

4.4.4 Processed Water Samples ...... 99

4.4.5 Comparison of Water Chemistry Eﬀects ...... 101

4.4.6 Kinetic Rates of Dissolution ...... 101

4.5 Implications ...... 104

4.5.1 Eﬀects of Water Chemistry on ENP Transformation in Aerobic Systems ...... 105

4.5.2 Eﬀects of Water Chemistry on ENP Transformation in Anaerobic Systems ...... 107

4.5.3 Advancement of ENP Studies with the Use of spICP-MS ...... 108

CHAPTER 5 CONCLUSIONS ...... 109

5.1 Feasibility of detection nanoparticulate silver using spICP-MS ...... 109

5.2 Optimization and Comparison of nanometrology techniques ...... 110

5.3 Case Study: Application of spICP-MS to study dissolution kinetics of Ag NPs ...... 111

5.4 Collaborative eﬀorts and future work ...... 113

REFERENCESCITED ...... 117

APPENDIX A - AN INTRODUCTION TO FLOW FIELD FLOW FRACTIONATION AND COUPLING TO ICP-MS ...... 133

viii A.1 Introduction...... 133

A.2 FieldFlowFractionation ...... 133

A.3 FFFOperationandSeparationTheory ...... 134

A.4 CouplingFFFtoICP-MS ...... 138

APPENDIX B - COUPLING FLOW FIELD FLOW FRACTIONATION TO ICP-MS FOR THE DETECTION AND CHARACTERIZATION OF SILVER NANOPARTICLES ...... 140

B.1 Introduction ...... 140

B.2 Nanometrology ...... 141

B.3 Experimental ...... 142

B.3.1 Materials ...... 142

B.3.2 Instrumentation ...... 142

B.3.3 Daily Standards ...... 143

B.4 AnalyticalResults...... 145

B.4.1 Resoution and detection limit ...... 145

B.4.2 MixedmetalanalysiswithﬂowFFF-ICP-MS ...... 146

B.5 Conclusions ...... 148

APPENDIX C - FIELD-FLOW FRACTIONATION COUPLED WITH ICP-MS FOR THE ANALYSIS OF ENGINEERED NANOPARTICLES IN ENVIRONMENTAL SYSTEMS ...... 149

C.1 EngineeredNanomaterials ...... 149

C.2 Potential for Environmental Impact ...... 150

C.3 AnalyticalMethodologies...... 151

C.4 Field-Flow-Fractionation ...... 154

C.5 FFFCoupledwithICP-MS ...... 155

ix C.6 ParticleSizeReferenceStandards ...... 157

C.7 Calibration Strategies ...... 159

C.8 Recovery ...... 159

C.9 Detection Limits ...... 160

C.10 Conclusion ...... 163

APPENDIX D - SINGLE PARTICLE ICP-MS STANDARD OPERATING PROCEDURE...... 164

D.1 General lab practices and recording ...... 164

D.1.1 Scheduling and recording ...... 164

D.1.2 GenerallabpracticesforNexION ...... 164

D.2 Instrument start up procedure and tuning for sp-ICP-MS ...... 164

D.2.1 Preparting to start up ...... 165

D.2.2 Turningontheplasma ...... 166

D.2.3 DailyTuningofNexIONICP-MS ...... 166

D.2.4 Tuning for sp-ICP-MS (Optional) ...... 166

D.2.5 Standard calibration curve ...... 167

D.2.6 Gold nanoparticle eﬃciency calibration ...... 168

D.3 Analyzingsamplesinsp-ICP-MSmode ...... 168

D.3.1 Beginning a run ...... 169

D.3.2 Observations during a run ...... 169

D.4 Instrumentshutdown ...... 170

D.4.1 Cleaning tubing lines ...... 170

D.4.2 Shut down and ﬁnal checks ...... 170

x APPENDIX E - SUPPORTING AUTHOR PUBLICATIONS ...... 172

xi LIST OF FIGURES

Figure 2.1 Characterization of commercial colloidal Ag NP solution, ASAP . . . . . 33

Figure2.2 ConceptualdiagramforspICP-MS...... 35

Figure 2.3 Dataset manipulation to diﬀerentaite between dissolved and nanoparticulate silver ...... 37

Figure 2.4 Particle ionization eﬃciencyinICP-MSplasma...... 40

Figure 2.5 Success of ﬁltration as a preparation tequniqe for various sized Ag NPs . 41

Figure2.6 Rawdataoutput,sp-ICP-MSanalysis ...... 42

Figure2.7 CorrelationofpulsestoNPconcentration ...... 43

Figure2.8 Proofofconcept: AgNPscreatepulses ...... 45

Figure 2.9 Evidence of dissolved and nanoparticulate Ag in treated waters ...... 46

Figure3.1 Schematicofsp-ICP-MSdataprocessing ...... 56

Figure 3.2 Eﬀects of tuning ICP-MS prior to sp-ICP-MS analysis ...... 61

Figure 3.3 Analysis of dwell time choice for optimal sp-ICP-MS data collection parameteres ...... 62

Figure3.4 Typicalsp-ICP-MShistogramofrawdata...... 65

Figure3.5 Dynamicrangeandresolutionofsp-ICP-MS ...... 68

Figure 3.6 Concentration based detection limit, AF4-ICP-MS ...... 70

Figure3.7 Resolutionofparticlesizemixtures,spICP-MS ...... 72

Figure 3.8 Study of AF4 ﬂow rates for optimization of resolution ...... 73

Figure3.9 ResolutionofAgNPmixture,AF4-ICP-MS ...... 73

Figure 3.10 Analysis of Ag+ andAgNPmixtures,spICP-MS...... 75

xii Figure 3.11 Ag NP, Ag+, and BSA interaction analysis via AF4-ICP-MS ...... 77

Figure4.1 High(er)ConcentrationDissolutionOverTime ...... 92

Figure4.2 DissolutionviaPulseIntensityChange ...... 93

Figure4.3 DissolutionMassBalanceAnalysis...... 94

Figure 4.4 Dissolution Experiments in Laboratory Prepared Waters ...... 98

Figure 4.5 Dissolution Experiments in Environmental Waters ...... 100

Figure4.6 DissolutionExperimentsinProcessedWaters ...... 101

Figure4.7 ComparisonofDissolutioninVariousMatrices ...... 102

Figure4.8 KineticRatesofDissolutionforAgENPs ...... 104

FigureA.1 SchematicoftheAF4channel ...... 134

FigureA.2 CrosssectionofanAF4channel ...... 136

FigureA.3 ExampleofAF4calibration ...... 137

Figure A.4 Schematic of AF4-ICP-MS analysis with online addition of UV/Vis and ICP-MS analysis...... 138

FigureB.1 Analysisof35and100ppbnano-Agmixture ...... 146

FigureB.2 Multi-elementFFF-ICP-MSfractogram ...... 147

FigureC.1 FFFchannelandﬂowproﬁle ...... 155

Figure C.2 FFF UV/VIS fractogram ...... 156

FigureC.3 DiagramofFFF-ICP-MScoupling ...... 157

Figure C.4 Comparison of FFF Fractogram with ICP-MS and UV/VIS as online detector ...... 158

FigureC.5 RecoveryfortheFFFtechnique ...... 161

FigureC.6 Multi-elementanalysisviaFFF-ICP-MS ...... 162

xiii LIST OF TABLES

Table2.1 ICP-MSoperatingconditions...... 34

Table 3.1 Analysis of Ag+ andAgNPmixtures,spICP-MS ...... 76

Table4.1 WaterChemistryComposition ...... 87

Table 4.2 EﬀectofLightTreatmentsonParticleDissolution ...... 92

Table4.3 DissolutionMassBalanceSummary–DIWater ...... 95

Table 4.4 Dissolution Mass Balance Summary – Clear Creek Water ...... 96

Table4.5 SummaryofExperimentalDissolutionRates ...... 103

TableB.1 FFFParameters...... 144

TableB.2 ICP-MSParameters...... 144

xiv LIST OF ABBREVIATIONS

Asymmetrical Flow Field Flow Fractionation ...... AF4

Bovine serum albumen ...... BSA

CarbonNanotube ...... CNT

Carrierﬂuid...... CF

Counts per second ...... cps

Detection limit ...... DL

Diﬀerential Centrifugal Sedimentation ...... DCS

Dissolved Oxygen ...... DO

Dissolved organic matter ...... DOM

Dynamic Light Scattering ...... DLS

Engineerednanoparticle ...... ENP

Fieldﬂowfractionation...... FFF

Inductively Coupled Plasma - Mass Spectrometry ...... ICP-MS

Inductively Coupled Plasma Atomic Emissions Spectroscopy ...... ICP-AES

Nanomaterial ...... NM

Nanoparticle ...... NP

NanoparticleTrackingAnalysis ...... NTA

NationalInstituteofStandardsandTechnology ...... NIST

Naturalorganicmatter...... NOM

Quantum Dot ...... QD

xv Quartzcyrstalmicrobalance ...... QCM

Reactive Oxygen Species ...... ROS

SuwaneeRiverhumicacid ...... SRHA

TransmissionElectronMicroscopy ...... TEM

Ultravioletabsorbance(signal) ...... UV

UnitedStatesEnvironmentalProtectionAgency ...... USEPA

WastewaterTreatmentPlant ...... WWTP parts per billion ...... ppb parts per million ...... ppm parts per trillion ...... ppt single particle ICP-MS ...... sp-ICP-MS

xvi ACKNOWLEDGMENTS

IfeelincrediblyfortunatetohavehadtheopportunitytostudyandliveinColorado,and to have managed a good balance between work and play in such an inspiring environment. So, first and foremost, I would like to thank everyone these last years who has helped me achieve this; to be curious about science, creative, active, healthy, and adventurous. I am very grateful to both Dr. Jim Ranville and Dr. Chris Higgins, who encouraged me endlessly, were supportive and understanding, helped me find my focus, gave me frank advice about science and all my future endeavors, and indulged me with opportunities to explore new avenues (and places!). I was happy to have supportive colleagues in the research group, some for many years, including Valerie Stucker, Rob Reed, Evan Gray, Angie Barber, Dr. Emily Lesher, Dr. Heather Pace, and Manuel Montano. The day-to-day challenges of grad school life were made easier by my office mates, Steph Carr and Andy Glossner, who provided some great advice in addition to welcome distractions. To the foundation of my scientific interests and patience with my initial forays into research I thank my undergraduate advisor, Dr. Sandor Kadar. For his encouragement to pursue my varied interests and giving me new perspectives on innumerable issues I am grateful; a good mentor and a good friend. For all those too numerous to mention that have made Golden feel like home, if even for a short while, thank you. Especially, Brian Quarnstrom, Julieta Giraldez, Nick Venechuck, Karen and Dave Bobella, Robert Jinkerson, Halley Eydal, and Elizabeth Easley. To my friends now abroad, thank you for exploring new environments with me and for your hos- pitality during my travels; I will see you again soon as I embark on my next chapter as a postdoc. And finally, to Aaron Pugmire, my best friend and strongest supporter: thank you for sharing your adventurous spirit, always putting things into perspective, and your adopted

xvii love of listening about nanometrology these past years.

xviii CHAPTER 1 INTRODUCTION

New technologies often have unanticipated or unwanted consequences. Asbestos, coal, leaded gasoline, and freons are all examples of useful substances with unforeseen effects. Notably, when DDT and other organic pesticides were used, unacceptable consequences to the health of the environment and people were seen; the key event that started an environmental revolution. From the past we have learned that caution must be taken to ensure full benefits from scientific advancements without detrimental effects. Today, nanotechnology, the control of matter at dimensions of roughly 1 to 100 nanometers, has great potential in both industrial and commercial sectors. Nanomaterials (NMs) are already recognized as useful products for society either when used alone or when integrated into larger products (e.g. consumer goods, foods, pesticides, pharmaceuticals, personal care products, among others). With the list of applications growing rapidly, human exposure and environmental prevalence of NMs is expected to increase substantially. Detection and characterization of NMs is an essential part of understanding the potential benefits as well as the potential risks of the application of nanoparticles (NPs). One of our great challenges lies in the innate difficulties of NP detection in complex matrices such as environmental, biological, and food samples. Many of the current detection and characterization methods that have been used to study other pollutants are not adequate. The systems of interest are complex and involve heterogeneous matrices which may contain very low (parts per trillion) levels of NPs. The biological impacts of NP and biokinetics of nanoparticle toxicity are dependent on size, chemical composition, surface structure, solubility, shape, and aggregation. Confounding the issue is that metal NPs are postulated to be linked to the release of dissolved metal, which confers a different form of toxicity. Lastly, the final fate of NPs is highly dependent on a host of environmental and biological parame-

1 ters: concentration of ligands, interactions with organic matter, ionic strength, and pH. The objective of the proposed research is to evaluate current analytical detection approaches and subsequently improve these techniques to develop new avenues enabling toxicologists and legislators to establish the risk potential of NPs. The US National Science and Technology Council identified top priority research needs for NPs in the national nanotechnology initiative (2008) which included development of instrumentation, metrology, and analytical methods to detect NPs in biological matrices, the environment and the workplace. In response to this, the US Environmental Protection Agency (EPA) noted in their NM research strategy in 2009 that analytical detection and method development underscores five of seven categories of comprehensive environmental assessment. Likewise, EFSA, the European Food and Safety Authority, concluded in 2009 that actions should be taken to develop methods to detect and measure engineered NPs in food/feed and biological tissues as a prerequisite to assess exposure and carry out toxicological studies. Adequate oversight of new technologies will depend on our ability to forecast the possible risk the technologies pose. As outlined in a report by the Woodrow Wilson Project on Emerging Nanotechnologies in 2009, international cooperation is paramount in making effective technology oversights as the combination of a worldwide economy and international trade will move products across country and continent borders. It is imperative that the scientific community begin to address the issue of risks, such as the emergence of NPs, in order to strike the balance between the harm that could be done by proceeding with an innovation and the harm that could be done by not proceeding

1.1 There’s plenty of room at the bottom: the rise of the nanotechnology industry

Many suggest that Nobel lauriate Richard Feynman ushered in the era of nanotechnology when he gave a presentation entitled, “There’s Plenty of Room at the Bottom”, at Caltech in December 1959. Feynman considered a number of interesting ramiﬁcations of a general ability to manipulate matter on a atomic scale, yet the ﬁeld did not truly come into its own

2 until decades later. With the publication of his 1986 book, Engines of Creating: The Coming Era of Nanotechnology, K. Eric Drexler brought the nanotechnology field into the limelight. In subsequent years, scientific journals gave more attention to the topic with the journal Nanotechnology launching in 1989 and Science printing a special issue on nanotchnology in November 1991. Today, nanotchnology is one of the fasting growing and most promising technologies in our society. Nanotechnology can be defined as the control of matter at dimensions of roughly 1 to 100 nanometers where unique phenomena arising from its size enable novel applications. These particles can take on a variety of different structures and shapes, can be organic or inorganic in composition (or both), and can be manufactured to have various surface functionalities. While there are innumerable possibilities to the diversity of nanomaterials, a key character- istic shared by all structures on this scale is that they are extremely small, representing the transition between individual atoms and bulk material. At this size, the solid state physics of bulk materials no longer apply. Unique, size-dependent properties arise from the large surface area to volume ratio. This ability to control matter at the nanometer scale is leading to technological advances in many areas, including energy, medicine, advanced materials, food stuffs, electronics, and the environment. These advances are based on the fact that at the nanoscale, materials have different properties than in macroscopic or bulk form. These changes at the nanoscale, in turn, can result in different chemical, physical, electrical and biological characteristics than materials with a larger structure. The applications of current research in the nano-realm may soon effect almost every area of human activity. Nanotechnology innovations are now being discovered in medicine, food, clothing, defense, national security, environmental clean-up, energy generation, electronics, computing and construction are among the leading sectors to benefit from these breakthroughs. From a few recent reports by the Woodrow Wilson Project of Emerging Nanotechnologies, below are outlined a few examples of practical applications that may be available in the next 15 years[28, 148]. In the field of smart drugs, it is predicted that there

3 will be nanotechnologies devoted to cancer detection and cures, while other drugs can function inside the body and serve as drug delivery systems with specific targets. Enhansing water purification via nano-products is also on the horizon. Nano-engineered membranes and filtration devices may be used to detect and remove viruses and other pollutants that are difficult to trap using current technology, for example, pharmaceuticals, or NPs may be sprinkled into a polluted body of water to remove PCBs and dioxins. There is also much being done in the area of energy security, with potential for nanotechnology to influence efficient energy conversion, efficient energy storage, efficient energy transmision, and efficient energy use. However, not much is known so far about the amounts of nanomaterials that are actually being produced currently or being implemented into commercial products available on the market today. There are some studies in determining material-flow modeling from products to the environment and the prediction of environmental exposure, though an important input variable is the production amount [48]. One study has specifically aimed to survery com- panies using nanomaterials to estimate worldwide production of ten different nanomaterials

(TiO2, ZnO, FeOx, AlOx,SiO2,CeO2, Ag, quantum dots, CNT, and fullerines) [121, 130]. It should be noted, though, that not only does the production matter but the release of the material and its relative toxicity in nanoparticulate/dissolved ionic/complexed form(s) that should inform decisions as to the potential risks these materials pose to environmental health. Typically, the evaluation of new hazards to the environment are reactionary, meaning that characterization and quantiﬁcation of the contaminant is only fully understood once the pollutant is detected in the biome. The implicit challenge of this research is to thoroughly and eﬀectively develop the key methods to quantify and characterize NPs in complex matrices before they can be detected in the environment. In this way, the goal of this research is to provide powerful analytical techniques that can be implemented with a variety of environmental and biological samples when NPs become as widespread in the environment

4 as predicted.

1.2 Silver nanotechnology and the environment: bactericidal eﬀects and consequences of nanosilver

By some estimates, nanotechnology promises to far exceed the impact of the Industrial Revolution and is projected to become a 1 trillion dollar market by 2015 [116]. The Woodrow Wilson International Center for Scholars compiled a list of more than 1500 consumer products that claim to include some form of engineered nanoparticle [102], a list which as been growing exponentially since. Of these products, about 20% contain silver nanoparticles which is used to control bacterial growth in a variety of applications. Socks, paints, bandages, food containers, among many other manufactured goods incorporate nanosilver to exploit its antimicrobial properties [13]. Commercial products that generate silver ions or contain nanosilver are one of the most rapidly growing classes of nanoproducts. Most exploit silver’s eﬀectiveness in killing a wide range of bacteria (thus the term broad-spectrum biocide), including some of the strains that have proven resistant to modern antibiotics. Perhaps most important, nanosilver particles deliver toxic silver ions in large doses directly to sites where they can most eﬀectively attack microbes [96].

1.2.1 Ionic versus particle silver behavior

Nanosilver as an environmental hazard has been relatively recent. In earlier years, the primary source of silver pollution had been photographic waste material including film and processing reagents [20] but this source has waned as of recently with the advent and popular- ity of digital photography. Despite the rapid progress and early acceptance of nanotechnology, the potential for adverse effects to the biome has not yet been established. However, the environmental prevelance of nanomaterials is expected to increase substantially in the future. Thus, nanomaterials, including nanosilver, are currently being investigated more closely to determine what possible detrimental effects may occur, and subsequently what regulatory precautions must be taken to prevent serious contamination before it is widespread. In

5 particular, the behavior of nanoparticles inside cells is still an enigma, and little about the metabolic and immunological responses induced by these particles is understood so far [3]. The apparent toxicity of silver is related to individual silver species rather than total silver concentration. Indeed, some studies suggest that nanomaterials are not inherently benign and they aﬀect biological behaviors at the cellular, sub-cellular, and protein levels [116]. Complicating the situation further, the persistence of the particle itself is likely to be in issue. At least some formulations of silver nanoparticles dissolve or degrade in slightly acidic conditions and at temperatures not much above room temperature. The presence of chloride or dissolved organic matter in the water could change the rate of nanoparticle dissolution if these ligands are abundant [96]. Thus, nanosilver may not be entirely stable and form (through oxidative dissolution) the more biologically toxic free silver ion (Ag+); while the free silver is known to bind to a variety of environmental ligands so it would no longer be bioavailable. Clearly, determining silver speciation and relative concentrations that would induce toxicity in a natural system is not a simple matter. The aqueous concentrations of Ag+ are typically low in the natural environment or in wastewater treatment systems because of its strong complexation with various ligands such as

-9.75 -49 chloride (Ksp=10 ), sulfide (Ksp=10 ), thiosulfate, and dissolved organic carbon [149]. Because of its cationic nature and its strong association with various ligands in natural waters; the toxicity of Ag+ ions depends largely on the strength and amount of the ligands present. As a result, silver toxicity to microorganisms is not always observed. Nanosilver has remarkably different physiochemical characteristics such as increased optical, electromagnetic and catalytic properties from either ionic silver or the bulk material [168]. Silver in the form of nanoparticles could be therefore more reactive with its increased catalytic properties and become more toxic than the bulk counterpart. Subsequently, there is a strong possibility that the composition of nanosilver would invoke a different toxicity mechanism in biological systems. Furthermore, toxicity is presumed to be size- and shape-dependent, because small size nanoparticles (e.g. < 10 nm) may pass through cell membranes and the accumulation

6 of intracellular nanoparticles can lead to cell malfunction [25].

1.2.2 Toxicity of (nano) silver: a cellular approach

It is widely known that free silver ion (Ag+)ishighlytoxictoawidevarietyoforganisms including bacteria, yet the bactericidal mechanism is only partially understood. Metal toxicity to planktonic species such as algae and bacteria is often governed by the concentrations of aqueous free metal species (i.e. Ag+) [22]. The inhibitory eﬀect of Ag+ is believed to be due to its sorption to the negatively charged bacterial cell wall, deactivating cellular enzymes, disrupting membrane permeability, and ultimately leading to cell lysis and death [139]. It has been proposed that ionic silver strongly interacts with thiol (-SH) groups of vital enzymes and inactivates them [53, 101]. In the cell, silver ions may deactivate cellular enzymes and DNA by reaction with electron-donating groups such as thiol groups and generate reactive oxygen species (ROS) [25].

The reactive oxygen species (ROS) including singlet oxygen (O2), superoxide (O2), hydrogen peroxide (H2O2), and hydroxyl radical (OH.) are generated constantly through exogenous (extracellular) and endogenous (intracellular) processes as part of aerobic life [23]. Recent studies suggested that ROS generation by Ag NPs to Ag+ ions is responsible for the strong bactericidal activity [75], although quantitative estimation was not carried out in that study. It must be noted that ROS forms as a natural byproduct of the normal aerobic metabolism, although its levels can increase dramatically under environmental stress [116]. By examining the correlation between inhibition of bacterial species and ROS, Choi et al. found that the toxicity of Ag nanoparticles could be related to the intracellular ROS mediated cell death process. At the same level of intracellular ROS or the same total Ag concentration, however, Ag nanoparticles appeared to be more toxic than Ag+ ions alone, suggesting factors other than intracellular ROS also aﬀect nanosilver toxicity. Many natural waters possess characteristics that could potentially reduce silver toxicity, reducing the availability of free silver ions through complexation. Competing cations (e.g. Ca2+) prevent binding of free silver ions to the reactive surfaces of organisms [139]. Solubility

7 of the Ag compound and the presence of complexing agents (e.g. thiosulfate or chloride), DOC, and competing ions are all important. The biological impacts of nanomaterials and the biokinetics of nanoparticles are dependent on nanoparticle size, chemical composition, surface structure, solubility, shape, and aggregation. In addition, these parameters can modify cellular uptake, protein binding, translocation from portal of entry to target site, and the possibility of causing tissue injury [122]. An example of how nanosilvers’ catalytically active surface can lead to toxicity is the interaction of electron donor or acceptor active sites (chemically or physically activated) with molecular O2. Just as with the free silver ion, electron capture can lead to the formation of a superoxide radical, which through dismutation or Fenton chemistry, can generate additional ROS on the surface of the nanosilver [116]. Uncontrolled generation of free radicals can attack membrane lipids and lead to a breakdown of membrane function [75]. It is suspected that the antimicrobial mechanism of Ag NPs is related to the formation of these free radicals and subsequent free radical-induced membrane damage. Several studies have explored how nanosilver damages the membrane and subsequently affects the cell health. Choi et al. obtained high-resolution electron micrographs that showed nanosilver absorbed to the microbial surfaces, probably causing cell wall pitting [23]. Fur- thermore, Morones et al. showed through the use of STEM analysis that the nanoparticles are not only found on the surface of the cell membrane, but also inside the bacteria [111]. The nanoparticles were found distributed all throughout the cell; they were attached to the membrane and were also able to penetrate the bacterium. The changes created in the membrane morphology may produce a significant increase in its permeability and affect proper transport through the plasma membrane [151], which could in turn explain the considerable numbers of Ag NPs found inside the bacteria. Previous research demonstrated that Ag NPs attach to the surface of cell membranes, causing the change of membrane permeability, dissipation of the ATP pool and proton motive force, and finally cell death [94, 111]. However, the results from a bacterial viability test by Choi et al. indicated that there is

8 no evidence of cell membrane leakage caused by any Ag species at 1 mg/L Ag. Although electrophoresis studies indicated no direct effect of Ag NPs on intracellular DNA or protein expression, other results demonstrated that inhibition by the nanosilver might be attributed to the accumulation of intracellular ROS. The observation of Ag NPs attached to the cell membrane and inside the bacteria is fundamental in the understanding of the bactericidal mechanism. As stated previously, silver will tend to have a higher affinity to react with phosphorus and sulfur compounds. The membrane of the bacteria is known to contain many sulfur-containing proteins [132]; these might be preferential sites for the Ag NP attachment. On the other hand, NPs found inside will also tend to react with other sulfur-containing proteins in the interior of the cell, as well as with phosphorus containing compounds such as DNA [111]. In addition it is plausible that the intercellular environment could facilitate the dissolution of nanosilver into ionic silver. In closing, changes to morphology presented in the membrane of the bacteria as well as internal damage to DNA and proteins caused by the NPs affect cellular processes such as the electron transport chain, cell division, and DNA fidelity, which often result in cell death [132]. Several studies indicated that the antimicrobial effects of Ag NPs may be associated with characteristics of certain bacterial species [23, 75, 111]. Gram-positive and Gram-negative bacteria have differences in their membrane structure, the most distinctive of which is the thickness of the peptidoglycan layer. Gram-negative bacteria exhibit only a thin peptidoglycan layer (about 2 - 3 nm) between the cytoplasmic membrane and the outer membrane. In contrast, Gram-positive bacteria lack the outer membrane but have a peptidoglycan layer of about 30 nm thick [111]. It still must be determined if one classification of bacteria are more susceptible to membrane damage due to nanosilver.

1.2.3 Nanosilver, bacteria, and wastewater treatment plants

The ubiquitous use of commercial products containing nanosilver could potentially compromise the health of many ecosystems. For example, Benn and Westerhoﬀgive a realistic

9 scenario of the household washing of clothing containing nanosilver released into sewer systems. Since more than 70% of the U.S. population is served by public sewers, much of the nanosilver from consumer products released from these materials would enter a municipal wastewater treatment plant (WWTP). The nanosilver present in the sewage may partition onto wastewater biomass and be removed at a WWTP, only to re-enter the environment via agricultural land application of wastewater treatment biosolids. If nanosilver proves to be difficult to remove in a wastewater treatment system, nanosilver remaining in the treated effluent stream may enter surface water environments, potentially causing disruption among numerous ecosystems. The main process controlling silver removal appears to be sorption (partitioning) of Ag to particles that are removed within the plant by settling and/or filtration process. High levels of sulfur in the WWTP waters may partially explain these partition trends of Ag. As was demonstrated previously, the Ag speciation and size (dissolved versus nanoparticluate) plays a large role in the severity and mechanism of toxicity of organisms and so must be addressed when considering the effectiveness of silver removal in these plants. Through testing several WWTPs, Shafer et al. consistently found a significant portion of the total Ag in large colloid and submicron particle size classes, with the dissolved fraction (< 0.05 or 0.1 m) typically representing less than 15% of total Ag [149]. In fact, a strong linear relationship exists between effluent DOC levels and the fraction of Ag dissolved, suggesting that DOC is able to compete effectively with particulate phases above a certain threshold DOC level. Therefore, the concentration of Ag in the biosolids is worthy of some consideration. In one model presented by Benn and Westerhoff, if the influent concentration to a WWTP was

180 g/L Ag, the Ag concentration in the biosolids would exceed the 5mg/L Toxicity Charac- teristic Leaching Procedure (TCLP) set forth by the USEPA. An increase in consumer use of nanosilver could therefore restrict municipal wastewater treatment facilities from exporting their biosolids as fertilizer for agricultural lands.

10 In unperturbed systems, geochemical processes regulate background Ag to low ng/L levels. However, there is concern that point discharges of silver from WWTPs may produce acute and/or chronic toxic responses in aquatic organisms in the receiving streams [96]. In view of the proportion of Ag associated with colloids in WWTP effluents, the fate of colloidal Ag in the receiving stream is likely to be an important consideration in the overall impact of Ag in fresh waters. The presence of Ag in colloidal material is also relevant to the potential bioavailability of Ag and the fate of Ag in effluent-receiving systems. Ag is removed by incorporation into stream sediments. Measurements on stream sediments confirm Ag accumulation but also indicate downstream movement of Ag via sediment transport [149]. Thus, sediments “store” a large reservoir of potentially bioavailable Ag, yet many benthic organisims also ingest sediments, providing direct exposure to the bound Ag [96]. Although both of the above scenarios are likely to occur to some extent in affecting both sludge application to agricultural lands and NP containing effluent released into receiving streams, it is also possible to imagine that the nanosilver would affect the bacteria that control the wastewater treatment plant itself. Little work has been done to evaluate the inhibition of microbial growth by different Ag species, especially Ag NPs in wastewater treatment systems. Both autotrophic and heterotrophic microorganisms are important in wastewater treatment [23, 24]. While heterotrophs are responsible for organic and nutrient removal, autotrophs are responsible for nitrification. This is considered as the controlling step in biological nitrogen removal because of the slow growth rate of nitrifying organism and their sensitivity to temperature, pH, dissolved oxygen (DO) concentration, and toxic chemicals [64, 65]. Nitrate can have serious health effects when it enters drinking water wells and is consumed, as well as having detrimental effects on the environment. For this reason, many alternative technologies have been designed to remove total nitrogen from wastewater.

These technologies use bacteria to convert ammonia to nitrate to gaseous nitrogen, N2,which as an inert gas is then released to the air [52].

11 In testing the growth rate of a number of bacterial species, Choi et al. found that nanosilver had the highest inhibition on nitrifying bacterial growth. Nitrifying bacteria have remarkably complex internal membrane systems where ammonia monooxygenase (AMO, responsible for ammonia oxidation to produce hydroxylamine, NH2OH) is located, whereas hydroxylamine oxidoreductase (HMO) is located in the periplasm [100]. It is speculated that nanosilver may have a direct impact on nitrifying cell membranes where key ammonia oxidation enzymes are located. The results of nanosilver toxicity to environmentally sensitive nitrifying microorganisms suggest that stringent regulations of nanosilver entering WWTPs are necessary. Nitrifying microorganism involved in nitrification are critical to biological nutrient removal in modern wastewater treatment [23]. In contrast, some bacterial species are known to accumulate silver with no adverse effects on their vitality. Several species were found by Charley et al. that were not only highly tolerant of silver but also had the capacity for its bioaccumulation [20]. Microorganism having such properties might be exploitable in the development of a recovery process for Ag from industrial effluents and waste materials. Furthermore, these cultures may even be used as a “pre-step” for WWTP if they are affected by nanosilver killing nitrifiles in the system. Moreover, multispecies communities tend to be more tolerant of toxic metals than monospecies cultures [20]. The silver-acclimated community in the above study was able to remove 75.8 ± 5.6% of the silver supplied to a chemostat population when the feed concentration was 1 mM Ag+. Although it has yet to be determined where the Ag is being “stored” in these organisms, the study determined that is unlikely the Ag is simply accumulating on the cell surface. Further research is needed to continue this line of thinking, but the controlled removal or recovery of toxic heavy metals by microorganisms operating under process conditions has been used in other heavy metal treatment [20]. The persistence of NPs, including nanosilver, on timescales of environmental relevance is unknown. The environmental fate of nanosilver will depend upon the nature of the NP. NPs that aggregate and/or associate with dissolved or particulate materials in nature will

12 likely end up deposited in sediments or soils [41, 96]. The bioavailability of these materials will be determined by their uptake when ingested by organisms. There are several possible mechanisms by which Ag nanoparticles can inhibit microbial growth which include particle attachment to cell membranes, causing the changes of membranes permeability and redox cycle in the cytosol, intracellular radical accumulation, and dissipation of the proton motive for for ATP synthesis [64]. There are no examples of adverse eﬀects of nanosilver technologies occurring in the environment at the present, but environmental surveillance is a critical requirement for a future risk management strategy because Ag NPs are rapidly proliferating through the consumer marketplace [96]. Existing knowledge provides a powerful baseline from which to identify research priorities and to begin making scientiﬁcally defensible policy decisions about nanosilver, but it will be necessary to integrate new research information about how this new technology will manifest itself in the environment.

1.3 Potential transformations of silver nanoparticles in natural and treated water systems

The unique characteristics of nanomaterials bring a new dimension, and complexity, to environmental effects testing. The chemical composition of the material matters, as does the size, shape, coating, and subsequent particle transformations (aggregation, dissolution, sorption), which can be dependent on a suite of environmental parameters such as pH, ionic strength, and natural organic matter (NOM) concentration, to name a few. This can make environmental monitoring of nanoparticles (NPs) difficult, especially when considering regulations regarding environmental health and safety. Despite existing laboratory research, which aims to understand nanomaterial fate, the ability to accurately estimate exposure concentrations is complex, which leaves high levels of uncertainty in estimating risk. Fur- thermore, it is reasonable to expect that nanomaterials will behave differently than soluble chemicals in the environment, and so many of the traditional characterization methods prove inadequate and do not adapt well to NP research. It is therefore important to note that although the field of nanotechnology is expanding rapidly, with billions of dollars being in-

13 vested annually, research on the environmental health and safety of nanomaterials receives less than 5% of the funding [78]. Regardless of this, the scientiﬁc consensus is that producing, using, and disposing of nanomaterials will lead to environmental releases [49, 71, 112, 119]. Initial studies have shown the release of a wide range of NPs from a variety of consumer products including TiO2 leaching from housing paint [71]; Ag NP release from textiles into water [13, 14], washing liquid [44, 68], and washing machines [42], along with a myriad of other NPs released from abrasion of coatings, textiles, and personal care products [164].

1.3.1 Potential scenarios for nanoparticle release and subsequent transformation in the environment

Recent reviews attempt to tackle some of the most pressing concerns for those who examine nanomaterials in the environment, including standardizing test methods [55], potential release scenarios [121], and transformations and alterations in the aquatic environment and biota [41, 161]. To date, many of those who study fundamental aspects of ecotoxicity, such as lethal concentrations, or the environmental fate and behavior of nanomaterials, have developed customized protocols to match their research objectives. In doing so, researchers have consequently used a variety of methods for characterization and detection. Addition- ally, many research groups synthesize their NPs in house, which makes their specific results difficult, if not impossible, to replicate. Much of the fundamental research is also being conducted at NP concentrations that are at least an order of magnitude above expected environmental concentrations (g/L vs. ng/L), in part, due to the limitation of applicable detection methods. This approach may alter the extent and type of interactions of the particles and environmental constituents. To date, the studies in the literature tend to address only one or two environmental parameters at a time (such as effects of ionic strength, dissolved oxygen, or NOM). As a result, it is very difficult to determine which of these factors contribute the most to particle transformations, or which constituent may trump another when it comes to controlling the final fate of the NPs. While it is true that the complexity of natural systems make it difficult to account for all likely scenarios NPs may encounter, these

14 basic concepts must be better understood before definitive fate and transport predictions can be made. Critical to determining the potential adverse effects of NPs upon introduction to aqueous systems is knowledge regarding whether the particles remain stable in solution or whether they aggregate and subsequently deposit. Metal-containing NPs form a particularly prominent group of engineered NPs (ENPs), specifically Ag NPs, as they are the fastest growing category of ENPs. Ionic silver (Ag+) release from Ag NP bearing plastics and textiles may be substantial enough to cause considerable concern. However, this dissolution process, in addition to any aggregation or NP alteration, is likely dependent on the given environmental

- 3- parameters that can result in Ag NP transformations: electrolyte composition (Cl ,PO4 ,

2- 2- SO4 ,andS ), ionic strength, redox environment, pH, and the potential stabilizing effects of NOM. In general terms, it is unlikely that primary particles of any sort will persist in environmental conditions as they were manufactured. The important aspects of solution chemistry and coating layer (capping agent) have been explored for a variety of NPs (Ag, Au,

TiO2,FeO2)byseveralrecentpublications[8,67,89],butallwereconductedathigherthan expected environmental NP concentrations. Using dynamic light scattering (DLS), asymmetrical flow field flow fractionation (AF4), in some cases coupled to inductively coupled plasma mass spectrometry (ICP-MS), or imaging techniques such as transmission electron microscopy (TEM), these studies worked in the g-mg/L range. Under these artificially-high concentrations, the mechanisms for either toxicity, ionic release, particle coating degradation, or surface modifications may be altered. Therefore, it is prudent to conduct these experiments under more realistic conditions, using approaches of detection and characterization that are suited to work in the ng/L range. Currently there are few ideal methods meeting these criteria in complex, environmental samples, leading to significant gaps in the assessment of these products. Therefore, owing to the lack of classical techniques, it is key for researchers to further develop analytical methods for the sensitive detection of NPs directly in aqueous samples. Single particle (sp) ICP-

15 MS, is an emerging technique that can both size and count metal-containing NPs at the expected low environmental concentrations. Indeed, although some aspects of this method need improvement, such as decreasing the smallest detectable particle size and developing software so data processing can be more intuitive to newer users, those who work in the area of NP metrology have begun to see the potential of the technique [79, 161]. With advancements in data collection and processing in recent years Laborda et al. [80], Mitrano et al. [107, 109], the method is prime to be used for applications now that basic development has been completed. For these reasons, sp-ICP-MS can be implemented in a series of experiments to better understand the transformations Ag NPs undergo when introduced into simulated environmental water and wastewater samples.

1.3.2 Alterations of silver nanoparticles in complex media

Understanding the behavior and role of NPs in aquatic systems is a task of enormous complexity. This includes understanding the manifold of physical-chemical properties of how the NPs were designed, and taking into account that there is a large suite of geogenic, biogenic, and anthropogenic influences. It is probable that substantial alteration of the primary particle in situ will occur. On one hand, the theoretical fundamentals of colloid science are still valid for many ENPs, yet on the other, there is the added difficulty of tailored (functionalized) NPs, mainly consisting of a core material and a synthetic surface modification (coating, functional groups) [33]. This surface modification might be altered or might become lost in environmental aquatic systems, which will lead to significant deviations from the initial properties of the NP. To begin to unravel these complex systems, there is a need to start with simple model systems and gradually increase the complexity. A variety of these simple systems have been studied in regards to Ag NP alterations in recent years. Ag NPs are not likely to persist in oxygenated environments. The dissolution (i.e. ion release kinetics) of Ag NPs is important because it is Ag+ that is responsible for much of the ecotoxicity seen in bacterial studies [23, 151]. Ag NPs are highly sensitive to oxygen, resulting in formation of partially oxidized Ag NPs with chemisorbed Ag+ at levels that may

16 be suﬃcient to be bactericidal [94]. As Ag NPs are introduced into a water solution, they undergo the following redox reaction and release Ag+ ions [91]:

+ + 2 Ag (s)+1/2 O (aq)+2H 2 Ag (aq)+H O (aq) 2 → 2

There is evidence that a peroxide intermediate may be formed [91]. If so, H2O2 is a more powerful oxidant than O2 and so the Ag NP/H2O2 reaction would be fast and the rate limiting step would be the initial oxidation stage of the Ag NPs and dissolved O2.

In another study, decomposition of H2O2 was found to generate hydroxyl-radicals which were responsible for quantum dot dissolution [105]. Removing dissolved oxygen completely inhibits dissolved Ag release, indicating the essential role of particle Ag surface oxidation in this process [91]. In this way, anoxic and anaerobic conditions exert low redox potentials, which may inhibit Ag+ release and other process kinetics. These ion release rates increase with temperature in the range of 0 – 37 oC. At lower pH, Ag+ release is a cooperative

+ oxidation process requiring both protons and dissolved O2. However, the released Ag ion reaches equilibrium where ions have the possibility to complex with matrix constituents, or rejoin existing nanoparticles, as the dissolution follows first-order kinetics under relatively short time periods (less than 48 hours) at low (<1mg/L)concentrations[84].Inshort,since Ag NPs will likely be dispersed into various environments (rivers, deep soils, sediments, and underground water) where the redox conditions may vary, the ion release of Ag NPs, and the associated bioavailability and toxicity of Ag NPs, can be dependent on the redox conditions [40]. Also critical to determining the potential adverse effects of NPs upon introduction to aqueous systems is particle stability; if dissolution or aggregation occurs. In these cases, other solution components than the redox potential can equally affect the stability of Ag NPs. The presence of simple electrolytes (i.e. mono- and divalent ions) and the pH can cause NP instability. This can be well interpreted by the classic Derjaguin-Landau-Verwey-Overbeek

17 (DLVO) theory, which includes aggregation by screening the surface charge. In ion-free oxygenated systems, dissolution of the oxide layer at the particle surface is hindered by Ag+ adsorption. It has been suggested that only when an electrolyte is introduced into the system does the dissolution of the surface oxide layer begin, due to an increase in the solubility of the surface oxide with increasing ionic strength as well as redistribution and/or replacement of sorbed Ag+ by the electrolyte ion [89]. As the protective layers of the particle degrade, this will expose the metallic core of the particle, which is readily oxidized. Increased NP aggregation is seen with increasing ionic strength [143], and with divalent (e.g. Ca2+,Mg2+) vs. monovalent cations in solution [67]. Initial aggregation in oxygenated, ionic systems is evident, but as the surface coatings degrade, particles will inevitably ionize to some extent, but may not completely dissolve under natural conditions [173]. Conversely, natural organic matter (NOM) and other existing colloids may slow the dissolution of NPs via several mechanisms. Surface adsorption of NOM to block Ag NP

+ oxidation sites [37], reversible reaction of released Ag to Ag0 with humic/fulvic substances as reductants [144], and oxidation of NOM by H2O2, where NOM serves as a competitive sink for H2O2 [165], have all been suggested. Additionally, NOM may simply provide steric stabilization when there is an increase of charged particles in a high ionic strength solution [153]. However, this may instead lead to flocculation of NP and thus form larger particle communities, especially in the presence of divalent cations [34]. In general, NOM tends to adsorb rather quickly onto NP surfaces and stabilize the suspensions as a result of steric repulsion forces, whereas suspensions of NPs in water with moderate ionic strength are more prone to aggregation, and their behavior can be predicted by DLVO theory of colloidal stability [99]. In a study with a variety of coated Au NPs, evidence was found that Suwanee River Humic Acid (SRHA) enhanced particle stability by substituting and/or over-coating the original stabilizer on the particle surface, thus affecting surface charge and chemistry [36]. This effectively negated the chemistry of the original particle coating.

18 The coating (capping agent) can play two large roles in NP fate: colloidal stability and resistance to dissolution. The potential fate and transport of NPs can be closely associated with the chemistry of the capping agent. In simple electrolyte solutions, there is a high potential for particles that are uncoated or electrostatically stabilized to be unstable (aggre- gating or settling out of suspension) in various environments such as landfills, wastewater, soils, surface, and groundwater [10]. Huynh et al. found that while citrate coated Ag NPs aggregated in excellent agreement with DLVO theory, PVP coated NPs appeared much more stable, likely due to the steric repulsion imparted by the surface-bound PVP molecules [67]. Likewise, Li et al. found that non-ionic stabilizers significantly enhanced stability in addition to physically protecting the integrity of particles, as dissolution was significantly reduced [89]. Of note, particles did not aggregate in ionic solutions over an order of magnitude more concentrated when steric stabilizers, such as PVP, were employed. However, as alluded to above, it is still unknown if coating the surface of NPs by NOM will override the initial capping agent chemistry, resulting in NPs with different coatings potentially behaving similarly after being altered and transformed in the environment. There have been few comprehensive studies involving natural water samples, and even these have involved bacterial inhibition as an endpoint rather than characterizing NP alterations/transformations outright. In their work dispersing ENPs in river water samples with varying chemical composition, Gao et al. collected Suwanee River water at three locations which had differing pH, alkalinity, DOC, and ionic strength [43]. Although Ag complexed by dissolved organic compounds had a much lower toxic effect, both Ag NPs and Cu NPs were found to inhibit organisms regardless of the water matrix tested. The authors note that although several laboratory experiments have determined the individual critical coagulation concentrations (CCC) of common electrolytes such as NaCl, MgCl2,andCaCl2 in rather simple aqueous solutions, the complexity of natural water matrices, in which several electrolytes co-occur, could result in different and possibly lower CCC values. Again, DOC and redox environment may also contribute handily towards altering the CCC of NPs sus-

19 pended in individual electrolytes. In contrast, results from a more recent study of capped Ag NPs by Das et al. indicated that impacts of nanomaterials in natural aquatic ecosystems would likely be minimized by dilution, and that bacterial populations would be expected to recover unless in a small body of water or near a point source [27]. Thus, Ag NP additions at concentrations of 10 g/L are unlikely to impact biogeochemical cycles and aquatic food webs in natural aquatic environments. However, one notable exception was clear inhibition at the predicted Ag NP concentrations in treated municipal wastewater and in leachate from wastewater treatment-derived agricultural fertilizers into aquatic ecosystems. This is of high relevance, as the broad application of Ag NPs may aﬀect the performance of wastewater treatment plants owing to their antimicrobial activity. It is therefore important to continue to understand the themes of NP alteration as they enter wastewater collection and treatment systems.

1.3.3 Occurance of silver nanoparticles in wastewater treatment plants and subsequent transformation

Both thermodynamic calculations and kinetic measurements indicate that Ag NPs will not persist in realistic environmental compartments containing dissolved O2. However, the oxidative processes described above are likely to be slow under most environmental conditions, requiring months to reach completion. We then must consider how the particles will be used in practice in order to better determine realistic transformations and ﬁnal fate. One popular application of Ag NPs is to impregnate fabrics so as to resist microbes, but they are then released during the wash cycle [13, 14], with various physicochemical changes from pH, surfactants, and oxidizing agents [44]. Additionally, some commercially available washing machines release (nano) silver into the wash cycle, with up to 11 g/L Ag detectable in the discharge, which is directly plumbed into the waste water system [42]. In simulated clothes washing conditions, Ag NPs were converted to AgCl signiﬁcantly faster when exposed to a hypochlorite/detergent solution than to a NaCl salt solution [68][9]. The authors from that study suggest that, when washed, Ag-containing materials that make their way to a waste

20 water treatment plant (WWTP) will most likely be in the form of AgCl, a much more insoluble and less reactive form than elemental Ag. In contrast, Ag NPs were recently shown to transform rapidly under anaerobic WWTP conditions into insoluble silver sulfides. Due to the significant complexation of Ag ions with ligands, microorganisms are unlikely to be ef- fected in an appreciable way in these wastewater systems. Little is known about the fate and transport of Ag NPs in the wastewater treatment process, particularly their complexation and alterations. First and foremost, the variability and complexity of aqueous waste streams[54, 104], as well as the number of insoluble salts and Ag complexes, make the speciation of Ag in wastewater and resulting biosolids variable and complex. Although studies have shown that sorption to wastewater biomass removed >90% of both total Ag and Ag NPs [77, 149, 157] for final deposition in sludge, the Ag that managed to stay suspended in the supernatant was in the NP form [157], implying closer consideration should still be given to the longer- term impact of their release. Interaction is inevitable between Ag NPs and common ligands in the wastewater treatment stream such as sulfate, sulfide, chloride, and phosphate, as well as carboxylic acids, polyalcohols, and amines found in humic substances [24]. Many of these ligands are known to either complex directly with Ag NPs or with the Ag+ that is released during NP oxidation. Despite this complexity, silver ion released from Ag NPs react immediately with large amounts of chloride present in wastewater to from silver chloride, though dissolution may be slowed (or nonexistent) as Ag NPs are coated with sulfide, which is 200 to 300 times in excess of the total Ag in wastewater [1]. In this way, the fate of Ag in this matrix is likely to be dominated by sulfide chemistry [161] in the treatment plant, with some considerations for chloride complexation before introduction into the wastewater system. In practice, the formation of silver-sulfide complexes were shown to not inhibit nitrification activity from bacteria in activated sludge in contrast to either AgCl or Ag ions

[19]. Due to the insolubility of Ag2S, it had negligible impact on anarobic digestion and methanogenic populations in the anaerobic chamber [172]. In several pilot and operational

21 WWTPs, the discovery of silver sulfide in the sludge confirmed this expected behavior for Ag NPs to migrate towards attaching to biomass and settle in these systems [19, 72, 74]. However, under these anaerobic, rich sulfide environments, it still not entirely clear if particles dissolve and subsequently form (new) Ag2S particles, or if Ag2S NPs are formed outright with excess S on the surface [74].

1.4 Nanometrology: analytical techniques for nanoparticle detection

A wide range of analytical tools are available to examine systems containing NPs, and many carry great promise but also limitiations inherent to either the physical, chemical, or even biological principles on which they are based, or the current state of technology. Under- standing nanoparticle fate in the environment is difficult, due in part to challenges associated with detecting small amounts in complex environmental and biological matrices. These complex, heterogeneous matrices that may confound detection of very low (g/L or less) levels of engineered nanomaterials. Spectrometry techniques, including a very new method, single particle ICP-MS (sp-ICP-MS), has been applied for the detection of NP determining both concentration and size simultaneously [107, 127, 158]. However, the most commonly used detection and characterization methods available to assess particle concentration and size distributions are not adequate for the study of NPs in complex systems or at low concentrations [156]. These include microscopy [85], chromatography [152], centrifugation [97], laser light scattering [133], and filtration [2, 63]. One particular analytical challenge is distinguishing engineered nanoparticles from other constituents of the matrix such as natural particles, humic substances, and debris [82]. Perhaps the major problem identified with most techniques relates to method sensitivity, which is generally insufficient compared to environmentally and toxicologically relevant concentrations (g/L to ng/L ranges) [59, 73]. One of the most commonly used approaches to study nanoparticles is microscopy, either Transmission Electron Microscopy (TEM) or scanning probe microscopy. Theoretically, microscopy offers the ultimate sensitivity, with the ability to detect/image a single nanoparticle; however, accomplishing this practically is equivalent to the proverbial ‘needle in a haystack’.

22 These techniques create images of single particles but their shortcomings include unrepresen- tative sampling, changes during the preparative process (i.e. agglomeration), and inability to find particles in very dilute samples. Another common approach that has long been used to study colloidal solutions is Dynamic Light Scattering (DLS), which measures the particle hydrodynamic diameter. However, limitations for the study of nanoparticles are numerous including: poor sensitivity at dilute concentrations, nonselective material detection, failure to differentiate between nanoparticles and other matrix components, and the inability to re- liably quantify the relative proportions of particle or aggregate sizes in multi-modal distributions. Multi-modal populations are particularly problematic for DLS as intensity-normalized results will characteristically be disproportionately skewed to the larger particles/aggregates in suspension even if smaller sizes predominate. Furthermore, many analytical techniques have difficulty distinguishing organic coatings that may be present on nanoparticles. Field-Flow Fractionation (FFF) consists of a suite of high resolution sizing techniques which allows separation and sizing of macromolecules, submicron colloids, and nanoparticles of 1 – 100 nm, depending on the type of field applied and mode of operation [136]. The separation process is similar to chromatography except that it is based on physical forces (e.g. diffusion) as opposed to chemical interactions. Particle separation is performed in a thin channel with laminar flow under the influence of a perpendicular field. Depending on the type of analysis that is being performed, a different member of the FFF family would be chosen to achieve optimal separation results. The three techniques that are the most commercially available and thus most commonly used include thermal, sedimentation and flow FFF. Applications of FFF have become increasingly diverse in the recent years to include separation and characterization of proteins [92], polymers [103], cells [138], natural nanoparticles [21] and more recently manufactured nanoparticles [81]. Flow FFF (Fl-FFF) is the most widely used subset of FFF techniques for environmental analysis and is highly versatile for a range of both natural and manufactured NPs. As outlined in Baalousha et al. [6], the increased use of FlFFF can be related to the wide

23 size range that can be fractionated either of natural colloids (1-1000 nm) or natural and manufactured NP (1 - 100 nm) [11]. Many FFF techniques, including Fl-FFF, are also adaptive, allowing the ability to change carrier solutions with respect to pH and ionic strength so as to match the carrier solution with sample composition [110, 142], and the possibility of both on-line hyphenation to a wide range of detectors as well as collection of sample fractions for further off-line analysis [6, 45, 46, 69]. UV absorbance is a common non-destructive detector for on-line processes; therefore, FFF-UV has been used extensively. However, with the sensitivity limited to particle concentrations in the mg/L range, FFF-UV is not suitable for many environmental studies since aggregation behavior may be vastly different at low g/L levels where the probability of particle-to-particle collisions is lower. Additionally, the UV detector lacks specificity, even when multiple wavelengths are employed. Interfacing FFF with ICP-MS or ICP-AES, however, allows element specific detection at trace levels [39] when studying metal-containing NP [47]. Several reviews have been published discussing the broad range of environmental [58], biological [141, 142], and nanoparticle [9, 17, 171] applications for FFF-ICP-MS [79]. Furthermore, the capability of multi-metal analysis is an added benefit when using these detection techniques. Though literature is scarce to engineered nano-specific studies, there is a growing potential for the use of FFF in nanoecotoxicity studies with increasing interest concerning characterization methodology for environmental and biological risk evaluation. Notably, recent studies to characterize quantum dots [125] and NP [131? ]inbiologicalmediabefore and after exposure, as well as environmental samples [33, 34]have shown promising results when using FFF-ICP-MS and FFF-ICP-AES. The resultant hyphenated techniques of FFF- ICP-MS and FFF-ICP-AES described in Bednar et al, for example, provide nanoparticle detection, sizing, and compositional analysis capabilities at the g/L level for multiple elements present within the nanoparticle [12], which is critical to environmental and toxicological investigations of nanomaterials [63]. The methodology developed in the previously mentioned

24 work provides detection capabilities for multiple components of nanoparticle systems, including core-shell gold-silver nanoparticles, Cd-Se Zn-S quantum dots, and silver nanoparticles after sulfidation. These methods can be used for more descriptive nanoparticle fate and characterization. One of the newest techniques in the nanometrology toolbox is sp-ICP-MS. The fundamental assumption behing sp-ICP-MS is that each pulse represents a single particle event, which depends on short dwell times, constant flow rate, and a sufficiently low particle number concentration. If these parameters are met, then the frequency of pulses is directly related to the number concentation of particles (particle number per volume), and the intensity of the pulse (i.e. height) is related to partize size (mass). In single particle analysis, the smallest peak height that can be distinguished from the background determines the smallest detectable single particle mass (volume, size); therefore, the detection limit determines the lower particle size threshold of the system. The limitations of this technique arise from the inherent sensitivity of the given ICP-MS being used. For example, theoretically sector- field instruments provide smaller particle-size thresholds than quadrupole based systems , although there has not been much work in this specific area. The best reported size limits for quadrupole ICP-MS systems are approximately 20 nm diameter, and even then for elements with excellent sensitivity (comprising 100% of the particle mass), absence of major interferences, and low dissolved background, such as Au NPs [161]. However, because the technique takes little sample preparation, uses relatively common/available laboratory instrumentation, and shows much promise to detect a wide range of NPs in a variety of complex media, it is expected for many new publications to be emerging shortly using this method.

1.5 Purpose and Signiﬁcance

One of the biggest challenges in quantifying the environmental risk posed by nanomaterials lies in detection. Thus, it is imperative to develop techniques capable of measuring and characterizing exposures while dealing with the innate diﬃculties of nanoscale detection in environmental samples: dilution, complex matrices, transformations and interactions, and

25 false positives associated with natural analogs. Here we hope to provide promising methods to differentiate between dissolved and nanoparticluate silver as well as determining the concentration of each. Furthermore, we hope to show that surface complexation or aggregation of the NP can be monitored over time. It is expected that the methods developed here will provide the analytical capability to monitor Ag NPs while conducting fate and transport, stability, and toxicity studies with small investments of time and capital. Rapid throughputs and relatively simple calibration and quality assurance protocols are additional benefits to these proposed nanometric tools; sp and AF4-ICP-MS. While there are a myriad of available analytical techniques that can broadly characterize NPs, there are few that can provide quantitative results that many researchers feel are most important. These include detecting particles at very low (ng/L - g/L) concentrations while accurately sizing NPs. Furthermore, the ability to determine NP and associated dissolved metal concentrations. Specifically, spICP-MS was developed intensively in the Ranville research group, in addition to comparing AF4-ICP-MS to other, more traditional, detection techniques for the detection of nanomaterials. The advancements to these methods as a whole are recognized by collaboration with several other research groups who wished to implement the methods developed here in their “real world” studies. These studies ranged from detection of NPs in waste water treatment plants, complexation of NPs with natrual organic matter (NOM), simulated environmental releases in lake mesocosms, release of NPs and carbon nanotubes from plastics and other consumer products, and detection of NPs in food and animal feed. To some extent, in each case above, the spICP-MS (and in some cases AF4-ICP-MS) technique(s) proved more capable than traditional techniques to more easily determine key characteritics of the particles in each matrix. The improvements to both the spICP-MS and AF4-ICP-MS methods have already and will continue to prove useful to many researchers as the community continues to determine what transformations NPs undergo both in environmental and biological contexts.

26 1.6 Organization of This Work

This project can be split into two major aspects for research effort: 1) analytical method development of spICP-MS and AF4-ICP-MS for the characterization of metal-bearing nanomaterials, and 2) use of these (improved) methods for quantitative, environmentally relevant characterization and transformations of Ag NPs. As sp-ICP-MS is a relatively new technique, we first conducted several proof of concept studies, including correlating number of pulses in the data set to increasing particle concentrations. In addition, we acidified and filtered samples to demonstrate increased background (dissolved metal) counts and dis- appearance of pulses from the dataset, respectively, to further establish the link between pulses and NPs. Finally, we conducted a pilot study to determine if we had the capability to monitor Ag NPs in waste water treatment influent and effluent (Chapter 2). We then compared detection characteristics of spICP-MS with another emerging technique, AF4-ICP-MS. These parameters included detection limit (both particle size and concentration), resolution, and multi-form analysis. Multi-form analysis highlighted some of the unique capabilities of each technique, such as ability to simultaniously quantitatively determine dissolved and NP constituents in a sample (spICP-MS), analyze NP complexes and multiple metals analysis (AF4-ICP-MS). The results of this comparison, as well as the capabilities and pitfalls of each technique, can be found in Chapter 3. The second, parallel objective of this reseach project was to implement our newly defined analytical capabilities in the hopes of determining the transformations (Ag) NPs undergo when subjected to environmentally relevant conditions. Namely, the chief improvement over other, similar, enviromental NP studies was that we could conduct experiments with particle concentrations orders of magnitude lower because of the capability of the sp-ICP-MS technique. These lower NP concentrations were in the same range as predicted envionmental concentrations based on modeling from product usage reports, which gave us confidence we could more accurately describe the transformation of particles (dissolution and/or aggregation) in various water chemistry matrices, as discussed in Chapter 4.

27 CHAPTER 2 INVESTIGATIONS FOR THE FEASIBILLITY OF DETECTING NANOPARTICULATE SILVER USING SINGLE PARTICLE INDUCTIVELY COUPLED PLASMA-MASS SPECTROMETRY

Amethodforquantitativelyinvestigatingmetallicnanomaterials,speciﬁcallysilvernanopar- icles, in complex matrices is presented here. This chapter is reproduced with permission and some modiﬁcation from Mitrano, D.M., Lesher, E.K., Bednar, A.J, Monserud, J., Higgins, C.P., Ranville, J.F. (2012). Detection of Nanoparticulate Silver Using Single-Particle Induc- tively Coupled Plasma-Mass Spectrometry. Environmental Toxicology and Chemistry, 31, 115-121. Copyright 2012 Society Environmental Chemistry and Toxicology.

2.1 Abstract

The environmental prevalence of engineered nanomaterials (ENMs), particularly silver nanoparticles (AgNP), are expected to increase substantially in the future. The ubiquitous use of commercial products containing AgNPs may result in their release to the environment, where potential for ecological effects are unknown. Detection of ENMs is one of the greatest challenges in quantifying their risks. Thus, it is imperative to develop techniques capable of measuring and characterizing exposures, while dealing with the innate difficulties of nanomaterial detection in environmental samples, such as low ENM concentrations, aggregation of particles, and detection in complex matrices. Here, it is demonstrated that the use of inductively coupled plasma–mass spectrometry (ICP–MS), operated in a single particle counting mode (spICP-MS) can detect and quantify AgNPs. In the present study, two AgNP products were measured by spICP-MS, including one of precisely manufactured size and shape, as well as a commercial AgNP containing health food product. Serial dilutions, filtration, and acidification were applied to confirm that the method detected particles. Differentiation of dissolved and particulate silver is a feature of the technique. Analysis of two wastewater

28 samples demonstrated the applicability of spICP-MS at ng/L silver Ag concentrations. In this pilot study, AgNPs were found at 100 to 200 ng/L in the presence of 50 to 500 ng/L dissolved silver Ag. The method provides the analytical capability to monitor silver Ag and other metal and metal oxide nanoparticles in fate, transport, stability, and toxicity studies using a commonly available laboratory instrument. Rapid throughput and element speci- ﬁcity are additional beneﬁts of spICP-MS as a measurement tool for metal and metal oxide engineered nanoparticles.

2.2 Introduction

Nanoparticulate silver (AgNP) is currently incorporated into a myriad of industrial and medical products with more manufacturer-identified consumer products than any other nanomaterial [4]. The aim of these commodities is to take advantage of the antiseptic properties of silver (Ag), as it is a wide spectrum biocide. Many of these products bring AgNP directly in contact with the human body [3], while all have the potential to disperse AgNP to the environment during and after their manufacture and use [13, 44, 139]. Despite the rapid progress and early acceptance of nanotechnology, adverse ecosystem effects from the inadvertent release of AgNP have not yet been established. To provide adequate oversight, and forecast the risk AgNP may pose, environmental exposure concentrations must be established. Mueller and Nowack [112] initially concluded from a materials flow analysis that AgNPs would not likely harm aquatic ecosystems. However, an updated approach concluded the ratio of predicted environmental concentration to predicted no-effect concentration, the risk quotient, for nano-Ag AgNP will be much greater than one in wastewater treatment plant effluents in Europe, the United States, and Switzerland, and slightly greater than one in surface waters in Europe and Switzerland [51]. Focusing on a different mechanism of impairment, Blaser and colleagues found that ionic silver (Ag+) released from AgNP-bearing plastics and textiles may also be substantial enough to result in predicted environmental concentration to predicted no-effect concentration ratios greater than one [15]. Given that AgNPs are commonly used in products that are washed or dis-

29 posed of in wastewater, wastewater treatment plant effluent is a logical matrix for developing methods for environmental monitoring for the emergence of AgNP. In freshwater systems, Ag+ has long been recognized as a tracer of anthropogenic pollution [145] that is toxic to a wide variety of organisms [41], including bacteria [115, 151]. In fact, AgNP are used because of their bactericidal properties. Less is known about the occurrence and direct toxicity of engineered AgNP. The formation of reactive oxygen species (ROS) may be one mechanism by which AgNPs are also toxic [22, 116]. However, there are several possible mechanisms by which AgNPs can inhibit microbial growth that may vary from those of Ag+, such as impairment to cell membrane architecture [129, 151]. The biological impacts of nanomaterials and the biokinetics of nanoparticles [116] are dependent on nanoparticle size [5, 66, 111], chemical composition, surface structure [26], solubility, shape [129], and aggregation [40]. Confounding the issue is that AgNP toxicity is postulated to be linked to the release of Ag+ [91] and to be highly dependent on a host of environmental parameters, including concentration of ligands, interactions with organic matter, ionic strength, and pH [10]. These factors are considered important because of their influence on aggregation and settling, but it has been suggested that toxicity mechanism can be directly affected [57, 114]. Anumberofrecentreviews[18,41,59,120]haveidentifiedanimportantissue:the lack of appropriate detection, characterization, and quantification methods for inorganic nanomaterials in environmental samples. Separation methods for particles and ions, which include dialysis, filtration, and ultracentrifugation [59, 117], are often coupled with analysis methods. Filtration is the most common, albeit problematic, approach. Augmentation of available analytical techniques to detect and quantify AgNPs and released dissolved Ag in both environmental and biological media is highly warranted. Ionic silver is highly reactive and will readily complex with sulfide (log K = 50.1), chloride (log K = 9.7), and other ligands which may mitigate toxicity [24]. For example, nanosized Ag sulfide (α-Ag2S) has recently been identified in final stage sewage sludge materials [74] as had been previously hypothesized

30 due to the high binding constant for the surfaces of treatment plant particles. Impellitteri et al. found that Ag+ leaching from consumer products may be scavenged by chloride present to form the much less reactive form AgCl [68]. The formation of these nanoparticulate forms is a complicating factor if the goal is to determine the fate of engineered elemental Ag NP released from products. The aggregation behavior and hydrologic transport of all forms of Ag NPs are likely to be similar and will depend on surface composition and particle size. However, toxicity and solubility of the various forms may be quite different and requires further investigation. Because of its element-specific detection and high sensitivity, inductively coupled plasma mass spectrometry (ICP–MS) is ideal for the study of AgNPs and other inorganic nanoparticles. The operation of the ICP-MS in the single particle mode (spICP-MS) provides a means of detecting individual nanoparticles. Single particle ICP-MS relies on the extremely sensitive metal detection capability of ICP-MS, but in contrast to traditional ICP analysis techniques, thousands of individual intensity readings are acquired, each with a very short dwell time (10 - 20 ms). Instead of measuring metal concentration representative of the bulk sample, the intensity readings can be collected as a function of time, where pulses above the background represent the measurement of an individual nanoparticle. Versions of the sp-ICP-MS concept have been used to measure metals in airborne particles [117] and to directly measure aerosols [118]. Degueldre and colleagues measured lab-synthesized colloids in suspension and additionally developed a theoretical link between intensity and particle size [29–32]. This relationship involves a number of poorly known parameters—most important, is the fraction of sample actually reaching the plasma. This is a topic of ongoing work; whereas, the primary objective of this present study was to collect data to illustrate the use of spICP-MS to detect Ag NPs in unprocessed environmental waters. The specific aims were as follows: to ensure that nanoparticles can be analyzed without prior acid digestion by ICP-MS; to examine the deficiencies of the traditional filtration

31 approach for Ag NP characterization; to deﬁnitively demonstrate that pulses in spICP-MS data are quantitatively related to number and size nanoparticles; and to test the applicability of the method to measure the presence of Ag NPs in wastewater treatment plant eﬄuent.

2.3 Materials and Methods

Silver nanoparticles were acquired in sizes of 20, 60, and 100 nm (Nanocomposix, NanoX- ACT). Suspensions were supplied at a nominal concentration of 20 mg/L Ag and were stabilized in aqueous 2 mM citrate per the manufacturer. Accompanying size information (dynamic light scattering and transmission electron microscopy) veriﬁed these particles to be monodisperse with these sizes being 20 ± 1.9 nm, 60 ± 5.3 nm, and 100 ± 6.3 nm. ASAP™, a colloidal Ag consumer product marketed as a dietary supplement (American Biotech Labs), was also used. The product was polydisperse in size, as demonstrated by transmission electron microscopy characterization Figure 2.1. Total Ag concentration was 10.3 mg/L and contained a variable dissolved Ag fraction dependant on lot; but appeared to be very stable over time and with dilution. Nanoparticulate silver suspensions were made by diluting the stock solutions with 18.3 M-ohm nanopure water to ﬁnal concentrations ranging from 10 to 1000 ng/L. Dissolved Ag standards (high quality standards; QC-7-M), used for calibration, were diluted in 1% nitric acid (optima grade) to concentrations ranging from

0.1 to 1 g/L. For the filtration comparisons, disposable filter membranes (25 mm) including Target® (National Scientific Company) nylon (0.2 mand0.45m pore sizes) and Whatman Anotop alumina (0.02 m, 0.1 m, and 0.2 m pore sizes) (Whatman International) were used. Samples were hand filtered by filling polypropylene 30 ml syringes (National Scientific) with the sample and pushing it into polystyrene 15 ml centrifuge tubes (BD Falcon). Two wastewater effluents were investigated during the course of the present study. The first was collected from a municipal treatment plant in Black Hawk, Colorado, USA. Nei- ther dissolved nor AgNPs were detected (estimated detection limit = 0.023 g/L) and these samples were used for spiking experiments. A second treatment plant, located in Boulder, Colorado, USA, was also sampled. Preliminary sampling and historical effluent data span-

32 ! Figure 2.1: Characterization of commercial colloidal Ag NP solution, ASAP Characterization of the ASAP Ag NP suspensions was done using a Philips CM200 Transmission Electron Microscope. A drop of each suspension was placed onto a 300 mesh copper Holey Carbon grid, the sample was left to dry over the course of two hours. Each sample was placed into a single tilt stage for analysis. For analysis 60 to 100 images were taken at 19600 x magnification. The images were analyzed for average length (diameter) using the pixel ruler via the ImageJ® freeware program. From the analysis number distributions were produced to characterize the Nanoparticles in suspension. For ASAP a broad distribution was observed with an average diameter of about 70 nm. ning the last 20 years demonstrated the presence of Ag in the effluent. Both primary influent and final effluent samples were collected on March 12, 2009. Grab samples were taken and transported to the laboratory where they were stored unfiltered at 4°Cuntilanalyzedby spICP-MS. A PerkinElmer Elan 6100 ICP- MS was used for all Ag analyses. Operating conditions can be seen in Table 2.1. There was an online addition of 1% HNO3 prior to nebulization into the spray chamber and into the plasma. Mass 107Ag was monitored for detection, with an integration dwell time of 20 ms per reading. Consequently, a typical scan of 10,450 measurements corresponded to a run time of approximately four minutes. The dwell time per reading, readings per replicate, and total analysis time can be varied but may be limited by computing power and software capability. Longer analysis times give more data points and uses more sample volume. The length of the dwell time contributes significantly to the minimum detectable particle size obtained by the method. More details for optimization of dwell time length can be found in 3.4.1. Intensity data were recorded using PerkinElmer

33 Elan software (Elan, version 2.4) and exported to Excel® (Microsoft) for data handling and processing.

Table 2.1: ICP-MS operating conditions

Parameter Group Parameter Setting ICP-MS model Elan 6100 (Perkin ICP-MS components Elmer) Nebulizer cross-flow (Perkin Elmer) Nebulizer gas flow 0.72 to 0.8 L/min rate Sample flow rate 12 rpm (1.2 mL/min) ICP-MS settings and Plasma RF power 1000 to 1300 W method details Pulse stage detector 1500 W voltage Sample flush, read 45 s; 15 s; 45 s delay, and rinse time Instrument dwell 20 ms time Analyte monitored 107Ag Sweeps/reading 1 Readings/replicate 10450 Total analysis time 4minutes per sample Sample Concentration of Ag 10 to 1000 ng/L characteristics NPs total Ag Diluted with DI water

Instrument calibration was achieved by analysis of a blank and four dissolved Ag solutions ranging from 0 to 1 g/L in the single particle mode. The 107Ag intensity output of Ag for each solution was then averaged from the entire length of the standard analysis. No internal standard was employed, as only 107Ag was detected during the run. To ensure the absence of signiﬁcant instrumental drift over time, a 100 ng/L Ag dissolved standard was run in single particle mode for every 10 AgNP samples analyzed. In traditional ICP-MS analysis, a solution containing a dissolved metal will give a stable intensity versus time signal, the magnitude of which is proportional to the concentration of

34 the metal. For dissolved metals, this is still the case when performing an analysis in SP- ICP-MS mode. In contrast, when a metal-bearing particle is ablated in the plasma during spICP-MS analysis, a pulse of ions is generated that results in a signal greater than the constant dissolved background. Thus, the signal should be steady at (or near) baseline until aparticlepassesthroughtheplasmaandcreatesadeviationduringasingledwelltime that is above the dissolved Ag background. The concept, and a comparison with traditional ICP-MS analysis, is illustrated in Figure 2.2. Observation of pulses is evidence of either Ag-bearing nanoparticles or Ag+ in association with other (not primarily Ag) nanoparticles.

!"#$%&' (%")#"' *&)+%,'

-./),"/,')01/"%' 56'

70#&' 2/30403+"%'' $+%)&)'

56'

70#&'

Figure 2.2: Conceptual diagram for spICP-MS Conceptual diagram for the single particle inductively coupled plasma–mass spectrometry (spICP-MS) method. Samples containing dissolved metals will produce a constant stream of charged ions after passing through the plasma. The detector will then have a relatively constant intensity versus time signal for each dwell time. Conversely, a sample containing inorganic nanoparticles at a suﬃciently low concentration will create a pulse of charged ions after passing through the plasma. Here, a resulting spike in intensity versus time will occur for dwell times that contained nanoparticulate metal.

35 Dissolved Ag, regardless of concentration, will produce a stable signal with few pulses. Conversely, there will typically be a range of intensities in a sample containing Ag NP, some being only slightly greater than the background. Distinguishing between dissolved background and pulses is not a trivial task. To qualify a given intensity as a pulse, we developed an iterative algorithm. The 3σ value of all the data is first calculated and added to the average. Data points having values greater than this value are considered to be due to nanoparticles and are consequently removed. This process is repeated with the remaining data until no more pulses can be differentiated. The intensities of the pulses can be integrated to determine the concentration of Ag NP. The values remaining in the data set are averaged and are considered dissolved Ag. Dataset manipulation to differentiate between dissolved and nanoparticluate silver can be seen in Figure 2.3. Data collected include: an instrument blank (1a), dissolved Ag standards (1b), sample background/dissolved Ag concentration (2), 3-sigma distinction between background (Ag+) and nano-Ag (3), and nano-Ag pulse (4). After calibration with dissolved standards (Step 1), one can directly determine the dissolved silver background in a given sample by comparing the background intensity to the calibration curve (Step 2). To establish this background level, an iterative process is used to converge on the dissolved silver concentration from the full dataset. Through data manipulation, the full data set is averaged and those pulses above 3 sigma are removed. This process is subsequently repeated until the baseline concentration is refined to a stable value. Lastly, once the baseline has been established, pulses over 3 sigma are defined as nanoparticluate silver (Step 3). More work is needed to also be able to size the nanoparticles that appear as pulses (Step 4). It is recognized by the author, however, that the term dissolved Ag is used operationally in this context to refer to both Ag+ and any Ag NP that are smaller than those that can be distinguished as Ag NP by the spICP-MS method. From the previous discussion, it can be seen that determining the amount of Ag in the dissolved and nanoparticulate form is rather direct. However, quantifying nanoparticle number, concentration, and size is more difficult. Calculations rely on quantifying the efficiency

36 ! Figure 2.3: Dataset manipulation to differentaite between dissolved and nanoparticulate silver of the ICP-MS sample introduction system, primarily that of the spray chamber. While the concentration of particles is qualitatively proportional to the number of pulses observed during a run, to be more quantitative, the number of pulses must be divided by the efficiency. Although the efficiency will depend on both the nebulizer/spray chamber design and the instrumental operating conditions, it is generally less than 10%. Continuous signals (dissolved) are not adjusted as the efficiency is already captured by the fact that the dissolved calibration standards are presumably affected by the same sample introduction system efficiency. However, only about 1 in 10 particles in a sample will be detected, and thus the nanoparticulate data (number, total nanoparticle mass) should be multiplied by approximately 10. By comparing pulse intensity measurements to a calibration curve derived from dissolved standards, run under the same conditions, intensity readings are converted to Ag concentration of the pulse. The relationship between the measured pulse concentration and mass of metal in the nanoparticle is: (1) where C is measured pulse concentration (g/L), mp is mass of metal in the nanoparticle (g), Vd is volume analyzed during one dwell time (td), χ is the

37 fraction of analyte metal in the nanoparticle (unitless), ρ is the density of the analyte metal

(g/L), Vp is the volume of the nanoparticle (L), q is the sample ﬂow rate (L/s), and td is the dwell time (s). After the concentration in each Ag pulse is determined, the background is then subtracted from the pulse signal to determine the concentration in only the pulse. m χ ρ V C = p = × × p (2.1) V q t d × d Assuming spherical geometry for Ag NPs and rearranging Equation 2.1to solve for d, Equation 2.2 allows size to be estimated from the measured concentration and other known parameters. As in the computation of particle number, the instrument eﬃciency must be taken into account. In this case, when preparing the dissolved Ag calibration curve, the amount of Ag actually entering the plasma is only roughly 10% of the actual concentration. Because the entire nanoparticle enters the plasma, the intensity of the pulse should be compared to the corrected dissolved Ag data. For natural samples, the other important parameter is χ. This will depend on the composition of the Ag-containing nanoparticle. For metallic Ag NP, Ag2S, or AgCl, the most likely forms of Ag NP, χ values are 1, 0.87, and 0.75, respectively. The Ag2S nanocrystals may have excess S on the surface depending on the environment leading to variable Ag to S ratios ranging from 1.7 to 1.1 [29]. However, negating any associated surface S, a Ag2S nanocrystal would be approximately 13% larger in diameter, i.e., 11.49 nm compared to 10 nm. Likewise, AgCl nanocrystals would be 13.3 nm versus a 10 nm Ag NP particle. The amount of Ag associated with other natural nanoparticle and colloidal phases can be determined, but the actual size of the particle will not be determinable from the Ag signal alone and will depend on additional information to characterize the other constituent.

1 C q t /3 d = × × d 2(2.2) 4/3π χ ρ × × ×

38 2.3.1 Accompanying Studies

To better understand the concept of ablation eﬃciency, and to ensure NPs can be analyzed directly by ICP-MS, we compared analysis of acid digested versus nondigested

NanoComposix NP. For digested samples, we added trace metals grade concentrated HNO3 directly to AgNPs, to a concentration of 67%, the digestates were further diluted prior to analysis using 1% HNO3.Digestionwasperformedatroomtemperatureforaminimumof 30 minutes. Non-digested samples were prepared by diluting AgNPs in both 2 mM phosphate buffer and DI water. For filtration studies, we aimed to test two main aspects of the filtering process including the ability to recover NP smaller than the nominal size of the filter and filter material.

2.4 Results and Discussion

Below is a description of results and discussion of the implications of the experiments outlined above.

2.4.1 ICP-MS Recovery

The results of the comparison of digested and non-digested AgNPs are given in Figure Figure 2.4. There is no difference in the amount of Ag detected in samples dispersed in deionized water, suggesting that intact AgNPs are effectively measured. Size was not a factor for the range considered, although it did appear that the 100 nm NanoComposix stock solution was less concentrated than that reported by the manufacturer. Ionization was likely inhibited when diluting with 2 mM phosphate buffer (Figure 2.4), resulting in lower apparent Ag concentrations. An approximate 40% decline was observed for all AgNP sizes, and a 20% decrease was observed for dissolved Ag. Given the results of the digestion experiment, the most likely explanation for the relatively lower recovery of the Ag NP in the phosphate buffer as compared to the Ag+ is that aggregation is occurring, either in the sample container or within the ICP-MS sample introduction system. This is somewhat unusual, as phosphate is the dispersing agent used in other stock solutions produced by

39 6+

7)8.(2.9+

7.)$#)0.9+:"2./+ ;<$('<"2.+5=>./+ %$#1.#2/"3$#+4''5

!"#$%$&'$()*+(),-./+()0.+

Figure 2.4: Particle ionization efficiency in ICP-MS plasma Comparison of acid-digested and nondigested nanoparticulate silver (AgNP) in both deionized water and 2 mmol phosphate buffer. Concentration in ppb (mg/L Ag) is measured by ICP-MS response. the same manufacturer. This highlights the importance of considering not only ablation efficiency of the ICP-MS but also the need for consideration of particle stability. An initial hypothesis was that the smaller decrease in dissolved Ag recovery was suspected to be due to the precipitation of an insoluble Ag phosphate (log Ksp = -17). However, through modeling with Visual Minteq, we determined that free Ag ion was at a level several orders of magnitude lower that the predicted saturation concentration, suggesting that precipitation of Ag3PO4 was not responsible for the decrease in Ag analyzed.

2.4.2 Filtration

We examined serial filtration as a means of determining particle size of our monodisperse Ag NP samples. The limitations of filtration arise from the possibility of artifacts due to particle-filter interactions, surface coagulation, and pore blockage and Ag+-membrane interactions. Despite these problems, filtration is often used for natural colloids with broad size

40 distributions [35]. Perhaps not surprisingly, filtration was found to be a problematic technique. Normalizing the amount of Ag NP passing through the filter to the raw (unfiltered) samples of each size particle enabled determinations as to which variables most affected the process and avoidance of possible complications such as the matrix effects as discussed previously. First, it was found that NPs smaller than the nominal filter size did not always pass through the filter (Figure 2.5). The smallest of the NPs (20 nm) were able to pass through

! Figure 2.5: Success of filtration as a preparation tequniqe for various sized Ag NPs Filtration proved to a variable and unreliable technique. NP did not necessarily pass through a filter even if it was smaller than the nominal pore size. A significant amount of the 20 nm NP passed through the smallest filter. The composition of the filter membrane highly affected the proportion of NP which pass through. Nylon filters retained all sizes of NP while aluminum oxide filters (Anotop) did not. all the aluminum oxide (Anotop) filter sizes, while the 100 nm NPs were completely blocked by even the largest pore sizes. Second, the most important factor in filtration appeared to be the filter membrane material. Nanoparticulate silver did not pass through the nylon membranes, regardless of NP size or filter pore sizes.

41 2.4.3 Analysis of Commercial Colloidal Silver Suppliment (ASAP) via Sp-ICP- MS

An ICP-MS output of 107Ag intensity versus time for blank, dissolved (500 ng/L) Ag, and nanoparticulate (100 ng/L ASAP) Ag samples is illustrated in Figure 2.6. The dissolved Ag solutions, regardless of concentration, produced a stable signal with few spikes. With the AgNP suspensions, numerous spikes were observed, consistent with the mechanism of single particle detection illustrated in Figure 2.3. To conﬁrm that the pulses were

30/*4% 566%771%8.9% :66%771%8.;<% -".*/0%"*1$*'"12%

!"#$%&'$()*+',%

Figure 2.6: Raw data output, sp-ICP-MS analysis Single-particle inductively coupled plasma–mass spectrometer output with different silver (Ag) solutions: blank, 500 ng/L dissolved Ag, and nanoparticulate silver (AgNP; 100 ng/L AgNP). truly arising from the presence of nanoparticles, we examined the effects of concentration, filtration, and acidification. The number of pulses should be directly correlated with the number of nanoparticles in the solution. Several concentrations of ASAP ranging between 10 to 1000 ng/L Ag were analyzed via spICP-MS, and increasing numbers of pulses were

42 observed with higher concentrations of ASAP (Figure 2.7). This is mainly because at low concentrations each nanoparticle causes a unique incidence in a given dwell time. At higher concentrations, the relationship becomes nonlinear. This could be because at higher concentrations, it is more likely that two or more nanoparticles may be analyzed simultaneously in a dwell time. An alternative explanation is that as total Ag concentration increases, so does the background signal. This makes it increasingly difficult to detect AgNP that produces a signal near the background. In essence, at higher concentrations, small AgNPs are increasingly lost to the background. This effect may also explain the slight decrease in pulses with increasing concentration observed at low concentrations in the filtered and acidified samples.

8(9* :;6&%'%<* =$;<;>%<* 0123%'*"4*5167%7*

!"#$%#&'()"#*+#,-./*

Figure 2.7: Correlation of pulses to NP concentration Pulses positively correlate with concentration of nanoparticulate silver (AgNP). Unprocessed (raw) AgNP shows increasing number of pulses with increasing concentration. After ﬁltering or acidifying AgNP samples, the number of pulses dropped dramatically.

Filtered samples were prepared with 0.02 m Whatman Anotop alumina oxide ﬁlters. The

ASAP was diluted in 1% HNO3 and was shaken for approximately five hours. As is evident in Figure 2.7, both filtration and acidification dramatically decrease the number of pulses observed in ASAP samples, strengthening the argument that pulses are indeed a result of

43 AgNPs. Incomplete digestion of acidified samples may explain the observed increase in pulse numbers with increasing concentration. The acidification approach is a common preparation step for ICP-MS samples. Results would likely be different if an approach designed to completely digest solid material was used. To determine if the results we observed in simple matrices could be seen in environmental samples, Blackhawk wastewater effluent samples were spiked with 1 g/L Ag using ASAP solution. A matrix of four experiments Figure 2.8 was devised to affirm that pulses seen were truly Ag NPs and not an artifact of analysis such as random noise, colloid-bound Ag+,etc. First, both ASAP and wastewater were filtered with 0.02 and 0.45 micron filters, respectively, and combined prior to analysis to ensure no pulses would be detected Figure 2.8A. Second, only ASAP was filtered and spiked into raw wastewater to show that the dissolved (Ag+) portion of ASAP would not bind to particulates in the wastewater and register as a pulse Figure 2.8B. Lastly, when raw (unfiltered) ASAP was spiked into both filtered Figure 2.8C and raw wastewater Figure 2.8D, no statistical differences in the number of pulses were observed, indicating that no additional spikes were obtained when adding Ag to wastewater. The higher baseline for the samples spiked with raw ASAP as compared to filtered ASAP suggests that a significant fraction of the Ag NPs are too small to be differentiated from the background.

2.4.4 Wastewater Inﬂuent and Outﬂuent Samples

The results of an analysis of samples collected from a wastewater treatment plant located in Boulder, Colorado, are shown in Figure 2.9. The influent to the plant, untreated sewage, and the final effluent were sampled. Samples were measured without filtration; however, samples were allowed to settle for several hours prior to analysis. As a proof of concept, dissolved Ag was quantified from the baseline Ag level. Nanoparticulate concentrations were determined by summing all observed pulses and applying a 10% efficiency correction. In the influent, we found 520 ng/L dissolved Ag and 200 ng/L AgNP, with the final effluent measuring 60 ng/L dissolved Ag and 100 ng/L AgNP (Figure 2.9). These total Ag concentrations

44 4"01$5$+%6-678%901$5$+%:/'1$:/1$5% ;/:%6-678%901$5$+%:/'1$:/1$5% -".*/0%"*1$*'"12%&()3*1',%

4"01$5$+%6-678%5/:%:/'1$:/1$5% ;/:%6-678%5/:%:/'1$:/1$5% -".*/0%"*1$*'"12%&()3*1',%

!"#$%&'$()*+',% !"#$%&'$()*+',%

Figure 2.8: Proof of concept: Ag NPs create pulses spICP-MS data output for matrix of AgNP spiked wastewater showing the effects of filtration. First, AgNP and wastewater were filtered with 0.02- and 0.45-mm filters, respectively, and combined prior to analysis to ensure that no pulses would be detected (A). Second, only nanoparticulate silver (Ag NP) was filtered and spiked into raw wastewater to show that the Ag+ portion of AgNP would not bind to particulates in the wastewater and register as a pulse (B). Finally, when raw (unfiltered) Ag NP was spiked into both filtered (C) and raw (D) wastewater, no statistical differences in the number of pulses were observed, indicating that no additional spikes were obtained when adding Ag to wastewater.

45 are consistent with historical data kept by the treatment plant for the last 20 years and are in the range predicted for wastewater predicted from materials flow analysis [50]. Wastew- ater treatment facilities are likely places for Ag NP to begin to be observed. Therefore, use of spICP-MS to monitor effluents may provide a better understanding of how Ag NPs are affecting downstream aquatic environments.

3*45$*1% 675$*1% -".*/0%"*1$*'"12%

!"#$%&'$()*+',%

Figure 2.9: Evidence of dissolved and nanoparticulate Ag in treated waters Boulder, CO, USA, waste water inﬂuent and eﬄuent. Evidence of dissolved silver (Ag) and nanoparticulate silver can be seen by the elevated background ionic silver (Ag+)andpulses (nanosilver [Ag NP]) in the data set.

At present, the spICP-MS method cannot distinguish between engineered Ag NP and other colloid-bound Ag+ ions or precipitates (AgCl, AgS). In a previous study of ﬁve Colorado streams, ﬁltration indicated that 60% of Ag was found to be in the colloidal form, presumably adsorbed onto natural organic matter particles [167]. The present study was performed before the rapid increase in the use of engineered AgNP. Whether colloidal natural organic matter in streams carry enough Ag+ to register a discrete pulse in spICP-MS is not known, but

46 should be considered. It has been said that in the absence of a careful and complete description (characterization) of the nanoparticle-type being evaluated, the results of nanotoxicity experiments will have limited value or significance [166]. The technique demonstrated here is the use of spICP-MS to screen for nanosilver in environmental samples. The ultimate goal is to develop spICP-MS to characterize the particle size distribution and particle number concentration for screening of environmental waters and the exposure solutions of ecotoxicity tests. While highly promising, as tested, this method may be limited to analyzing particles larger than 40 nm (H. Pace, unpublished data). This may in fact exclude a fraction of Ag NPs having the most ecological interest [170]. However, spICP-MS may be a useful tool in a quiver of methods for examining colloidal materials that could potentially be an indicator of contaminants. With further refinement of the method and our understanding of the controlling variables, such as transport in the ICP-MS sample introduction system, matrix effects, etc., we expect considerable improvement in this technique.

47 CHAPTER 3 SILVER NANOPARTICLE CHARACTERIZATION USING SINGLE PARTICLE (SP)-ICP-MS AND ASYMMETRICAL FLOW FIELD FLOW FRACTIONATION (AF4)-ICP-MS

Analytical method development and optimization of single particle (sp)ICP-MS and Asymmetrical Flow Field Flow Fractionation (AF4)-ICP-MS is discussed. Additionally, we compare and contrast the highlights and pitfalls of each technique. Discussion of unique aspects of each method are explored to analyze a wide variety of samples. This chapter is reproduced with permission and some modiﬁcation from Denise M. Mitrano, Anglea Barber, Anthony Bednar, Paul Westerhoﬀ, Christopher P. Higgins, and James F. Ranville (2012). Silver nanoparticle characerization using Single Particle ICP-MS (spICP-MS) and Asymmet- rical Flow Field Flow Fractionation ICP-MS (AF4-ICP-MS). Journal of Analytical Atomic Spectroscopy, 27, 1131-1142. Copyright 2012 RSC.

3.1 Abstract

Methods to detect, quantify, and characterize engineered nanoparticles (ENPs) in environmental matrices are highlighted as one of the areas of highest priority research needs with respect to understanding the potential environmental risks associated with nanomaterials. More specifically, techniques are needed to determine the size and concentration of ENPs in a variety of complex matrices. Furthermore, data should be collected at environmentally- and toxicologically-relevant concentrations. Both single particle inductively coupled plasma mass spectrometry (spICP-MS) and asymmetrical flow field flow fractionation (AF4)-ICP-MS of- fer substantial advantages for detecting ENPs and assessing many of the above parameters in complex matrices over traditional characterization methods such as microscopy, light scattering, and filtration. In this study, we compared the ability of two emerging techniques to detect well characterized, monodisperse silver ENPs and examined their overall applicability

48 to environmental studies specifically with respect to their: A) size and concentration detection limits, B) resolution and C) multi-form elemental analysis. We find that in terms of concentration detection limit (both, on a mass basis and particle number basis) spICP-MS was considerably more sensitive than AF4-ICP-MS (ng/L vs. g/L, respectively), and offers the unique ability to differentiate dissolved and nanoparticulate fractions of total metal. With a variety of optimization parameters possible, AF4-ICP-MS can detect a much smaller NP size (2 nm vs. 20 nm for spICP-MS), providing the possibility for greater size resolution.

3.2 Introduction

Nanomaterials have great potential in both industrial and commercial sectors, becoming useful products for society either when used alone or when integrated into larger products (e.g. consumer goods, foods, pesticides, pharmaceuticals, and personal care products, among others). With the list of applications growing significantly, the ubiquitous use of goods containing nanomaterials may compromise the health of many ecosystems. A number of life cycle assessments concluded that a significant amount of nanomaterials may enter aquatic systems in both the United States and Europe [51, 112, 119]. Metal-containing nanoparticles (NPs) form a particularly prominent group, specifically Ag NPs, as they are the fastest growing category of engineered NPs (ENPs). Furthermore, ionic silver (Ag+)releasefrom Ag NP bearing plastics and textiles may also be substantial enough to cause considerable concern [15]. However, the potential biological impacts of Ag NPs are not solely due to their release of Ag+ alone, as NP size [5, 111], chemical composition, surface structure [26], solubility, shape [129], and aggregation [96] have been documented as important factors controlling their biokinetics. Confounding the issue, the release of Ag+ from Ag NPs is likely dependent on the given environmental parameters: concentration of ligands, interactions with organic matter, ionic strength, and pH [10]. Our lack of understanding of potential environmental impacts of nanomaterials is, in part, due to the difficulties of nanoscale detection that exist in environmental and biological samples. The systems are complex and involve heterogeneous matrices, all of which may

49 contain very low levels of ENPs. There are nearly universal calls within the nanotechnology community for improvements in regards to nanometrology, and the need to fill in many existing knowledge gaps in detection, characterization, and quantification of nanomaterials. Although there have been recent strides in the analysis of ENPs in complex matrices [161], the most common detection and characterization methods used for assessing particle concentration and size distributions, namely microscopy [85], chromatography [152], centrifugation [97], laser light scattering [133], and filtration [2, 17] are inadequate for the study of NPs in complex systems [156]. One particular analytical challenge is distinguishing NPs from other constituents of the matrix such as natural particles, humic substances, and debris 82. Another problem is that method detection limits are higher for many techniques than expected exposure concentrations [59]. Differentiating dissolved and nanoparticulate forms of the metal, as well as possible NP coatings and aggregates, are also key aspects that should be investigated when conducting nano-ecotoxicological studies. Here, we compare and contrast two methods that address a number of challenges for nanometrology: single particle inductively coupled plasma mass spectrometry (spICP-MS) and asymmetrical flow field flow fractionation ICP-MS (AF4-ICP-MS). Of note, NP shape is also thought to influence toxicity, but this is a parameter that these methods do not currently address. The operation of the ICP-MS instrument in the single particle mode provides a means of detecting individual NPs. The technique is outlined in brief here, as several recent publications delve into the specific methods and theory more completely Laborda et al. [80], Mitrano et al. [107], Reed et al. [140]. spICP-MS relies on the extremely sensitive elemental detection capability of ICP-MS, but in contrast to traditional ICP analysis techniques, thousands of individual intensity readings are acquired, each with a very short dwell time (˜10 msec). When analyzing an unacidified, NP-containing solution, instead of measuring elemental concentration representative of the bulk sample, the intensity readings can be collected as a function of time, where pulses above the background represent the measurement of an individual NP. The spICP-MS technique may enable simultaneous determination both of dissolved metal

50 concentrations as well as NP concentration and size (to as low as 20 nm), all at very low (ng/L) concentrations. The technique requires little sample preparation and little additional method development for a given matrix and/or analyte. However, the technique is highly dependent on the signal to noise ratio of a given ICP-MS, which may significantly hinder analysis of smaller sized Ag NPs. Field Flow Fractionation consists of a suite of high resolution elution techniques which, depending on the type of field applied and mode of operation, allows separation and sizing of macromolecules, submicron colloids, and particles of 2 nm-100 nm [136]. The separation process is similar to chromatography except that the separation is based on physical forces as opposed to chemical interaction. All separation is performed in a thin channel with laminar flow under the influence of a perpendicular field. AF4 was chosen for this study because it is the most widely used subset of FFF techniques for environmental analysis and is highly versatile for a range of both natural and manufactured NPs. As outlined in Baalousha et al. [6], the increased use of flow-FFF can be related to: (i) the wide size range that can be fractionated either of natural colloids (1-1000 nm) or natural and manufactured NP (1-100 nm) [11]; (ii) the ability to change carrier solutions with respect to pH and ionic strength as to match the carrier solution with sample composition [16, 110]; and (iii) the possibility of both on-line hyphenation to a wide range of detectors as well as collection of sample fractions for further off-line analysis [9, 45, 46, 69]. Several reviews have been released touting the broad range of environmental [58], bio [141, 142], and nanoparticle [9, 17, 59, 171] applications for FFF-ICP-MS. Furthermore, the capability of multi-element analysis is an added benefit when coupling FFF with mass spectrometry. Though literature is scarce for engineered nano-specific studies, there is a growing use of FFF for nanoecotoxicology, with increasing interest concerning characterization methodology for environmental and biological risk evaluation. Notably, recent studies to characterize quantum dots [125] and NPs Poda et al. [131], Schmidt [147][154] in biological media before and after exposure, as well as environmental samples [9, 33, 34, 153], have shown promising results when using FFF-ICP-

51 MS. The goal of this present study is to compare the advantages and limitations of the two techniques. When comparing spICP-MS and AF4-ICP-MS, there are a number of factors that were considered to determine which technique would be appropriate for a given set of experimental conditions. To make the comparison simple and objective, a list of criteria were made on which to determine each techniques’ capabilities. These criteria include: A) size and concentration detection limit, B) size resolution and C) multi-form metal analysis (such as distinguishing NP vs. dissolved constituents, NP complexes and aggregates, and multi-metals analysis). By using well characterized, monodisperse Ag NPs, we used systematic dilution, mixture analysis (both multi-size NP and Ag+/Ag NP mixtures), and spiked complex matrices, to assess the smallest particle size and concentration determined by each technique, the ability of the techniques to diﬀerentiate between sample constituents, and their abilities to track NP transformations.

3.3 Materials and Methods

Below a description is provided of materials used and method development parameters.

3.3.1 Materials

Silver nanoparticles (NanoXact) were acquired in sizes of 20, 40, 60, 80 and 100 nm diameters (Nanocomposix, San Diego, CA). Suspensions were supplied at a nominal concentration of 20 mg Ag/L and were stabilized in aqueous 2 mM citrate per the manufacturer. Accompanying size information (Dynamic Light Scattering and Transmission Elec- tron Microscopy) veriﬁed these particles to be monodisperse with the nominal sizes being: 20 ± 1.9 nm, 40 ± 3.6 nm, 60 ± 5.3 nm, 80 ± 6.8 nm, and 100 ± 9.4 nm though further characterization using a disc centrifuge showed the size of the 100 nm particles to be 91.3 ± 0.6nm with an associated secondary particle with an equivalent diameter of 109.7 ± 0.8 nm37. Ag NP suspensions were made by diluting the stock solutions with 18.3 M- ohm Nanopure water to ﬁnal concentrations ranging from 2 to 500 ng Ag/L. Dissolved Ag

52 standards (High-Purity Standards; QC-7-M), used for calibration, were diluted in 1% nitric acid (Optima grade) to concentrations ranging from 0.1 to 1 g/L. This standard was also used as the dissolved Ag fraction when studying Ag+/AgNP mixtures. Bovine serum albumen (BSA; Sigma Aldrich) was used for particle stabilization in studies of dissolved Ag complexation using AF4-ICP-MS.

3.3.2 Instrumentation - sp-ICP-MS

A Perkin Elmer NexION 300Q was used for single particle (sp) analysis. Operating conditions were optimized to produce maximum Ag intensity by modifying the sample introduction rate and changing the nebulizer gas flow. 107Ag was continuously monitored for detection, with integration dwell times ranging from 0.1 to 20 ms per reading. The length of dwell time was found to contribute significantly to the quality of data, where 10 ms was optimal. Intensity data were recorded using the ICP-MS software and were exported to Excel (Microsoft) for data handling and processing. Instrument calibration was achieved by analysis of a blank and four dissolved Ag solutions ranging from 0 to 1 g/L with data collected in the sp mode. The 107Ag intensity of Ag for each solution was then averaged from the entire length of the standards analysis (typical times). No internal standard was employed, as only 107Ag was quanitfied during the run. To ensure the absence of significant instrumental drift over time, a single 100 ng/L Ag dissolved calibration check standard was run in SP mode for every ten Ag NP samples analyzed.

3.3.3 Instrumentation - AF4-ICP-MS

For AF4-ICP-MS analysis, both a Perkin Elmer Elan 6100 and a Perkin Elmer Nex- IONQ were used. Standard operating and tuning procedures were used in maintaining and calibrating the instrument. Only one Ag isotope was monitored for detection (107Ag), with an integration time of 2000 ms, alternating with a Bi internal standard (with a dwell time of 1000 ms), resulting in data being collected at approximately one reading every 3 s for the entire length of the fractogram, which depending on experimental conditions, ranged

53 from 40 to 80 minutes. An asymmetrical FFF, AF 2000 AT, from Post Nova Analytics (Salt Lake City, UT) was used with a 10 kDa regenerated cellulose membrane, changed approximately every 25 runs, and with a carrier fluid consisting of 0.025% FL-70 surfactant and 0.01% sodium azide as an antibacterial agent. A 100 Linjectionloopwasusedto load samples onto the channel, and flushed continuously throughout analysis. The AF4 was directly plumbed into the ICP-MS. The channel flow conditions allowed direct connection of the AF4 effluent to the ICP-MS nebulizer without a flow splitter. The AF4 separation conditions varied through method development, but were predominately a 10 min relaxation period (focusing step), followed by 40 - 80 min elution (0.7 – 1. 0 mL/min cross flow and 1.0 mL/min detector flow), and a 10 min flush (field-off, 1.0 mL/min) between each experimental run.

3.3.4 Data collection, conversion to particle size, and quality of analysis - spICP-MS

The theoretical basis of single particle detection has been outlined by Degueldre et al. [29– 31] with further refinement and development in several recently published articles Laborda et al. [80], Mitrano et al. [107], Reed et al. [140] that discuss specific aspects and applications of the technique. The fundamental assumption behind spICP-MS is that at a sufficiently short dwell time and low particle number concentration, a pulse will represent a single particle event. If this holds true, the number of pulses can be directly correlated to the number concentration of particles (particle number per volume) and the intensity of the pulse (i.e. height) can be related to the particle size through particle mass, by making assumptions about particle geometry. Converting pulse height to particle diameter involves several assumptions, but particularly hinges on the calculation of an efficiency factor (η,nebulization/transporteffi- ciency) for the ICP-MS. This can be measured using a standard, well-characterized metal

NP such as Au NP, where η is the percentage of detected particles in sp-ICP-MS mode versus theoretical (calculated) particle number as determined by the elemental concentration, size, and density. Sizing the “unknown” analyte particles then takes several steps (Figure 3.1),

54 where we relate NP pulse height to NP mass, and subsequently size. First, a calibration curve is made by plotting concentration vs. instrument response of the dissolved analyte metal. Mass flux, is then calculated by relating the dissolved calibration curve to the analyte mass that actually enters the plasma during each reading. Here, we relate signal intensity to a total mass transported into the plasma in a given dwell time through the transport efficiency (η)whichisinstrument-specificandshouldbedeterminedonthedayofanalysis . Finally, pulse occurrences are determined for the NP sample, and individual pulses from this dataset can be transformed using the mass flux calibration curve to particle mass, which can then be converted to particle diameter, when assuming a spherical geometry. In this way, a consistent method to size any particle is used even if no monodisperse standards for the given particle exist. For specifics in determining either particle number concentration or particle size, we refer to the recently published studies mentioned above. For all sp-ICP-MS analysis, raw intensity data were plotted as intensity of pulse versus number of events to create a pulse distribution histogram. Very low intensity readings were considered to be instrument background, or, for slightly higher intensity values, dissolved metal. After background/dissolved metal was subtracted from the pulse intensity, NPs were sized using the process previously outlined. Pulses that register at higher intensities are associated with larger diameter NPs, which plot with approximate Poisson distribution around a mean as a function of NP size. Deviation from this shape may be an indication of particle coincidence in a given dwell time, or a polydisperse sample set, and thus would require further sample dilution and characterization to differentiate between these two occurrences, as will be discussed. Dwell time was optimized by analyzing both a 40 nm and a 100 nm Ag NP sample (100 ng Ag/L) at 0.1, 1, 5, 10, 15, and 20 ms dwell times. Similarity to expected peak shape, separation from the background, no evidence of incomplete particle analysis, and blank counts registering at a low intensity value were considered in choosing the best dwell time. Finally, the effect of tuning of the ICP-MS to optimize for a given metal, in this case Ag, on the performance of the sp-ICP-MS technique was investigated by adjusting

55 41H).*%"I-*=$/)(*0*) 9$/&'-#"$/)0$)+*##)J8BE)8#"/1)'K."'/.?) !"##$%&'()#*+,%'#),-$&"(') 0&&" 1((" .$/#0*/0)#"1/*%)23"%')41567#) $&&" +((" %&&" !"#"$%&'()"*"&'(&%$" *((" .-'*0'),8%#'#) +,"#"&'----$" /&&" 0((" !"#"$%&$$'"&"()$(*+" ,-"#"()....+"

.&&" /(("

9$%%'.0)03$8#*/(#)$:) 34564785!"9:;<457=" 23453674!"89:;346<" (&&" $((" LM46N) !"##$%&'()O0()) 67)O*+,%')) "/("&"(8*%)(*0*),$"/0#)*0);<) (" &" (" +%2$(" $%2(." $)+%2(." /%2(." /)+%2(." 0%2(." 0)+%2(." *%2(." +#)(2'%%)=+') &" &'." &'%" &'0" &'1" (" ('." =:39534>?@:3"8AAB<" >?77"@A<'"9

#!" '#!" '#" 9*/)$,=+">'),3?#".*%)#'@/1#) '!!" '!" $-).$%%'.=$/),*-*+'0'-#):$-)*) &#" &!" &!" %!" 1"&'/)#*+,%')A*/*%?0')$-) %#" %!" $!" +*0-"BC) $#" $!" #!" D*2)(*0*)$80,80E)#0*/(*-(#) -./01/234"56"7/89:2;" )*+,-+./0"12"3+456.7" #" !"

*/()#*+,%'#) !" *" #$" )$" &$" +*" %*" **" '()" ''$" '$$" '#*" #" '%()" +'()" $%()" %'()" ,%()" *'()" (!" '!%()" '#'()" '+%()" $(!" %(!" &(!" '(!" 8-9:+";.<+.:6<0"=/1-.<:>" <:8=/>/."?2=@" F8/G/$2/)F)67)#*+,%'E) F8/G/$2/)F)67)#*+,%'E)-*2)(*0*) .$/&'-#"$/)0$)("*+'0'-)

Figure 3.1: Schematic of sp-ICP-MS data processing Here, we relate pulse height to nanoparticle size. The ICP-MS is a mass-based technique where particle size is determined by relating the pulse intensity to an elemental mass. Following the figure above, as in traditional ICP-MS analysis, the first step in this process is to create a dissolved standard calibration curve. This relates the ICP-MS detector signal intensity to the concentration of the analyte entering the plasma. The next step is to determine the relationship between the concentration of the dissolved analyte and the total analyte mass (mp)thatenterstheplasmaduringeachdwelltime.Thisrelationship between analyte concentration and the mass observed per event is mass flux, which is highly dependent on the transport efficiency (ηn). The transport efficiency must be calculated for each instrument and under the given run conditions for the mass flux to be accurate. In this way, the mass flux calibration curve relates signal intensity (counts/event) to a total mass transported into the plasma in a given dwell time. Finally, the intensity of each individual pulse can then be transformed using the mass flux calibration curve to determine the particle mass. The ionization efficiency of the NP is presumed to be 1 (particles are 100 percent ionized) in most circumstances. From here, one can convert the mass to particle diameter, d (nm), using the particle density ρ (g/cm3), and assuming 1) spherical geometry of the particle, and 2) capturing the complete ion cloud from the NP in one dwell time.

56 ICP-MS parameters for maximum analyte sensitivity. We then analyzed 100 nm Ag NP (100 ng Ag/L) and compared the sp-ICP-MS data with standard tuning to optimized tuning approaches for Ag NP pulse intensities.

3.3.5 Data collection, converstion to particle size, and quality of analysis - AF4- ICP-MS

Size fractionation in AF4 takes place in a thin channel, which is constructed using a polyester spacer (0.25 - 0.5 m) enclosed between two plexiglass blocks, with one porous block (frit) on the lower side. The laminar channel (tip) flow, which carries the sample through the system, creates a parabolic flow velocity profile. A perpendicularly-applied fluid cross flow pushes particles against the lower (accumulation) wall, which consists of a semi-permeable membrane on top of the lower ceramic frit. After the sample is injected, a focusing step occurs and the sample is concentrated near the entrance of the channel. After a set focusing time, the separation of the particles occurs during the elution phase of analysis. During the elution, the cross flow pushes particles against the membrane, while diffusion causes particles to move away from the wall and into higher velocity flow. The interplay between these two forces causes smaller particles to interact with the faster part of the parabolic flow, resulting in size dependent elution from the channel. In this way, retention time can be directly related to particle size where shorter retention times coincide with smaller particle diameters. In theory, the balance of these forces can only be changed by varying the flow conditions or spacer thickness. Due to interactions between the analytes and the membrane, this balance can also be affected by the membrane composition and/or the carrier solution composition (surfactant, pH, or ionic strength). Manipulation of all physical and chemical variables can optimize fractionation with respect to resolution and sample recovery. However, presuming all particles in the sample are of regular (spherical) shape and have equal interaction with the membrane, the interpretation of the elution profile (fractogram) is straight-forward. FFF theory or a simple linear calibration from retention time versus particles of known size (excluding the void peak), can be used to size all analyte

57 particles that have similar behavior within the FFF channel. The AF4-ICP-MS fractograms were evaluated for sample recovery, reproducibility, absence or minimal height of the void peak, and distance between void and analyte peaks (retention ratio), and separation between analyte peaks (resolution). We determined suitable parameters for the eﬀective fractionation over the size range of Ag NP of interest in this study, but did not necessarily ﬁnd the best possible run conditions, as the complete optimization of all parameters was not the objective.

3.3.6 Size, detection limit, and resolution experimental parameters

The AF4-ICP-MS fractograms were evaluated for sample recovery, reproducibility, absence or minimal height of the void peak, and distance between void and analyte peaks (retention ratio), and separation between analyte peaks (resolution). We determined suitable parameters for the effective fractionation over the size range of Ag NP of interest in this study, but did not necessarily find the best possible run conditions, as the complete optimization of all parameters was not the objective. For sp-ICP-MS analysis of size, detection limit, and resolution, 40, 60, 80 and 100 nm Ag NPs were analyzed either individually, or in mixtures, at concentrations ranging from 2 to 500 ng/L (as Ag). Size detection limit was defined by separation of the NP histogram from the instrumental/dissolved metal background where the smallest detectable NP creates a pulse intensity greater than the mean background intensity plus at least three times its standard deviation. A concentration detection limit was defined as the lowest possible mass-based concentration that still produced a clearly defined histogram for a given NP size. Resolution in sp-ICP-MS was determined by finding the difference in the histogram means between the particles of interest, in counts, and dividing by the average width of the peaks in corresponding units, as seen in 3.1 below, where Ir is the separation between peaks (in units of counts), and wav is the average width of the two peaks. The resolution itself is instrument dependent, as instrument sensitivity dictates the pulse intensity for a given NP size, and so with an increase in instrument sensitivity, there will be a larger intensity difference between particle sizes.

58 I R = r (3.1) wav A mixture of 20 and 40 nm Ag NPs, each at 100 g/L (as Ag), was analyzed by AF4- ICP-MS, at the beginning and end of each day to ensure that membrane conditions did not change during the course of the day. This would be noted by a shift in elution time or a change in peak area (percent recovery). Subsequent studies on detection limit and resolution were performed using mixtures of 20 to 80 nm Ag NPs in concentrations ranging from 2.5 to 100 g/L. A size detection limit for this technique was obtained by referring to literature, where particles as small as 2 nm have been sized. A mass based detection limit was deﬁned as the lowest possible concentration that still produced a fractogram where both particles sizes were distinguishable from the background using a typical injection volume

(20 - 100 L). Resolution was determined by determining the diﬀerence in peak maxima, in retention time (seconds), and dividing by the average width of the two peaks, in similar fashion to the resolution equation above. The results presented here can be considered typical of AF4-ICP-MS, but as noted, resolution can be easily altered by changing ﬂow conditions and physical parameters, such as spacer thickness.

3.3.7 Multi-form analysis

Determining the effectiveness of differentiating between dissolved and nanoparticulate elements in a sample by sp-ICP-MS was accomplished by varying the concentrations of 0, 50, and 100 ng/L of combinations of Ag+ and 100 nm Ag NPs. Once the determination between dissolved and NP fractions had been made, each fraction was quantified using the appropriate procedure. For the dissolved fraction, this was performed via a direct comparison to the dissolved calibration curve. However, for the NP fraction, once the mean dissolved background intensity was subtracted from the pulse intensity, the intensity was converted to mass by accounting for the transport efficiency, which subsequently enabled calculations of NP diameter and mass/number concentrations.

59 To test the capability of detecting NP surface modiﬁcations, NP aggregation, and dissolved complexes, we performed an experiment using AF4-ICP-MS, on protein-containing solutions. Bovine serum albumen (BSA), a protein that has numerous biochemical applications, was chosen as a surface modiﬁer for Ag NPs because it is also reported to help disperse and stabilize NPs in complex media 50. A solution containing 1.5 mg/mL BSA was added directly to an Ag NP mixture of 20 and 40 nm particles, at 100 g/L each, and was allowed to equilibrate for 5 min before dilution (1000-fold) with DI water. The samples were sonicated for 5 min before analysis by AF4-ICP-MS, at 60 min intervals. The NP size was then compared to the daily standard (unaltered) NP mixture to determine if the BSA had adhered to the NP surface. Furthermore, association of Ag+ to BSA was tracked by the increase in small particle size over time. A comparison of Ag NP aggregation was made between unaltered and BSA coated particles.

3.4 Results and Discussion

Desciption of results and discussion of the implications of the experiments is provided below.

3.4.1 Optimization for sp-ICP-MS

Instrument tuning has a clear impact on the sensitivity of sp-ICP-MS. After tuning optimization, the pulse intensity is significantly higher for the same 100 nm Ag NP solution (Figure 3.2). Optimization consisted of tuning the instrument for maximum sensitivity of 107Ag by adjusting sample introduction flow rate and increasing nebulizer gas flow rate. The outcome gave operational conditions that were not necessarily optimal, as per standard daily ICP-MS tuning, such as higher than normal oxide levels, but these conditions yielded asignificantincreaseinsensitivityduringsp-ICP-MSanalysis,withnoapparentanalytical pitfalls. Additionally, through dwell time optimization analysis (Figure 3.3), we found that 10 ms dwell times consistently provided the most accurate results for a variety of sample concentrations and NP sizes.

60 Figure 3.2: Effects of tuning ICP-MS prior to sp-ICP-MS analysis It is clear that after tuning optimization, the pulse intensity is significantly higher for the same 100 nm Ag NP solution. Optimization consisted of tuning the instrument for maximum sensitivity of analyte metal by adjusting sample introduction flow rate and increasing nebulizer gas flow rate. The outcome gave parameters that were not necessarily optimal during a standard daily tune for traditional ICP-MS analysis, such as higher than normal oxide levels, but a significant increase to sensitivity during sp-ICP-MS analysis was gained, with no apparent pitfalls.

61 #!!" #!!" '&!" #!"?:" '&!" '%!" !" '%!" #" '!"?:" '$!" @A.99" '$!" @A.99" '#!" B" 809:.";17.1:<73"=24017:>" #!!" #!!" '&!" '&!" '%!" $" '?:" '%!" %" !?'"@:" '$!" @A.99" '$!" AB.99" '#!" '#!" C<@." '!!" B" 809:.";17.1:<73"=24017:>" &!"

%#" &" $!"B="CD.<<"E9B." %!" $"B="CD.<<"E9B." $#"

$!"

#" ,-./0.123"45"6.7891:"

!" $!" '!" $$!" $'!" %$!" %'!" &$!" &'!" ($!" ('!" #$!" #'!" '$!" ''!" )$!" )'!" *$!" *'!" +$!" +'!" ;0<=.">1?.1=9?3"@2401?=A"

Figure 3.3: Analysis of dwell time choice for optimal sp-ICP-MS data collection parameteres Through dwell time optimization analysis, we found that 10 ms dwell times consistently provided the most accurate results for a variety of sample concentrations and NP sizes. Using 40 nm Ag NPs, Figure 1 A-D show the change in histogram shape by varying the data collection parameters. For long dwell times, 20 ms, (A), it is clear that coincidence of particles in a single dwell time is masking the mono-disperse nature of the sample. Conversely, a 0.1 ms dwell time that is too short (D); particles are not being completely analyzed as ions cannot be transported from the plasma to the detector in such a short time. When looking at the two moderate dwell times with the 40 nm particles (B and C), it is less clear which is optimal. By using 100 nm particles at the 10 and 1 ms dwell times (E) we note that the incomplete analysis of particles at 1 ms dwell times becomes more evident with the appearance (increase) of smaller, intermediate particle sizes that were not originally present.

62 3.4.2 Method Optimization for AF4-ICP-MS

AF4 generally provides high-resolution separation for NPs, but there are a number of factors that may compromise separation, including particle aggregation within the channel and particle membrane interactions, among others. These factors must be taken into account during method optimization by choices of: 1) the carrier solution; 2) the membrane; and 3) the applied field (cross flow). To avoid altering NP properties, such as surface charge, double layer thickness, particle aggregation/disaggregation and dissolution, the chosen carrier solution should mimic the matrix in which the NPs are suspended (i.e. pH, ionic strength, etc.) [6, 9]. A bactericide (such as sodium azide) is a common addition [7]. The two most popular membranes used in AF4 are regenerated cellulose and polyethersulphone (PES) with molecular weight cut-offs ranging from 300 - 10000 Da. A choice between these focused on minimizing particle-membrane interaction and maximizing sample recovery, which generally occurs by selecting a membrane with same charge as the particles or relying on the surfactant to create uniformly charged NPs and membrane. In preliminary experiments with fresh membranes, similar results were observed for a variety of size cut offs and both membrane compositions. However, the 10 kDa regenerated cellulose gave the most reproducible results with the longest membrane life at approximately 25 fractionations per membrane, and so was used in all subsequent analyses. The cross flow may be considered the key component of analysis as it determines the resolution and quality of separation. The choice of cross flow should be made so that there is a high percent recovery while still having a good separation between void and analyte peaks as well as between different analyte peaks (sizes of particles). As a general rule, high cross flows are used when separating smaller particle sizes while lower cross flows are applied when fractionating larger sizes.

63 3.4.3 spICP-MS and AF4-ICP-MS comparison: Detection limit, NP size

Size detection limit, here defined as a particle distinguishable as a pulse at least three times the standard deviation above the background, is highly dependent on the ICP-MS being used and to the element of interest (isotopic ratio of mass being measured). To maximize the applicability of the sp-ICP-MS technique to its fullest extent, the sensitivity of the instrument itself is paramount. Newer, more sensitive instruments that have been used for spICP-MS (Perkin Elmer NexION 300Q, Perkin Elmer NexION 300D, Thermo X-Series 2) had the ability to detect NPs as small as 20 nm Ag, as part of the 40 nm particle distribution tail (Figure 3.4), while less sensitive instruments, such as the Perkin Elmer Elan 6100, could detect no Ag NP smaller than 80 nm. However, we are undertaking research to develop statistically-based methods to deconvolute smaller sized particles from the instrumental and/or dissolved metal background. The many interchangeable parameters and optimization procedures involved with any FFF method development will allow for variable run conditions. A highlight of the AF4 technique is the capability to separate particles within a 10 to 20-fold size range. The size range (2 nm [9] to 50 nm[137]) and separation capability can be altered by varying flow rates and operating conditions [45, 46, 146]. The purpose of this study was not to determine the smallest possible size detectable by the technique but rather to highlight, in a more general sense, that smaller size fractions could be detected than by the sp-ICP-MS technique. Differences in the size detection limits of the two techniques warrant further discussion. When using sp-ICP-MS, the pulse registered would be the primary particle size as only the element of interest is being detected. Pulse integration can lead directly to elemental mass, so this negates the possibility of determining if the element is present as an NP bound to larger particles in solution, adhered to humic substances, or is present as an altered particle such as

Ag2S or AgCl. These are important considerations for environmental studies and somewhat limit the interpretations made by sp-ICP-MS analysis. Conversely, since separation in AF4 is related to hydrodynamic diameter, the retention time is based on particle size but may

64 #!!" &!9/A"'!"9BCD" +!" (!9/A"#!!"9BCD" *!" #!!9/A"#!!"9BCD" )!" (!" '!" &!" %!" $!"

,-./012345"6.47849:;" #!" !" !" '!" #!!" #'!" $!!" $'!" %!!" %'!" &!!" &'!" '!!" ''!" (!!" ('!" )!!" <81=4"29>49=2>;"?:-89>=@"

Figure 3.4: Typical sp-ICP-MS histogram of raw data Pulse intensity versus particle number (normalized to the highest frequency) determined by Perkin Elmer NexION 300Q ICP-MS. Each sized NP gives an average pulse intensity with a distribution around the mean, with larger diameter particles registering as higher pulses. For the 40 nm Ag NP distribution, particles of approximately 20 nm are considered the approximate detection limit because lower intensities blend into the instrumental background counts.

65 lead to misleading results if the particles become signiﬁcantly coated or aggregate (such as in high ionic strength solution or in the presence of larger organic particles). Furthermore, interaction with the membrane may delay particle elution from the channel, and in this way may be incorrectly sized. As with any sizing technique, a second form of measurement should be used to ensure correct measurements were made.

3.4.4 spICP-MS and AF4-ICP-MS comparison: Dynamic range, NP concentration

Though both sp and AF4-ICP-MS rely on the very sensitive nature of ICP-MS, the applicable concentration ranges for the two techniques are quite different. In particular, spICP-MS requires very low particle number concentrations to avoid NP coincidence. As the number of particles for a given mass concentration (i.e., ng of nanoparticles per L) increases significantly with decreasing particle size, it is particularly important to ensure that smaller NPs are analyzed via spICP-MS at much lower mass-based concentrations. Using spICP-MS, one could theoretically detect a single particle in as much volume of sample as one is willing to pump through the ICP-MS, though one has to additionally consider the occurrence of false positives. To make any data meaningful, however, a sufficient number of particles should be analyzed to provide a distribution of particle sizes. As a practical matter, it is useful to define a minimum percent of readings that should contain a NP to enable the development of a NP size distribution. Similarly, a maximum percent of readings is useful to ensure, or at least minimize, the likelihood of NP coincidence. As an illustration of this, we determined an approximate concentration detection limit of 2.5 ng/L for 40 nm Ag NPs (Figure 3.5A), or approximately 7.11 X 106 particles per L, which is in line with other studies Pace et al. [124]. With a standard data collection time of 120 seconds (ttotal), a dwell time (dt)of0.01secondsperreading,andinstrumentsettling time of approximately 0.0001 seconds between each reading, the total number of readings (events) collected is 11,770. The number of readings containing pulses corresponding to NPs is the product of the particle number concentration of NPs (Np;particles/L),thesampleflow

66 rate (qliq;L/s),ttotal,andthetransporteﬃciencyofNPsintheparticularICP-MSsystem

(ηn; unitless). When combined, the fraction of readings (F) that should contain NPs can be calculated as: readings containing NP s N q t η F = = p × liq × total × η (3.2) total readings t d total \ t Using the typical values of ttotal and dt listed above, a 5% transport efficiency, a typical flow rate of 1 mL/min, the fraction of readings that should be particles is approximately 5.92%. Using this, we can then theoretically determine the concentration based detection limit for other particles sizes. For example, under these same conditions, the detection limit of 100 nm particles would be approximately 40 ng/L. This is, of course, for simple, monodisperse systems. To avoid coincidence in polydisperse samples and not degrade the quality of distribution obtained from sp-ICP-MS analysis, one could analyze the sample for a longer period of time so as to capture enough NP pulse events to populate a statistically significant distribution. Additionally, increasing the particle transport efficiency will increase the sensitivity of the spICP-MS technique. As spICP-MS is primarily a particle counting technique, there are instances where a sample contains too many particles. Under these circumstances, multiple smaller particles enter the plasma during one dwell time resulting in fewer, larger pulses, equating to particles of larger size. The upper dynamic range will be dictated by the likelihood of coincident NPs entering the plasma and being detected as a single particle. This is illustrated in Figure 3.5B, the 40 nm particles at 100 ng/L are too concentrated (resulting in a large coincidence of particles), while they are sized correctly when analyzed at 25 ng/L. Under somewhat different conditions (ttotal of 30 seconds, ηn of 9%, and qliq of 0.16 mL/min), Pace et al. calculated that a solution containing 1.2 X 108 particles/L equated to a 5% probability for particle coincidencePace et al. [126]. Converting this to a maximum fraction of readings using the above equation, this translates to 12% of the total readings collected. Thus, as long as the fraction of readings containing NPs is approximately 5 - 15% of the total readings, one should be able to collect sufficient data to develop an appropriate size distribution as

67 &!!" &!!" &!!" &!!"

%#!" %#"0=>?" %#!" $!"0=>?" %#!" #"0=>?""" %#!" %=#"0>?@" !" '!0:"@="AB" '!0:"@="AB" '!0:"@="AB" '!0:"A>"BC" %!!" %!!" %!!" %!!"

$#!" $#!" $#!" $#!"

$!!" $!!" $!!" $!!" +,-./-012"34"-5-06" +,-./-012"34"-5-06" +,-./-012"34"-5-06" #!" #!" #!" +,-./-012"34"-5-06" #!"

!" !" !" !" %!" %#" &!" &#" '!" '#" #!" ##" (!" (#" )!" )#" *!"%!" %#" &!" &#" '!" '#" #!" ##" (!" (#" )!" )#" *!"%!" %#" &!" &#" '!" '#" #!" ##" (!" (#" )!" )#" *!"%!"%#" &!"&#"'!"'#" #!"##"(!"(#" )!")#"*!"

789:-6-,";0:<" 789:-6-,";0:<" 789:-6-,";0:<" 789:-6-,";0:<" '#!" #" '!!" '!/9"$!!"/<=>?"$!!/9"%!!"/<=>" &#!" '!/9"%#"/<=>?"$!!/9"%!!"/<=>" &!!" %#!" %!!" $#!" $!!" *+,-.,/01"23",4,/5" #!" !" !" %!" '!" (!" )!" $!!" $%!" $'!" $#!" 6789,5,+":/9;"

Figure 3.5: Dynamic range and resolution of sp-ICP-MS A) Dilution scheme of 40 nm Ag NP, using ap-ICP-MS, with approximate detection limit of 2.5 ng/L. B) 40 and 100 nm Ag NP mix. 40 nm particles at 100 ng/L are outside sp-ICP-MS dynamic range, observed coincidence of particles registering as larger sized particles. The 40 nm at 25 ng/L and 100 nm particles are within the acceptable range for the spICP-MS technique.

68 well as minimize the probability of NP coincidence. As an illustration of the importance of NP size in determining this maximum fraction of readings, Figure 3.5 also shows correctly- sized 100 nm particles run at 200 ppt, which is within the correct dynamic range. Note that these “100 nm particles” were analyzed using diﬀerential centrifugation and found to be 90 nm, with a secondary associated particle of 110 nm, and so are sized correctly through this technique. The dynamic range, as calculated above, should guide one in analyzing unknown samples. For unknown samples where NP size and concentrations are initially unknown, determining the appropriate dynamic range for spICP-MS is somewhat problematic. However, one could determine if the particle number concentration is too high through serial dilution of the sample. With dilution, if the shape of the particle histogram changes with concentration (as illustrated in Figure 3.5B, 40 nm particles), then coincidence is likely occurring at the higher concentrations. Conversely, if the sample was truly polydisperse, the histogram should retain its general shape, though the number of events would decrease proportionally with dilution.

With AF4-ICP-MS, analyzed with an injection volume of 100 L, the detection limit is amuchhighermassconcentration,atapproximately5g/L (Figure 3.6). This is mainly because of the dilution that takes place within the AF4 channel during separation. It is noted, however, that in the given example, a shorter than optimal relaxation time was used, which explains a large amount of NPs eluting in the void (first) peak of the fractogram. By decreasing the elution time, at the sacrifice of resolution, the degree of dilution could be reduced. Other methods of FFF operation, such as post-channel concentration (split flow) or large volume sample introduction (on-channel concentration) may provide a more sensitive analysis, with some studies showing successful characterization with concentration factors as high as 105 achieved [83]. Although AF4 is subject to overloading effects like any other analytical separation method, it is highly unlikely that this upper limit for particle concentration would be reached when dealing with NPs in environmental systems.

69 '%!!!" '!!">?@A"" '$!!!" )!">?@A"" '#!!!" #)">?@A"" '!!!!" '!">?@A"" &!!!"

%!!!"

$!!!" *+,-./"01234521"

#!!!"

!" !" '!" #!" (!" $!" )!" %!" 67891:1;"<59=" Figure 3.6: Concentration based detection limit, AF4-ICP-MS 20 and 40 nm Ag NP mix dilution scheme with AF4-ICP-MS, detection limit approximately 5mg/L.ICP-MS:PerkinElmer6100.Crossﬂow0.7mL/min,Detectorﬂow1mL/min

3.4.5 spICP-MS and AF4-ICP-MS comparison: Resolution

The resolution of the spICP-MS technique is fixed by the sensitivity of the ICP-MS instrument, although tuning plays a role as previously described in the discussion of detection limit. Since the pulse intensity generated from the detection of the NP is directly correlated with the NP mass, the differentiation between two differently sized NP is dependent on the ICP- MS detector sensitivity. Furthermore, the resolution between particles is not linear with NP diameter, but rather, with NP mass. The total mass of smaller diameter NPs are more similar than larger diameter NPs. In this way the resolution, in theory, will be better for a larger set of particles, i.e. 60 and 80 nm particles would confer better resolution than 40 and 60 nm particles in a given solution. However, this is assuming that the particle distribution for each size is similar, which may not be the case. In Figure 3.7, we show good resolution (R=0.702) between the 40 and 80 nm particle mixture (panel A), yet the 60 and 80 nm particles mixture (panel B) was not as well-resolved (R=0.492) due to the breadth of the 80 nm particle dis-

70 tribution. These mixtures of equal particle number (7.11 X 107 particles/L), correspond to 25 ng/L, 85 ng/L, and 200 ng/L for 40, 60, and 80 nm NPs, respectively. When analyzing a mixture of these three particle sizes, and decreasing the concentration of each NP constituent to 3.55 X 107 particles/L to keep the total particle number within the dynamic range of the sp-ICP-MS technique, similar resolution is achieved between the 60 and 80 nm (R= 0.499) as the two-particle mixtures (Figure 3.7C), with good resolution between the 40 and 60 nm particles (R= 1.05). For AF4-ICP-MS analysis there is a linear relationship between NP diameter and retention time so there is no differential resolution improvement due to size as with spICP- MS. However, as discussed previously, there are a multitude of choices that will directly affect the separation quality in the AF4 channel (Figure 3.8). For example, when analyzing 20 and 40 nm particles of equal concentration, although peak breadths at 1 mL/min cross flow, 1 mL/min detector flow are narrower, the run conditions 1.25 mL/min cross flow, 0.5 mL/min detector flow provide superior resolution. For the present study, conditions were chosen to be able to achieve nearly baseline resolution for 20 and 40 nm Ag NP for comparison to spICP-MS, with 1 mL/min detector flow and 0.7 mL/min cross flow. This mixture was also used as a daily standard to test membrane quality and flow conditions, resulting in our “standard run conditions” resolution of 1.13. This resolution is slightly reduced when analyzing a mixture of 40, 60, and 80 nm (40:60 mix R= 0.876; 60:80 mix R= 0.744) particles of equal concentration (20 g/L, each constituent) shown in Figure 3.9, where the cross flow was increased to 1 mL/min to account for particle sizes being more similar. In contrast to spICP-MS, each analyte peak can be quantified individually through the increased resolution. However, it is noted that the increased breadth of the 80 nm peak, in relation to the other particle sizes, is maintained from the sp results. Furthermore the decrease in recovery with particle size is not unusual in AF4 analysis.

71 '#!" '!!" !" '#!" &#!" '!!" A &!!" &#!" %#!" &!!" %!!" %#!" $#!" %!!" $!!" $#!" ,-./0.123"45".6.17" #!" $!!" ,-./01"23"040567" !" #!" &#!" %!" &!" '!" #!" (!" )!" *!" +!" $!!" $$!" $%!" !" &!!" 89:;.7.-"<1;=" %!" &!" '!" #!" (!" )!" *!" +!" #" &#!" $!!" $$!" $%!" %#!" &!!" B 89:.0601";5.<" %!!" %#!" $#!" %!!" $!!" $#!"

,-./0.123"45".6.17" #!" $!!"

!" ,-./01"23"040567" #!"

#!!" %!" &!" '!" #!" (!" )!" *!" +!" $!!" $$!" $%!" '&!" !" 89:;.7.-"<1;="

$" %!" &!" '!" #!" (!" )!" *!" +!" '%!" #!!" $!!" $$!" $%!" '$!" '&!" 89:.0601";5.<" '#!" '%!" C '!!" '$!" &!" '#!" %!" '!!" $!" &!" ,-./0.123"45".6.17" #!" %!" !" $!" ,-./01"23"040567" #!" (!" $!" )!" %!" *!" &!" +!"

'!!" ''!" '#!" #!" 89:;.7.-"<1;=" !" #!" (!" $!" )!" %!" *!" &!" +!" '!!" ''!" '#!" Figure 3.7: Resolution of particle size mixtures, spICP-MS 89:.0601";5.<" Detection of Ag NP mixtures using SP-ICP-MS. Equal particle number (7.11 X107 particles/L, panels A and B) for each constituent. Mixtures of 40 and 80 nm particles (A), and 60 and 80 nm particles (B), conferring a resolution of 0.702 and 0.492 respectively (C) Equal particle number mixture (3.55 X107 particles/L) of 40, 60, and 80 nm particles with similar resolution to the two particle samples.

72 %!!!!" $"94>982"(?1//@"$"94>982"A.B.;B1?""

$C%#"94>982"(?1//@"!C#"94>982"A.B.;B1?" $#!!!" $C#"94>982"(?1//@"$"94>982"A.B.;B1?"

$!!!!"

'()*+,"-./012/." #!!!"

!" !" #!!" $!!!" $#!!" %!!!" %#!!" !"#$%&'(&&!!!" &#!!" 345612"789.":/.;12

'$!!!!"

'#!!!!"

'!!!!!"

&!!!!"

%!!!!"

"+,-./0"+12314526" $!!!!" '!* #!!!!" ()"

!" !" #!" $!" %!" &!" '!!" '#!" 7589323:";19<" Figure 3.9: Resolution of Ag NP mixture, AF4-ICP-MS Mixture of 40, 60, and 80 nm Ag NPs (20 mg/L, each constituent), with flow conditions of 1mL/min detector flow and 1ml/min cross flow. ICP-MS: Perkin Elmer NexIONQ

73 3.4.6 Multi-form analysis: Dissolved versus NP constituents

For differentiating dissolved from nanoparticulate elements there is a clear advantage in using spICP-MS over AF4-ICP-MS, as any dissolved constituent moves through the membrane and cannot be quantified in the latter technique. However, distinguishing between dissolved background and NP pulses in spICP-MS is not a trivial task. We have suggested in previous work [107], using an iterative algorithm to qualify a NP pulse as at least three sigma above the dissolved background. Although this method may still be theoretically valid in some instances, we found that with more sensitive instruments pulses correlating to NPs register at higher intensities and therefore skew the distribution of readings so that the three sigma qualification to register a pulse as a NP is set too high, and NPs are therefore incorrectly classified as background. In these circumstances, the iterative method is abandoned and the raw data are plotted as pulse intensity as a function of pulse number, where any values below the first minimum in the histogram were considered background/dissolved constituent, and values higher than the first minimum were considered NP pulses. It is noted, however, that the term dissolved metal is used operationally in this context to refer to both Ag+ and any Ag NP that is smaller than can be distinguished as a NP by the spICP-MS method at this time. The concentrations of Ag+ and NP can be independently quantified in a given sample. An increasing concentration of Ag+ can be recognized by an increase in intensity of the lower intensity counts. Once the dissolved Ag is distinguished from the Ag NP pulses, the background intensity can be directly compared to the calibration curve to quantify the concentration of dissolved metal in the sample. In solutions containing both dissolved and Ag NPs, the Ag NP pulse would register as the summation of Ag+ background and Ag NP pulse intensities, i.e. the intensities are additive. Since the increase of dissolved metal not only presents itself by increasing the background concentration at low counts, but also by shifting the NP pulses to higher intensities, Ag NP can be correctly sized only after subtract- ing the background intensity from the Ag-NP pulse intensity before sizing. In Figure 3.10, we

74 show the addition of increasing Ag+ (0, 50, and 100 ng/L) to 100 nm Ag NP (100 ng/L). As the dissolved constituent increases, we note the background moves to proportionally higher intensities. In addition, the NP histogram shifts to higher intensities as well. As shown in Table 3.1, regardless of any increase in Ag+ background, the adjusted pulse intensity is the same in each analysis, and so the NP can be independently sized regardless of Ag+ concentration in the sample.

%#" '#!!" !"&'()"*'+,"$!!"&'()"*'-." %!" !"189:";8<="$!!"189:";8>?" '!!!" $#" $!" #!"189:";8<="$!!"189:";8>?" &#!!" #"

$!!"189:";8<="$!!"189:";8>?" !!"# !"$# !%"# &"$# &%"# "$$# "!"# ""$# !" &!!!" %#" #!"&'()"*'+,"$!!"&'()"*'-." %!"

%#!!" $#"

$!" %!!!" #"

$#!!" !!"# !"$# !%"# &"$# &%"# "$$# "!"# ""$# !" / !"$ )##$*+,-$.+ 0$)##$*+,-$.+12$ !"#$%#&'()*+)#,#&-) !#$

,-./0.123"45".6.17" $!!!" )"$

#!!" )#$ "$ !" #$ !" $!" %!" &!" '!" #!" (!" )!" *!" +!" !"#$ %##$ %"#$ &##$ &"#$ "##$ ""#$ '##$ '"#$ (##$ ("#$ $!!" $$!" $%!" $&!" .%/0#)1&-#&01-()2'*%&-03)

Figure 3.10: Analysis of Ag+ and Ag NP mixtures, spICP-MS Binned pulse intensity versus frequency for addition of dissolved Ag+ and 100 nm Ag NP mixtures. Left panel depicts increasing background concentration, right three panels show increasing pulse intensity for the 100 nm Ag NPs with increasing background intensity, where grey bar indicates median particle number.

75 Table 3.1: Analysis of Ag+ and Ag NP mixtures, spICP-MS Adjustment of background corection to NP pulse intensity to corectly size Ag NP with increased concentrations of dissolved Ag. Sample Average Conc. Avg. NP Background Calculated Back- (ng/L) Pulse Corrected Diameter ground Intensity Pulse (nm) (counts) Intensity 0 ng/L Ag+; 6 ± 2.3 6.3 459 453 83 ± 1.2 100 ng/L 100 nm NP 50 ng/L Ag+; 44 ± 7.5 53.2 502 458 82 ± 1.3 100 ng/L 100 nm NP 100 ng/L Ag+; 78 ± 10.8 94.5 561 483 84 ± 1.6 100 ng/L 100 nm NP

3.4.7 Multi-form analysis: NP complexes

The AF4 technique has an advantage over sp-ICP-MS when studying small sized complexes or, possibly, for the study of NP aggregation. In a 20, 40 nm Ag NP mixture (100 g/L) dispersed in DI water, we were able to detect surface modifications to Ag NPs with the addition of BSA to solution (Figure 3.11). We noticed an initial increase of particle diameter (approximately 5 nm) from the daily standard (20, 40 nm Ag NP mixture in DI water, yellow trace) in the first analysis of Ag NP coated with BSA (red trace). Analysis over the following hours (green and purple traces) showed the slight dissolution of NP, indicated by the decrease in particle diameter. The freed Ag+ released from the NP was subsequently complexed to the excess BSA in solution, as evidenced by the increasing formation of small sized particles (first peak) over time. In this way, we were able to quantify protein bound free silver ions in solution.

3.4.8 Multi-form analysis: Multiple metals analysis

By design, spICP-MS is best suited to analyze one metal at a time in an individual NP. Therefore, multiple metal analysis is not preferable, at least not in one run. However, the

76 %#!!!" 576=>".974?7:?" @69A"B.C";D/E.F<" %!!!!" @69A"B.C";$"A:"3=?<" @69A"B.C";%"A:"3=?<" $#!!!"

$!!!!"

)*+,-."/0123410" #!!!"

!" !" $!" %!" &!" '!" #!" (!" 5678090:";48<"

Figure 3.11: Ag NP, Ag+, and BSA interaction analysis via AF4-ICP-MS AF4-ICP-MS fractogram depicting daily 20 and 40nm Ag NP standards in DI water (yellow trace). Coating of Ag NP with BSA (blue trace) and subsequent dissolution of Ag NP and tracking of accumulation of protein bound Ag+ (red, green, and purple traces) capability of multi-metal analysis by spectroscopy is an added beneﬁt when combined with the continuous fractionation of AF4. The resultant hyphenation of AF4-ICP-MS provides nanoparticle sizing, detection, and compositional analysis capabilities at the parts per billion level, which is critical to environmental and toxicological investigations of nanomaterials. Furthermore, an increased knowledge about size-dependent variations in composition and trace element interactions may interest those working in areas concerning methods for characterization of NP interactions in risk evaluation. This advantage is highlighted particularly well by Pace et al. [125], in a nanoecotoxicity study.

3.5 Conclusions

There is a need to determine NP behavior, aggregation, complexation, and dissolution because different fate and transport predictions for environmental and biological effects depend on the NP state when exposure occurs. It is unlikely that NPs will remain as they were when manufactured during their entire lifecycle (creation, use, transport, final fate) [120, 121]. The transformations that occur, as well as when and under what conditions they

77 occur, should be evaluated so as to enable the appropriate human and ecological risk analysis. Here, spICP-MS and AF4-ICP-MS were able to describe a number of NP properties that are specifically relevant to environmental and toxicological studies, such as size, concentration, associated dissolved constituents, coatings, and complexation. However, each technique has specific strengths, which prove valuable for defining a given set of NP characteristics, as well as limitations that are inherent to each technique.

3.5.1 Advantages and limitations of using spICP-MS in NP characterization

As primarily a counting and sizing technique, the main advantage of spICP-MS over other techniques is its high sensitivity. Size and particle number data can easily and accurately be obtained at sub g/L concentration levels for a variety of metallic NPs. Furthermore, the technique can differentiate the particle of interest from other incidental particles of the same size, but different composition, by its element specificity. This is true even in complex systems. Distinguishing dissolved from nanoparticulate constituents of a given metal is another distinct advantage of spICP-MS. Additionally, spICP-MS can provide better resolution than many of the currently available instruments for sizing, especially when considering the environmentally relevant concentrations at which the method operates. From a practical standpoint, the data collection parameters are similar to traditional ICP-MS operational procedures and so require little additional training. Furthermore, as spICP-MS utilizes a relatively standard laboratory instrument and no additional equipment, laboratories would not incur extra costs in performing this type of analysis. On the other hand, presently, the smallest detectable particle may be larger than many particles of interest and so may exclude this technique as a viable option under some experimental conditions. However, efforts are currently underway to deconvolute smaller sized NP from background intensities. Certainly, spICP-MS is a promising technique for nanoparticle metrology that has the ability to address many of the current analytical challenges that are faced in characterizing nanomaterials in anumberofcomplexmatricesincludingenvironmental,biological,andfoodsamples.

78 3.5.2 Advantages and limitations of using AF4-ICP-MS in NP characterization

Since its conception, the development of field flow fractionation has progressed to have numerous applications that demonstrated the versatility of the technique, yet until recently had not found widespread use for quantitative environmental applications. Today, researchers in the fields of nanoscience, nanotechnology, and biotechnology are eager to try new metrologies and analytical methods that can size particles in the range and with the precision that FFF can perform. AF4 can detect very small particles over a wide size range, with superb resolution, and when coupled to ICP-MS has element specific capabilities, including mixed metals analysis. The ability to detect small size changes allows one to validate coating thickness on the surface of NPs as well as determine the extent of complexation and aggregation of NP in a variety of matrices. However, the lengthy method development process and extra cost may be a drawback to some laboratories. The detection limit, although improved with the hyphenation to ICP-MS than with previous couplings such as UV-VIS, DLS, or fluorescence detectors, is high due to the dilution that takes place within the channel. Some concentration techniques may be applied either before analysis or by loading the channel, yet this is beyond the scope of the present study. Nevertheless, FFF (and consequently AF4-ICP-MS) has become more of a mainstream analytical technique for separating and characterizing analyte species for physiochemical changes for many particles, including NP. This will, in time, provide the method development and method refinement necessary for making the implementation of FFF a less time consuming task.

3.5.3 Future development in the areas of spICP-MS and AF4-ICP-MS

For those in the nanotechnology community, there are many current knowledge gaps in regards to the release, environmental transformation, and potential toxicity of engineered NPs. New analytical techniques are under development that will enable more rapid, sensitive, and speciﬁc detection for an array of these products, both in the lab and in more complex ﬁeld and biological samples. Many of the method details of spICP-MS have been developed

79 so that, at least, researchers can now use the technique to be implemented beyond the laboratory and should now focus on application to real world samples. To strengthen SP now, a more sensitive ICP-MS, such as the single element sector field HR-ICP-MS, may enable detection of smaller sized NPs [161]. Development of time-of-flight (TOF-ICP-MS) could overcome the current spICP-MS limitation of only single element detection. There has been suggestion of coupling sp-ICP-MS with chromatography-like techniques, such as FFF, as well as hydro-dynamic chromatography, to pre-sort constituents of complex samples. Although there is a large dilution factor with FFF separation, this is not a concern for spICP- MS as dilute samples are a prerequisite for analysis. Indeed, as a stand-alone technique or coupled to FFF, spICP-MS shows particular promise in the field of nanometrology, both for manufactures of nanoparticles and nanoproducts to conduct quality assurance tests, and as the demand to characterize more complex and varied samples arise in the quest to determine the environmental and biological impacts of nanotechnology.

80 CHAPTER 4 TRACKING TRANSFORMATIONS OF SILVER NANOPARTICLES IN SYNTHETIC, NATURAL, AND PROCESSED WATERS USING SINGLE PARTICLE (SP)ICP-MS

Accurate data on engineered nanoparticle (ENP) environmental behavior and the interplay between ENP size, surface area, and dissolution rate is critical to appropriately characterize the risks these novel materials may pose to environmental health. The advancement of the single particle ICP-MS (spICP-MS) technique is a great benefit for the study of ENPs in natural systems at environmentally relevant (ng/L) concentrations. Previous studies may have obscured environmentally-relevant dissolution rates because of the artificially high ENP concentrations used in the experiments, an insufficient variety of particle types, or incomplete assessment of water chemistry factors. Here, we analyzed 60nm and 100nm Ag ENPs (citrate, tannic acid, and polyvinylpyrrolidone (PVP) surface coatings) for changes in size (diameter) as a function of time. Both short term (< 24 hour) and intermediate term (1 week) dissolution was examined, with rates of dissolution slowing over an order of magnitude (one log unit) after 24 hours. Dissolution was measured directly as dissolved Ag and by computation from the reduction in measured particle diameter as a function of time. The effects of environmentally-relevant water chemistry parameters such as chloride, sulfide, and dissolved organic carbon were examined, where higher concentrations (1 mg/L Cl-,S2- and 20 mg/L DOC) showed negligible Ag ENP dissolution over 24 hours while in dilute concentrations (equimolar Cl-,S2- and 2 mg/L DOC) approximately a 10% decrease in particle diameter was detected. Additionally, we investigated ENP behavior in both natural (moderately hard reconstituted laboratory water, Clear Creek water) and processed (tap, sludge supernatant) waters. Water chemistry was the most significant factor to affect stability over time, ranging from near complete dissolution in tap water within several hours to particle persistence in waters containing chloride, sulfide, or dissolved organic carbon, where negli-

81 gible dissolution was detected. Particle capping agents were variably effective in decreasing the dissolution rate. These findings can be compared to previous studies at higher particle number concentrations, to elucidate the importance of working at far-from-equilibrium conditions to isolate kinetic effects.

4.1 Introduction

The unique characteristics of nanomaterials bring a new dimension, and complexity, to environmental and biological effects testing. Over the past decade, scientists have made significant progress in understanding factors that influence the environmental fate and transport of nanomaterials including qualitative risk assessment[60, 162], quantitative exposure modeling[51, 134], industrial production[61, 130], release from products[14, 42, 70, 164], and behavior and effects in the environment[135, 169]. Recent reviews tackle some pressing concerns for nanomaterials in natural systems, including standardizing test methods[56, 143], potential discharge scenarios, and transformations and alterations in the aquatic environment and biota[41, 95, 161]. Generally, it is believed that producing, using, and disposing of nanomaterials and nano-imbedded products will lead to environmental releases[51, 112, 119]. Much of the fundamental research is being conducted at engineered nanoparticle (ENP) concentrations that are at least an order of magnitude above expected environmental concentrations (g/L vs. ng/L), in part, due to the limitation of applicable detection methods. This approach may alter the extent and type of interactions of the particles with each other and environmental constituents. This problem can potentially be alleviated by new techniques, such as spICP-MS, which allow for detection and characterization of trace levels of ENPs in complex samples. Particle stability is critical to determining behavior of ENPs in aqueous systems. It is unlikely that pristine particles will persist under environmental conditions, and substantial ENP alteration may occur in situ. Transformative processes, such as dissolution and aggregation, as well as physical parameters such as sedimentation, hydrologic dispersion, and advective transport, are likely to be dependent on a few key features. Important water

82 - 3- 2- 2- chemistry parameters are electrolyte composition (Cl ,PO4 ,SO4 ,andS ), ionic strength, redox environment, pH, and natural organic matter (NOM) present as dissolved organic carbon (DOC). Understanding the manifold of physical-chemical properties of ENP design, and the large suite of potential geogenic[88], biogenic[159], and anthropogenic influences[95] is a complex task. The fundamentals of colloid science are valid for many ENPs, and yet, there is an added difficulty for tailored (functionalized) ENPs, which consist of a core material and a synthetic surface modification[33]. This coating may be altered after particle release, leading to significant deviations from initial ENP properties. The capping agent can play two major roles in ENP fate: colloidal stability and resistance to dissolution. In one study by Huynh et al., citrate coated Ag ENPs aggregated readily while PVP coated ENPs appeared more stable, likely due to the steric repulsion imparted by the surface-bound PVP molecules[67]. Like- wise, Li et al. found that non-ionic stabilizers significantly enhanced stability in addition to physically protecting the integrity of particles, as dissolution was significantly reduced[89]. Dissolution is a particularly important parameter which influences ENP mode of action (e.g. antimicrobial properties, toxicity, medicinal applications and environmental impact) [23, 62, 106]. Both thermodynamic calculations and kinetic measurements indicate that Ag

ENPs will not persist in realistic environmental compartments containing dissolved O2[88– 90]. This high sensitivity to oxygen results in the formation of partially oxidized Ag ENPs with chemisorbed Ag+ or in Ag+ release [41, 91]. The released Ag+ may rejoin existing ENPs,

- 2- form secondary precipitates with complexing species (e.g. Cl and SO4 )[24, 43] and/or interact with DOC. Dissolution has been shown to follow ﬁrst-order kinetics under relatively short time periods (less than 48 hours) at low (< 1g/L) concentrations[84]. Preferential dissolution with smaller particle sizes has been suggested[98], yet may not always be the case: mass normalized dissolution rates nearly independent of particle size have been observed[35]. Natural organic matter may slow the dissolution of ENPs via several mechanisms. Surface adsorption of DOC blocking Ag ENP oxidation sites[37], reversible reaction of released Ag+

83 0 to Ag with humic/fulvic substances as reductants[144], and oxidation of DOC by H2O2, where DOC serves as a competitive sink[165], have all been suggested. Additionally, DOC may simply provide steric stabilization[99] when in a high ionic strength solution[76, 153]. However, this may instead lead to flocculation of ENPs[33]. Ag speciation in natural waters will be strongly dependent on redox conditions, and under aerobic conditions, AgCl species are predicted to form[88]. In simulated clothes washing conditions, Ag ENPs were converted to AgCl(S)[68]. Many common ligands in natural systems, such as sulfate, sulfide, chloride, and phosphate, carboxylic acids, polyalcohols, and amines found in humic substances[24, 88], are known to either complex directly with Ag ENPs or with the Ag+ that is released during ENP oxidation[87, 93]. (Oxy)sulfidation is suggested as another likely avenue for Ag ENP alteration in the environment[87, 91]. Ag ENPs were shown to transform rapidly under anaerobic wastewater treatment plant (WWTP) conditions into insoluble silver sulfides. Although studies showing sorption to wastewater biomass removed > 90% of both total Ag and Ag ENPs [77, 149, 157] resulting in their presence in the sludge phase, some particles were detected in the effluent stream[108, 157]. Finally, Ag ENPs could become coated with sulfide, which is 200 to 300 times in excess of the total Ag in wastewater[1]. Any of these scenarios may either slow or halt dissolution. Solution chemistry and ENP capping agent interactions have been explored for a variety of ENPs (Ag, Au, TiO2,FeO2)[9, 67, 90], but all were conducted at higher than expected environmental ENP concentrations. Using dynamic light scattering, asymmetrical flow field flow fractionation (AF4), in cases coupled to ICP-MS, or imaging techniques such as transmission electron microscopy, these studies worked in the g-mg/L range[98, 158]. It is not immediately evident whether the same processes affecting ENP behavior at these artificially- high concentrations (i.e. higher than predicted environmental concentrations) are the same at lower, environmentally-relevant concentrations[150]. Therefore, it is prudent to conduct experiments under more realistic conditions, using approaches of detection and characterization that are suited to work in the ng/L range. Currently there are few ideal methods

84 meeting these criteria in complex, environmental samples. In this current study, pristine manufactured Ag ENPs of two sizes (60 and 100nm) and three capping agents (citrate, tannic acid, and PVP) were suspended in various synthetic, natural, and processed waters to discern factors responsible for particle stability, and in particular, particle dissolution. Deionized, tap, stream, and EPA moderately hard waters were used to test the oxidizing effects of chlorine, the stabilizing effects of DOC, and the influence of ionic strength and anion composition. Additionally, particles were suspended in 1 mg/L and mass equimolar concentrations of both Cl- and S2- were added to independently study the effects of chloride and sulfide complexation. Finally, ENPs were suspended in filtered WWTP anaerobic sludge supernatant to examine dissolution in this extremely complex matrix. These studies were conducted at predicted environmentally relevant ENP concentrations (ng/L range) using spICP-MS, and, to the best of our knowledge, the first study of its kind in this regard. We hypothesize that ENP transformation (namely, dissolution), is expedited at these lower particle mass concentrations due to an initial stronger driving force for the particles to disperse and, therefore, faster dissolution.

4.2 Materials

Ag ENPs (NanoXact) were acquired in sizes of 60 and 100 nm diameters (Nanocom- posix, San Diego, CA). For each size, three surface coatings (capping agents) were tested including citrate, tannic acid and polyvinylpyrrolidone (hereafter TA and PVP, respectively). Suspensions were supplied at a nominal concentration of 20 mg Ag/L. Accompanying size information (Dynamic Light Scattering and Transmission Electron Microscopy) verified these particles to be monodisperse with the nominal sizes being: 60 ± 5.3 nm and 100 ± 9.4 nm. In house characterization by both asymmetrical flow field flow fractionation and differential centrifugation of the particles suggested the “100nm” particle to be 91.3 ± 0.6 nm[127]. Ag ENP suspensions were made by diluting stock solutions with the appropriate water composition to a final concentration of 50 ng Ag/L. Dissolved Ag standards (High-Purity Standards; QC-7-M), used for calibration, were diluted in 2% nitric acid (Optima grade) to concentra-

85 tions ranging from 0.1 to 1 g/L. Several water chemistries were investigated including deionized water (DI,18.3M-ohm Nanop- ure), tap water (Colorado School of Mines campus), surface water, EPA moderately hard reconstituted laboratory water, and WWTP anaerobic sludge supernatant. The surface water sample (Clear Creek water), collected in June 2012 from Clear Creek in Golden, CO, was collected beneath the water surface, approximately 1m from shore, and passed through a0.45micronﬁlter.Thesamplewasstoredinapolyethylenebottleat4°Cpriortouse. WWTP centrate collected from Boulder, CO municipal WWTP on October 18, 2012, was stored in polyethylene bottles and frozen at 5°C until analysis. When used, centrate was thawed in a water bath, passed through a 0.45 micron ﬁlter (aerobically), and placed in the anaerobic chamber where ENPs were spiked into the solution. Samples were kept in the anaerobic chamber until analysis. Concentration of common elements were measured by ICP-AES (Perkin Elmer 5300). Anion concentrations were determined by ion chromatography, Dionex ICS-90 ion chromatography system, while total organic carbon was measured by a Sievers model 5310C

TOC analyzer. Tap water samples contained 1.13 ± 0.04 mg/L residual free chlorine, as determined through historical records at the Golden, CO water treatment facility. Complete water chemistry information is provided below (Table 4.1). When studying eﬀects of chloride on Ag NP stability, NaCl was added to DI water at 1 mg/L and equimolar (50 ng/L) Cl-.Sulﬁdationexperimentswereconductedinananaerobic chamber, with samples being maintained in the chamber until ten minutes prior to analysis. Na2S was added at 1 mg/L and equimolar S- concentrations, prepared inside the chamber, with DI water being sparged with N2 prior to salt addition. Suwanee River DOM (90% fulvic acid), purchased from the IHSS, was added to DI water at 2 and 20 mg C/L concentrations.

4.3 Methods

Methods to study the dissolution of Ag ENP are described below.

86 Table 4.1: Water Chemistry Composition Water chemistry composition of various samples, mg/L TOC and IC Results Analyte Name DL (mg/L) Tap Water Clear Creek WWTP Centrate DOC 0.92 0.42 2.49 0.02 166 ± ± F 0.05 0.47 0.44 1.15 Cl 0.1 9.08 6.35 971.7 Br 0.1 BDL BDL BDL NO3 0.1 0.95 0.43 0.7 PO4 0.5 BDL BDL 79.21 SO4 1 52.1 24.56 5.69 ICP-OES Results Analyte Name DL (mg/L) Tap Water Clear Creek WWTP Centrate Ag 0.0002 BDL BDL 0.0002 Al 0.0138 0.0232 0.0405 BDL Ca 0.0041 27.6590 14.6438 40.0532 Fe 0.0004 0.0022 BDL 0.0390 K 0.0315 2.7639 1.3960 65.8117 Mg 0.0003 6.8457 3.2773 2.6462 Na 0.0211 23.3971 6.4513 43.4222 P 0.0197 BDL BDL 12.1176 S 0.0061 29.9759 8.5558 2.9709 Si 0.2195 2.0525 2.5983 0.3664 Zn 0.0003 1.2883 0.0246 0.1146

87 4.3.1 Instrumentation

A Perkin Elmer NexION 300Q was used for single particle analysis (spICPMS). 107Ag was continuously monitored for detection, with integration dwell times of 10 ms. Instrument calibration was achieved by analysis of a blank and four dissolved Ag solutions (0 to 1 g/L) with data collected in the sp mode. Standards were made both in 2% HNO3 acid and in matrix matched to the water chemistry being studied on a given day. The acidified sample served as a check standard through the day and a measure of sensitivity of the instrument, where the latter calibration curve was used for particle sizing. No internal standard was employed, as only 107Ag was quantified during the run. To ensure the absence of significant instrumental drift over time, a 100 ng/L Ag dissolved calibration check standard was run in sp mode for every ten Ag ENP samples analyzed. If drift in the standard signal was detected (generally due to the salinity of the samples salting the ICP-MS cones), the particle sizing equation was adjusted accordingly for the decrease in sensitivity. If check standards drifted more than 30% in intensity for a given day, data was not used because 1) pulses would often become too small to interpret for smaller size particles, and 2) reproducibility was poor under these conditions.

4.3.2 Data Collection, Conversion to Particle Size, and Quality Assurance

The theoretical basis of spICP-MS detection, has been well studied in recent years[29, 80, 107, 123]. Further refinement and development of the technique has been explored to discuss co-detection of dissolved and nanoparticulate metal[107, 109], polydisperse NP samples[140], use of dissolved standards in spICPMS analysis[127], and comparison to other sizing methods[109, 128]. For spICP-MS analysis, raw intensity data were plotted as pulse intensity as a function of pulse number, where any values below the first minimum in the histogram were considered background/dissolved, and values higher than the first minimum were considered ENP pulses. In solutions containing both dissolved and Ag ENPs, the background/dissolved counts were

88 subtracted from the pulse intensity, and ENPs were sized using spICP-MS theory[109, 127]. To avoid particle coincidence concentrations were used whereby no more than 15 percent of the measurements were ENP pulses[109, 140]. In many cases determining ENP size is a more sensitive metric than accurately determining dissolved concentration. This is especially true in more complex water chemistries, where the sensitivity of the instrument is inhibited. This is because the dissolved counts are typically very near the detection limit, and have larger standard deviation, than the pulse incidents. This confers a greater source of error when comparing the dissolved count to the calibration curve, where a few counts difference may be several ng/L difference in dissolved concentration whereas a few counts difference on a pulse height, after transformation to particle size, related to fractions of a nm difference in particle diameter.

4.3.3 Characterization of Particle Stability and Silver Release

The stability and dissolution of the particles were studied over a period of seven days. All time points and replicates in a “dissolution set” were analyzed in a single day by staggering the start of dissolution for each time point. Each dissolution set consisted of a given water chemistry and one particle size, inclusive of all particle coatings and time points, performed in triplicate. Time points included 0, 1, 2, 4, 8, 12, 24, 96, and 168 hours, with some dissolution sets excluding later time points. Samples were shaken by hand prior to analysis. Size detection limit was defined by separation of the ENP histogram from the instrumental/dissolved metal background, where the smallest detectable ENP created a mean pulse intensity beyond the first minimum in the ENP histogram, i.e. we were not determining particle size from a “shoulder” of the pulse distribution, but rather, only defined particles as clearly separate from the background, typically 25 – 30 nm in size. Once the determination between dissolved and ENP fractions had been made, each fraction was quantified using the appropriate procedure. For the dissolved fraction, this was performed via a direct comparison to the dissolved calibration curve. For the ENP fraction, once the mean dissolved background intensity was subtracted from the pulse intensity, the intensity was converted to mass, which subsequently

89 enabled calculations of ENP diameter and mass/number concentrations[109]. Contributing factors that tended to increase the minimum size detection limit included decreased ICP-MS sensitivity, signal suppression due to the matrix, salting of the cones with high(er) saline solutions, and increased concentrations of dissolved Ag in the background.

4.3.4 Dissolution Rate Kinetics

Using the instantaneous average particle diameter, as determined by spICP-MS, we calculated the mass of Ag lost from the original particle. Normalizing by surface area, we plotted the Ag mass lost per surface area (mol/cm2)versustime,withtheresultingslopeasthe dissolution rate constant. For experimental data sets that contained the longer time points (i.e. 96 and 168 hours), we calculated two rates: one for the initial dissolution stage of time points under 24 hours and one for time points 24 hours and beyond. Rate constants were calculated for all systems where the particles showed change in size over the experimental time period, and experimental average rate constants were calculated as the average of the log-transformed rate constants for individual experiments.

4.3.5 Preliminary Control Studies

A number of simple, preliminary control studies were conducted to ensure consistent and representative analysis over the course of many dissolution sets. These included, 1) effect of light on ENP dissolution, 2) effect of ENP concentration on dissolution, and 3) ability to ensure that an effective mass balance was achieved via spICP-MS calculations. All dissolution sets were aged on the laboratory bench top. To ensure minimal contribution to dissolution due to light, we aged 100nm citrate capped Ag ENPs in DI water for twelve hours with various light treatments including: natural light (from window), laboratory bench, UV lamp, dark, and refrigerated. Size analysis via sp-ICP-MS was performed to determine if preferential dissolution occurred during any treatment. As an example of decreased rate of ENP dissolution at higher concentrations, we made asuspensionof50g/L 60nm TA capped Ag ENPs in DI water (3 orders of magnitude

90 higher concentration than other dissolution sets). At hour time intervals (up to 12 hours), we analyzed the sample by 1) direct injection of concentrated sample via AF4-ICP-MS and 2) dilution to 50 ng/L in DI water, with immediate analysis by spICPMS. For both methods we determined ENP size and calculated relative change in ENP diameter over time. Mass balance was determined for all dissolution sets to include, 1) dissolved Ag+ concentration, 2) particle concentration as calculated by summation of particle mass analyzed, and 3) total Ag as measured by addition of ionic and ENP mass fractions. In some cases where particle number decreased signiﬁcantly with time (e.g. surface water), we ﬁxed initial particle number concentration and calculated an “adjusted total Ag concentration” for comparison with the measured total Ag concentration.

4.4 Results and Discussion

Here the results and discussion of the dissolution of Ag ENPs is presented.

4.4.1 Preliminary Control Studies

No significant size difference was detected regardless of light treatment after twelve hours (Table 4.2). This ensures us that, at least over the time scales of the majority of our experiments and under these laboratory conditions, there was no preferential dissolution due to photo-degradation. There was no detectable dissolution of 60 nm TA capped Ag ENPs at a concentration of 50 g/L (Figure 4.1) over a twelve-hour time period. This was confirmed both by AF4-ICP-MS and spICP-MS. However, over a relatively short time scale, this persistence of particle size is in direct contrast to dissolution results at lower concentrations, as will be discussed. Determining the change in ENP size is a good measure of dissolution and the best metric from which to calculate the release of Ag+ from the ENP surface. Change in pulse intensity, subsequently converted to particle diameter, is both sensitive and reliable. In solutions where particles were dissolving, even without converting data to diameter, it was clear from the raw data that pulse intensity decreased dramatically over time. An example of this with

91 Table 4.2: Eﬀect of Light Treatments on Particle Dissolution Particle size as determined by spICP-MS. 100nm citrate coated Ag ENPs suspended in DI water, aged 12 hours in various light treatments. Treatment Avg. Particle STDEV (nm) Size (nm) T=12 hr Lab 78 1.4 Dark 82 0.7 Fridge 82 0.2 UV Light 79 2.4 Window 78 3.2

>?'@ABC@D,">/0154-4" '!!!!" EF!" &#!!!" !" &!!!!" EF'"

%#!!!" EF$%" %!!!!" $#!!!" $!!!!" #!!!"

",-./01"-/23/4-25"6789/24:" !" %!" &!" '!" #!" (!" )!" *!" +!" $!!" ;323/<8/"<=3"6437:" 7=>?@A>BC"D5:EF797" #!!" '&!" GH!" '%!" #" GH$" '$!"

'#!" GH'#" '!!" &!" %!" $!" ,-./01"23"040567" #!" !" #!" (!" $!" )!" %!" *!" &!" +!" '!!" 89:.0601";5.<"

Figure 4.1: High(er) Concentration Dissolution Over Time 60 nm TA coated Ag ENPs, initial concentration 50 g/L. Analyzed over time (0, 4 and 12 hours). A) Sample direct injection for AF4-ICP-MS analysis, B) Sample dilution to 50 ng/L for spICP-MS analysis.

92 100nm TA capped Ag ENP, 50 ng/L, in tap water (Figure 4.2).

'!" &!" %!" <=!" $!" <=$"

7+89.:"5;"45+01-" #!" <=&" !"

!" <=>" '!"

#!!" <=#$" #'!" $!!"

$'!" <=$&" %!!" %'!"

*+,-."/01.0-/12" &!!" &'!" <=?(" 345+01-6" '!!" ''!"

(!!" <=#(>" ('!" )!!"

Figure 4.2: Dissolution via Pulse Intensity Change Dissolution observed at 50 ng/L by spICP-MS. 100 nm TA Ag ENP.

Silver mass balance, (ENP plus ionic), was not observed for all dissolution sets. Specif- ically, DI water or systems that promoted rapid ENP dissolution achieved a good mass balance whereas systems containing DOC, such as Clear Creek water, did not (Figure 4.3). Panels A and C, DI and Creek waters, picture measured ionic and ENP concentrations, with total Ag being a summation of these two fractions. Decrease in particle number over time is observed (panels B and E, DI and creek water respectively). In creek water, this difference is more dramatic, with loss of particle number and lack of Ag+ detected. Adjustment of particle number over time to fix it to the initial particle number, and with subsequent calculation of the ENP and ionic fractions of Ag (panels C and F, DI and creek water), a steady Ag concentration over time is achieved. In the latter samples, there was a deficit in total Ag concentration over time. We suggest two responsible factors, 1) decrease in ENP number analyzed over time (panels B and E), and 2) inability to detect Ag+. However, since in later studies we are determining change in ENP size instead of change in ENP concentration, we

93 feel that the decrease in pulse number analyzed does not directly aﬀect particle sizing. In Figure 4.3 panels C and F, after ﬁxing the particle number over time to the initial particle number and calculating Ag+ from the decrease in particle diameter, we achieved a steady concentration of Ag over time in both DI and creek samples (Table 4.3andTable 4.4).

)!" '#!" )!" 9-13?"*+",-./0.1234-." @A",-./0.1234-." (!" (!" !" B3/C+2-=.D",-./0.1234-." '!!" $" #" '!" '!" &!" &!" &!" %!" *+,-./0+"=3/7A"*1"234504/67834" %!" %!" *+,-./0+"BC"234504/67834" $!" 27A5-A7/0+"*1"D34"234504/67834" $!" $!"

#!" #!" #!" *+",-./0.1234-."5.+678"

!" !" *+,-./0+"*1"234504/67834"941:;<" !" *+,-./0123"4.,56/2"7+86289,.09+8" !" '" #!" #'" $!" $'" !" #(" (!" )(" '!!" '#(" '(!" ')(" !" $'" '!" )'" #!!" #$'" #'!" #)'" 9:;0"5<-=2>8" :0-2";<+=,>?" =>?0"9@3-6.<" )!" '#!" )!" 9-13?"*+",-./0.1234-." *+,-./0+"=3/7A"*1"234504/67834" @A",-./0.1234-." (!" (!" *+,-./0+"BC"234504/67834" B3/C+2-=.D",-./0.1234-." '!!" %" &" '" 27A5-A7/0+"*1"D34"234504/67834" '!" '!" &!" &!" &!" %!" %!" %!" $!" $!" $!"

#!" #!" #!" *+",-./0.1234-."5.+678"

!" !" *+,-./0+"*1"234504/67834"941:;<" !" *+,-./0123"4.,56/2"7+86289,.5+8" !" '" #!" #'" $!" $'" !" #(" (!" )(" '!!" '#(" '(!" ')(" !" $'" '!" )'" #!!" #$'" #'!" #)'" 9:;0"5<-=2>8" :0-2";<+=,>?" =>?0"9@3-6.<"

Figure 4.3: Dissolution Mass Balance Analysis 100 nm TA Ag ENPs. Ionic and ENP concentrations in A) DI water and D) Creek water. Observed decrease in particle number, panels B and E, DI and creek water respectively. Adjustment of particle number over time by ﬁxing to initial particle number, subsequent calculation of ENP and ionic fractions (panels C and F, DI and creek water).

4.4.2 Eﬀects of Water Chemistry on Ag NP Stability - Mechanistic Studies

Comparison of results in simple laboratory-prepared waters, including DI and additions of chloride, sulfide, or DOC, can be found in Figure 4.4. Dissolution of 60 and 100 nm particles in DI water is seen in Figure 4.4A, where the largest decrease in particle diameter is within the first 24 hours. A dissolution rate of approximately -11.5 mol/cm2s-1 was calculated, with no induction period noted. Slower but steady subsequent decline over the remaining 144 hours of the experiment is observed, with a rate an order of magnitude (one log unit) lower. Without additional time points, it is difficult to ascertain if the rate has reached steady

94 Table 4.3: Dissolution Mass Balance Summary – DI Water

DI Water Samples, Raw Data Time Diameter Individual Particle Particle Particle Mass Calculated Total (h) (nm) Particle Num- Num- Concentra- Ag+ Loss Ag mass (ug) ber/mL ber/L tion (ng/L) (ng/L) (ng/L) 0 91 4.11E-9 513 5.13E+5 52.7 0 52.7 1 84 3.25E-9 492 4.92E+5 40.0 10.5 50.5 2 81 2.94E-9 467 4.67E+5 34.3 13.7 48.0 4 81 2.90E-9 455 4.55E+5 33.0 13.8 46.7 8 78 2.64E-9 459 4.59E+5 30.2 16.9 47.1 12 70 1.88E-9 469 4.46E+5 22.1 26.2 48.2 24 61 1.23E-9 392 3.96E+5 12.2 28.4 40.6 96 48 5.92E-10 328 3.28E+5 4.8 28.8 33.6 168 36 2.65E-10 309 3.09E+5 2.0 29.8 31.8 DI Water Samples, Adjusted Particle Number Time Diameter Individual Fixed Particle Particle Mass Calculated Adjusted (h) (nm) Particle Particle Num- Concentra- Ag+ Loss Total mass (ug) Num- ber/L tion (ng/L) Ag ber/mL (ng/L) (ng/L) 0 91 4.11E-9 515 5.15E+5 52.9 0.0 52.9 1 84 3.25E-9 515 5.15E+5 41.9 11.0 52.9 2 81 2.94E-9 515 5.15E+5 37.8 15.1 52.9 4 81 2.90E-9 515 5.15E+5 37.3 15.6 52.9 8 78 2.64E-9 515 5.15E+5 34.0 18.9 52.9 12 70 1.88E-9 515 5.15E+5 24.2 28.7 52.9 24 61 1.23E-9 515 5.15E+5 15.9 36.9 52.8 96 48 5.92E-10 515 5.15E+5 7.6 45.2 52.9 168 36 2.65E-10 515 5.15E+5 3.4 49.5 53.0

95 Table 4.4: Dissolution Mass Balance Summary – Clear Creek Water

Clear Creek Water Samples, Raw Data Time Diameter Individual Particle Particle Particle Mass Calculated Total (h) (nm) Particle Num- Num- Concentra- Ag+ Loss Ag Mass (ug) ber/mL ber/L tion (ng/L) (ng/L) (ng/L) 0 86 3.50E-9 539 5.39E+5 47.1 0.0 47.1 1 86 3.44E-9 486 4.86E+5 41.8 0.6 42.4 2 85 3.40E-9 487 4.87E+5 41.3 1.2 42.5 4 85 3.35E-9 444 4.44E+5 37.2 1.6 38.8 8 84 3.25E-9 404 4.04E+5 32.9 2.5 35.3 12 82 3.07E-9 373 3.73E+5 28.6 3.9 32.6 24 80 2.85E-9 323 3.23E+5 23.1 5.2 28.2 96 84 3.30E-9 173 1.73E+5 14.3 0.8 15.1 168 72 2.05E-9 76 7.60E+4 3.9 2.7 6.6 Clear Creek Water Samples, Adjusted Particle Number Time Diameter Individual Fixed Particle Particle Mass Calculated Adjusted (h) (nm) Particle Particle Num- Concentra- Ag+ Loss Total mass (ug) Num- ber/L tion (ng/L) Ag ber/mL (ng/L) (ng/L) 0 86 3.50E-9 539 5.39E+5 47.1 0.0 47.1 1 86 3.44E-9 539 5.39E+5 46.4 0.7 47.1 2 85 3.40E-9 539 5.39E+5 45.7 1.4 47.0 4 85 3.35E-9 539 5.39E+5 45.1 2.0 47.1 8 84 3.25E-9 539 5.39E+5 43.8 3.3 47.1 12 82 3.07E-9 539 5.39E+5 41.3 5.7 47.1 24 80 2.85E-9 539 5.39E+5 38.4 8.7 47.1 96 84 3.30E-9 539 5.39E+5 44.4 2.6 47.1 168 72 2.05E-9 539 5.39E+5 27.6 19.4 47.0

96 state dissolution after 168 hours. We attribute the particle dissolution to surface oxidation and interpret the slowing of dissolution to a build up in Ag+ ions on the surface, as these ions may kinetically inhibit further oxidation of the particle. Dissolution rates depend on particle surface coating, ranging up to 0.5 log units in a given dissolution set. In most instances, the TA capping agent shows faster dissolution (up to 0.9 log units) compared to both the citrate and PVP coatings, which are similar. In the first 24 hours, the 100 nm TA capped ENPs had decreased up 15% more in diameter, while after 168 hours the diameter was nearly 40% smaller. When adding Cl– (Figure 4.4B) we detect little to no change in ENP size over time and thus no rate could be determined. While the more concentrated 1 ppm Cl- hinders dissolution more than the equimolar Cl-:Ag+ experimental set; approximately 10% difference in diameter after 24 hours was observed in the latter experiment. Although the particles sized slightly larger than the nominal diameter in the sulfide sets, the trend is clear that there is no dissolution over the length of the experiment (Figure 4.4C.) This may indicate that the anaerobic conditions and/or complexation with sulfide completely suppressed particle diameter change, which is strong evidence that particle surface oxidation is an important factor in the dissolution of these particles. Finally, the addition of DOC, at both 2 and 20 mg/L also slowed the dissolution of particles (Figure 4.4D), with the higher DOC essentially halting dissolution and the 2 mg/L concentration allowing the particles to decrease less than 10% in diameter over 24 hours. Because of minimal dissolution in these sets, no kinetic dissolution rates were calculated. We suspect that DOC adhering onto the particle surface provided a physical barrier which prevented oxidation of the Ag core, though without further data this is conjecture. Many studies investigating the complexation of DOC and ENP where electron microscopy was employed showed general aggregate complexes and did not discern the effect of DOC on individual particles[9, 36, 89], though this may be because either the higher concentration of particles induced aggregation more readily in these systems or imaging techniques are not poised to capture individual ENP transformations. We suspect that if aggregation had

97 +""#""$ '"$72$>045143$ '"$72$91770?$@?0A$ *"#""$ !" '"$72$.B.$ +""$72$>045143$ )"#""$ +""$72$91770?$@?0A$ ("#""$ +""$72$.B.$

'"#""$

&"#""$ -.$/0123435$6728$ %"#""$

!"#""$ "$ ,&$ &"$ (&$ +""$ +,&$ +&"$ +(&$ 9023$6:;<5=8$ *""# *""# *""#

)"# )"# )"#

("# ("# ("#

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

Figure 4.4: Dissolution Experiments in Laboratory Prepared Waters Summary of Ag NP dissolution, 50 ng/L, in simple systems and mechanistic studies. A) Comparison of 60 nm and 100 nm NPs in DI water, B) Addition of 1 mg/L and equimolar Cl-, C) Addition of 1 mg/L and equimolar S2-, and D) Addition of 20 ppm and 2 ppm NOM. Note that the approximate detection limit of 30 nm was reached for the 60 nm TA sample

98 occurred, it would be evidenced by either increased pulse intensities or multiple, adjacent readings containing ENP pulses.

4.4.3 Environmental Systems

Only slight variations in size of the nominal 100 nm Ag ENPs were detected, regardless of capping agent, over a seven day period (Figure 4.5A) in environmentally relevant waters. Because similar trends were observed between the 60 nm and 100 nm particles in other water chemistry variations, we did not complete dissolution sets for the 60 nm particles in EPA moderately hard water. In creek water, Figure 4.5B, we observe slightly more dissolution than in the EPA moderately hard water, where both the 60 nm and 100 nm samples decrease just over 10% in particle diameter. Because minimal to no dissolution was observed, no kinetic measurements were derived for these sets.

4.4.4 Processed Water Samples

Dissolution in chlorine-containing tap water was faster than all other solutions examined (Figure 4.6A), averaging an increase of 0.4 log units and 0.92 log units compared to DI water and creek water, respectively. For the 60 nm particles, only data points within the ﬁrst eight hours are represented, as particles from latter time points were no longer detectable; a combination of decreased pulse intensity and increased background concentration made the signal to noise ratio too small. We hypothesize that the free residual chlorine in the tap water was responsible for the increased rate of dissolution for all particles. For the WWTP anaerobic sludge supernatant dissolution set, only the TA coated particles were examined. Due to the high salt and DOC of this matrix, preliminary studies resulted in the cones of the ICP-MS becoming caked with debris causing a drop in ICP-MS sensitivity. In this study, when running the triplicate TA samples at hourly intervals, there was an approximate 10-15% decrease in sensitivity, which recovered after ﬂushing the system with 2 % nitric acid (rinse) for the remainder of the hour until the next time point. As TA-coated particles appeared to be most sensitive to dissolution, this particle was chosen as a potential

99 *""#

)"#

("# '"# #" &"#

%"#

,-#./012324#5617# *""#61#=/34032#

$"# *""#61#8066/>#?>/@#

*""#61#-A-# !"# "# +%# %"# '%# *""# *+%# *%"# *'%# 8/12#59:;4<7# *""# &"#61#=/34032# &"#61#8066/>#?>/@# )"# &"#61#-A-# !" *""#61#=/34032# *""#61#8066/>#?>/@# ("# *""#61#-A-#

'"#

&"#

%"# ,-#./012324#5617#

$"#

!"# "# +%# %"# '%# *""# *+%# *%"# *'%# 8/12#59:;4<7#

Figure 4.5: Dissolution Experiments in Environmental Waters Summary of Ag NP dissolution, 50 ng/L, in environmentally relevant water systems: A) EPA moderately hard water and B) Clear Creek water, Golden CO

100 indicator as to the maximum amount of dissolution likely for a given water chemistry. As seen in Figure 4.6B, there is evidence of dissolution WWTP sludge supernatant. This is unexpected, as the experiment was conducted in the anaerobic chamber (albeit with an initial aerobic filteration step), contained sulfide and significant amounts of DOC; all factors that previously had minimized the dissolution trend.

*""# &"#61#=/34032# *""#

&"#61#8066/>#?>/@# *""#61#8066/=#>=/?# )"# !" &"#61#-A-# )"# #" *""#61#=/34032# ("# *""#61#8066/>#?>/@# ("# *""#61#-A-# '"# '"#

&"# &"#

%"# %"# ,-#./012324#5617# ,-#./012324#5617# $"# $"#

!"# !"# "# +# $# &# (# *"# *+# *$# *&# *(# +"# ++# +$# "# +# $# &# (# *"# *+# *$# *&# *(# +"# ++# +$# 8/12#59:;4<7# 8/12#59:;4<7#

Figure 4.6: Dissolution Experiments in Processed Waters Summary of Ag ENP dissolution, 50 ng/L, in processed waters: A) tap water and B) TA Ag ENPs in sludge supernatant. Limited time points collected for 60 nm tap water samples as they decreased below the method detection limit at latter time points.

4.4.5 Comparison of Water Chemistry Eﬀects

We can elucidate several similarities between data sets where water chemistry was similar. For example, in Figure 4.7, we note that in both the 2 mg/L DOC simple experiment and the Clear Creek experiment, which measured approximately 2 mg/L DOC, the dissolution rate is similar, within standard deviation, over the 24 hour time period.

4.4.6 Kinetic Rates of Dissolution

In Figure 4.8A, the calculated mass loss of Ag from the 100 nm particle in DI water attains a constant rate consistent with steady-state dissolution immediately, until 24 hours where the dissolution rate slows. Similarly, as shown in Figure 4.8B, the instantaneous Ag

101 *""#

)"#

("#

'"#

&"#

%"# ,-./0102#34/5# ,;#<.102# $"# =>0.2#=200?#<.102# +#@@/#ABC# !"# "# +# $# &# (# *"# *+# *$# *&# *(# +"# ++# +$# 6-/0#37892:5#

Figure 4.7: Comparison of Dissolution in Various Matrices Creek and NOM spiked water, both with approximately 2 ppm NOM, compared to DI water. Apparent similar dissolution rates due to NOM. dissolution rates is steady for the first 24 hours. A summary of all dissolution experiments and results is provided in Table 4.5, with separates rates of dissolution for time points before and after 24 hours. With respect to specific characteristics of both the particles and/or water chemistry, there are several trends that can be garnered from Table 4.5: ENP size did not appear to be a controlling factor in the dissolution rate, as the mass released per surface area was not significantly different in most dissolution sets. This suggests that for mass normalized rates can, at first approximation, be particle size independent. Variable dissolution rates were observed for different particle coatings. The TA particle coating seemed to be the least resistant to dissolution, while citrate and PVP provided nearly equal stability to the Ag NPs. Water chemistry was the most important factor in Ag NP dissolution in this study. A number of experimental sets (i.e. 1 mg/L chloride and sulfide, 20 mg/L DOC, and EPA moderately hard water) did not show any significant change in size over 24 hours (i.e. less than 10% diameter difference). Therefore, these experimental sets are not included in the

102 Table 4.5: Summary of Experimental Dissolution Rates

Nominal Water Surface t < 24 hr average t > 24 hr average Diameter Chemistry Coating rate Log rate Log (nm) r/(mols/cm2/s) r/(mols/cm2/s) 100 DI Water Citrate -11.72 0.07 -12.59 0.15 ± ± 100 DI Water Tannic Acid -11.62 0.11 -11.92 0.072 ± ± 100 DI Water PVP -11.78 0.13 -12.71 0.07 ± ± 100 Tap Water Citrate -11.32 0.05 ± 100 Tap Water Tannic Acid -11.18 0.06 ± 100 Tap Water PVP -11.56 0.21 ± 100 Clear Creek Citrate -12.14 0.09 -13.07 1.03 ± ± 100 Clear Creek Tannic Acid -12.49 0.10 -13.34 0.56 ± ± 100 Clear Creek PVP -12.33 0.03 -12.03 0.18 ± ± 100 2ppmDOC Citrate -12.46 0.09 ± 100 2ppmDOC Tannic Acid -12.57 0.13 ± 100 2ppmDOC PVP -12.16 0.07 ± 100 WWTP Tannic Acid -12.01 0.01 Centrate ± 60 DI Water Citrate -12.23 0.09 -13.12 0.29 ± ± 60 DI Water Tannic Acid -11.80 0.04 ± 60 DI Water PVP -12.62 0.07 -12.57 0.26 ± ± 60 Tap Water Citrate -11.64 0.13 ± 60 Tap Water Tannic Acid -11.76 0.11 ± 60 Tap Water PVP -11.88 0.09 ± 60 Clear Creek Citrate -12.71 0.14 -13.14 0.16 ± ± 60 Clear Creek Tannic Acid -12.38 0.10 -12.88 0.12 ± ± 60 Clear Creek PVP -12.36 0.12 -13.03 0.16 ± ±

103 ,"&#'!*% !"%$%& ,"&#'!*% "%%&=6&>

8% !"%$#& + ,"!#'!*% +"&#'!*% "%%&=6&GHGB&CD&EA?4@& +"!#'!*% !" #" 79:& !""$%& 8% ) & )"&#'!*% 12/%3456174 $ )"!#'!*% !""$#&

+"!#'!(% /0 &"!#'!(% !")$%& !"!#$!!% ,!!%=4%>:?@A?;B%CD%EA?;@% !% +!!!!% -!!!!% .!!!!% (!!!!% )!!!!!% *"!#'!(% 9:4;%3<8% ,!!%=4%9/B%CD%EA?;@% !")$#&

,!!%=4%FGFB%CD%EA?;@% !"($%&

12/%3456174 )"!#'!(% $ !"($#&

/0 !"'$%& &"!#'!(% ./0&12/34&56/2786 !"'$#&

!"!#$!!% !"#$%& !% ,!!!!!% &!!!!!% -!!!!!% )!!!!!% .!!!!!% *!!!!!% (!!!!!% %& "%%%%& )%%%%& (%%%%& '%%%%& #%%%%& *%%%%& +%%%%& ,%%%%& -%%%%& "%%%%%& 9:4;%3<8% ;<64&59:& Figure 4.8: Kinetic Rates of Dissolution for Ag ENPs Temporal variation and instantaneous rates during representative dissolution set, DI water. Blue, red, and green traces represent citrate, TA, and PVP coatings respectively. A) Geometric surface area normalized calculated mass of Ag released from particles over time (inset, 24 hour time period) and B) instantaneous dissolution rates. list of results in Table 4.5. Likewise, solutions containing lower concentrations of DOC, such as the 2 mg/L DOC, creek water, and WWTP centrate, had signiﬁcantly reduced dissolution rates compared with DI water, up to 1 log unit. For both the 60 nm and 100 nm particles, regardless of capping agent, tap water promoted the fastest dissolution. We hypothesize this trend may be attributed to the residual chlorine, which would promote surface oxidation on the particles and subsequently lead to increased dissolution. Dissolution rate slowed after 24 hours compared to the initial rate in any given set. This trend could be explained by a number of factors, depending on the solution chemistry. In waters where there is DOC, both the initial rate of dissolution and latter time points may generally be slower due to the sorption of organic carbon onto the particle surface, forming aprotectivebarrierfromoxidationandsubsequentdissolution.

4.5 Implications

Dissolution has the potential to be a key component of the screening process for catego- rizing ENPs with common hazard potential based on their potential to release ionic species. However, simple correlations between dissolution and physico-chemical properties of ENPs

104 are often difficult to make because of the myriad of particle types, capping agents, and complexity of the system being studied. Often, a change in one parameter will significantly change the extent of dissolution and determining how multiple factors might co-vary is exceedingly difficult. Protocols for addressing the most wholistic yet reasonably accomplishable measure of particles in this way has yet to be addressed. Generally, environmental exposures are likely to be chronic and at low concentration, likely over long time scales dictating slow or multiple transformations of the ENPs over time. A number of analytical techniques will likely have to be used in conjunction with extrapolating calculations to feasibly understand the multitude of ENP interactions with exposure media. However, the advancement of the spICP-MS technique is a great benefit for the study of ENPs in natural systems at trace concentrations. Used as a screening tool, this method could elucidate trends in ENP behavior accurately and relatively quickly.

4.5.1 Eﬀects of Water Chemistry on ENP Transformation in Aerobic Systems

Ag+ ion-release rate at high concentrations of Ag ENPs is retarded because Ag+ rapidly reaches a critical concentration (i.e. equilibrium). In comparison, Lee et al. suspended 100 nm citrate Ag ENPs in air-saturated DI water in concentrations of 0.05 mg/L, 0.1 mg/L, and 1 mg/L; calculating rate constants (ln[Ag]/hr) of 0.0734/h, 0.0709/h, and 0.0278/h, respectivelyLee et al. [84]. Under similar conditions, Liu and Hurt examined approximately 2 nm citrate Ag ENPsLiu & Hurt [91], measuring Ag+ accumulation in solution until a seemingly steady state was achieved (between 6 and 125 days, depending on concentration), yet no kinetic rate of dissolution was calculated. These examples draw emphasis to the point that variations in both experimental conditions, particles used, and reporting of results make comparison between studies exceedingly difficult. However, in relation to our present study, these groups also found first order dissolution kinetics over the first 24 hours, with slowed release both at longer time points and in more complex environments. Using approximately 5 nm Ag ENPs at concentrations ranging from 0.001 mM – 0.1 mM, Ho et al. determined pseudo-first-order rate constants for Ag ENP dissolution [62]. A

105 dramatic decrease in Ag ENP dissolution rate was observed in the presence of one equivalent of Cl- (0.05 mM) and the dissolution tended to zero when 1 mM Cl- was used. The authors suggest halide binding to the surface slowed down the etching of Ag ENPs by blocking added

H2O2, which limited dissolution. Although at orders of magnitude higher concentration, this is similar to our observations. The study of sulfidation processes is crucial from an environmental perspective because the majority of Ag ENPs may end up in WWTP effluent and biosolids. Liu et al. have suggested two different mechanisms of sulfidation depending on sulfide concentration[93].

At high sulfide concentrations, by direct conversion of Ag-ENPs to Ag2S NPs, or at lower concentrations by oxidative dissolution of Ag ENPs followed by sulfide precipitation. In the presence of a small amount of sulfur (enough needed to form a monolayer of Ag2Son the ENP surface) solubility was reduced by about 7-fold[87]. Furthermore, these complexes were not oxidized after a prolonged period of aeration (18h), suggesting that after Ag2S shell formation, the possibility of further oxidative dissolution is negligible. In comparison to our present studies, we also determined that there was essentially no Ag ENP dissolution at higher (1 mg/L) concentrations of S2-,thoughaslightdecreaseindiameterwasdetected at equimolar S2-:Ag concentrations. This latter observation may be explained by a pre- oxidation step, where Ag+ forms on the surface and reacts with sulfide homogeneously to form distinct Ag2Sparticles[93]. DOC has been shown to increase particle stability, even in high ionic strength solutions, which is postulated to be due to both providing a physical barrier to oxidation and changing the surface charge[34]. This effect has also been reported in the study of gold ENPs[36]. Likewise, we have observed decreased or negligible dissolution rates of Ag ENPs in solutions with 2 and 20 mg/L DOC, respectively. In natural systems, which are inherently complex, difficulties arise in direct comparisons. In one example, Li and Lenhart studied the aggregation and dissolution of various capped (citrate or Tween 80) Ag ENPs in river water over 15 days[89]. Similar to our studies,

106 the sterically stabilized particle (Tween 80 and TA, for each study), showed a rapid release of dissolved silver. Additionally, they noted substantial aggregation of the citrate capped particles after only six hours, in contrast to our results, though aggregation may have been induced due to the higher particle number concentrations (5.2x108 particles/mL, i.e. approximately 1.5 mg/L). To our knowledge, there are no studies that attempt to characterize Ag ENPs in tap water outright. However, studies have been performed to understand the washing behavior of materials embedded with Ag ENPs, in which tap water was used, such as adding a hypochlorite/detergent solution[68], particle release due to simulated washing[14], or characterization of the eﬄuent from a nanosilver producing washing machine[42].

4.5.2 Eﬀects of Water Chemistry on ENP Transformation in Anaerobic Systems

Formation of a relatively insoluble metal-sulfide shell on the particle surface can potentially 1) inhibit further dissolution or 2) alter the surface charge and induce aggregation[87]. While Ag ENP interaction with sulfur atoms can also occur with sulfur-containing biomacro- molecules and inorganic sulfur in sediments, soils, and air, the excess of sulfide in WWTP sludge is strongly suggested as a major sink of Ag ENPs. In the Targeted National Sewage Sludge Survey Statistical analysis, all 74 WWTPs tested contained Ag in the sludge, ranging from 1.94 - 856 mg/kg (on a dry weight basis)[160]. Kim et al. suggested particles formed in situ during WW treatment by reacting Ag ENPs with reduced S under anaerobic conditions [74], finding loose aggregates of 5 – 20 nm sized particles with excess surface S. This is corroborated by Choi et al. where indication that the surface of Ag ENPs (oxidized to Ag+)reactsrapidlywithsulfideions,therebypreventingor slowing further oxidation[24], which inhibits further dissolution[172]. The transformation of

Ag2Soccursinlessthan2hrsinanaerobicenvironments[72].Thisiscontradictorytoour findings, where ENP diameter decreased nearly 10% over the course of 24 hours. This may be because 1) initial aerobic filtration caused initial oxidation of the particles, or 2) variations in behavior due to the capping agent differences between these studies is substantial.

107 4.5.3 Advancement of ENP Studies with the Use of spICP-MS

The complications regarding environmental ENP analysis, and subsequent ambiguity caused by the plethora of various tests, is in part due to the lack of adequate (and standardized) analytical methods rather than due to the lack of powerful analytical instruments. Only a limited number of methods can be directly applied to aqueous samples, especially considering expected ENP concentrations (ng/L range for surface waters and mg/kg range for sewage treatment sludge)[33]. The available techniques currently used are generally not capable of measuring transformation in situ (or in vivo) or these concentrations. In addition to ﬁlling this gap in metrology, we have shown that spICP-MS can track the rate and extend of transformations of nanomaterials under realistic conditions.

108 CHAPTER 5 CONCLUSIONS

The unique and rewarding aspect of analytical chemistry, and specifically analytical method development, is to watch the tools that are developed be implemented by many other researchers in a variety of applications. This has certainly proven true in the field of nanometrology, where tools developed to study colloidal behavior and ionic metal concentrations have been repurposed through method variations to study nanomaterials. This dissertation was designed with these principles in mind; to establish and/or further develop analytical nanometrology techniques, show their functionality, utility, and promise, and ultimately leave this work as a building block for other researchers to continue to study the behavior of nanomaterials in a wide variety of contexts. Examples of these applications can be seen throughout this work; both for proof-of-concept purposes and for unique particle characterization paradigms not yet before studied at low particle number concentrations. This work represents studies investigated in three distinct phases of research: 1) spICP-MS proof-of-concept and theory validation, 2) optimization and comparison of sp-ICP-MS and AF4-ICP-MS for the detection of nanoparticles, and 3) application of spICP-MS to study environmental transformations of nanomaterials. In this chapter, I present the overall conclusions, successes and shortcoming of this research in addition to collaborative efforts and future work using these techniques.

5.1 Feasibility of detection nanoparticulate silver using spICP-MS

The first part of this dissertation, Chapter 2, presents proof of concept studies in the development of spICP-MS. Working exclusively with Ag NPs as they are expected to be of particular prominence in environmental releases, definitive experiments were conducted to prove the detection of NPs via ICP-MS. The specific aims of this chapter were: to ensure that nanoparticles can be analyzed without prior acid digestion by ICP-MS; to examine the

109 deficiencies of the traditional filtration approach for Ag NP characterization; to definitively demonstrate that pulses in spICP-MS data are quantitatively related to nanoparticles; and to test the applicability of the method to measure the presence of Ag NPs in wastewater treatment plant influent and effluent. Particles were determined to be completely ionized after reaching the plasma. Although we have yet to analyze a recalcitrant metallic nanoparticle where the ionization efficiency is less than 100%, it is recommended to ensure complete particle ablation when working with a new particle type. As anticipated, filters proved a poor way to separate NPs because, 1) particles smaller than the nominal filter pore size did not always pass through the membrane, and 2) some membrane compositions appeared to sorb copious amounts of silver to the surface. These two pitfalls made filtration an unattractive option for sample pre-treatment, regardless of downstream processing and analysis. Additionally, it was shown that pulses positively correlated with concentration of Ag NPs, and the number of pulses decreased dramatically after sample acidification. This brought us to the conclusion that pulses did in fact signify NPs. Finally, in a proof of concept study, particles were detected in both influent and effluent of the Boulder, CO wastewater treatment facility.

5.2 Optimization and Comparison of nanometrology techniques

Extensive method development for both sp-ICP-MS and AF4-ICP-MS was performed (Chapter 3), with subsequent comparisons of the techniques to determine highlights and pitfalls of each method of analysis. Optimization generally was split into two facets for both the sp and AF4 methods: physical (mechanical) optimization and data manipulation. With regards to spICP-MS, of the two optimizations, improvement to data processing protocol was most important where special eﬀort was made to elucidate the diﬀerence between background/dissolved ionic counts and NP pulses. To date, processing of the spICP-MS dataset remains the most time consuming and cumbersome aspect of the process, and is most likely to inhibit new users from quickly familiarizing themselves with the technique. Conversely, optimization in regards to AF4-ICP-MS was essentially changing the physical parameters

110 of ﬂow rates, carrier ﬂuid composition, and membrane composition to ensure reproducible results with a high percent recovery across a wide NP size range. The seemingly endless combinations of these parameters make sample analysis tedious, and too time consuming for high throughput types of analyses. Coupled with the (relatively) long analysis time per sample, the time investment to AF4-ICP-MS is the greatest downfall. Analytical capabilities were divided into the following criteria for comparison of sp and AF4-ICP-MS: A) size and concentration detection limit, B) size resolution, and C) multi- form metal analysis (e.g. distinguishing NP vs. dissolved constituents, NP complexes and aggregates, and multi-metals analysis). AF4 was easily able to detect smaller particle size (2 nm vs. 3 nm), while sp analysis could detect a lower particle concentration

(ng/L vs g/L Ag). However, the dynamic range of spICP-MS is limited, as high particle number creates coincidence and incorrect sizing. Resolution in spICP-MS is fixed by ICP-MS sensitivity, as more sensitive instruments will discriminate between smaller changes in mass better. AF4 can be “tuned” for a given resolution by changing the operating conditions, including carrier fluid and flow rate(s). In terms of multi-form analysis, spICP-MS has the distinct advantage of simultaneously determining NP size (and concentration) in addition to dissolved ion concentration. On the other hand, AF4 excels in determining NP complexes (e.g. surface modifications for increased particle size) and multiple metals analysis (i.e. bi-metallic nanoparticles).

5.3 Case Study: Application of spICP-MS to study dissolution kinetics of Ag NPs

In Chapter 4, the spICP-MS technique was used to study the kinetic rate of dissolution of Ag NPs. To date, there is not extensive work in this area because, as mentioned previously, the complexity of all factors involved including particle size, capping agent, dimensional differences (i.e. spherical, cubic, etc.) and water chemistry variations. Additionally, the studies that have been performed are at very high particle concentrations (mg/L) because of the lack of sensitive measurement techniques available. This may mask or change the behav-

111 ior the particles if they were at a more dilute concentration, which are more in agreement with typical projected environmental concentrations. Our goal was to address some of this complexity, while at the same time perform the experiments at lower particle concentrations than has been done before. When testing Ag NPs, there were signiﬁcantly higher rates of dissolution at lower particle number concentration. Essentially no change in particle size was observed at 50 g/L in DI water over a twelve hour time period, yet at three orders of magnitude lower, 50 ng/L, particles decreased up to 30% in diameter. Capping agents had variable resistance to dissolution, with tannic acid being less stable than both citrate and PVP. Many water chemistry variations altered the initial dissolution rate observed in DI water. All waters with NOM decreased the dissolution rate, and in water with either Cl- or S-,in1mg/Lorequimolar concentrations, there was no dissolution observed. Simulated natural water samples, such as EPA mod hard water, and environmental samples, such as Clear Creek water, also conferred limited change in NP size over a seven day time period. However, it should be noted that the dissolution rate was consistently faster in the ﬁrst 24 hours than from 24 hours to seven days. This may indicate Ag ion sorbed to the particle surface prohibiting further oxidation, or, in samples with NOM, NOM binding to the surface where there are no free sites to be oxidized and subsequently dissolve. The major exception to this stabilizing trend was tap water, where all particle sizes and capping agents had very rapid dissolution rates, and analysis was not possible after 24 hours. This was attributed to the free residual chlorine present in tap water, which acted as an oxidizing agent on the particle surface. Previously, it was noted that a highlight of the sp-ICP-MS technique was the ability to independently discern dissolved background concentration and NP size and concentration. It was expected, and in some cases true, that the accumulation of Ag+ could be observed over time as the particles decreased in size. However, during the course of the dissolution studies, not all mass balance measurements were complete. In systems containing NOM, for example, no Ag+ was observed and the frequency of pulses decreased dramatically in

112 latter time points. We suspect that this was due to NOM complexing the Ag (ionic or nanoparticulate) and adhering to the side of the plastic container. While a surfactant or other additive to the solution may help avoid this problem, this addition may aﬀect the NP behavior in the matrix, which is not desirable. For these experiments, particle number was not the main objective, and so this was not too much of a concern. However, for analysis in varied and more complex media, for example tissues, sediments, or wastewater sludge, this problem may need more attention in the future to ensure a representative and accurate analysis.

5.4 Collaborative eﬀorts and future work

There are a number of studies that benefit from the NP characterization abilities that both spICP-MS and AF4-ICP-MS provide. To date, there have been a number of collaborative efforts to use these techniques in new and interesting ways, with the help of colleagues from other institutions. In addition to the body of work in this thesis, the breadth of collaborative projects shows the utility of the methods advanced during the course of this work. These projects include sp-ICP-MS detection of NPs in waste water at Arizona State Univer- sity; spICP-MS and AF4-ICP-MS analysis of the complexation of NPs and NOM at King Abdullah University of Science and Technology; spICP-MS analysis for the detection of NPs in food-stuffs at RIKILT, institute of food safety in the Netherlands; spICP-MS detection of Ag NPs in large scale outdoor mesocosms from Trent University; and a spICP-MS round- robin study hosted by RIKILT for numerous laboratories to analyze test samples across the world; among others. In addition, there has been work in parallel to mine being conducted at the Colorado School of Mines in the Ranville research group to advance the spICP-MS technique. Two significant advancements include methods to analyze NPs in tissue and the analysis of carbon nanotubes released from plastics. These efforts were conducted by Evan Grey and Rob Reed; respectively. In terms of future work, there has been increasing interest in the capabilities of the AF4 instrumentation. Currently, work is progressing in the area of AF4-ICP-MS in terms of mak-

113 ing the method development easier and faster in the Von der Kammer lab at the University of Vienna. There, the group has conducted extensive work in determining a matrix of optimal flow rates, carrier fluids, etc. with which one could select the “pre-optimized condition” for the particle and system of interest. This type of approach will eventually expediate the method development process for other researchers and is very necessary so that AF4 might have the capability to become a more standard analysis. The major benefits of AF4 would be the small particle size detection limit at relatively low number concentrations and the capability to analyze multiple elements which can be associated with a NP distribution. However, as it stands, with the investment of time and money (for the AF4 machine itself), the expansion of this technique may be slower than others. I personally believe that with the current capabilities and promising application abilities of spICP-MS, it is the superior technique for NP analysis. Partly because of the ability to detect particles at lower concentrations, have little sample preparation, and utilize instrumentation that is generally available for groups studying metals analysis. However, there are several aspects that could be improved, such as: 1) streamlining the data processing steps, 2) improving the software for data collection, 3) further optimization of physical ICP-MS parameters for the analysis of NPs, and for the analysis of complex matrices, and 4) the continuation of method development to extract NPs from complex matrices. As it now stands, the arduous data processing for spICP-MS is the most difficult aspect for new users of the technique. Not only are there multiple inputs that may need to be changed on a regular basis, such as calibration curves and sampling efficiency, but there are numerous factors that an individual must manually select, leaving the data processing somewhat subjective. These factors include selecting background intensity, distinction between background and particle pulses (i.e. smallest NP pulse cut-off), and then determining if the final data set is sufficient in terms of particle number (i.e. lack of coincidence). Admittedly, while a number of spreadsheets were developed for this work that make the process semi- automated, there is still a significant amount of manual spreadsheet manipulation to process

114 each data set. Ideally, a program could be developed that easily allows the user to input variables (calibration curve and efficiency) and import complete data sets, which then could be processed automatically with algorithms differentiating between the ionic background and particle pulses, determine if the sample was in sp dynamic range, etc. Second, and closely related to the data processing software, is improving the software for data collection. There are several aspects that can be improved, namely eliminating the settling time (dead time between readings) and acquiring data at a faster rate, on the order of 0.1 millisecond dwell times. In conjunction, these two improvements would allow one to capture multiple points in a single ion cloud that represents one NP. By integrating the peak, we can then achieve more accurate sizing. In addition, for systems with high(er) ionic background concentrations, better definition of the pulse may improve our signal to noise ratio in comparison to the longer dwell times, where essentially we would be eliminating extraneous analysis times where we are not analyzing a particle event. In fact, work is now underway in conjunction with the ICP-MS manufacturer Perkin Elmer on this front. Although a number of physical ICP-MS parameters were optimized during the course of this doctoral work for spICP-MS such as, sample introduction rate and volume, nebulizer gas flow rate, and integration dwell time, there are still other factors that could be investigated to both increase NP signal and decrease matrix effects. Some of these factors are specific to the newest version of ICP-MS acquired by the Ranville lab, the Perkin Elmer NexION 300q. These include adjusting the torch position (especially in the z-axis), changing the voltages on the quadrupole ion deflector, and adjusting the RF frequency. In addition, sample ablation may also be optimized by adding auxiliary gas(es) to the plasma. While not specific to

NP analysis, the addition of gases such as N2, He, O2, methane, ethane and ammonia to the argon plasma has been research thoroughly in ICP-MS literature. The eﬀects are many and varied, but of note, methane and nitrogen are known to promote ionization of elements and the addition of oxygen helps eliminate excess carbon from depositing on the interface cone(s). These two facets are of particular interest to NP analysis to both increase the signal

115 of the NP pulse and reduce salting of the cones from complex matrices (e.g. wastewater, tissues, soils, etc), respectively. Finally, the sp analysis of NPs in solids, such as foods and tissues, hinges on the capability of digesting the matrix without dissolving the metallic NPs. This excludes many of the more traditions avenues, such as acidiﬁcation, for the digestion of these samples. Continued development (and standardization) of sample preparation, therefore, is necessary in this and other NP characterization paradigms.

116 REFERENCES CITED

[1] Adams, N. W. H., & Kramer, J. R. 1999. Silver speciation in wastewater eﬄuent, surface waters, and pore waters. Environmental toxicology and Chemistry, 18,2667–2673.

[2] Akthakul, A., Hochbaum, A. I., Stellacci, F., & Mayes, A. M. 2005. Size fractionation of metal nanoparticles by membrane ﬁltration. Advanced Materials, 17(5), 532–535.

[3] AshaRani, P. V., Mun, G. L. K., Hande, M. P., & Valiyaveettil, S. 2009. Cytotoxicity and Genotoxicity of Silver Nanoparticles in Human Cells. ACS Nano, 3(2), 279–290.

[4] Atiyeh, B. S., Costagliola, M., Hayek, S. N., & Dibo, S. A. 2007. Eﬀect of silver on burn wound infection control and healing: review of the literature. Burns: journal of the International Society for Burn Injuries, 33(2), 139.

[5] Auﬀan, M., Rose, J., Bottero, J. Y., Lowry, G. V., Jolivet, J. P., & Wiesner, M. R. 2009. Towards a deﬁnition of inorganic nanoparticles from an environmental, health and safety perspective. Nature Nanotechnology, 4(10), 634–641.

[6] Baalousha, M., & Lead, J. R. 2007. Characterization of natural aquatic colloids 5 nm by flow-field flow fractionation and atomic force microscopy. Environmental Science & Technology, 41(4), 1111–1117.

[7] Baalousha, M., Kammer, F. V. D., Motelica-Heino, M., & Le Coustumer, P. 2005. Natu- ral sample fractionation by FlFFF-MALLS-TEM: Sample stabilization, preparation, pre- concentration and fractionation. Journal of Chromatography A, 1093(1-2), 156–166.

[8] Baalousha, M., Manciulea, A., Cumberland, S., Kendall, K., & Lead, J. R. 2008. Ag- gregation and surface properties of iron oxide nanoparticles: inﬂuence of pH and natural organic matter. Environmental toxicology and Chemistry, 27(9), 1875–1882.

[9] Baalousha, M., Stolpe, B., & Lead, J. R. 2011. Flow ﬁeld-ﬂow fractionation for the analysis and characterization of natural colloids and manufactured nanoparticles in environmental systems: A critical review. Journal of Chromatography A, 1218(27), 4078–4103.

[10] Badawy, Amro M. El, Luxton, Todd P., Silva, Rendahandi G., Scheckel, Kirk G., Suidan, Makram T., & Tolaymat, Thabet M. 2010. Impact of Environmental Condi- tions (pH, Ionic Strength, and Electrolyte Type) on the Surface Charge and Aggregation of Silver Nanoparticles Suspensions. Environ Sci Technol, 44(4), 1260–1266.

117 [11] Beckett, R., & Hart, B. T. 1993. Use of ﬁeld-ﬂow fractionation techniques to characterize aquatic particles, colloids and macromolecules. Environmental particles, 2,165–205.

[12] Bednar, A., Poda, A. R., Mitrano, D. M., Kennedy, A. J., Gray, E. P., Ranville, J., Hayes, C. A., Crocker, F. H., & Steevens, J. A. 2012. Comparison of On-Line Detectors for Field Flow Fractionation Analysis of Nanomaterials. Talanta.

[13] Benn, T. M., & Westerhoﬀ, P. 2008. Nanoparticle silver released into water from commercially available sock fabrics. Environ Sci Technol, 42(11), 4133–4139.

[14] Benn, T. M., B, Cavanagh, K, Hristovski, JD, Posner, & Westerhoﬀ, P. 2010. The release of nanosilver from consumer products used in the home. Journal of Environmental Quality, 39(1875-1882).

[15] Blaser, S. A., Scheringer, M., MacLeod, M., & Hungerbhler, K. 2008. Estimation of cumulative aquatic exposure and risk due to silver: Contribution of nano-functionalized plastics and textiles. Science of the total environment, 390(2-3), 396–409.

[16] Bolea, E., Jimenez-Lamana, J., Laborda, F., & Castillo, J. R. 2011. Size characterization and quantification of silver nanoparticles by asymmetric flow field-flow fractionation coupled with inductively coupled plasma mass spectrometry. Analytical and Bioanalytical Chemistry,1–10.

[17] Bouby, M., Geckeis, H., & Geyer, F. W. 2008. Application of asymmetric flow field- flow fractionation (AsFlFFF) coupled to inductively coupled plasma mass spectrometry (ICPMS) to the quantitative characterization of natural colloids and synthetic nanoparticles. Analytical and Bioanalytical Chemistry, 392(7), 1447–1457.

[18] Boxall, A. B. A., Tiede, K., & Chaudhry, Q. 2007. Engineered nanomaterials in soils and water: how do they behave and could they pose a risk to human health? Nanomedicine, 2(6), 919–927.

[19] Burkhardt, M., Zuleeg, S., Kaegi, R., Sinnet, B., Eugster, J., Boller, M., & Siegrist, H. 2010. Verhalten von Nanosilber in Klaobranlagen und dessen Einfluss auf die Nitri- fikationsleistung in Belebtschlamm. Umweltwissenschaften und Schadstoff-Forschung, 22, 529–540.

[20] Charley, R. C., & Bull, A. T. 1979. Bioaccumulation of silver by a multispecies community of bacteria. Archives of Microbiology, 123(3), 239–244.

[21] Chiana, T., Assidjo, N. E., & Cardot, P. J. P. 2000. Sedimentation ﬁeld-ﬂow- fractionation: emergence of a new cell separation methodology. Talanta, 51(5), 835–847.

118 [22] Choi, O., & Hu, Z. 2008. Size dependent and reactive oxygen species related nanosilver toxicity to nitrifying bacteria. Environ Sci Technol, 42(12), 4583–4588.

[23] Choi, O., Deng, K. K., Kim, N. J., Ross Jr, L., Surampalli, R. Y., & Hu, Z. 2008a. The inhibitory eﬀects of silver nanoparticles, silver ions, and silver chloride colloids on microbial growth. Water Research, 42(12), 3066–3074.

[24] Choi, O., Clevernger, Thomas E., Deng, Baolin, Surampalli, Rao Y., Ross Jr, L., & Hu, Z. 2009. Role of sulﬁde and ligand strength in controlling nanosilver toxicity. Water Research, 43(7), 1879–1886.

[25] Choi, Okkyoung, Deng, Kathy Kanjun, Kim, Nam-Jung, Ross Jr, Louis, Surampalli, Rao Y., & Hu, Zhiqiang. 2008b. The inhibitory eﬀects of silver nanoparticles, silver ions, and silver chloride colloids on microbial growth. Water Research, 42(12), 3066–3074.

[26] Christian, P., Von der Kammer, F., Baalousha, M., & Hofmann, T. 2008. Nanoparticles: structure, properties, preparation and behaviour in environmental media. Ecotoxicology, 17(5), 326–343.

[27] Das, Pranab, Xenopoulos, Marguerite, Williams, Clayton, Hoque, Md Eshnaul, & Met- calfe, C. 2012. Eﬀects of silver nanoparticles on bacterial activity in natural waters. Environmental toxicology and Chemistry, 31(1), 122–130.

[28] Davies, J. Clarence. 2009. Oversight of Next Generation Nanotechnology. Woodrow Wilson International Center for Scholars, Washington, DC,.

[29] Degueldre, C., & Favarger, P. Y. 2003. Colloid analysis by single particle inductively coupled plasma-mass spectroscopy: a feasibility study. Colloids and Surfaces A: Physico- chemical and Engineering Aspects, 217(1-3), 137–142.

[30] Degueldre, C., & Favarger, P. Y. 2004. Thorium colloid analysis by single particle inductively coupled plasma-mass spectrometry. Talanta, 62(5), 1051–1054.

[31] Degueldre, C., Favarger, P. Y., & Wold, S. 2006a. Gold colloid analysis by inductively coupled plasma-mass spectrometry in a single particle mode. Analytica Chimica Acta, 555(2), 263–268.

[32] Degueldre, C., Favarger, P. Y., Rossa, R., & Wold, S. 2006b. Uranium colloid analysis by single particle inductively coupled plasma-mass spectrometry. Talanta, 68(3), 623–628.

[33] Delay, M., & Frimmel, F. H. 2012. Nanoparticles in aquatic systems. Anal Bioanal Chem, 402,583–592.

119 [34] Delay, M., Dolt, T., Woellhaf, A., Sembritzki, R., & Frimmel, F. H. 2011. Interactions and stability of silver nanoparticles in the aqueous phase: Inﬂuence of natural organic matter (NOM) and ionic strength. Journal of Chromatography A, 1218(27), 4206–4212.

[35] Diedrich, T., Dybowska, A., Schott, J., Valsami-Jones, E., & Oelkers, E. H. 2012. The Dissolution Rates of SiO2 Nanoparticles As a Function of Particle Size. Environmental Science & Technology, 46(9), 4909–4915.

[36] Diegoli, S., Manciulea, A. L., Begum, S., Jones, I. P., Lead, J. R., & Preece, J. A. 2008. Interaction between manufactured gold nanoparticles and naturally occurring organic macromolecules. Science of the total environment, 402(1), 51–61.

[37] Dubas, Stephan, & Pimpan, Vimolvan. 2008. Humic acid assisted synthesis of silver nanoparticles and its application to herbicide detection. Materials letters, 68,2661–2663.

[38] Dubascoux, S., Von Der Kammer, F., Le Hecho, I., Gautier, M. P., & Lespes, G. 2008. Optimisation of asymmetrical flow field flow fractionation for environmental nanoparticles separation. Journal of Chromatography A, 1206(2), 160–165.

[39] Dubascoux, S., Hecho, I. L., Hassellov, M., Kammer, F. V. D., Gautier, M. P., & Lespes, G. 2010. Field-ﬂow fractionation and inductively coupled plasma mass spectrometer coupling: History, development and applications. Journal of Analytical Atomic Spectrometry, 25(5), 613–623.

[40] Fabrega, J., Fawcett, S. R., Renshaw, J. C., & Lead, J. R. 2009. Silver nanoparticle impact on bacterial growth: Eﬀect of pH, concentration, and organic matter. Environmental Science & Technology, 43(19), 7285–7290.

[41] Fabrega, J., Luoma, S. N., Tyler, C. R., Galloway, T. S., & Lead, J. R. 2011. Silver nanoparticles: Behaviour and eﬀects in the aquatic environment. Environment Interna- tional, 37(2), 517–531.

[42] Farkas, J., Peter, H., Christian, P., Gallego Urrea, J. A., Hasselloev, M., Tuoriniemi, J., Gustafsson, S., Olsson, E., Hylland, K., & Thomas, K. V. 2011. Characterization of the eﬄuent from a nanosilver producing washing machine. Environment International.

[43] Gao, J., Youn, S., Hovsepyan, A., Llaneza, V. L., Wang, Y., Bitton, G., & Bonzongo, J. C. J. 2009. Dispersion and toxicity of selected manufactured nanomaterials in natural river water samples: Eﬀects of water chemical composition. Environmental Science & Technology, 43(9), 3322–3328.

[44] Geranio, L., Heuberger, M., & Nowack, B. 2009. The Behavior of Silver Nanotextiles during Washing. Environ Sci Technol, 43(21), 8113–8118.

120 [45] Giddings, J. C. 1993. Field-ﬂow fractionation: analysis of macromolecular, colloidal, and particulate materials. SCIENCE, 260(5113), 1456.

[46] Giddings, J. C., Yang, F. J., & Myers, M. N. 1976. Flow-ﬁeld-ﬂow fractionation: a versatile new separation method. SCIENCE, 193(4259), 1244.

[47] Gimbert, L. J., Andrew, K. N., Haygarth, P. M., & Worsfold, P. J. 2003. Environ- mental applications of flow field-flow fractionation (FIFFF). TrAC Trends in Analytical Chemistry, 22(9), 615–633.

[48] Gottschalk, F., & Nowack, B. 2011. The release of engineered nanomaterials to the environment. Journal of Environmental Monitoring, 13(5), 1145–1155.

[49] Gottschalk, F., Sonderer, T., Scholz, R. W., & Nowack, B. 2009a. Modeled environmental concentrations of engineered nanomaterials (TiO2, ZnO, Ag, CNT, fullerenes) for diﬀerent regions. Environ. Sci. Technol, 43(24), 9216–9222.

[50] Gottschalk, F., Scholz, R. W., & Nowack, B. 2009b. Probabilistic material ﬂow modeling for assessing the environmental exposure to compounds: Methodology and an application to engineered nano-TiO2 particles. Environmental Modelling & Software.

[51] Gottschalk, Fadri, Sonderer, Tobias, Scholz, Roland W., & Nowack, Bernd. 2009c. Modeled Environmental Concentrations of Engineered Nanomaterials (TiO2, ZnO, Ag, CNT, Fullerenes) for Diﬀerent Regions. Environmental Science & Technology, 43(24), 9216–9222.

[52] Grady, C. P. L., Daigger, G. T., & Lim, H. C. 1999. Biological wastewater treatment. CRC Press.

[53] Gupta, A., Maynes, M., & Silver, S. 1998. Eﬀects of halides on plasmid-mediated silver resistance in Escherichia coli. Applied and Environmental Microbiology, 64(12), 5042.

[54] Hammer, & Jr, Hammer. 2012. Water and Wastewater Technology, Seventh Edition. Columbus: Prentice Hall.

[55] Handy, R., Cornelis, G., Fernandes, T., Tsyusko, O., Decho, A., Sabo-Attwood, S., Metcalfe, C., Steevens, J., Klaine, S. J., Koelmans, A. A., & Horne, N. 2012a. Ecotoxicity test methods for engineered nanomaterials: Practical experiences and recommendations fromt he bench. Environmental toxicology and Chemistry, 31(1), 15–31.

[56] Handy, R., Cornelis, G., Fernandes, T., Tsyusko, O., Decho, A., Sabo-Attwood, S., Metcalfe, C., Steevens, J., Klaine, S. J., Koelmans, A. A., & Horne, N. 2012b. Ecotoxicity test methods for engineered nanomaterials: Practical experiences and recommendations fromt he bench. Environmental toxicology and Chemistry, 31(1), 15–31.

121 [57] Handy, R. D., Owen, R., & Valsami-Jones, E. 2008. The ecotoxicology of nanoparticles and nanomaterials: Current status, knowledge gaps, challenges, and future needs. Ecotoxicology, 17(5), 315–325.

[58] Hassellov, M., der Kammer, F., & Beckett, R. 2007. Characterisation of Aquatic Colloids and Macromolecules by Field Flow Fractionation. Environmental Colloids and Particles, 223–276.

[59] Hassellov, M., Readman, J. W., Ranville, J. F., & Tiede, K. 2008a. Nanoparticle analysis and characterization methodologies in environmental risk assessment of engineered nanoparticles. Ecotoxicology, 17(5), 344–361.

[60] Hassellov, M., Readman, J. W., Ranville, J. F., & Tiede, K. 2008b. Nanoparticle analysis and characterization methodologies in environmental risk assessment of engineered nanoparticles. Ecotoxicology, 17(5), 344–361.

[61] Hendren, C. O., Mesnard, X., Droage, J., & Wiesner, M. R. 2011. Estimating production data for ﬁve engineered nanomaterials as a basis for exposure assessment. Environmental Science & Technology, 45(7), 2562–2569.

[62] Ho, C. M., Yau, S. K. W., Lok, C. N., So, M. H., & Che, C. M. 2010. Oxidative dissolution of silver nanoparticles by biologically relevant oxidants: A kinetic and mechanistic study. ChemistryˆaAn Asian Journal, 5(2), 285–293.

[63] Howell, K. A., Achterberg, E. P., Tappin, A. D., & Worsfold, P. J. 2006. Colloidal metals in the tamar estuary and their inﬂuence on metal fractionation by membrane ﬁltration. Environmental Chemistry, 3(3), 199–207.

[64] Hu, Zhiqiang, Chandran, Kartik, Grasso, Domenico, & Smets, Barth F. 2002. Eﬀect of Nickel and Cadmium Speciation on Nitriﬁcation Inhibition. Environmental Science & Technology, 36(14), 3074–3078.

[65] Hu, Zhiqiang, Chandran, Kartik, Grasso, Domenico, & Smets, Barth F. 2003. Impact of Metal Sorption and Internalization on Nitriﬁcation Inhibition. Environmental Science & Technology, 37(4), 728–734.

[66] Hussain, S. M., Hess, K. L., Gearhart, J. M., Geiss, K. T., & Schlager, J. J. 2005. In vitro toxicity of nanoparticles in BRL 3A rat liver cells. Toxicology in vitro, 19(7), 975–983.

[67] Huynh, K. A., & Chen, K. L. 2011. Aggregation kinetics of citrate and polyvinylpyrrolidone coated silver nanoparticles in monovalent and divalent electrolyte solutions. Envi- ronmental Science & Technology, 45,5564–5571.

122 [68] Impellitteri, C. A. T., Scheckel, T. M., & Kirk, G. 2009. The Speciation of Sil- ver Nanoparticles in Antimicrobial Fabric Before and After Exposure to Hypochlo- rite/Detergent Solution. Journal of Environmental Quality, 38,1528–1530.

[69] Isaacson, C. W., & Bouchard, D. 2010. Asymmetric flow field flow fractionation of aqueous C60 nanoparticles with size determination by dynamic light scattering and quantification by liquid chromatography atmospheric pressure photo-ionization mass spectrometry. Journal of Chromatography A, 1217(9), 1506–1512.

[70] Kaegi, R., B, Sinnet, Zuleeg, S., Hagendorfer, H., H, Simmler, S, Brunner, H, Vonmont, M, Burkhardt, & M, Boller. 2008. Synthetic TiO2 nanoparticle emission from exterior facades into the aquatic environment. Environmental Pollution, 156,233–239.

[71] Kaegi, R., Sinnet, B., Zuleeg, S., Hagendorfer, H., Mueller, N. C., R, Vonbank, Boller, M., & M, Burkhardt. 2010. Release of silver nanoparticles from outdoor facaces. Envi- ronmental Pollution, 158,2900–2905.

[72] Kaegi, R., Voegelin, A., Sinnet, B., Zuleeg, S., Hagendorfer, H., Burkhardt, M., & Siegrist, H. 2011. Behavior of Metallic silver nanoparticles in a Pilot wastewater treatment plant. Environmental Science & Technology, 45,3902–3908.

[73] Kennedy, A. J., Hull, M. S., Bednar, A. J., Goss, J. D., Gunter, J. C., Bouldin, J. L., Vikesland, P. J., & Steevens, J. A. 2010. Fractionating nanosilver: importance for determining toxicity to aquatic test organisms. Environmental Science & Technology, 44(24), 9571–9577.

[74] Kim, Bojeong, Park, Chee-Sung, Murayama, Mitsuhiro, & Hochella Jr., Michael F. 2010. Discovery and Characterization of Silver Sulﬁde Nanoparticles in Final Sweage Sludge Products. Environ Sci Technol, 44(19), 7509–7514.

[75] Kim, Jun Sung, Kuk, Eunye, Yu, Kyeong Nam, Kim, Jong-Ho, Park, Sung Jin, Lee, Hu Jang, Kim, So Hyun, Park, Young Kyung, Park, Yong Ho, Hwang, Cheol-Yong, Kim, Yong-Kwon, Lee, Yoon-Sik, Jeong, Dae Hong, & Cho, Myung-Haing. 2007. Antimicrobial eﬀects of silver nanoparticles. Nanomedicine: Nanotechnology, Biology and Medicine, 3(1), 95–101.

[76] King, S. M., & Jarvie, H. P. 2012. Exploring How Organic Matter Controls Structural Transformations in Natural Aquatic Nanocolloidal Dispersions. Environmental Science & Technology, 46(13), 6959–6967.

[77] Kiser, M., Ryu, H., Jang, H., Hristovski, K., & Westerhoﬀ, P. 2010. Biosorption of nanoparticles to heterotrophic wastewater biomass. Water Research, 44,4105–4114.

123 [78] Klaine, S. J., Koelmans, A. A., Horne, N., Carley, S., Handy, R., Kapustka, L., Nowack, B., & Von der Kammer, F. 2012. Paradigms to Assess the Environmental Impact of Manufactured Nanomaterials. Environmental toxicology and Chemistry, 31(1), 3–14.

[79] Krystek, P., Ulrich, A., Garcia, C. C., Manohar, S., & Ritsema, R. 2011. Application of plasma spectrometry for the analysis of engineered nanoparticles in suspensions and products. Journal of Analytical Atomic Spectrometry, 26,1701–1721.

[80] Laborda, F., Jimenez-Lamana, J., Bolea, E., & Castillo, J. R. 2011. Selective identiﬁ- cation, characterization and determination of dissolved silver (I) and silver nanoparticles based on single particle detection by inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom., 26(7), 1362–1371.

[81] Lead, J. R., & Smith, E. L. 2009. Environmental and human health impacts of nanotechnology. Wiley Online Library.

[82] Lead, J. R., & Wilkinson, K. J. 2006. Aquatic colloids and nanoparticles: current knowledge and future trends. Environmental Chemistry, 3(3), 159–171.

[83] Lee, H., Williams, S. K. R., & Giddings, J. C. 1998. Particle size analysis of dilute environmental colloids by flow field-flow fractionation using an opposed flow sample concentration technique. Analytical Chemistry, 70(13), 2495–2503.

[84] Lee, Yong-Ju, Kim, Jiwon, Oh, Jeehyun, Bae, Sujin, Lee, Sungktu, Hong, In Seok, & Kim, Sang-Ho. 2012. Ion-release kinetics and ecotoxicity eﬀects of silver nanoparticles. Environmental toxicology and Chemistry, 31(1), 155–159.

[85] Leppard, G. G., Mavrocordatos, D., & Perret, D. 2004. Electron-optical characterization of nano-and micro-particles in raw and treated waters: an overview. Water science and technology: a journal of the International Association on Water Pollution Research, 50(12), 1.

[86] Lesher, E. K., Ranville, J. F., & Honeyman, B. D. 2009. Analysis of pH Dependent Uranium (VI) Sorption to Nanoparticulate Hematite by Flow Field-Flow Fractionation- Inductively Coupled Plasma Mass Spectrometry. Environmental Science & Technology, 43(14), 5403–5409.

[87] Levard, C., Reinsch, B. C., Michel, F. M., Oumahi, C., Lowry, G. V., & Brown Jr, G. E. 2011. Sulﬁdation processes of PVP-coated silver nanoparticles in aqueous solution: impact on dissolution rate. Environmental Science & Technology, 45(12), 5260–5266.

[88] Levard, C., Hotze, E. M., Lowry, G. V., & Brown Jr, G. E. 2012. Environmental Transformations of Silver Nanoparticles: Impact on Stability and Toxicity. Environmental science & technology, 46(13), 6900–6914.

124 [89] Li, X., Lenhart, J., & Walker, H. W. 2012a. Aggregation Kinetics and Dissolution of Coated Silver Nanoparticles. Langmuir, 28(2), 1095–1104.

[90] Li, X., Lenhart, J., & Walker, H. W. 2012b. Aggregation Kinetics and Dissolution of Coated Silver Nanoparticles. Langmuir, 28(2), 1095–1104.

[91] Liu, J., & Hurt, R. H. 2010. Ion release kinetics and particle persistence in aqueous nano-silver colloids. Environ. Sci. Technol, 44(6), 2169–2175.

[92] Liu, J., Andya, J. D., & Shire, S. J. 2006. A critical review of analytical ultracentrifugation and ﬁeld ﬂow fractionation methods for measuring protein aggregation. The AAPS Journal, 8(3), 580–589.

[93] Liu, J., Pennell, K. G., & Hurt, R. H. 2011. Kinetics and mechanisms of nanosilver oxysulﬁdation. Environmental Science & Technology, 45(17), 7345–7353.

[94] Lok, C. N., Ho, C. M., Chen, R., He, Q. Y., Yu, W. Y., Sun, H., Tam, P. K. H., Chiu, J. F., & Che, C. M. 2007. Silver nanoparticles: partial oxidation and antibacterial activities. Journal of biological inorganic chemistry, 12(4), 527–534.

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

[96] Luoma, S. N. 2008. Silver nanotechnologies and the environment: old problemds or new challenges? edn. Woodrow Wilson International Center for Scholars, Project on Emerging Nanotechnologies, The PEW Charitable Trusts, Washington, DC.

[97] Lyon, D. Y., Adams, L. K., Falkner, J. C., & Alvarez, P. J. J. 2006. Antibacterial activity of fullerene water suspensions: Eﬀects of preparation method and particle size. Environmental Science & Technology, 40(14), 4360–4366.

[98] Ma, R., Levard, C., Marinakos, S. M., Cheng, Y., Liu, J., Michel, F. M., Brown Jr, G. E., & Lowry, G. V. 2011. Size-controlled dissolution of organic-coated silver nanoparticles. Environmental Science & Technology, 46(2), 752–759.

[99] MacLaughlin, Julianne, & Bonzongo, Jean-Claude. 2012. Eﬀects of natural water chemistry on nanosilver behavior and toxicity to ceriodaphnia dubia and pseudokirchnerilla subcapitata. Environmental toxicology and Chemistry, 31(1), 168–175.

[100] Madigan, M. T., EDS, JMM, BERLANGA, M., & ED, DS. 2005. Brock Biology of Microorganisms, 11th edn. INT. MICROBIOL, 8(2).

125 [101] Matsumura, Y., Yoshikata, K., Kunisaki, S., & Tsuchido, T. 2003. Mode of bactericidal action of silver zeolite and its comparison with that of silver nitrate. Applied and Environmental Microbiology, 69(7), 4278.

[102] Maynard, A., & Michelson, E. 2006. The Nanotechnology Consumer Product In- ventory. Woodrow Wilson International Center for Scholars, Washington, DC, accessed March, 23.

[103] Messaud, F. A., Sanderson, R. D., Runyon, J., Otte, T., Pasch, H., & Williams, S. 2009. An overview on ﬁeld-ﬂow fractionation techniques and their applications in the separation and characterization of polymers. Progress in Polymer Science, 34(4), 351–368.

[104] Metcalf, & Eddy. 1991. Wastewater Engineering. Treatment, Disposal, and Reuse. New York: McGraw-Hill.

[105] Metz, K. M., Mangham, A. N., Bierman, M. J., Jin, S., Hamers, R. J., & Pedersen, J. A. 2009. Engineered nanomaterial transformation under oxidative environmental conditions: development of an in vitro biomimetic assay. Environmental Science & Technology, 43(5), 1598–1604.

[106] Misra, S. K., Dybowska, A., Berhanu, D., Luoma, S. N., & Valsami-Jones, E. 2012. The complexity of nanoparticle dissolution and its importance in nanotoxicological studies. Science of the total environment, 438,225–232.

[107] Mitrano, D. M., Lesher, E. K., Bednar, A. J., Monserud, J., Higgins, Christopher P., & Ranville, J. F. 2012a. Detection of nano-Ag using single particle inductively coupled plasma mass spectrometry. Environ Toxicol Chem, 31,115–121.

[108] Mitrano, D. M., Lesher, E. K., Bednar, A. J., Monserud, J., Higgins, Christopher P., & Ranville, J. F. 2012b. Detection of nano-Ag using single particle inductively coupled plasma mass spectrometry. Environ Toxicol Chem, 31,115–121.

[109] Mitrano, D. M., Barber, A., Bednar, A., Westerhoff, P., Higgins, C., & Ranville, J. 2012c. Silver nanoparticle characterization using single particle ICP-MS (SP-ICP-MS) and asymmetrical flow field flow fractionation ICP-MS (AF4-ICP-MS). J. Anal. At. Spectrom., 27,1131–1142.

[110] Moon, M. H., Kang, D., Jung, J., & Kim, J. 2004. Separation of carbon nanotubes by frit inlet asymmetrical flow field flow fractionation. Journal of separation science, 27(9), 710–717.

[111] Morones, J. R., Elechiguerra, J. L., Camacho, A., Holt, K., Kouri, J. B., Ramirez, J. T., & Yacaman, M. J. 2005. The bactericidal eﬀect of silver nanoparticles. Nanotechnology, 16(10), 2346.

126 [112] Mueller, N. C., & Nowack, B. 2008. Exposure modeling of engineered nanoparticles in the environment. Environ. Sci. Technol, 42(12), 4447–4453.

[113] National Nanotechnology Initiative Workshop, 2009. 2011. Nanomaterials and the Environment & Instrumentation, Metrology, and Analytical Methods.

[114] Navarro, E., Baun, A., Behra, R., Hartmann, N. B., Filser, J., Miao, A. J., Quigg, A., Santschi, P. H., & Sigg, L. 2008. Environmental behavior and ecotoxicity of engineered nanoparticles to algae, plants, and fungi. Ecotoxicology, 17(5), 372–386.

[115] Neal, A. L. 2008. What can be inferred from bacterium/nanoparticle interactions about the potential consequences of environmental exposure to nanoparticles? Ecotoxicology, 17(5), 362–371.

[116] Nel, A., Xia, T., Madler, L., & Li, N. 2006. Toxic Potential of Materials at the Nanolevel. SCIENCE, 311(5761), 622–627.

[117] Noble, C. A., & Prather, K. A. 2000. Real-time single particle mass spectrometry: A historical review of a quarter century of the chemical analysis of aerosols. Mass Spectrom- etry Reviews, 19(4), 248–274.

[118] Nomizu, T., Kaneco, S., TANAKA, T., YAMAMOTO, T., & KAWAGUCHI, H. 1993. Determination of femto-gram amounts of zinc and lead in individual airborne particles by inductively coupled plasma mass spectrometry with direct air-sample introduction. Analytical Sciences, 9(6), 843–846.

[119] Nowack, B. 2009. Is anything out there? What life cycle perspectives of nano-products can tell us about nanoparticles in the environment. Nano Today, 4(1), 11–12.

[120] Nowack, B., & Bucheli, T. D. 2007. Occurrence, behavior and eﬀects of nanoparticles in the environment. Environmental Pollution, 150(1), 5–22.

[121] Nowack, B., Ranville, J. F., Diamond, S., Gallego-Urrea, J. A., Metcalfe, C., Rose, J., Horne, N., Koelmans, A. A., & Klaine, S. J. 2012. Potential scenarios for nanomaterial release and subsequent alteration in the environment. Environmental toxicology and Chemistry, 31(1), 50–59.

[122] Oberdaurster, G., Oberdaurster, E., & Oberdaurster, J. 2005. Nanotoxicology: an emerging discipline evolving from studies of ultraﬁne particles. Environmental health perspectives, 113(7), 823.

[123] Olesik, J. W., & Gray, P. J. 2012. Considerations for measurement of individual nanoparticles or microparticles by ICP-MS: determination of the number of particles and the analyte mass in each particle. Journal of Analytical Atomic Spectrometry, 27(7), 1143–1155.

127 [124] Pace, H. E., Rogers, N. J., Jarolimeck, C., Coleman, V. A., Higgins, C. P., & Ranville, J. F. Validation of single particle inductively coupled plasma-mass spectrometry for determining inorganic nanoparticle size.

[125] Pace, H. E., Lesher, E. K., & Ranville, J. F. 2010. Inﬂuence of stability on the acute toxicity of CdSe/ZnS nanocrystals to Daphnia magna. Environmental toxicology and Chemistry, 29(6), 1338–1344.

[126] Pace, H. E., Rogers, N. J., Jarolimek, C., Coleman, V. A., Higgins, C. P., & Ranville, J. F. 2011a. Determining transport eﬃciency for the purpose of counting and sizing nanoparticles via single particle inductively coupled plasma mass spectrometry. Analytical Chemistry, 83(24), 9361–9369.

[127] Pace, H. E., Rogers, Nicola J., Jarolimeck, Chad, Coleman, Victoria A., Higgins, Christopher P., & Ranville, J. F. 2011b. Determining transport eﬃcieny for the purpose of counting and sizing nanoparticles via single particle inductively coupled plasma-mass spectrometry. Analytical Chemistry, 83,9361–9369.

[128] Pace, H. E., Rogers, N. J., Jarolimek, C., Coleman, V. A., Gray, E. P., Higgins, C. P., & Ranville, J. F. 2012. Single particle inductively coupled plasma-mass spectrometry: A performance evaluation and method comparison in the determination of nanoparticle size. Environmental Science & Technology, 42(22), 12272–12280.

[129] Pal, S., Tak, Y. K., & Song, J. M. 2007. Does the antibacterial activity of silver nanoparticles depend on the shape of the nanoparticle? A study of the gram-negative bacterium Escherichia coli. Applied and Environmental Microbiology, 73(6), 1712.

[130] Piccinno, F., Gottschalk, F., Seeger, S., & Nowack, B. 2012. Industrial production quantities and uses of ten engineered nanomaterials in Europe and the world. Journal of Nanoparticle Research, 14(9), 1–11.

[131] Poda, A. R., Bednar, A. J., Kennedy, A. J., Harmon, A., Hull, M., Mitrano, D. M., Ranville, J. F., & Steevens, J. 2011. Characterization of silver nanoparticles using flow-field flow fractionation interfaced to inductively coupled plasma mass spectrometry. Journal of Chromatography A, 1218(27), 4219–4225.

[132] Pommerville, J. C., & Alcamo, I. E. 2004. Alcamo’s fundamentals of microbiology. Jones & Bartlett Publishers.

[133] Powers, K. W., Palazuelos, M., Moudgil, B. M., & Roberts, S. M. 2007. Characteri- zation of the size, shape, and state of dispersion of nanoparticles for toxicological studies. Nanotoxicology, 1(1), 42–51.

128 [134] Praetorius, A., Scheringer, M., & Hungerbuhler, K. 2012. Development of Environ- mental Fate Models for Engineered Nanoparticles A Case Study of TiO2 Nanoparticles in the Rhine River. Environmental Science & Technology, 46(12), 6705–6713.

[135] Quik, J. T. K., Vonk, J. A., Hansen, S. F., Baun, A., & Van De Meent, D. 2011. How to assess exposure of aquatic organisms to manufactured nanoparticles? Environment International, 37(6), 1068–1077.

[136] Ranville, J. F., Chittleborough, D. J., Shanks, F., Morrison, R. J. S., Harris, T., Doss, F., & Beckett, R. 1999. Development of sedimentation ﬁeld-ﬂow fractionation- inductively coupled plasma mass-spectrometry for the characterization of environmental colloids. Analytica Chimica Acta, 381(2-3), 315–329.

[137] Ratanathanawongs, S. K., & Giddings, J. C. 1993. Particle-Size Analysis Using Flow Field-Flow Fractionation. PROCESS-RENNES-, 521,13–13.

[138] Ratanathanawongs, S. K., & Lee, D. 2006. Field-ﬂow fractionation of proteins, polysac- charides, synthetic polymers, and supramolecular assemblies. J. Sep. Sci, 29,1720–1732.

[139] Ratte, H. T. 1999. Bioaccumulation and toxicity of silver compounds: a review. Env- iron Toxicol Chem, 18(1), 89–108.

[140] Reed, R., Higgins, C., Westerhoﬀ, P., Tadjiki, S., & Ranville, J. 2012. Overcoming challenges in analysis of polydisperse metal-containing nanoparticles by single particle inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom.

[141] Reschiglian, P., Zattoni, A., Roda, B., Michelini, E., & Roda, A. 2005. Field-ﬂow fractionation and biotechnology. Trends in Biotechnology, 23(9), 475–483.

[142] Roda, B., Zattoni, A., Reschiglian, P., Moon, M. H., Mirasoli, M., Michelini, E., & Roda, A. 2009. Field-ﬂow fractionation in bioanalysis: A review of recent trends. Analytica Chimica Acta, 635(2), 132–143.

[143] Romer, I., White, T. A., Baalousha, M., Chipman, K., Viant, M. R., & Lead, J. R. 2011. Aggregation and dispersion of silver nanoparticles in exposure media for aquatic toxicity tests. Journal of Chromatography A, 1218(27), 4226–4233.

[144] Sal’nikov, D. S., Pogorelova, A. S., Makarov, S. V., & Vashurina, I. Yu. 2009. Silver ion reduction with peat fulvic acids. Russian Journal of Applied Chemistry, 82(4), 545–548.

[145] Sanudo-Wilhelmy, S. A., & Flegal, A. R. 1992. Anthropogenic silver in the Southern California Bight: a new tracer of sewage in coastal waters. Environ Sci Technol, 26(11), 2147–2151.

129 [146] Schimpf, M. E., Caldwell, K., & Giddings, J. C. 2000. Field-Flow Fractionation Hand- book. New York: Wiley-Interscience.

[147] Schmidt, C. W. 2009. Nanotechnology-related environment, health, and safety research: examining the national strategy. Environmental health perspectives, 117(4), A158.

[148] Schmidt, K. F. 2007. NanoFrontiers. Visions for the future of nanotechnology, Woodrow Wilson International Center for Scholars Project on Emerging Nanotechnologies, Wash- ington DC.

[149] Shafer, M. M., Overdier, J. T., & Armstong, D. E. 1998. Removal, partitioning, and fate of silver and other metals in wastewater treatment plants and eﬄuent receiving streams. Environmental toxicology and Chemistry, 17(4), 630–641.

[150] Siy, J. T., & Bartl, M. H. 2010. Insights into Reversible Dissolution of Colloidal CdSe Nanocrystal Quantum Dots. Chemistry of Materials, 22(21), 5973–5982.

[151] Sondi, Ivan, & Salopek-Sondi, Branka. 2004. Silver nanoparticles as antimicrobial agent: a case study on E. coli as a model for Gram-negative bacteria. Journal of Colloid and Interface Science, 275(1), 177–182.

[152] Song, Y., Jimenez, V., McKinney, C., Donkers, R., & Murray, R. W. 2003. Estimation of size for 1-2 nm nanoparticles using an HPLC electrochemical detector of double layer charging. Analytical Chemistry, 75(19), 5088–5096.

[153] Songsilawat, K., Shiowatana, J., & Siripinyanond, A. 2011. Flow field-flow fractionation with off-line electrothermal atomic absorption spectrometry for size characterization of silver nanoparticles. Journal of Chromatography A, 1218(27), 4213–4218.

[154] Tadjiki, S., Assemi, S., Deering, C. E., Veranth, J. M., & Miller, J. D. 2009. Detection, separation, and quantification of unlabeled silica nanoparticles in biological media using sedimentation field-flow fractionation. Journal of Nanoparticle Research, 11(4), 981–988.

[155] Thomas, C. R., George, S., Horst, A. M., Ji, Z., Miller, R. J., Peralta-Videa, J. R., Xia, T., Pokhrel, S., Ma dler, L., & Gardea-Torresdey, J. L. 2011. Nanomaterials in the Environment: From Materials to High-Throughput Screening to Organisms. ACS Nano, 5(1), 13–20.

[156] Tiede, K., Boxall, A., Tear, S., Lewis, J., David, H., & Hassellov, M. 2008. Detection and characterization of engineered nanoparticles in food and the environmental review. Food Additives and Contaminants, 25(07), 795–821.

130 [157] Tiede, K., Boxall, A., Wang, X., Gore, D., Tiede, D., Baxter, M., David, H., Tear, S., & Lewis, J. 2010. Application of hydrodynamic chromatography-ICP-MS to investigate the fate of silver nanoparticles in activated sludge. J. Anal. At. Spectrom., 25,1149–1154.

[158] Ulrich, A., Losert, S., Bendixen, N., Al-Kattan, A., Hagendorfer, H., Nowack, B., Adl- hart, C., Ebert, J., Lattuada, M., & Hungerbhler, K. 2012. Critical aspects of sample handling for direct nanoparticle analysis and analytical challenges using asymmetric ﬁeld ﬂow fractionation in a multi-detector approach. Journal of Analytical Atomic Spectrome- try.

[159] Unrine, J. M., Colman, B. P., Bone, A. J., Gondikas, A. P., & Matson, C. W. 2012. Biotic and Abiotic Interactions in Aquatic Microcosms Determine Fate and Toxicity of Ag Nanoparticles. Part 1. Aggregation and Dissolution. Environmental Science and Technology-Columbus, 46(13), 6915.

[160] USEPA. Targeted National Sweage sludge Survey Sampling and Analysis TEchnical Report. http://water.epa.gov/scitech/wastertech/biosolids/upload/2009 01 15 biosolids tnss- tech.pdf.

[161] von der Kammer, F., Ferguson, P. L., Holden, P. A., Masion, A., Rogers, K. R., Klaine, S. J., Koelmans, A. A., Horne, N., & Unrine, J. M. 2012a. Analysis of engineered nanomterials in complex matrices (environment and biota): General considerations and conceptual case studies. Environmental toxicology and Chemistry, 31(1), 32–49.

[162] von der Kammer, F., Ferguson, P. L., Holden, P. A., Masion, A., Rogers, K. R., Klaine, S. J., Koelmans, A. A., Horne, N., & Unrine, J. M. 2012b. Analysis of engineered nanomterials in complex matrices (environment and biota): General considerations and conceptual case studies. Environmental toxicology and Chemistry, 31(1), 32–49.

[163] Von der Kammer, Frank, Legros, Samuel, Larsen, Erik H., Lauschner, Katrin, & Hof- mann, Thilo. 2011. Separation and characterization of nanoparticles in complex food and environmental samples by ﬁeld-ﬂow fractionation. TrAC Trends in Analytical Chemistry, 30(3), 425–436.

[164] Vorbau, M., Hillemann, L., & Stintz, M. 2009. Method for the characterization of the abrasion induced nanoparticle relase into the air from surface coatings. Journal of Aerosol Science, 40,209–217.

[165] Wan, G. S., Liao, C. H., & Wu, F. J. 2001. Photodegradation of humic acid in the presence of hydrogen peroxide. Chemosphere, 42(4), 379–387.

[166] Warheit, D. B. 2008. How meaningful are the results of nanotoxicity studies in the absence of adequate material characterization? Toxicological Sciences, 101(2), 183.

131 [167] Wen, L. S., Santschi, P. H., Gill, G. A., & Tang, D. 2002. Silver concentrations in Colorado, USA, watersheds using improved methodology. Environ Toxicol Chem, 21(10), 2040–2051.

[168] Wenseleers, W., Stellacci, F., Meyer-Friedrichsen, T., Mangel, T., Bauer, C. A., Pond, S. J. K., Marder, S. R., & Perry, J. W. 2002. Five orders-of-magnitude enhancement of two-photon absorption for dyes on silver nanoparticle fractal clusters. J. Phys. Chem. B, 106(27), 6853–6863.

[169] Westerhoﬀ, P., & Nowack, B. 2012. Searching for Global Descriptors of Engineered Nanomaterial Fate and Transport in the Environment. Accounts of Chemical Research.

[170] Wiesner, M. R., Lowry, G. V., Jones, K. L., Hochella Jr, M. F., Di Giulio, R. T., Casman, E., & Bernhardt, E. S. 2009. Decreasing Uncertainties in Assessing Environmen- tal Exposure, Risk, and Ecological Implications of Nanomaterials. Environ. Sci. Technol, 43(17), 6458–6462.

[171] Williams, S. K. R., Runyon, J. R., & Ashames, A. A. 2011. Field-Flow Fractionation: Addressing the Nano Challenge. Analytical Chemistry, 83(3), 634–642.

[172] Yang, Y., Chen, Q., Wall, J., & Hu, Z. 2012. Potential nanosilver impact on anaerobic digestion at moderate silver concentrations. Water Research, 46,1176–1184.

[173] Zhang, W., Yao, Y., Li, K., Huang, Y., & Chen, Y. 2011. Inﬂuence of dissolved oxygen on aggregation kinetics of citrate-coated silver nanoparticles. Environmental Pollution, 159,3575–3762.

132 APPENDIX A - AN INTRODUCTION TO FLOW FIELD FLOW FRACTIONATION AND COUPLING TO ICP-MS

D. M. Mitrano, James F. Ranville, and Kenneth Neubauer. An Introduction to Flow Filed Flow Fractioantion and Coupling to ICP-MS. Perkin Elmer Application Note, Copyright 2011. PerkinElmer, Inc. Shelton, CT USA

A.1 Introduction

Inductively coupled plasma-mass spectrometry (ICP-MS) is the method of choice for analysis of most elements across the periodic chart. Its multi-element capability, low detection limit (ppt), and wide dynamic range (109 orders of magnitude) also make it ideal for the measurement of inorganic engineered nanoparticles (ENPs). While ICP-MS can be used directly to obtain concentrations of nanoparticulate-associated elements, more information on characteristics of ENPs can be obtained by first separating the particles by size prior to ICP-MS analysis. The most versatile size-separation technique is field flow fractionation (FFF). By introducing size- fractionated material into the ICP-MS, the size and elemental composition of complex, polydisperse and chemically heterogeneous ENPs can be determined. Furthermore, the similar flow conditions required by both ICP-MS and FFF make interfacing relatively simple.

A.2 Field Flow Fractionation

Field ﬂow fractionation (FFF) consists of a suite of high-resolution elution techniques which can size separate nanoparticles in the 1 - 100 nm range and colloids up to 1 micron. By use of either FFF theory or calibration with size standards, the technique can be utilized to determine particle size. The separation process is similar to chromatography except that the separation is based on physical forces as opposed to chemical interactions. Depending on the type of analysis that is being performed, a diﬀerent member of the FFF family

133 can be chosen to achieve optimal separation results. The three FFF techniques that are commercially available, and thus most commonly used, include thermal, sedimentation, and ﬂow. Flow FFF can be used in either the symmetrical or asymmetrical modes and is the most widely used subset of techniques for environmental applications. It is highly versatile for a range of both natural and manufactured NPs [9, 38, 163]. The combination of FFF and ICP-MS provides size, detection, and compositional analysis at the parts per billion (ppb) levels, which is critical to environmental and toxicological investigations of nanomaterials.

A.3 FFF Operation and Separation Theory

Fractionation in flow FFF takes place in a thin channel, which is constructed using a polyester spacer (200 - 500 m) enclosed by one (asymmetrical FFF, Figure A.1) or two (symmetrical FFF) porous blocks (frits). The laminar channel (tip) flow, which carries the sample through the system and to the ICP-MS, creates a parabolic flow-velocity profile across the channel. A perpendicularly applied fluid cross flow pushes particles against the lower (accumulation) wall, which consists of a semi-permeable membrane on top of the lower ceramic block.

Figure A.1: Schematic of the AF4 channel

134 In Figure A.2, we show a typical separation of two differently sized particles. After the sample is injected (A), a focusing step occurs in which the sample is concentrated near the entrance of the channel (B) prior to analysis. After a set focusing time, the separation of the particles occurs during the elution phase of analysis. During the elution, the cross flow continuously pushes particles against the membrane, while Brownian diffusion causes particles to move away from the wall and into a higher velocity flow (C). The balance between these two forces causes smaller particles to interact with the faster part of the parabolic flow, resulting in their being eluted more quickly from the channel (D). . In theory, the balance of these forces can only be changed by varying the flow conditions or spacer thickness. Due to some interactions between the analytes and the membrane, this balance can also be affected by the membrane composition and/or the carrier solution composition (composition, pH or ionic strength). Manipulation of all physical and chemical variables can optimize fractionation with respect to resolution and sample recovery. The lower size limit of analyte particles is established by the cross flow applied and the molecular weight or size cutoffof the membrane, which also governs the highest cross flow rate that can be used before the system over-pressurizes. Analyzing standards of a known size with little particle/ membrane interactions, such as polystyrene beads, can determine the particle size versus retention time (Figure A.3). Asimplelinearcalibrationcanthenbemadebyplottingretentiontimeofparticleversus particle size (excluding the void peak), which can then be used for analyte particles that would have similar behavior within the FFF channel. The particle interactions within the channel directly affect the elution profile. The driving force (F) exerted on a single particle includes the friction coefficient (f), viscosity (η), and the hydrodynamic diameter (d), and is represented by the following two equations:

135 "#$#%&'%()*%)+,#)+-$#)*.)/%/0-1'1)+,/+)'1)2#'%()$#3.*34#&5)/) &'..#3#%+)4#42#3)*.)+,#)666)./4'0-)7/%)2#)7,*1#%)+*)/7,'#8#) *$+'4/0)1#$/3/+'*%)3#190+1:);,#)+,3##)666)+#7,%'<9#1)+,/+) /3#)7*44#37'/00-)/8/'0/20#5)/%&)+,91)4*1+)7*44*%0-)91#&5) '%709&#)+,#34/05)1#&'4#%+/+'*%5)/%&).0*=:)60*=)666)7/%)2#) 91#&)'%)#'+,#3)+,#)1-44#+3'7/0)*3)/1-44#+3'7/0)4*)/%&)'1) +,#)4*1+)='�-)91#&)1921#+)*.)+#7,%'<9#1).*3)#%8'3*%4#%+/0) /$$0'7/+'*%1:)>+)'1),'(,0-)8#31/+'0#).*3)/)3/%(#)*.)2*+,)%/+93/0) /%&)4/%9./7+93#&)?@1:ABC);,#)7*42'%/+'*%)*.)666)/%&)>D@BEF) $3*8')1'G#5)&#+#7+'*%5)/%&)7*4$*1'+'*%/0)/%/0-1'1)/+)+,#) $/3+1)$#3)2'00'*%)H$$2I)0#8#015)=,'7,)'1)73'+'7/0)+*)#%8'3*%4#%+/0) /%&)+*J'7*0*('7/0)'%8#1+'(/+'*%1)*.)%/%*4/+#3'/01:

!!!"#$%&'()*+"'+,"-%$'&'()*+"./%*&0 63/7+'*%/+'*%)'%).0*=)666)+/K#1)$0/7#)'%)/)+,'%)7,/%%#05) =,'7,)'1)7*%1+397+#&)91'%()/)$*0-#1+#3)1$/7#3)H!LLBMLL)N4I) #%70*1#&)2-)*%#)H/1-44#+3'7/0)6665)6'(93#)AI)*3)+=*)) H1-44#+3'7/0)666I)$*3*91)20*7K1)H.3'+1I:);,#)0/4'%/3)7,/%%#0)) H+'$I).0*=5)=,'7,)7/33'#1)+,#)1/4$0#)+,3*9(,)+,#)1-1+#4) /%&)+*)+,#)>D@BEF5)73#/+#1)/)$/3/2*0'7).0*=B8#0*7'+-)$3*.'0#) /73*11)+,#)7,/%%#0:)O)$#3$#%&'790/30-)/$$0'#&).09'&)73*11) Figure A.2: Cross section of an AF4 channel .0*=)$91,#1)$/3+'70#1)/(/'%1+)+,#)0*=#3)H/779490/+'*%I)=/005) Figure 2. Cross section of an asymmetrical field flow fractionation (AF4) The schematicchannel: above separation shows: and A) elution injection process. of particles4 The schematic onto the above AF4 shows: channel; A) B) focusing =,'7,)7*%1'1+1)*.)/)1#4'B$#34#/20#)4#423/%#)*%)+*$)*.)+,#)(concentration)injection of of analyte particles at onto the topthe AF4 of the channel; channel B)prior focusing to analysis;(concentration) C) elution of beginning 0*=#3)7#3/4'7)20*7K:) with theanalyte focusing at ﬂowthe top ending of the and channel analyte prior group to analysis; moving C) down elution the beginning channel; with and D) small andthe focusing large particles flow ending separating and analyte over group time asmoving they down move the down channel; the channel. and D) small and large particles separating over time as they move down the channel.

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igure 1. Schematic of the asymmetrical flow FFF (AF4) channel. /01*)(*8#3%1)+,#),'(,#1+)73*11).0*=)3/+#)+,/+)7/%)2#)91#&) 2#.*3#)+,#)1-1+#4)*8#3B$3#1193'G#1:

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

! where U is essentially the cross ﬂow velocity, k is Boltzmann’s constant, and T is absolute temperature. The force exerted on the particle (F) can be directly related to the retention time (tr) through the following equation:

t w F w r = = | | t0 6 l 6 kT

where t0 is the void time, w is the channel thickness, and l is the layer thickness (sample cloud thickness). Assuming all particles in the sample are of similar shape, understanding the elution proﬁle is simple, as the retention time (tr)canbeattributedtothediameterof the particle (d), and the fractogram (graphical output) reﬂects the size distribution of the sample.

Figure A.3: Example of AF4 calibration Overlay of triplicate FFF-UV fractograms of polystyrene bead calibration standards. FFF separation conditions were 1.0 mL/min channel ﬂow and 0.75 mL/min cross ﬂow. UV absorbance detection is at 254 nm wavelength. Insert: linear regression calibration function using 20, 50, and 100 nm polystyrene bead standards. Error bars represent standard deviation of the triplicate retention times obtained from UV absorbance data at maximum absorbance.[131]

137 A.4 Coupling FFF to ICP-MS

By combining the high-resolution separation of FFF with the sensitivity and specificity of ICP-MS (FFF-ICP-MS), a powerful approach for ENP characterization is created. In addition to ICP-MS, UV/Vis spectrophotometry is often used as an additional online instrument in order to either verify ICP-MS results or for polystyrene standard calibration, as noted earlier. A schematic representation of the FFF-ICP-MS is given in Figure A.4. Here, the carrier (surfactant) solution that is pumped through the system must first be degassed, as oxygen and other dissolved gasses affect the accuracy of FFF. The removal of gasses from the solvent improves the reliability of the analysis by alleviating pressure fluctuations, as well as the detection efficiency, by ensuring no disturbances during analysis.

Figure 4. Schematic of AF4-ICP-MS analysis with online addition of UV/Vis analysis. Figure A.4: Schematic of AF4-ICP-MS analysis with online addition of UV/Vis and ICP-MS analysis. !"#$%&'()*+(#&#,-)+("&)"()*+(...(/*#&&+,(%0($"&+()*1"2'*(#&( .(/(0($"(- %&3+/)%"&(4#,4+5(6*%/*(%&()21&(%0(/"&&+/)+$()"()*+()%7(72879( S9(A2P#0/"2Z(B5(U[/*"(C5(U#00+,,\4(G5(]#88+1(.5(^#2)%+1(G5( :;)+1(8#&2#,(%&3+/)%"&5(;,"6(6%,,(/"&0)#&),-(8"4+()*1"2'*()*+( !+07+0(^9(.%+,$F;,"6(;1#/)%"&#)%"&(#&$(%&$2/)%4+,-(/"27,+$( There0#87,+(,""7(%&()*+($%1+/)%"&(";()*+(/*#&&+,()*1"2'*"2)()*+( are several pumps incorporated in FFF7,#08#(8#00(07+/)1"8+)+1(/"27,%&'_(U%0)"1-5($+4+,"78+&)( systems which control the ﬂow of carrier #&#,-0%09(:)()*+(+&$(";()*+(/*#&&+,5()*+(+,2+&)(<$+)+/)"1=(;,"6( #&$(#77,%/#)%"&09(!"#$%&'(")(*%&'+,-.&'(*,"/-.(012.,$"/2,$+( 8"4+0()*+(;1#/)%"&#)+$(#&#,-)+0()"(")*+1(%&0)128+&)0(;"1(( T`_XSYFXTY9 solution;21)*+1(#&#,-0%05(02/*(#0(>?@?%05($-%/(,%'*)(0/#))+1%&'(9B9(:18-(D"170(";(I&'%&++105(6*"5()*1"2'*('1#&)( ://+7)+$(G#&20/1%7)9 &28P+1(QRSTUVFWRFEFSWXY5(8#H+0(#(7"1)%"&(";()*%0(1+0+#1/*( 7"00%P,+9(( b9(O"2P-(G5(^+/H+%0(U5(^+-+1(.9(TWWc9(:77,%/#)%"&(";(#0-88+)1%/(( ;,"6(;%+,$F;,"6(;1#/)%"&#)%"&(<:0.,...=(/"27,+$()"(%&$2/)%4+,-(( 138 /"27,+$(7,#08#(8#00(07+/)1"8+)1-(

`9(E"$#(:5(O+$(:5(]+&&+$-(:5(U#18"&(:5(U2,,(G5(G%)1#&"( A5(f#&4%,,+(a5(B)++4+&0(a9(D*#1#/)+1%K#)%"&(";(0%,4+1(&#&"F 7#1)%/,+0(20%&'(;,"6F;%+,$(;,"6(;1#/)%"&#)%"&(%&)+1;#/+$()"( %&$2/)%4+,-(/"27,+$(7,#08#(8#00(07+/)1"8+)1-9(!"#$%&'(")( 34$"/&,"5$&14+(*6

;*#7,.<)'*#3$=.&6 EB-*F&3)1%*G)%11)* F?5)(?6/*H:*-,BI.*JG:* 0K*LC--M*NO,PB---*"%* LQ.M*,-RPE,IPBO-, 5556(*#7,.*)'*#6&"'

!"#$%$&"'()*+*$),-+,./$"0$"1#$/)"2%)$"00,&*-3$4,-,+$5556(*#7,.*)'*#6&"'89".+%&+:-

!"#$%&'()*+,-../*01%2&34561%/*7389*:55*%&'();*%1;1%<1=9*01%2&34561%>*&;*?*%1'&;)1%1=*)%?=16?%2*"@*01%2&34561%/*7389*:55*")(1%*)%?=16?%2;*?%1*)(1*#%"#1%)$*"@*)(1&%*%1;#18)&<1*"A31%;9 * -.--BCD-. Loading the analyte onto the FFF channel is done through an injection valve, which in turn is connected to the tip pump. After manual injection, ﬂow will constantly move through the sample loop in the direction of the channel throughout the analysis. At the end of the channel, the eluent (detector) ﬂow moves the fractionated analytes to other instruments for further analysis, such as UV/Vis, dynamic light scattering (DLS), and/or ICP-MS.

139 APPENDIX B - COUPLING FLOW FIELD FLOW FRACTIONATION TO ICP-MS FOR THE DETECTION AND CHARACTERIZATION OF SILVER NANOPARTICLES

D.M. Mitrano, James F. Ranville, and Kenneth Neubauer. Coupling Flow Fielf Flow Fractionation to ICP-MS for the Detection and Characterization of Silver Nanoparticles. Perkin Elmer Application Note, Copyright 2012. PerkinElmer, Inc. Shelton, CT USA

B.1 Introduction

Analysis of nanomaterials should include characterization of composition as well as size. Many techniques are capable of sizing nano-size particles, such as dynamic light scattering (DLS), UV/Vis spectrophotometry, and transmission electron microscopy (TEM), yet provide no information on the composition of the particle and/or are time intensive and costly. Inductively coupled plasma-mass spectrometry (ICP-MS), however, is a standard instrument in many analytical laboratories and is the method of choice for analysis of most elements across the periodic chart. The multi-element capability of the ICP-MS, low detection limit (ppt), and wide dynamic range (109 orders of magnitude) also make it ideal for application to the measurement of inorganic engineered nanoparticles (ENPs). While ICP-MS can be used directly to obtain concentrations of nanoparticulate-associated elements, more information on characteristics of ENPs can be obtained by coupling a size-separation step prior to ICP-MS analysis. The most versatile size-separation technique for this application is field flow fractionation (FFF). Although FFF is a powerful nanoparticle sizing technique, many common detectors used in conjunction with FFF do not provide the needed compositional information of the particles. Therefore, the resultant hyphenated technique of FFF-ICP-MS provides nanoparticle sizing, detection, and composition analysis capabilities at the parts per billion (ppb) level, which is critical to environmental investigations of nanomaterials. Fur- thermore, the similar flow conditions required by both ICP-MS and FFF make interfacing relatively simple.

140 B.2 Nanometrology

Nanotechnology has great potential in both industrial and commercial sectors, producing useful products for society either when used alone or when integrated with other material into products (e.g. consumer goods, foods, pesticides, pharmaceuticals, and personal care products, among others). Nanotechnology, defined as the control of matter between 1 and 100 nm where unique phenomena occur because of their small size, has seen great innovation and study in recent years. Several classes of ENPs contain metals that make them particularly suitable for characterization by ICP-MS methods. For example, quantum dots (QDs) often contain cadmium (Cd), selenium (Se), tellurium (Te) and zinc (Zn), among others. QDs are the smallest ENPs having semiconductor properties and are investigated for their use in transistors, solar cells, and LEDs. Despite rapid development, early public acceptance, and acknowledgement of probable release of nanoproducts to the environment, the potential for adverse environmental effects has not yet been established. In the case of QDs, most of the constituent elements can be toxic to organisms. This knowledge gap exists, in part, because of the innate difficulties of detection, characterization, and quantification of ENPs, particularly in environmental and biological samples. There are universal calls for improvements in nanometrology. Many techniques are capable of sizing nano-sized particles (nanoparticles and quantum dots) in simple laboratory systems, including dynamic light scattering (DLS), transmission electron microscopy (TEM), and disc centrifugation (DCS), among others. Yet these methods provide little or no information on the composition of the particle. These techniques are also often not sensitive enough to work at environmentally or biologically relevant concentrations (sub-g/L). Fi- nally, these techniques lack specificity, which means they are not able to distinguish ENPs from other matrix constituents, such as natural particles, humic substances, and cellular debris. Coupling FFF with ICP-MS (or ICP-OES/AES), however, garners element-specific

141 information at trace concentration levels when studying metal-containing NPs [47]. Further- more, the capability of multi-metal analysis is an added beneﬁt when coupling FFF with mass spectrometry.

B.3 Experimental

A selection of simple experiments to demonstrate the concepts of AF4-ICP-MS are shown below.

B.3.1 Materials

Silver nanoparticles of 20 and 40 nm (Nanocomposix, San Diego, CA, USA) were acquired in stock suspensions at a nominal concentration of 20 mg Ag/L and were stabilized in aqueous 2 mM citrate, per the manufacturer. Nano-Ag suspensions were made by diluting the stock solutions with 18.2 M-ohm Nanopure water to ﬁnal concentrations, ranging from

10 to 500 g/L. Aqueous Ag standards (High-Purity Standards, Charleston, SC, USA), used for calibration, were diluted in 1% nitric acid (Optima grade) to concentrations ranging from

1to100g/L. Red mercaptoundecanoic acid (MUA)-coated CdSe/ZnS quantum dots (NN-Labs, Fayettville, AR, USA) were investigated in the second study where the hydrodynamic diameter was 25 nm, with a metal core stated as 5 nm. Stock solutions were diluted approximately 1000- fold, using deionized water to concentrations ranging from 4.6 x 1013 to 1.8 x 1016 particles/L.

B.3.2 Instrumentation

An ELAN® 6100 ICP-MS (PerkinElmer, Shelton, CT, USA) was used for all analyses. Standard operating and tuning procedures were used. Only one silver isotope was monitored (107Ag) with a dwell time of 2000 ms, alternating with a Bi internal standard with a dwell time of 1000 ms, resulting in a data point being collected at a rate of approximately one every three seconds. The total number of readings per sample was chosen such that the data were collected for the entire length of the fractogram, which, depending on experimental

142 conditions, ranged from 40 to 60 minutes. An AF2000 asymmetrical FFF instrument (Postnova Analytics, Salt Lake City, UT, USA) was used for the silver experiments. A 10 kDa regenerated cellulose membrane was used and was replaced approximately every 25 runs. The carrier fluid consisted of 0.01% FL- 70 surfactant and 0.025% sodium azide (an antibacterial agent). The FFF instrument was directly plumbed into the ICP-MS. The channel flow conditions allowed direct connection of the FFF effluent to the ICP-MS nebulizer without a flow splitter. Asymmetrical flow field flow fractionation (AF4) runs were programmed to start with a 10 min relaxation period (focusing step), followed by 40 min elution (0.7 mL/min cross flow and 1.0 mL/min detector flow) with 10 min flush (field-off) between each experimental run. The detector flow can be diverted to a number of instruments, such as ICP-MS, for characterization after AF4 separation. The carrier fluid used to flush the channel after analysis is typically also analyzed in order to determine the unfractionated portion of analyte. Details of AF4 and ICP-MS run conditions are given in Table B.1 and Table B.2. Size characterization of quantum dot samples was accomplished by an F1000 symmetrical FFF instrument (Postnova Analytics) equipped with 1 kDa regenerated cellulose membrane. The ICP-MS (ELAN 6100, PerkinElmer) was used to measure the concentrations of 64Zn, 114Cd, and 82Se, with 209Bi as the internal standard. Carrier fluid also consisted of 0.01% FL-70 and 0.1 mM sodium azide. Pumps delivered the carrier fluid at a channel

flow rate of 1.0 mL/min and recirculated the cross flow at a rate of 0.9 mL/min. A 20 L injection loop was used for sample injection. The outlet flow from the FFF passed through a fluorescence detector and then to the ICP- MS. Details of FFF and ICP-MS run conditions are shown in Table B.1 and Table B.2. An example of the online addition of fluorescence and ICP-MS detectors is given in Figure A.4.

B.3.3 Daily Standards

For the silver nanoparticle study, to ensure the reproducibility of results from day to day, a daily standard was prepared for AF4-ICP-MS analysis that consisted of a mixture of

143 Table B.1: FFF Parameters

Parameter Ag Nanoparticles CdSe/Zn Quantum Dots Intsrument Postnova AF2000 Postnova F1000 (asymmetrical) (symmetrical) Chennel Size 355 x 60 x 40 mm 20 x 270 mm Membrane Type Regenerated cellulose Membrane Porosity 10 kDalton 1kDalton Spacer Width 500 m 254 m Sample Injection 100 L 20 L Volume Detector Flow 1mL/min Cross Flow 0.7 mL/min 0.9 mL/min Injection Delay 1min 15 sec Equilibration Time 10 min 2min Flush Time 10 min N/A Carrier Fluid 0.1% Fl-70, 0.025% 0.01% Fl-70, 1 mM NaN3 NaN3

Table B.2: ICP-MS Parameters

Parameter Ag Nanoparticles CdSe/Zn Quantum Dots Instrument Perkin Elmer ELAN 6100 ICP-MS Nebulizer Cross Flow Spray Chamber Scott Double Pass Neb Gas Flow Optimized for < 3% Oxides Sample Flow 1mL/min RF Power 1000 - 1300 W Dwell Time 3000 ms 4000 ms Analytes 107Ag; 109Bi 66Zn; 111Cd, 82Se; 109Bi Total Analysis Time 60 min 30 min

144 20 and 40 nm Ag nanoparticles at 100 g/L each. This sample was run at the beginning and end of each day, determining if there was a shift in retention time of the particles or change in ICP-MS response (percent recovery). If the retention times did not drift over the course of the day for the standard mixture, we presumed that sample runs were not aﬀected by matrix/membrane or particle/membrane interactions throughout the day. Although size can be directly computed from retention time using FFF theory, for this study we made a linear plot of particle size versus retention time. The linear equation from this plot could then be applied to sample runs to convert elution time to particle diameter.

B.4 Analytical Results

Results of simple experiments to demonstrate the concepts of AF4-ICP-MS are shown below.

B.4.1 Resoution and detection limit

There are a number of parameters in the AF4 method that contribute to both detection limit and resolution, the most important of which is the cross flow parameter of the FFF. Under the flow conditions used, we see nearly baseline separation between the 20 nm and 40 nm Ag nanoparticles (Figure B.1), with the void peak (unresolvable material) present on the far left of the fractogram. An increase in cross flow would allow better separation for smaller particles, down to as small as 3 - 5 nm. However, with increasing cross flow, the analysis time increases and there is a higher chance of particle/membrane interaction, which leads to a decrease in recovery. Therefore, a balance needs to be struck to achieve the best results when considering these multiple factors. Based on the data for 25 and 100 ppb Ag, the detection limit under the current run conditions is estimated to be approximately 5 ppb. The concentration detection limit was determined by running serial dilutions through the FFF-ICP- MS until no discernable analyte peaks were detected above the background.

145 )#*%+,($*%-./'&%,'

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igure 3. Symmetrical FFF-ICP-MS fractogram of red-emitting MUA coated CdSe/ZnS quantum dots (QDs). The first small peak is the void peak, which "4*#-*&"#(*&.;"&#$4*/#0)/&2+*(2/:#"4*&*#3.;"2';*#1%0")(&<# contains unfractionated materials. The larger analyte peak shows that all three N%&*+#)/#"4*#+%"%#1)(#KO#%/+#PLL#''-#9:,#"4*#+*"*0"2)/# metal signals are associated with the fluorescent signal (FL) from the QD. 4 ;232"#./+*(#"4*#0.((*/"#(./#0)/+2"2)/&#*&"23%"*+#")#-*# The particle size at the peak maximum was calculated to be 23 nm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

The results for the symmetrical flow FFF-ICP-MS characterization of a commercial *;*3*/"#2/"*(%0"2)/&#%"#*/D2()/3*/"%;;>#%/+#-2);):20%;;>## CdSe/ZnS/MUA2(3/0-2/,*%-)#*%+'('-4(,5-6%"4-66678!972: quantum dot is shown in Figure B.2. Here, it is demonstrated that the (*;*D%/"#0)/0*/"(%"2)/&<#9;"4).:4#3*"4)+#+*D*;)'3*/"## 0%/#%"#"23*&#-*#%#;*/:"4>#'()0*&&,#"4*#3.;"2".+*#)1#(./## manufacturer=4*#(*&.;")(#"4*#&>33*"(20%;#1;)$#???@!6A@7B#04%(%0"*(2C%"2)/# incorrectly described some quantum dot characteristics, specifically the metal# 0)/+2"2)/&,#&.04#%)$#(%"*&,#0%((2*(#1;.2+#0)3')&2"2)/,## content)1#%#0)33*(02%;#6+B*MQ/BM789#E.%/".3#+)"#2&#&4)$/#2/# in the dots [125]. It was found that the MUA coated quantum dots had a significantly %&#$*;;#%*3-(%/*#">'*#%/+#')()&2">,#;*/+"&*;1#")#"4*# higher(KIWTG*GTGKVKUFGOQPUVTCVGFVJCVVJGOCPWHCEVWTGT (9:1) Cd:Se ratio than the expected nearly 1:1 molar ratio. 1;*52-2;2">#/**+*+#")#1(%0"2)/%"*#%#D%(2*">#)1#'%("20;*&#./+*(## 2/0)((*0";>#+*&0(2-*+#&)3*#E.%/".3#+)"#04%(%0"*(2&"20&,#It was proposed, but not confirmed, that the excess Cd was associated with the MUA J %#/.3-*(#)1#0)/+2"2)/&<#=4*#./2D*(&%;2">#)1#"42&#$2;;# coating&'*02120%;;>#"4*#3*"%;#0)/"*/"#2/#"4*#+)"&< as part of an incomplete washing process during manufacturing.#!"#$%)./+#"4%"# This additional ./+).-"*+;>#&))/#3%I*#1;)$#???@!6A@7B#%/#2/"*:(%;#'%("## Cd"4*#789#0)%"*+#E.%/".3#+)"%+#%#&2:/2120%/";>#42:4*(# demonstrated much higher than expected toxicity. The symmetrical flow FFF-ICP-MS )1#"4*#&"%/+%(+#3*"4)+&#$2"4#$4204#")#&".+>#/%/)'()+.0"&# characterizationFRSPH#6+SB*#(%"2)#"4%/#"4*#*5'*0"*+#/*%(;>#PSP#3);%(#(%"2)<# proved integral in demonstrating that Cd was, at least initially, integral 2/#"42%&"#:()$2/:#12*;+<# to the quantum dot and not simply excess Cd in solution (Figure B.2). This element- specific information would be difficult, if not impossible, to acquire using chemical analysis approaches.

146

4 )#*%+,($*%-./'&%,'

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igure 3. SymmetricalFigure B.2: FFF-ICP-MS Multi-element fractogram FFF-ICP-MS of red-emitting fractogram MUA coated CdSe/ZnS quantum dots (QDs). The first small peak is the void peak, which "4*#-*&"#(*&.;"&#$4*/#0)/&2+*(2/:#"4*&*#3.;"2';*#1%0")(&<#Symmetrical FFF-ICP-MS fractogram of red-emitting MUA coated CdSe/ZnS quantum dots (QDs).contains The ﬁrst unfractionated small peak materials. is the void The peak, larger whichanalyte contains peak shows unfractionated that all three materials. N%&*+#)/#"4*#+%"%#1)(#KO#%/+#PLL#''-#9:,#"4*#+*"*0"2)/# The largermetal analytesignals are peak associated shows with that the all fluorescent three metal signal signals (FL) are from associated the QD. with the The particle size at the peak maximum was calculated to be 23 nm.4 ;232"#./+*(#"4*#0.((*/"#(./#0)/+2"2)/&#*&"23%"*+#")#-*# ﬂuorescent signal (FL) from the QD. The particle size at the peak maximum was %''()523%"*;>#O#''-<#=4*#0)/0*/"(%"2)/#+*"*0"2)/#;232"#$%&# calculated to be 23 nm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

!"#$%&'("#' 6)/0*(/2/:#*/:2/**(*+#/%/)'%("20;*&,#/%/)3*"();):>#2&#%# :()$2/:#12*;+#"4%"#2�*("%2/#")#-*/*12"#1()3#)/@;2/*#1;)$# ???@!6A@7B#%/%;>&2&<#=4*#0)3-2/%"2)/#)1#0)/"2/.).&## 1(%0"2)/%"2)/#.&2/:#???,#$2"4#"4*#&*/&2"2D*,#3.;"2@*;*3*/"%;## Figure 2. Analysis of 25 and 100 ppb nano-Ag mixture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

4 B.5 Conclusions

Concerning engineered nanoparticles, nanometrology is a growing field that is certain to benefit from on-line flow FFF-ICP-MS analysis. The combination of continuous fractionation using FFF, with the sensitive, multi-elemental capability of ICP-MS, will provide increased knowledge about size-dependent variations in composition and trace element interactions at environmentally and biologically relevant concentrations. Although method development can at times be a lengthy process, the multitude of run conditions, such as flow rates, carrier fluid composition, as well as membrane type and porosity, lends itself to the flexibility needed to fractionate a variety of particles under a number of conditions. The universality of this will undoubtedly soon make flow FFF-ICP-MS an integral part of the standard methods with which to study nanoproducts in this fast growing field.

148 APPENDIX C - FIELD-FLOW FRACTIONATION COUPLED WITH ICP-MS FOR THE ANALYSIS OF ENGINEERED NANOPARTICLES IN ENVIRONMENTAL SYSTEMS

This article provides an overview of the capabilities of field-flow fractionation coupled with inductively coupled plasma–mass spectrometry and demonstrates that the technique shows a great deal of promise to separate, detect, and quantitate nanoparticles in environmental matrices. This chapter is reproduced with permission and some modification from Denise M. Mitrano, James F. Ranville, Ken Neubauer, and Robert Thomas. Field-Flow Fractionation Coupled with ICP-MS for the Analysis of Engineered Nanoparticles in Envi- ronmental Samples (2012). Spectroscopy, 27(9), 36-44. Copyright 2012. The National Nanotechnology Initiative defines nanotechnology as the understanding and control of materials at dimensions of 1 – 100 nm, where unique properties enable novel applications to be carried out [113]. Gases, liquids, and solids can exhibit unusual physical, chemical, and biological properties at the nanoscale level, differing in critical ways from the properties of the bulk materials. Nanomaterials can occur in nature, such as clay minerals and humic acids; they can be incidentally produced by human activity such as diesel emissions, or weld- ing fumes; or they can be specifically engineered to exhibit unique optical, electrical, physical, or chemical characteristics. Depending on their chemical and physical characteristics, these engineered nanomaterials (ENMs) can be made to exhibit greater physical strength, enhanced mag- netic properties, conduction of heat or electricity, greater chemical reactivity, or size-dependent optical properties.

C.1 Engineered Nanomaterials

Most ENMs can be divided into two main classes: carbon- based nanomaterials, such as nanotubes and spherical bucky balls, and metal-containing ones such as Ag, Au, Fe or TiO2 nanoparticles. Of the carbon-based ENMs, many products incorporate carbon nanotubes

149 to improve strength. These ENMs are strongly bound to the matrix of the material and, as a result, are less likely to be released into the environment. Even though many metal- containing ENMs are also incorporated into a product matrix such as solar cells, a significant amount are used in dispersive applications. In these applications, they are intentionally released from the product, although incidental release can also be quite substantial. For example, fabrics containing silver nanoparticles used to kill bacteria release silver at varying rates during the washing cycle, depending on the type of fabric and the washing conditions. Another good example is a novel washing machine that “creates” Ag nanoparticles during the wash cycle by releasing silver nanoparticles and ions (Ag+) from a solid piece of Ag inside the machine. Therefore, it is clear that the use of ENMs in consumer, industrial, and agricultural products, as well as in environmental technology is growing rapidly. Often, the benefit of using nanomaterials stems from the increased surface area per unit mass of material, which increases with the inverse of the diameter. This results in faster rates of chemical reactions (for example, oxidative catalysis) occurring on material surfaces. Sometimes the benefit of nanomaterials arises from the quantum nature of energy states at the nanometer scale, as in the wavelength tuning of the fluorescence of quantum dots used in electronics applications. In the health sciences, the ability of surface-functionalized ENMs to bind to cell walls can be used for drug delivery. It has been estimated by the Project on Emerging Nanotechnologies, that there are currently more than 1000 products containing ENMs used for consumer, healthcare, and industrial applications[28].

C.2 Potential for Environmental Impact

The unique properties of ENMs have also created intense interest in the environmental behavior of these materials. Because of the increased use of nanotechnology in consumer products, industrial applications, and health care technology, nanoparticles are more likely to enter the environment. Therefore, to ensure sustainable development of nanotechnology products, there is clearly a need to evaluate the risks posed by these engineered nanoparticles

150 (ENPs), which will require proper tools to carry out exposure assessment studies. Current approaches to assess exposure include predictions based on computer modeling of ENP life cycles or alternatively, by direct measurement techniques. Although these are very different approaches, both require instrumentation and analytical methodologies. Prediction of environmental concentrations of ENPs through modeling is based on knowledge of how they are emitted into the environment using production volumes and life cycle assessment data and also by their eventual fate and behavior in the environment being studied (that is, soil, sediment, water, and air). Although we are now starting to understand the life cycles of ENPs during production, use, and disposal, there is very little known about their environmental fate and behavior. Different ENPs will have different properties and, therefore, will behave very differently when they enter the environment. The approach for prediction of environmental concentrations through life cycle assessment modeling requires validation through measurement of actual environmental concentrations. Extremely sensitive methods are required for ENPs that were only recently introduced into the environment. Although the direct measurement approach is not hampered by the underlying assumptions of exposure modeling, it is very important to ensure that direct observations are representative in time and space for the regional setting in which the observation was made. ENPs differ from most conventional ‘‘dissolved’’ chemicals in terms of their heterogeneous distributions in size, shape, surface charge, composition, and degree of dispersion. Therefore, it is not only important to determine their concentrations, but also other metrics such as shape, size distribution, and chemical composition.

C.3 Analytical Methodologies

The measurement and characterization of nanoparticles (nanometrology) is therefore critical to all aspects of nanotechnology. In the ﬁeld of environmental health and safety (EHS), it has become clear that “complete” characterization of nanomaterials is critical to interpreting the results of toxicological, human health, and environmental fate studies. Metal-containing ENPs form a particularly signiﬁcant class, as their use in consumer products and indus-

151 trial applications make them the fastest growing category of nanoparticles. Several life cycle assessments conclude that he predicted environmental concentrations (PEC) of some metal- bearing nanoparticle could exceed the predicted no effect concentration (PNEC), suggesting that ENPs can enter aquatic systems at potentially harmful concentrations. However, in most cases, these levels are typically at the part-per-trillion level. Many analytical techniques are available for nanometrology, only some of which can be successfully applied to nano-EHS studies [59]. These methods differ in part by the properties measured: average size, size distribution, surface characteristics, shape, and chemical composition. Methods for assessing particle concentration and particle size distributions include electron microscopy, chromatography, centrifugation, laser-light scattering, ultrafiltration, and spectroscopy. Difficulties generally arise because of a lack of sensitivity for characterizing and quantifying particles at environmentally relevant concentrations (low micrograms per liter). Furthermore, the lack of specificity of the technique is problematic for complex environmental matrices that may contain natural nanoparticles with polydisperse particle distributions, as well as heterogeneous compositions. Electron microscopy and dynamic light scattering (DLS) are the most commonly applied methods, but they each have advantages and disadvantages. Electron microscopy gives the most direct information on the size distribution and shapes of the individual nanoparticles. However, sample preparation steps such as drying or exposure to vacuum can induce an agglomeration (clustering) of the particles, thus making it difficult to define their size in the original media. In addition, organic coatings are not visible without staining, which can lead to errors in the measurement of the particle diameter. DLS measures the diameter of the particle while in motion (hydrodynamic diameter), and thus provides sizing of organically coated nanoparticles. Limitations of DLS include poor sensitivity at dilute concentrations, nonselective material detection, inability to distinguish mixtures in complex matrices, and difficulty in resolving the dominant size in polydisperse samples. The presence of a small number of aggregates can skew the effective diameter

152 toward a larger particle size distribution. However, despite its limitations, DLS remains a rapid technique to quickly determine average particle hydrodynamic diameter for a wide range of particle types. An emerging technique called single-particle inductively coupled plasma–mass spectrometry (sp-ICP- MS) has been developed for detecting and sizing metallic nanoparticles at environmentally relevant (nano- grams per liter) concentrations. Although this method is still in its infancy, it has shown a great deal of promise in several applications, including determining concentrations of silver nanoparticles in complex matrices such as wastewater effluent [107]. The method involves introducing nanoparticle-containing samples, at very dilute concentration, into the ICP-MS system and collecting time- resolved data. Integration times on the order of 10 ms are used to detect individual particles as pulses of ions after they are atomized by the plasma. Observed pulse number is related to the nanoparticle concentration by the nebulization efficiency and the total number of nanoparticles in the sample, and the mass, and thus the size of the nanoparticle, is related to the pulse intensity [29]. However, it should be emphasized that for this approach to work effectively at low concentrations, the speed of data acquisition and the response time of the ICP-MS detector must be fast enough to capture the time-resolved nanoparticle pulses, which typically last only a few milliseconds. Field-flow fractionation (FFF) analysis, the sizing technique highlighted in this article, is a powerful tool for sizing and separating ENPs. FFF, incorporating UV-absorbance detection, is generally limited to particle concentrations in the parts-per-million (milligrams-per- liter) range and lacks particle specificity. Furthermore, UV response is not a direct measure of particle mass concentration, but rather depends on particle size, shape, and optical characteristics. However, coupling FFF with a sensitive and selective multielement technique such as ICP-MS lowers detection capabilities by approximately three orders of magnitude, to the parts-per-billion (micrograms-per-liter) range, and provides direct information about particle mass concentration and composition [125]. For this reason, it is clear that because

153 of its elemental speciﬁcity, excellent resolution, and low detection limit, ICP-MS is perhaps becoming the most promising detection method for nano-EHS studies.

C.4 Field-Flow-Fractionation

FFF is a single-phase chromatography technique in which separation is achieved within averythinchannel,againstwhichaperpendicularforcefieldisapplied.Oneofthemost common forms of FFF is asymmetrical-flow FFF, in which the field is generated by a cross- flow applied perpendicular to the channel. The flow and sample are confined within a channel that consists of two plates separated by a spacer that is typically 100 – 500 mthick.The upper channel plate is impermeable and the bottom channel is made of a permeable porous frit material. A membrane covers the bottom plate to prevent the sample from penetrating the channel. Within the flow channel, a parabolic flow profile is created because of the laminar flow of the liquid. As a result, the stream moves slowly closer to the boundary edges than it does at the center of the channel flow. When a perpendicular force field, in this case fluid cross-flow, is applied to the f lowing, laminar stream, the analytes are driven toward the boundary layer of the channel. Diffusion, associated with Brownian motion, creates a counteracting motion. Smaller particles that have higher diffusion rates will reach an equilibrium position higher up in the channel where the longitudinal flow is faster. The smaller particles are transported much more rapidly along the channel than the larger particles, which results in the smaller particles being eluted before the larger ones. After a sample is injected through the inlet port of the FFF system, the flows are manipulated in such a way as to concentrate the particles into a narrow band. Following this period, the channel and cross-flows are applied to create the separation. The separated particles exit the outlet port, move into the detection system, and are displayed as a temporal signal called a fractogram (similar to a chromatogram in chromatographic separation techniques). The fundamental principles of this separation process are shown in Figure C.1. A

154 typical fractogram of a mixture of silver and gold particles using UV-absorbance detection is shown in Figure C.2[? ].

Figure C.1: FFF channel and flow profile Separation in an asymmetrical field-flow fractionataion (FFF) channel is a result of the imposition of a parabolic channel flow velocity profile on analytes that are positioned on the channel at heights that arise from the balance of the applied field (U) and the back diffusion (D).

C.5 FFF Coupled with ICP-MS

ICP-MS has become the dominant technique for ultratrace-level quantitation of metals in environmental matrices, with multielement capabilities similar to ICP-optical emission spectroscopy (ICP-OES), and detection limits an order of magnitude lower than graphite furnace atomic absorption spectroscopy (GFAAS). This makes the technique ideal for de-

155 38 Spectroscopy 27(9) September 2012 www.spectroscopyonline.com

the predicted environmental concentrations (PEC) of some metal- Channel flow in Focus flow in Channel flow out bearing nanoparticle could exceed the predicted no effect concentration (PNEC), suggesting that ENPs can enter aquatic systems at potentially harmful concentrations. However, in most cases, these levels are typically at the part-per-trillion level. Many analytical techniques are available for nanometrology, only some of which can be successfully applied to nano-EHS studies (3). These Ceramic frit methods differ in part by the properties measured: average size, size distribution, surface characteristics, shape, and chemical composition. Methods Cross flow out Channel spacer for assessing particle concentration and particle size distributions include electron microscopy, chromatography, centrifugation, laser-light scattering, ultrafiltration, and spectroscopy. Dif- ficulties generally arise because of a U D lack of sensitivity for characterizing and quantifying particles at environmentally relevant concentrations (low micrograms per liter). Furthermore, the lack of specificity of the technique Figure 1: Separation in an asymmetrical field-flow fractionataion (FFF) channel is a result of the is problematic for complex environ- imposition of a parabolic channel flow velocity profile on analytes that are positioned on the mental matrices that may contain channel at heights that arise from the balance of the applied field (U) and the back diffusion (D). natural nanoparticles with polydisperse particle distributions, as well as heterogeneous compositions. Electron microscopy and dynamic light scattering (DLS) are the most 19.1 nm commonly applied methods, but they 0.3 each have advantages and disadvantages. Electron microscopy gives the most direct information on the size 0.2 36.9 nm distribution and shapes of the individual nanoparticles. However, sample preparation steps such as drying 0.1 66.6 nm or exposure to vacuum can induce an agglomeration (clustering) of the particles, thus making it difficult to 0.0 define their size in the original media. In addition, organic coatings are not visible without staining, which can -0.1 Absorbance (mAU, 520 nm) lead to errors in the measurement of the particle diameter. DLS measures the diameter of the -0.2 particle while in motion (hydrody- 0 300 600 900 1200 1500 namic diameter), and thus provides Time (s) sizing of organically coated nanoparticles. Limitations of DLS include poor Figure 2: A fractogramFigure of the C.2: particle FFF size separation UV/VIS of a mixture fractogram of gold (Au) and silver (Ag) sensitivity at dilute concentrations, A fractogram ofparticles the using particle field flow sizefractionation separation with UV absorbance of a mixturedetection. Adapted of gold from reference (Au) 7. andnonselective silver (Ag) material detection, in- particles using field flow fractionation with UV absorbance detection. [131] tecting, quantifying, and characterizing metal nanoparticles with extremely high sensitivity and selectivity, as well as avoiding many of the known interferences associated with complex environmental samples. The added benefit of using ICP-MS is that it is a rapid multielement technique, so it can be applied to the analysis of metal salt semiconductors such as cadmium selenide quantum dots. Dissolution of the cadmium selenide core is inhibited by the addition of an outer shell of zinc sulfide or similar material. ICP-MS can detect all these metals, therefore allowing the study of size- dependent dissolution or aggregation of these kinds of multielement containing nanoparticles. ICP-MS is also relatively straight forward to couple to FFF because the sample flow rate of the ICP-MS sample introduction system is similar to the outlet flow rate of the FFF system (˜ 0.5 – 2.0 mL/min) [11]. However, some challenges still have to be overcome to quantify metal concentrations in fractionated samples, because some nanoparticles tend to stick to the internal membrane of the FFF system. Figure C.3 shows a typical instrumental setup for coupling an FFF system to an ICP-MS system. (Note: All the ICP- MS data published

156 www.spectroscopyonline.com September 2012 Spectroscopy 27(9) 39 in this study were generated on an ELAN DRC II ICP-MS system from PerkinElmer, Inc.). ability to distinguish mixtures in complex matrices, and difficulty in resolving the dominant size in poly- Recirculation or to waste Off-line mode Flow variable ≈ 0.5–3.5 mL/min disperse samples. The presence of a small number of aggregates can skew the effective diameter toward a larger Nondestructive particle size distribution. However, Cross flow flow-through Fraction collector despite its limitations, DLS remains a detector rapid technique to quickly determine average particle hydrodynamic diameter for a wide range of particle types. An emerging technique called single-particle inductively coupled F1-FFF channel plasma–mass spectrometry (SP-ICP- ICP-MS system MS) has been developed for detect- Channel flow ing and sizing metallic nanoparticles ≈ 1.0 mL/min at environmentally relevant (nano- On-line mode grams per liter) concentrations. Al- though this method is still in its Figure 3: A typicalFigure instrumental C.3: Diagramset-up for coupling of FFF-ICP-MS an FFF system coupling to an ICP-MS system. The infancy, it has shown a great dealA typicalof nondestructive instrumental detectors set-up can include for coupling light-scattering, an FFF UV–vis, system refractive to anindex, ICP-MS or fluorescence system. The promise in several applications, in-nondestructivetechniques. detectors can include light-scattering, UV–vis, refractive index, or cluding determining concentrations fluorescence techniques. of silver nanoparticles in complex ICP-MS system and collecting time- Observed pulse number is related to matrices such as wastewater effluent resolved data. Integration times on the nanoparticle concentration by the (4). The method involves introduc-C.6 Particlethe order Size of Reference10 ms are used Standards to detect nebulization efficiency and the total ing nanoparticle-containing samples, individual particles as pulses of ions number of nanoparticles in the sam- at very dilute concentration, intoFFF the theoryafter they is well-developed are atomized by forthe plasma. the separation ple, and and the sizing mass, of and polydisperse thus the size particles in simple matrices using retention times [46]. Interpreting peak areas to determine concentrations is somewhat more difficult. By using conventional detection, such as UV absorbance, refractive index, light scattering, or fluorescence, the technique can use stable dispersion calibration reference standards such as NIST-traceable polystyrene beads of known parti- Enwave provides a wide range of Raman solutions from low-cost routine Raman instrumentation to high sensitivity Ramancle instrument sizes. Unfortunately, with a variety the detectionof configurations of polystyrene to meet (carbon) your applications beads is not practical by ICP-MS. For that reason, when external size calibration is required, an in-line conventional, nondestructive detector before the ICP-MS detector is a relatively simple addition to the instrumental setup, as shown in Figure C.3. As new reference materials become more available, polystyrene bead standards can be replaced with metallic nanoparticles. Currently, NIST provides monodisperse (same size and shape) gold and silver nanoparticles suitable for sizing at trace levels. Laboratory Raman Field Portable Handheld Raman Process Raman Models for various appli- Research grade perfor- Easy to use and high High performance ³ PPM cations and budget mance in the field sensitivity sensitivity at affordable High performance/price Fully integrated, battery- 21 CFR Part 11 157compliant price ratio operated with integrated Ideal for incoming raw Reliability, Long term computer material identification in stability pharmaceutical, chemical, and other industries Enwave Optronics, Inc. ZZZHQZDYHRSWFRP_LQIR#HQZDYHRSWFRP_7HO--- The benefit of using ICP-MS as a detection system for FFF compared to UV absorbance can be seen in Figure C.4, which shows the separation of mixed gold and silver nanoparticles. It can be seen that using UV detection alone (inset graph), the fractogram does not differentiate between the silver and gold particles and only shows the FFF separated particles by size (10, 30, and 60 nm). The main fractogram obtained using ICP-MS detection clearly shows the separated particles and the signal intensities for both the silver and gold particles 40 Spectroscopy 27(9) September 2012 www.spectroscopyonline.com at their atomic masses of 107 amu and 197 amu, respectively [? ]. pulses, which typically last only a few milliseconds. 300,000 0.4 Field-flow fractionation (FFF) 17.7 nm 0.3 analysis, the sizing technique highlighted in this article, is a powerful 250,000 0.2 34.1 nm tool for sizing and separating ENPs. 0.1 FFF, incorporating UV-absorbance 200,000 37.3 nm 0.0 detection, is generally limited to

Absorbance (mAU, 520 nm) -0.1 particle concentrations in the parts-

150,000 17.8 nm 61.6 nm -0.2 per-million (milligrams-per-liter) 0 300 600 900 1200 1500 Time (s) range and lacks particle specificity. 68.8 nm Furthermore, UV response is not 100,000 ICP-MS response (cps) a direct measure of particle mass 107Ag concentration, but rather depends 197Au 50,000 on particle size, shape, and optical characteristics. However, coupling FFF with a sensitive and selective 0 0 200 400 600 800 1000 1200 1400 1600 multielement technique such as Time (s) ICP-MS lowers detection capabilities by approximately three orders of Figure C.4:Figure Comparison 4: FFF fractogram of FFF of the Fractogramseparation of mixed with silver ICP-MS and goldand nanoparticles, UV/VIS using as both online UV detectormagnitude, to the parts-per-billion FFF fractogramabsorbance of (inset the graph) separation and ICP-MS of detection. mixed silver Adapted and from gold reference nanoparticles, 7. using both(micrograms-per-liter) UV range, and absorbance (inset graph) and ICP-MS detection.[131] provides direct information about of the nanoparticle, is related to the tions, the speed of data acquisition particle mass concentration and pulse intensity (5). However, it should and the response time of the ICP-MS composition (6). For this reason, it Externalbe sizeemphasized calibration that alsofor this requires approach stable detector particle must dispersions be fast enough and minimal to cap- membraneis clear that because of its elemen- to work effectively at low concentra- ture the time-resolved nanoparticle tal specificity, excellent resolution, interactions. For that reason, a surfactant may be needed. In some cases the surfactantand most low detection limit, ICP-MS is perhaps becoming the most promis- suitable for metal-containing nanoparticles may not be compatible with polystyreneing stan- detection method for nano-EHS dards, again suggesting that other size calibration standards would be useful. Additionally,studies. when using ICP-MS on-line detection, other factors need to be considered with respectField-Flow to Fractionation FFF is a single-phase chromatog- choice of carrier solution. For example, the carrier solution can cause salt formationraphy on the technique in which separation is achieved within a very thin Tunable Light Sources channel, against which a perpen- t dicular force field is applied. One Maximizing throughput in Monochromator 158 Control Software of the most common forms of FFF the visible region of TLS Series is asymmetrical-flow FFF, in which Tunable Light Source the spectrum. the field is generated by a cross-flow t Used to study wavelength applied perpendicular to the chan- dependent chemical, nel. The flow and sample are con- biological, and physical fined within a channel that consists of two plates separated by a spacer changes or properties. that is typically 100–500 µm thick. t Also used in color analysis The upper channel plate is imperme- and reectivity measurements able and the bottom channel is made of products for aesthetic purposes. of a permeable porous frit material. A membrane covers the bottom plate Automated version: to prevent the sample from penetrat- scanning digital monochromator, lamp, control box and software ing the channel. Within the flow channel, a para- Manual version: digital monochromator, lamp and power supply bolic flow profile is created because www.optometrics.com t 978-772-1700 of the laminar flow of the liquid. As a result, the stream moves slowly closer to the boundary edges than it does at ICP-MS cones, leading to instability and a decrease in signal intensity over time. Another potential problem is that polyatomic spectral interferences can be created in the plasma by components in the carrier solution. For example, the use of chloride-based salts can generate interferences for elements such as vanadium, chromium, arsenic, and selenium, among others. So even though modern ICP-MS instrumentation often includes a collision–reaction cell to minimize these interferences[155], the optimization of FFF separation conditions must also consider the impact on the ICP-MS system.

C.7 Calibration Strategies

In addition, the correct implementation of both internal and external calibration standards is critical in ICP-MS. Typically, internal standards are prepared in 1 – 5% acid and added to all calibration standards and samples. Unfortunately, the introduction of a dissolved metal standard into a near-neutral-pH FFF mobile phase can result in precipitation of analytes resulting in inaccurate metal quantification of fractionated samples. An alternative approach is to use split flows that allow the introduction of acidified internal standards directly to the ICP nebulizer after elution from the FFF channel. Internal standards have also been mixed directly into the carrier fluid and simultaneously used for external calibration by comparison of elemental response factors to the internal standard. Alternative calibration approaches for FFF–ICP-MS studies include injecting a known mass of metal using a flow-injection technique and comparing the area of the known mass to the area of the unknown sample elution peak. External calibration can also be performed by analyzing a continuous flow (delivered by a flow rate–matched peristaltic pump) of known concentrations of metal, developing a calibration curve, and then converting the intensity of the fractogram reading to a concentration value [39].

C.8 Recovery

One of the inherent limitations with quantitative applications of FFF–ICP-MS can be low recoveries, which are attributed to several factors. Probably the most signiﬁcant prob-

159 lem area is the physical interaction of the analyte with the membrane by an adsorption mechanism, resulting in the particles sticking to the membrane and not being eluted. In addition, losses through the accumulation wall based on membrane cut-offvalues have been reported for samples containing dissolved and macromolecular components [161]. Analyte loss can also occur in the ICP-MS nebulizer, spray chamber, and sample tubing, but these losses are relatively small compared to membrane interactions. However, when the analytical conditions are favorable, good recoveries can be achieved as demonstrated by the silver nanoparticle data shown in Figure C.5. Recoveries of the three silver nanoparticle sizes tested (10, 40, and 70 nm) at different concentrations (200, 150, 100, and 50 ppb), are illustrated in the colored bar graph, which shows the integrated ICP-MS response signal (peak area) for the FFF cross-flow field off(blue); bypassing the FFF system entirely (brown); and with the FFF cross-flow field on (green). The inset graph shows the raw fractogram data for the different concentrations of the 10 nm particles, but as can be seen by the colored bar graphs for the 40 nm and 70 nm particles, the fractograms generated similar data. In fact, the recoveries for the four concentrations of the three different particle sizes of silver in this study were all in the range of 88 – 98%, based on integrated peak areas [? ].

C.9 Detection Limits

Traditional ICP-MS analysis generally has instrumental detection limits in the range of 1–100ng/L(partspertrillion),dependingonthespecificelementandtheabundanceof the isotope measured. However, in FFF applications, the mass of metal nanoparticles being detected is distributed over a size range that is diluted as the mass is spread out over the effluent volume. Spreading is a function of sample polydispersity, nonideal membrane interactions, and band broadening. For this reason, the detection limit is more appropriately defined as a mass, where the mass-based detection limit (mDL) is the product of the instrument detection limit and the peak width, as shown in Figure C.6A. This example shows the uranium elution profile during the measurement of uranium bound to monodisperse hematite

(Fe2O3)nanoparticlesalongwithfractogramsscaleddownbyfactorsof2and10.Ifthearea

160 42 Spectroscopy 27(9) September 2012 www.spectroscopyonline.com

salt semiconductors such as cadmium selenide quantum dots. Dissolution of the cadmium selenide core is inhib- Injection of 10-nm Ag particles into FFF system under various conditions Field off ited by the addition of an outer shell 450,000 A = Bypass membrane B = Field off of zinc sulfide or similar material. Bypass membrane C = 200 mg/L fractogram 200 µg/L D = 50 µg/L fractogram ICP-MS can detect all these metals, Fractogram A therefore allowing the study of size- 300,000 150 µg/L dependent dissolution or aggregation of these kinds of multielement-con- 100 µg/L taining nanoparticles. 150,000 B ICP-MS is also relatively straight-

ICP-MS response (cps) 50 µg/L C forward to couple to FFF because the D sample flow rate of the ICP-MS sam- 0 200 400 600 800 1000 1200 1400 1600 ple introduction system is similar to Time (s) the outlet flow rate of the FFF system (~0.5–2.0 mL/min) (9). However, some Integrated ICP-MS response challenges still have to be overcome to quantify metal concentrations in fractionated samples, because some nanoparticles tend to stick to the internal membrane of the FFF system.

50 ppb 50 ppb 50 ppb Figure 3 shows a typical instrumental 200 ppb 150 ppb 100 ppb 200 ppb 150 ppb 100 ppb 200 ppb 150 ppb 100 ppb setup for coupling an FFF system to 10 nm 40 nm 70 nm an ICP-MS system. (Note: All the ICP- MS data published in this study were Figure 5: ComparisonFigure of recoveries C.5: Recovery of three silver for the nanoparticle FFF technique sizes (10, 40, and 70 nm) at generated on an ELAN DRC II ICP- Comparisondifferent concentrations of recoveries (200, of three 150, 100, silver and nanoparticle 50 ppb), under sizes various (10, FFF 40, field and conditions. 70 nm) at Adapted different MS system from PerkinElmer, Inc.) concentrations (200, 150, 100, and 50 ppb), under various FFF field conditions. from reference 7. Particle Size the center of the channel flow. When a matogram in chromatographic sepa- Reference Standards perpendicular force field, in this case ration techniques). The fundamental FFF theory is well-developed for the fluid cross-flow, is applied to the flow- principles of this separation process separation and sizing of polydisperse ing, laminar stream, the analytes are are shown in Figure 1. A typical frac- particles in simple matrices using driven toward the boundary layer of togram of a mixture of silver and gold retention times (10). Interpreting the channel. particles using UV-absorbance detec- peak areas to determine concentra- Diffusion, associated with Brown- tion is shown in Figure 2 (7). tions is somewhat more difficult. By 161 ian motion, creates a counteracting using conventional detection, such motion. Smaller particles that have FFF Coupled with ICP-MS as UV absorbance, refractive index, higher diffusion rates will reach an ICP-MS has become the dominant light scattering, or fluorescence, the equilibrium position higher up in the technique for ultratrace-level quanti- technique can use stable dispersion channel where the longitudinal flow tation of metals in environmental ma- calibration reference standards such is faster. The smaller particles are trices, with multielement capabilities as NIST-traceable polystyrene beads transported much more rapidly along similar to ICP-optical emission spec- of known particle sizes. Unfortu- the channel than the larger particles, troscopy (ICP-OES), and detection nately, the detection of polystyrene which results in the smaller particles limits an order of magnitude lower (carbon) beads is not practical by being eluted before the larger ones. than graphite furnace atomic absorp- ICP-MS. For that reason, when ex- After a sample is injected through tion spectroscopy (GFAAS). This ternal size calibration is required, an the inlet port of the FFF system, the makes the technique ideal for detect- in-line conventional, nondestructive flows are manipulated in such a way ing, quantifying, and characterizing detector before the ICP-MS detec- as to concentrate the particles into a metal nanoparticles with extremely tor is a relatively simple addition to narrow band. Following this period, high sensitivity and selectivity, as the instrumental setup, as shown in the channel and cross-flows are ap- well as avoiding many of the known Figure 3. As new reference materi- plied to create the separation. The interferences associated with complex als become more available, polysty- separated particles exit the outlet environmental samples (8). The added rene bead standards can be replaced port, move into the detection system, benefit of using ICP-MS is that it is with metallic nanoparticles. Cur- and are displayed as a temporal signal a rapid multielement technique, so it rently, NIST provides monodisperse called a fractogram (similar to a chro- can be applied to the analysis of metal (same size and shape) gold and silver (mass) under the elution peak is compared with the area (mass) defined by the mDL, it is www.spectroscopyonline.com clear that the 0.1 scale fractogram is roughly the same areaSeptember as the 2012 mDL Spectroscopy and, as a result,27(9) 43 the mass of uranium would be very difficult to accurately quantitate[131]. nanoparticles suitable for sizing at trace levels. (a) (b) The benefit of using ICP-MS as a FractogramU fractogram 67 µg/L each of 3 size particles 0.1 40 nm detection system for FFF compared 0.5 scale 6.7 µg/L each of 3 size particles to UV absorbance can be seen in Fig- 0.1 scale 20,000 13.4 µg/L each of 3 size particles bkgd U = 2.8 ng/L 0.08 ure 4, which shows the separation of IDL = 2*bkgd 70 nm mDL 10 nm mixed gold and silver nanoparticles. 15,000 Peak mass = 5.7e-4 µg U It can be seen that using UV detec- 0.06 tion alone (inset graph), the fracto- 10,000

gram does not differentiate between 0.04 Ag response (cps) the silver and gold particles and only Uranium concentration (ppb) 5000 shows the FFF separated particles by 0.02 mDL = 1.3e-4 µg U size (10, 30, and 60 nm). The main fractogram obtained using ICP-MS 0 0 detection clearly shows the separated 150 350 550 750 950 0 200 400 600 800 1000 1200 1400 1600 particles and the signal intensities for Time (s) Time (s) both the silver and gold particles at Figure C.6: Multi-element analysis via FFF-ICP-MS Figure 6: (a) The detection limit in FFF–ICP-MS studies is defined as a mass, where, the mass- their atomic masses of 107 amu and (a) The detection limit in FFF–ICP-MS studies is defined as a mass, where, the mass- 197 amu, respectively (7). basedbased detection detection limit limit (mDL)(mDL) is is the the product product of the of theinstrument instrument detection detection limit and limit the andpeak the width. peak External size calibration also re- Adaptedwidth [86].from (b)reference In a three-particle14. (b) In a three-particle mixture, themixture, effect the of effect polydispersity of polydispersity degrades degrades the quires stable particle dispersions and the detection limit. Adapted from referencedetection 15. limit [131]. minimal membrane interactions. For that reason, a surfactant may be needed. ples.In particle Unfortunately, mixtures the the eff ectintroduction of polydispersity MS is evencan be more low dramatic. recoveries, Figure which C.6B are shows In some cases the surfactant most suit- of a dissolved metal standard into a attributed to several factors. Probably afractogramofamixtureofthreesizesofsilvernanoparticles(10,40,and70nm)containing able for metal-containing nanoparticles near-neutral-pH FFF mobile phase the most significant problem area is may not be compatible with polystyrene a totalcan Agresult concentration in precipitation of 201 ofg/L analytes (67 g/L eachthe physical of the three interaction sizes of particles), of the analyte together standards, again suggesting that other resulting in inaccurate metal quanti- with the membrane by an adsorption with serial dilutions of fivefold (purple peaks) and 10 - fold (red peaks) of the mixture. Of the size calibration standards would be use- fication of fractionated samples. An mechanism, resulting in the par- ful. Additionally, when using ICP-MS dilutedalternative samples, approach clearly only is the to fivefold use split dilu- tionticles sample sticking (total 40.2to theg/L membrane Ag) is far enoughand on-line detection, other factors need to flows that allow the introduction of not being eluted. In addition, losses above the background to allow it to be quantified with good accuracy and precision. With be considered with respect to choice of acidified internal standards directly through the accumulation wall based carrier solution. For example, the car- theto 10 the - fold ICP dilution nebulizer sample, after probably elution only theon 40 membrane nm particles havecut-off generated values a quantifi-have rier solution can cause salt formation ablefrom peak, the as FFF the 6.7channel.g/L (20.1 Internalg/L stan- total) three-particlebeen reported mixture for resultssamples in acontain- fractogram on the ICP-MS cones, leading to insta- dards have also been mixed directly ing dissolved and macromolecular bility and a decrease in signal intensity thatinto is only the slightlycarrier abovefluid theand background simultane- [86]. components (13). Analyte loss can over time. Another potential problem ously used for external calibration by also occur in the ICP-MS nebulizer, is that polyatomic spectral interfer- comparison of elemental response spray chamber, and sample tubing, ences can be created in the plasma by factors to the internal standard (12). 162but these losses are relatively small components in the carrier solution. For Alternative calibration approaches compared to membrane interactions. example, the use of chloride-based salts for FFF–ICP-MS studies include in- However, when the analytical con- can generate interferences for elements jecting a known mass of metal using ditions are favorable, good recoveries such as vanadium, chromium, arsenic, a flow-injection technique and com- can be achieved as demonstrated by and selenium, among others. So even paring the area of the known mass the silver nanoparticle data shown in though modern ICP-MS instrumenta- to the area of the unknown sample Figure 5. Recoveries of the three silver tion often includes a collision–reaction elution peak. External calibration nanoparticle sizes tested (10, 40, and cell to minimize these interferences can also be performed by analyzing 70 nm) at different concentrations (11), the optimization of FFF separa- a continuous flow (delivered by a (200, 150, 100, and 50 ppb), are illus- tion conditions must also consider the flow rate–matched peristaltic pump) trated in the colored bar graph, which impact on the ICP-MS system. of known concentrations of metal, shows the integrated ICP-MS re- developing a calibration curve, and sponse signal (peak area) for the FFF Calibration Strategies then converting the intensity of the cross-flow field off (blue); bypass- In addition, the correct implemen- fractogram reading to a concentra- ing the FFF system entirely (brown); tation of both internal and external tion value (13). and with the FFF cross-flow field on calibration standards is critical in (green). The inset graph shows the ICP-MS. Typically, internal standards Recovery raw fractogram data for the different are prepared in 1–5% acid and added One of the inherent limitations with concentrations of the 10-nm particles, to all calibration standards and sam- quantitative applications of FFF–ICP- but as can be seen by the colored bar C.10 Conclusion

This overview of the capabilities of field-flow fractionation coupled with ICP-MS has demonstrated that this hyphenated technique shows a great deal of promise to separate, detect, and quantitate nanoparticles in environmental matrices. FFF is a mature separation technique that has been used for more than 35 years and, when combined with detection techniques such as UV absorbance, has proved to be very capable of separating low concentrations of polydisperse particles. The recent coupling of ICP-MS with FFF has lowered its detection capability by more than three orders of magnitude and has allowed for multielement detection, which is proving absolutely critical when performing environmental risk assessments studies of different engineered nanoparticles. However, there are still challenges to overcome, both in the FFF system and in the ICP- MS system. One of the recognized problem areas is poor recoveries, which can be partly attributed to the nanoparticles sticking to the FFF membrane and not being eluted into the detection system. It is also critical to develop optimized FFF run conditions as well as external and internal ICP-MS calibration routines, which are each extremely important to achieve good accuracy, precision, and recoveries of metal concentrations in fractionated samples. As more studies are published, researchers in this field will have a better understanding of how to minimize these problem areas and it is only a matter of time before FFF–ICP-MS becomes a routine analytical technique for the measurement of ENPs in environmental samples.

163 APPENDIX D - SINGLE PARTICLE ICP-MS STANDARD OPERATING PROCEDURE

Updated January 2012

D.1 General lab practices and recording

For all ICP-MS users

D.1.1 Scheduling and recording

• Sign up in advance on the online-Gmail Lab 260 calendar. • If you are automatically recurring on the schedule, please plan at least two days in advance to remove yourself, if necessary, to allow other users to plan their experiments. If last minute cancelations occur, email group members who use the NexION of its availability.

• In the NexION logbook (now online), record time in/out, matrix, and general run information on the short log. For longer, more detailed information or problems, update the NexION log on the computer desktop with date, problem, and solution.

• Email those responsible for maintenance if a recurring error message persists or you encounter run problems that force you to not continue with planned experiments.

• Keep Daily Performance printouts in the logbook. For quick reference, note any error messages or major changes on the Daily Performance page.

D.1.2 General lab practices for NexION

• Leave the NexION area in clean condition and remove samples from the autosampler before ending your shift.

• Do not leave solution bottles empty for the next person to reﬁll (i.e. 2% nitric acid for rinsing, standardizing solutions, etc.)

D.2 Instrument start up procedure and tuning for sp-ICP-MS

General work instructions, can deviate depending on the particular analysis in question.

164 D.2.1 Preparting to start up

• Turn on the chiller. Chiller temperature should be about 15 oCandrunupto20oC when the instrument is in use. If it is out of this range, contact those responsible for ICP-MS maintenance.

• On the Windows desktop, open the NexION program. Open the appropriate workspace to begin the Daily Tuning procedure. o Note that the dataset should not be above 500 entries, as this will slow the computer and eventually cause problems with loading the program entirely. If the dataset becomes large, please create a new dataset for Daily Tuning.

• Connect the tubing for the peristaltic pump o For Daily Tuning, use the green/orange tubing as sample introduction tubing (auto-sampler sipper line). This will standardize counts/ﬂow rates so we can have a reliable instrument log of the performance of the machine over time o Use green/orange for the internal standard sipper line o / is the waste line and does not need to be changed on a regular basis.

• Before locking clamps over the tubing: from the “plasma” tab on the instrument software.

• Set the auto-sampler to rinse so that the sample introduction line uses the 2% nitric acid solution. Place internal standard line in 2% nitric acid solution to rinse. The speed should be -20 RPM.

• Allow the pump to run until there is smooth ﬂow and no bubbles appear in the sample lines or through the sample introduction system (i.e. autosampler line, internal standard line, at the T of analyte lines, interface with nebulizer, etc.) o Ensure that there is liquid ﬂowing in the waste line TO the waste container under the instrument o Tap out any bubbles that you see when starting the pump. o If bubbles continue to appear, check connection, check for clogs, or try changing tubing if it is aged.

165 D.2.2 Turning on the plasma

• Before the plasma is turned on, ensure there are no error messages and the vacuum pressure is at an acceptable level.

• Check the amount of Argon left in the dewer/cylinders: is there enough to complete your runs?

• Turn on the plasma using the NexION software

D.2.3 Daily Tuning of NexION ICP-MS

• Allow the plasma to warm up for twenty minutes • Keep internal standard line in 2% nitric acid solution throughout Daily Tuning • Place sample introduction line in pre-bottled daily standard solution. • Allow solution to reach the spray chamber before beginning the daily tune. • In the Daily Performance workspace, click on “Smart Tune” button on the top of the NexION software. Open CSM Daily Tune, which should include nebulizer optimization, autolens optimization, and daily performance.

• Start the sequence, the performance sheet will automatically print out after the last item. Ensure that oxides, doubly charged ions, and counts for all elements are within an acceptable range. Store this sheet in the log book, writing any anomalies with start up, error messages, or special notes on this page

• Take the sample introduction tubing out of the standard solution and place back into the autosampler holder. Set to rinse.

D.2.4 Tuning for sp-ICP-MS (Optional)

We have seen increased count intensities by optimizing run conditions for SP-ICP-MS for some particles of interest. Changing operating conditions may not be advised for all particles, as it may increase oxide levels and doubly charged ions.

• Open your workspace and method used for SP-ICP-MS.

166 • Prepare 100 ng/L dissolved sample of your analyte of interest in a large enough volume for several standard runs of SP-ICP-MS (50 mL) • Disconnect the T for internal standard line, only the sample line is used in SP mode.

• There are four main operating conditions that can be optimized: 1. Sample tubing. 2. Sample introduction flow rate. Run a full single particle run, and determine the average intensity. Try varying the flow rate between -5 and -25 RPM. To quickly approximate what the actual sample flow rate the tubing/pump speed combination equates to, refer to: http://robinsonscientific.co.nz/tools.php

• The optimal tubing and flow rate generally only needs to be investigated once. For Ag NP, it was found that black/black tubing at -20 RPM provided the best compromise between intensity and standard deviation. 3. Adjust the nebulizer gas flow rate. With the sipper in the 100 ng/L dissolved sample, run a SP run and watch the real-time data collection window. Manually increase the nebulizer gas flow 0.01 L/min at a time, collecting data for approximately 30 seconds to determine how this changes the signal. More than one run may be necessary. Choose the flow rate with the highest intensity value for your analyte. 4. Adjust the autolens. When the nebulizer gas flow is changed, the auto lens should be re-calibrated. Either 1) use the smart tune or 2) center the autolens on your mass of interest.

• Save conditions and note adjustments to the log book (specify adjustments were for SP mode)

D.2.5 Standard calibration curve

• Make at least ﬁve calibration standards (including a blank) for your analyte of interest from the concentrated stock solutions in dilute acid, 2% HNO3, that bracket the anticipated range of pulse intensities from your nanoparticle samples. Suggested curve will contain 0

(blank), 100 ng/L, 250 ng/L, 500 ng/L , and 1 g/L

167 • Keep these dilute standards no longer than 5 days, to ensure they remain accurate. Preferably, make fresh standards at the beginning of each day.

• If analysis is taking place in a more complex matrix (i.e. anything but DI water), make an additional calibration curve diluting with surfactant, wastewater, cell growth media, etc. These standards should be made fresh daily.

• Collect calibration curve data in real time mode, with the same parameters as your method, and create the calibration curve in excel or similar spreadsheet.

• Note changes to calibration curve slope over time, as this may indicate a “good running day”, or potential issues, such as dirty cones, sample introduction problems etc. which should be addressed before continuing with experiments

D.2.6 Gold nanoparticle eﬃciency calibration

• The eﬃciency calculations should be performed daily • Create a standard curve by diluting concentrated Au dissolved standard in acid. Use at least ﬁve points (0 (blank), 100 ng/L, 250 ng/L, 500 ng/L , and 1 g/L). Dilute standards should be made daily. Run each calibration curve point in real time mode, export data to excel for processing

• Nanoparticle eﬃciency should use 100nm, 100 ng/L Au NP, NanoComposix. • Tripli- cate SP runs should be conducted of a single NP sample. Find the average eﬃciency between these replicates for the day and for the given conditions.

• For more information on how to calculate eﬃciency, consult Pace et. al 2011.

D.3 Analyzing samples in sp-ICP-MS mode

General parameters are outlined below. Changes to analysis can (and should) be made depending on sample concentration, polydispersity of sample, and matrix being analyzed. However, detials below can serve as a general guideline for getting started.

168 D.3.1 Beginning a run

• Run samples after tuning and calibration is complete • Do not use internal standard line for SP runs, only the sample sipper. Set this to rinse while setting up your samples.

• Uncap samples and load onto the auto-sampler rack o You may wish to only upcap an hour’s worth of samples to prevent dust contamination or changes (aggregation, dissolution, etc.) of the NP sample while it is waiting to be analyzed

• Create your sample run list. o At minimum, a check standard should be run between every ten NP samples. This check standard should be a 100 ng/L dissolved sample of your analyte of interest, run in SP mode. This is to ensure no drift of the instrument over time, or other drops in signal such as salting the cones of the ICP-MS.

• Do not let your dataset go over 500 entries, as this will signiﬁcantly slow the software. Periodically check your workspace and make new datasets as necessary.

• To begin your run, click Analyze All. Individual run ﬁles will be sent to the Report Output folder, which you should move before the end of the day

D.3.2 Observations during a run

• Keep the Real Time observation window open at all times. Periodically check on your samples to visually check that you are counting the approximate correct number of pulses you expect (this visualization of frequency of pulses should become easier with time)

• Observe the results for all the check standards oDuringyourruns,calculatetheintensityofthe100ng/Lstandardtoensurethein- strument has not drifted. oIftheresponseforthestandarddrops25%below,itmaybeprudenttostopyourruns and determine the issue o If you decide to continue running, note that the NP mass and sizing calculation including the calibration curve should be adjusted accordingly over time to reﬂect the drop in

169 sensitivity.

• Periodically check the tubing for clogs and ﬂow smoothness o Line connections are likely places for bubbles to form oThejunctionofthesampleintothenebulizeroftenformsbubbles.Thesecancause sporadic results with NP samples. Ensure there are never bubbles in this area.

• If problems arise during a run, be sure to take detailed observations in case you need to contact someone

D.4 Instrument shut down

This is the same as the general procedures for the ICP-MS. Remember to complete the online log for how long the instrument was run, running conditions (including analytes and matrix) and if there were any problems during the days work (even if they were resolved).

D.4.1 Cleaning tubing lines

• Nitric Acid Flush – Set auto-sampler sipper to rinse and allow the pump to run, with plasma on, for at least three minutes

• Water Flush – Remove the auto sampler sipper from its holder arm and place it in a tube of DI water. Allow the pump to run with the plasma on for three mintues

• Air Flush – place the auto-sampler sipper back in its holder, but set to stand-by. Allow the pump to run until the samples lines have no liquid in them

• Peristaltic pump – turn oﬀthe peristaltic pump on the software • Disconnect all tubing from the peristaltic pump.

D.4.2 Shut down and ﬁnal checks

• Turn oﬀthe plasma using the NexION software. • Turn oﬀthe chiller for the NexION. • Open the ICP-MS and check sampler cone.

170 o If it is excessively dirty, clean the cone set for the next user. Note in the log book that the cones were cleaned and that alignment procedures are necessary before they use the ICP. oIfyouuseavery“dirty”matrix,suchaswastewater,cellgrowthmedia,concentrated surfactant, please clean the cones even if they visually do not appear to have signiﬁcant salting

• Make sure to sign out in the log book. • Remove all your data from the “Report Output” folder and put it into your own folder. • Generally leave the ICP-MS area tidy, i.e. tubes, caps, peristaltic pump tubing etc., that will get in the next users way

171 APPENDIX E - SUPPORTING AUTHOR PUBLICATIONS

Published works that were supported in part by the author of this thesis. A.R. Poda, A.J. Bednar. A.J. Kennedy, A. Harmon, M. Hull, D.M. Mitrano, J.F. Ranville, J. Steevens. Characterization of silver nanoparticles using flow-field flow fraction interfaced to inductively coupled plasma mass spectrometry. Journal of Chromatography A. 2001, 1218, 4219-4425 A.J. Bednar, A.R. Poda, D.M. Mirano, A.J. Kennedy, J.F. Ranville, F.H. Crocker, J.A. Steevens Comparison of On-Line Detectors for Filed Flow Fractionation Analysis of Nano- materials. Talanta. 2012.

172

Journal of Chromatography A, 1218 (2011) 4219–4225

Contents lists available at ScienceDirect

Journal of Chromatography A

jou rnal homepage: www.elsevier.com/locate/chroma

Characterization of silver nanoparticles using flow-field flow fractionation

interfaced to inductively coupled plasma mass spectrometry

a, a a a b,c d d a

A.R. Poda ∗, A.J. Bednar , A.J. Kennedy , A. Harmon , M. Hull , D.M. Mitrano , J.F. Ranville , J. Steevens

US Army Engineer Research and Development Center, Environmental Laboratory, 3909 Halls Ferry Rd., Vicksburg, MS 39180, USA

NanoSafe, Inc., 2200 Kraft Drive, Suite 1200 I, Blacksburg, VA 24060, USA

Department of Civil and Environmental Engineering, Institute for Critical Technology and Applied Science (ICTAS), International Center for the Environmental Implications of

Nanotechnology (iCEINT), Virginia Tech, Blacksburg, VA, USA

Department of Chemistry and Geochemistry, Colorado School of Mines, Golden, CO, USA

a r t i c l e i n f o a b s t r a c t

Article history: The ability to detect and identify the physiochemical form of contaminants in the environment is impor-

Available online 23 December 2010

tant for degradation, fate and transport, and toxicity studies. This is particularly true of nanomaterials

that exist as discrete particles rather than dissolved or sorbed contaminant molecules in the environment.

Keywords: Nanoparticles will tend to agglomerate or dissolve, based on solution chemistry, which will drastically

Nanoparticle characterization

affect their environmental properties. The current study investigates the use of ﬁeld ﬂow fractionation

FFF

(FFF) interfaced to inductively coupled plasma-mass spectrometry (ICP-MS) as a sensitive and selective

ICP-MS

method for detection and characterization of silver nanoparticles. Transmission electron microscopy

Comparison techniques

(TEM) is used to verify the morphology and primary particle size and size distribution of precisely engi-

Natural matrices

neered silver nanoparticles. Subsequently, the hydrodynamic size measurements by FFF are compared

to dynamic light scattering (DLS) to verify the accuracy of the size determination. Additionally, the sen-

sitivity of the ICP-MS detector is demonstrated by fractionation of ␮g/L concentrations of mixed silver

nanoparticle standards. The technique has been applied to nanoparticle suspensions prior to use in toxic-

ity studies, and post-exposure biological tissue analysis. Silver nanoparticles extracted from tissues of the

sediment-dwelling, freshwater oligochaete Lumbriculus variegatus increased in size from approximately

31–46 nm, indicating a signiﬁcant change in the nanoparticle characteristics during exposure.

Published by Elsevier B.V.

1. Introduction The bioavailability and potential toxicity of these materials

depend on their dispersion, transport and fate through the different

Nanotechnology is a rapidly developing ﬁeld, attracting sig- media encountered in the environment [3]. Nanomaterial aggre-

niﬁcant investment from government, industry, and academia gation, deposition, and dissolution behaviors factor into transport

[1]. Material applications have already yielded a variety of com- potentials and the subsequent environmental fate and ecotoxi-

mercially available products including cosmetics, antimicrobials, cological impacts of these materials. To quantify the stability of

suntan lotions, paints, stain-resistant clothing and remediation nanoparticles in the environment, the stability of their suspensions

products [2]. This increase in nanomaterial production poses and their tendency to aggregate and interact with other parti-

concerns for environmental safety with the potential release of cles must ﬁrst be determined [4]. Recent reviews have touched

nanomaterials into the environment. Understanding the envi- upon the challenges associated with characterizing nanomateri-

ronmental behavior of nanomaterials in different environmental als in environmental settings stressing the importance of not only

matrices is highly challenging. Due to their small size, nanomate- the material speciﬁc properties (size, shape, and chemical compo-

rials, exhibit physiochemical properties that differ from those of sition), but also the role that surface coatings play in determining

other bulk materials; hence their environmental fate and effects the reactivity, surface attachment and agglomeration properties

are largely unknown. of nanomaterials [5–7]. Therefore, no deﬁnitive conclusions on

nanoparticle fate can be made without sufﬁcient characterization

and a quantitative understanding of nanoparticle properties in rel-

evant environmental matrices.

A range of analytical techniques are available for provid-

∗ Corresponding author. Tel.: +1 601 634 4003; fax: +1 601 634 2742.

E-mail address: [email protected] (A.R. Poda). ing information on concentration and particle size distributions,

0021-9673/$ – see front matter. Published by Elsevier B.V. doi:10.1016/j.chroma.2010.12.076

173

4220 A.R. Poda et al. / J. Chromatogr. A 1218 (2011) 4219–4225

including microscopy approaches [8,9], chromatography [10,11], 80 nm as nominal 20 mg/L suspensions. The 10 nm silver particles

centrifugation [12], laser scattering [13] ﬁltration [14–16], spectro- were stabilized in 2 mM citrate buffer solutions, while the par-

scopic [17,18] and related techniques. Generally, difﬁculties arise ticles ranging in size from 20 to 80 nm were stabilized in 2 mM

due to a lack of analytical tools capable of characterizing and quan- phosphate buffer solutions as described by the manufacturer. Sec-

tifying particles at environmentally relevant concentrations (low ondly, a polyvinylpyrrolidone (PVP)-coated nanosilver, produced

ppb) or in complex environmental matrices that may induce het- by Luna Innovations (Blacksburg, VA, USA) by reduction of AgNO3

erodisperse particle size distributions [19]. To date there have in ethylene glycol (solvent and reducing agent) with PVP added for

been few measurements of manufactured nanoparticles in natural stabilization, was also utilized. The raw reaction product was dia-

waters or soils because of the extreme difﬁculty in detecting those lyzed against water to remove ethylene glycol, unbound PVP and

at environmentally relevant concentrations [20,21] while avoid- Ag that may have been present.

ing the potential interference of natural nanoparticles frequently

present in environmental samples [22]. 2.2. Transmission electron microscopy

It has been reported that the average size and size distribution of

nanoparticles can signiﬁcantly vary when comparing results from Transmission electron micrographs (TEM) of the dried silver

different techniques [23]. Each technique is not without limita- NanoXact silver particles were obtained by subsampling particle

tions and, therefore inaccurate predictions of material properties suspensions (10–20 ␮L) using a Zeiss 10CA TEM (Zeiss, Oberkochen,

and structure can result. Correct size measurements are difﬁcult, Germany) operating at 60 kV and equipped with an AMT Advantage

depending on the tool applied and the media in which the particles GR/HR-B CCD Camera digital imaging system. The longest dimen-

are dispersed. Electron microscopy (EM) and dynamic light scat- sion of all distinct particles ( 202 per material analyzed) that were

≥

tering (DLS) are the most commonly used techniques. Both have observed in each of 10 images was manually analyzed using Image-

advantages and disadvantages [24]. EM gives the most direct infor- Pro Plus software Version 7.0 (Media Cybernetics Inc., Bethesda,

mation on the size distribution and shapes of the primary particles, MD, USA). The scale bar from the TEM images was used to calibrate

however there is concern about artifacts introduced by the sample the software.

preparation step attributed to the lack of a representative sample.

In addition, organic coatings that are not visible in the electron 2.3. Dynamic light scattering

microscope (due to light elements, such as carbon) can lead to dis-

crepancies in sizing, especially when compared to sizing tools that Dynamic light scattering hydrodynamic size of the silver NanoX-

measure the hydrodynamic diameter of particles. Dynamic light act particles was obtained using a 90 Plus/BI-MAS (Brookhaven

scattering (DLS) measures the particle hydrodynamic diameter, but Instruments, Holtsville, NY, USA) instrument applying a 660 nm

limitations include: poor sensitivity at dilute concentrations, non- laser oriented at 90◦ relative to the sample. The software was

selective material detection, inability to distinguish mixtures or optimized to report summary statistics based upon the intensity

complex matrices and little capability to count particles to resolve of light scattered. Two mL sample volumes from each nanosil-

the dominant size in multi-modal particle or aggregate size dis- ver dispersion (10 mg/L nominal) were loaded into glass cuvettes

tributions. With DLS, the presence of a relatively small number (supplied by manufacturer) and summary statistics were obtained

of large aggregates will skew the effective diameter of a distribu- using triplicate 3 min analyses (total analysis time = 9 min). Instru-

tion of predominantly smaller particles toward a larger particle size ment performance was veriﬁed using a polymer reference standard

distribution. known to be 92 3.7 nm (NIST traceable diameter, Duke Scientiﬁc,

For studies of nanoparticles, ﬁeld ﬂow fractionation (FFF) has 3090A, Palo Alto, CA, USA).

been advocated, in particular, a variation called flow field flow frac-

tionation (FFFF) [25,26]. FFF is a family of separation techniques 2.4. FFF–ICP-MS

designed to separate particles based on diffusion coefﬁcient, and

when coupled to an elemental speciﬁc detector, such as ICP-MS, The instrument used for all studies was an F-1000 symmetrical

particle composition as a function of hydrodynamic size can be flow field flow fractionation (FFF) system from Postnova Analytics

determined. This paper describes the development and application (Salt Lake City, UT), interfaced to a PerkinElmer Elan DRC II ICP-MS

of an FFF–ICP-MS method for the characterization of silver nanopar- using a MiraMist pneumatic nebulizer. An Agilent 1100 variable

ticle mixtures. It has been applied to two types of particles known wavelength detector was placed in-line between the FFF and ICP-

to have stable aqueous suspensions. The primary advantages over MS systems to collect UV absorption data, primarily for detection

DLS and EM are demonstrated with element/particle speciﬁc detec- of polystyrene bead size standards. UV absorbance data was not

tion and the ability to size particle mixtures. Furthermore, the collected for the dilute nanosilver particles measured due to the

addition of the sensitive and selective ICP-MS detector allows for limited absorbance of the silver nanoparticles at the low concen-

determination of silver nanoparticles at environmentally relevant trations (␮g/L) studied. The FFF system was equipped with a 10kDa

concentrations (low ppb). Furthermore, the technique is applied regenerated cellulose membrane. The mobile phase consisted of

to biological media to characterize silver nanoparticles before and a 0.025% sodium azide and 0.025% FL-70 surfactant dissolved in

after exposure to the freshwater oligochaete, Lumbriculus variega- deionized water with a resistivity of 18.3 M! cm. Separation of the

tus. particles under investigation was achieved using a channel ﬂow

of 1.0 mL/min and a cross ﬂow of 0.75 mL/min. The channel ﬂow

conditions allow direct connection of the FFF efﬂuent to the ICP-

2. Materials and methods MS nebulizer without a ﬂow splitter. Additional details of the FFF

separation conditions are listed in Table 1.

2.1. Nanosilver particles The ICP-MS was operated in standard mode due to the lack

107 109

of interferences on the 2 isotopes of silver ( Ag and Ag).

Two sources of silver nanoparticles were investigated in the The plasma was operated at 1250 W and the nebulizer ﬂow at

current study. Aqueous NanoXact silver nanoparticle suspensions 0.8 mL/min. Both silver isotopes were monitored for detection and

ranging in size from 10 to 80 nm were supplied by Nanocomposix conﬁrmation, each had an integration dwell time of 500 ms, result-

(San Diego, CA). These particles were generally monodisperse in ing in a data point being collected at a rate of approximately 1 per

size and were acquired in 10 nm increments ranging from 10 to second. The number of readings per replicate was chosen such that

174

A.R. Poda et al. / J. Chromatogr. A 1218 (2011) 4219–4225 4221

Table 1

some loss of particles can be expected due to interactions with

Analytical instrumentation parameters used for separation and characterization of

the FFF membrane, minimal loss occurs to the ICP-MS sample

silver nanoparticles by FFF–ICP-MS.

introduction system, although only about 5–10% of the sample

FFF system Postnova F-1000 Symmetrical is actually aspirated into the plasma, due to known nebulizer

Membrane 10 kDa regenerated cellulose efﬁciencies.

Channel and cross ﬂow 1.0 and 0.75 mL/min,

respectively

Injection volume 50 ␮L 2.7. Biological exposure

Load time 15 s

Relaxation time 3.2 min

A freshwater sediment (Browns Lake, Vicksburg, MS, USA) was

Approximate fractogram time 25 min

(100 nm elution) nominally spiked at 100 mg Ag/kg (measured = 70 mg/kg) with the

UV absorbance detector Agilent 1100 VWD PVP-coated silver nanoparticle (described above) and aged for two

Wavelength monitored 254 nm

weeks. Following the aging period, the freshwater oligochaete L.

Integration time 0.4 s

variegatus was exposed to the sediment for 28 days following stan-

ICP-MS PerkinElmer Elan DRC II dard method guidance [32]. Organisms were removed from the

Plasma power 1250 W

sediment, depurated as speciﬁed by method guidance [32] com-

Nebulizer, spray chamber, and MiraMist, Double Pass Scott,

posited from each experimental replicate, and frozen. Prior to

ﬂow 0.8 mL/min

107 109 FFF analysis, 1.0 g of frozen tissue was added to 10 mL of deion-

Masses monitored Ag, Ag

Dwell time per AMU 500 ms ized water and sonicated for 1 h and then centrifuged to remove

Readings per replicate 1600

biological debris. The supernatant was then analyzed by FFF–

ICP-MS.

data were collected for the entire length of the fractogram, usually

2.8. Statistical analysis

for about 25 min. Table 1 also lists the operating conditions for the

variable wavelength detector and ICP-MS.

Data normality (Kolmogorov–Smirnov’s test), homogeneity

(Levene’s test), and one-way ANOVA and Tukey’s test were deter-

2.5. Calibration

mined at the ˛ = 0.05 level. The results from the different particle

sizing techniques were compared by Pearson product moment cor-

The theory behind FFF separation and sizing is well developed

relation. All analyses were performed using SigmaStat Software

[27–29]. One of the advantages of ﬂow FFF for particle size deter-

(SSPS, Chicago, IL, USA).

mination is that elution time under identical processing conditions

(cross-ﬂow and channel ﬂow settings, carrier solution, etc.) is solely

related to particle size, and follows a linear correlation [30]. In this 3. Results and discussion

paper, ﬂow FFF was used to determine mean particle size as a func-

tion of fractogram elution time using NIST-traceable polystyrene 3.1. Transmission electron microscopy

bead size standards obtained from Postnova Analytics (Art. Nr. z-

PS-POS-000-0 (02:05:1)). A three bead mixture (20, 50, and 100 nm) The TEM images shown in Fig. 2 demonstrate the spherical

was created by dilution of the single-size stock standards (1% solids shape and narrow particle size distribution of the NanoXact sil-

in 15 ml) in deionized water to a ﬁnal concentration of approxi- ver nanoparticles. Manual size analysis of the individual particles

mately 80 mg/L for each particle size. over a range of TEM magniﬁcations is listed in Table 2. The aver-

The fractograms of the polystyrene bead mixture shown in Fig. 1 age primary particle size is very close to the nominal size reported

illustrate this size-distribution of particles is well-resolved. Data by the manufacturer. There is some indication that the 60 nm size

from three replicate injections approximately 24 h apart was col- particles are slightly larger than the reported (67 nm by TEM vs.

lected for Fig. 1 showing excellent reproducibility. The small peak nominal 60 nm), yet overall agreement within 1–5 nm is observed

at about 275 s is the “void peak” representing the material not (except the 60 nm) with similar size standard deviations for each

retained by the ﬁeld. The elution time at maximum absorbance particle size.

was related to the mean particle size of the polystyrene stan-

dards. Retention time from the UV absorbance fractogram of the 3.2. Dynamic light scattering

polystyrene standards was then used to establish a linear response

function of size vs. elution time, as shown in Fig. 1 inset. Typical Hydrodynamic effective diameters of the particles measured by

correlation coefﬁcients from the three point calibration are greater DLS are listed in Table 2. As expected, the hydrodynamic diameter

than 0.9999. This linear response function was used in conjunction is slightly larger than the primary particle size, indicative of a sur-

with the ICP-MS data to determine the mean particle size for the face layer of the stabilizing agent and/or the hydration sphere. The

nanosilver examined in this study. DLS effective diameter ranges reported above may be indicative

of some occurrence of particle aggregation in aqueous suspension.

2.6. Quantitative analysis The autocorrelation function indicated acceptable data capture for

all analyzed particles, the baseline index ranged from 6.0 to 9.5

Quantitative analyte recovery experiments designed to deter- (exceptions: NC10 = 0; NC20 = 1.3) and the data retention ranged

mine the amount of nanoparticle loss to the FFF separation system from 93% to 100%. The count rate (kcps) ranged from 136 to 490,

and ICP-MS sample introduction system were performed. This anal- although substantially lower (16) for nominal 10 nm particles. It

ysis addresses concerns over nanosilver analysis, namely loss of is noteworthy that the comparisons between TEM primary particle

NPs due to adhesion to physical surfaces of the membrane, tub- size, FFF and DLS are very close in this study partially due to the tight

ing and spray chamber. Recoveries of the three silver nanoparticle particle size distribution of the NanoXact materials. In cases where

sizes tested (10, 40, and 70 nm) with the cross ﬂow ﬁeld on, cross much more polydisperse primary particle (or aggregate) sizes are

ﬂow ﬁeld off, and bypassing the FFF entirely yielded recover- found in suspension [35], larger discrepancies between FFF and DLS

ies of 88–98% based on integrated peak areas, which is deemed outputs are likely to be observed due to differences in how particle

excellent recovery for any traditional metals analysis [31]. While sizes are determined and how summary parameters are weighted

175

4222 A.R. Poda et al. / J. Chromatogr. A 1218 (2011) 4219–4225

Fig. 1. Overlay of triplicate FFF-UV fractograms of polystyrene bead calibration standards. FFF separation conditions were 1.0 mL/min channel ﬂow and 0.75 mL/min cross

ﬂow. UV absorbance detection is at 254 nm wavelength. Inset: Linear regression calibration function using 20, 50, and 100 nm polystyrene bead standards. Error bars represent

standard deviation of the triplicate retention times obtained from UV absorbance data at maximum absorbance.

Fig. 2. TEM panels showing, in order from a to h, 10, 20, 30, 40, 50, 60, 70, and 80 nm particles. All particles were imaged at 125k magniﬁcation. Scale bars denote 100 nm.

Table 2

Size determinations by three independent analytical techniques for the NanoXact silver nanoparticles. Ranges are provided in parentheses.

Nominal

10 nm 20 nm 30 nm 40 nm 50 nm 60 nm 70 nm 80 nm

Mean TEM Std. dev. (size 9 1 (2–20) 20 1 (5–33) 32 4 (5–48) 42 4 (16–60) 55 5 (35–74) 67 4 (38–88) 72 3 (16–89) 84 5 (48–112) ± ± ± ± ± ± ± ± ±

range)

DLS effective diameter (size 22 (11–84) 29 (13–90) 41 (15–124) 51 (35–113) 54 (14–121) 67 (32–133) 74 (64–104) 86 (58–142)

range)

FFF–ICP-MS mean 26 31 40 52 61 75 76 86

hydro-dynamic diameter

176

A.R. Poda et al. / J. Chromatogr. A 1218 (2011) 4219–4225 4223

(e.g., DLS intensity analysis is weighted toward the larger particles

in the dispersion).

3.3. FFF separation and sizing

Size fractionation of the NanoXact particles by serial ﬁltration

was examined prior to the FFF separation method development.

Serial ﬁltration is generally an appealing approach because of its

low cost and ease of use. However ﬁltration size resolution is lim-

ited by the available ﬁlter pore sizes. Of more concern, however, is

that separation results were highly variable and dependant not only

on the ﬁlter pore size but also on the composition of the ﬁltration

membrane [33]. Therefore, the need to develop the FFF separation

method was critical.

The overlays shown in Fig. 3 are FF fractograms of the individ-

ual NanoXact particles obtained under the standardized processing

conditions (Table 1) by FFF–ICP-MS. The concentration of silver in

each Injection was 200 ␮g/L. The size data obtained from the FFF

analysis listed in Table 2 agrees well with the DLS size results. In

both cases the size measurements are slightly larger than the TEM

Fig. 3. Individual FFF–ICP-MS fractograms overlain of NanoXact silver nanoparti-

results and are reﬂective of measurement techniques speciﬁc to

cles. Each peak represents 200 ␮g/L total silver as nanosilver particles. FFF separation

the measurement of the hydrodynamic diameter rather than the conditions were 1.0 mL/min channel ﬂow and 0.75 mL/min cross ﬂow with ICP-MS

107

primary particle. Under the ﬂow conditions outlined in Table 1, detection using Ag.

baseline resolution of nanomaterials that vary in size by 10 nm was

not obtained. However, sufﬁcient resolution was achieved for siz-

ing the subject particles based on maximum peak intensity. The signals appears to increase with the nanoparticle size. It is hypoth-

nominal 60 and 70 nm particles graphed in Fig. 3 are in agreement esized that the larger nanoparticles may result in more ‘spikes’ in

with the DLS results with sizes reported in Table 2 that are nearly the ICP-MS signal due to the delivery of larger amounts of silver

the same. This is clearly demonstrated as the fractograms nearly into the plasma/detector system per particle unit. This phenom-

overlap, indicating the similar size of these two nanoparticles. ena is described in the use of ICP-MS as a ‘single particle counter’

To demonstrate the separation and detection potential of the currently being developed by several research groups [34].

FFF–ICP-MS method more clearly, Fig. 4 shows a mixture of the Also shown in Fig. 5 is a 1:5 dilution of the 67 ␮g/L sample, which

nominal 10, 40, and 70 nm silver nanoparticles, each particle yields a total silver concentration for each particle of approximately

present at a total silver concentration of 67 ␮g/L. The particles 13.4 ␮g/L. Although the fractogram peaks are quite small, they are

produced clearly deﬁned peaks under these separation conditions, still clearly deﬁned at this concentration level. To further test the

although baseline resolution was not achieved. The noise of the sensitivity of the ICP-MS detector, the inset in Fig. 5 shows a 1:10

Fig. 4. FFF–ICP-MS fractograms of a mixture of 10, 40, and 70 nm silver particles at 67 and 13.4 ␮g/L each. FFF separation conditions were 1.0 mL/min channel ﬂow and

107

0.75 mL/min cross ﬂow with ICP-MS detection using Ag. Inset: FFF–ICP-MS fractogram of a mixture of 10, 40, and 70 nm silver particles at 6.7 ␮g/L each. FFF separation

107

conditions were 1.0 mL/min channel ﬂow and 0.75 mL/min cross ﬂow with ICP-MS detection using Ag.

177

4224 A.R. Poda et al. / J. Chromatogr. A 1218 (2011) 4219–4225

Fig. 5. FFF–ICP-MS fractogram of a mixture of 10, 40, and 70 nm silver parti-

Fig. 6. FFF–ICP-MS fractogram of the stock PVP-coated silver nanoparticle in

cles at 67 ␮g/L each. FFF separation conditions were 1.0 mL/min channel ﬂow and

deionized water. FFF separation conditions were 1.0 mL/min channel ﬂow and

107

1.1 mL/min cross ﬂow with ICP-MS detection using Ag. 107

0.75 mL/min cross ﬂow with ICP-MS detection using Ag.

dilution of the 67 ␮g/L mixture, yielding a total silver concentration

nominal vs. TEM primary particle size) and hydrodynamic diameter

of 6.7 ␮g/L for each particle size. At this concentration, the ICP-MS

(DLS vs. FFF) had intercepts approximating zero; 0.75 and 2.891,

signal is quite noisy, yet three peaks are still sufﬁciently deﬁned −

respectively.

in the fractogram to characterize the particle size, which suggests

the method is applicable to detection and characterization of silver

3.5. Application to biological exposures

nanoparticles at concentrations less than 10 ␮g/L.

Improved separation of the 10, 40, and 70 nm particle mix-

The above described FFF–ICP-MS technique was subsequently

ture would be ideal, therefore, as shown in Fig. 5, increasing the

applied to characterization of nanoparticles after exposure to a

cross ﬂow to 1.1 mL/min, further increases separation with minimal

biological receptor. Speciﬁcally, the freshwater oligochaete Lum-

reduction in sensitivity and minor peak broadening. Under these

briculus variegates, was exposed to PVP-coated nanosilver spiked

ﬂow conditions, the separation is improved with similar sensitivity,

sediment as described above. In order to determine the effect

and only a minimal sacriﬁce in analytical time.

of environmental exposure on the nanoparticle physiochemical

3.4. Comparison of size measurements form, the stock nanoparticles were ﬁrst analyzed by FFF–ICP-MS,

as shown in Fig. 6. Comparison to the polystyrene size calibration

Pearson correlations of all of the size measurement techniques indicates the PVP-silver particles produce a stable dispersion and

(manufacturer nominal size, TEM, DLS, FFF) resulted in very strong have an average hydrodynamic size of about 31 nm.

and signiﬁcant correlations (r > 0.99; p < 0.001). Comparisons of After the 28-day biological exposure, the tissues were extracted

the slopes of linear regressions of the measurement techniques with deionized water using sonication, with the resultant super-

(Table 3) indicated that the measures of hydrodynamic diameter natant analyzed by FFF–ICP-MS, as shown in Fig. 7. The silver

(DLS and FFF) had the best correlation (slope = 1.000), followed nanoparticles extracted from the tissue have an average hydro-

by nominal size vs. TEM primary particle size (slope = 1.075) and dynamic size of approximately 46 nm, compared to the original

nominal size vs. FFF (slope = 0.904). The smallest slope (0.830) was 31 nm. This size increase may indicate coating of the particles

observed for TEM primary particle size vs. DLS, which is only par- with proteins or other biological molecules. However, because no

tially due to the relatively larger than expected effective diameter data are available on the coating from this analysis, it is possible

of the 10 nm particle (Table 2). This may be primarily explained by that biological mechanisms, or abiotic reactions in the soil expo-

the DLS measuring of the hydrodynamic diameter of the particles, sure medium, have removed the stabilizing PVP coating, resulting

which is by deﬁnition larger than the primary particle size (note in aggregation of the resulting destabilized silver particles. Addi-

that TEM vs. FFF also results in a relatively small slope for the same tional work is underway to discern the exact mechanism, yet these

reasons). This is further supported by y-intercepts being closest to preliminary results demonstrate that exposure of nanoparticles

zero for more similar measurement techniques, i.e., the measure- to environmental media (sediment or biological) can change the

ment techniques that elucidate primary particle size (manufacturer particle physiochemical form and FFF–ICP-MS can be successfully

used to characterize nanomaterials extracted from such complex

media. Unfortunately, recovery of particle vs. total silver loading

Table 3

cannot be determined for the biological sample as the particles

Results of linear regression analysis of the different particle sizing techniques

employed in this investigation. were extracted using a deionized water and sonication method,

2 which has an unknown extraction efﬁciency. Total silver analysis

Methods compared R y-Intercept Slope

is routinely accomplished with aggressive acid digestion meth-

Nominal vs. TEM 0.995 0.75 +1.075

− ods which completely destroy the sample matrix and likely any

Nominal vs. DLS 0.992 +12.607 +0.898

particle-form information, resulting in a difference between par-

Nominal vs. FFF 0.985 +15.214 +0.904

TEM vs. DLS 0.986 +13.458 +0.830 ticle analysis by FFF–ICP-MS and total silver analysis by digestion

TEM vs. FFF 0.992 +15.79 +0.842 and ICP-MS being a result of the extraction procedures rather than

DLS vs. FFF 0.979 +2.891 +1.000

the analytical techniques.

178

A.R. Poda et al. / J. Chromatogr. A 1218 (2011) 4219–4225 4225

USAERDC. Permission was granted by the Chief of Engineers to

publish this information. The ﬁndings of this report are not to be

construed as an ofﬁcial Department of the Army position unless

so designated by other authorized documents. The authors also

thank Frances Hill and Gail Blaustein of the USACE for their editorial comments.

References

[1] B. Karn, T. Kuiken, T.A. Otto, Environ. Health Perspect. 117 (2009) 1823.

[2] Y. Tian, B. Gao, C. Silvera-Batista, K.J. Ziegler, J. Nanopartic. Res. (2010),

doi:10.1007/s11051-010-9912-7.

[3] M.R. Wiesner, G.V. Lowry, K.L. Jones, M.F. Hochella Jr., R.T. DiGiulio, E. Casman,

E.S. Bernhardt, Environ. Sci. Technol. 43 (17) (2009) 6458.

[4] C.E. Mackay, M. Johns, J.H. Salatas, B. Bessinger, M. Perri, Integr. Environ. Assess.

Manage. 2 (2006) 293.

[5] B. Nowack, Environ. Pollut. 150 (1) (2007) 5.

[6] N. Saleh, K. Sirk, Y.-Q. Liu, T. Phenrat, B. Dufour, K. Matyjaszewski, R.D. Tilton,

G.V. Lowry, Environ. Eng. Sci. 24 (2007) 45.

[7] H.F. Lecoanet, J.-Y. Bottero, M.R. Wiesner, Environ. Sci. Technol. 38 (2004) 5164.

[8] D. Mavrocordatos, D. Perret, J. Microsc. 191 (1998) 83.

[9] G.G. Leppard, D. Mavrocordatos, D. Perret, Water Sci. Technol. 50 (2004) 1.

[10] W.A. Scrivens, J.M. Tour, K.E. Creek, L. Pirisi, J. Am. Chem. Soc. 116 (1994) 4517.

[11] Y. Song, V. Jimenez, C. McKinney, R. Donkers, R.W. Murray, Anal. Chem. 75

Fig. 7. FFF–ICP-MS fractogram of a PVP-coated silver nanoparticle extracted from (2003) 5088.

Lumbriculus tissue by sonication in deionized water. FFF separation conditions were [12] D.Y. Lyon, L.K. Adams, J.C. Falkner, P.J.J. Alvarez, Environ. Sci. Technol. 40 (2006)

1.0 mL/min channel ﬂow and 0.75 mL/min cross ﬂow with ICP-MS detection using 4360.

107Ag. [13] K.W. Powers, M. Palazuelos, B.M. Moudgil, S.M. Roberts, Nanotoxicology 1

(2006) 42.

[14] A. Akthakul, A. Hochbaum, F. Stellacci, A.M. Mayes, Adv. Mater. 17 (2005) 532.

[15] K.A. Howell, E.P. Achterberg, A.D. Tappin, P.J. Worsfold, Environ. Chem. 3 (2006)

4. Conclusions 199.

[16] J. Jamison, et al., Nanotechnology 20 (2009) 355702.

[17] C.J. Powell, J. Vac. Sci. Technol. A 8 (1980) 735.

FFF–ICP-MS provides a powerful tool for nanoparticle char-

[18] J.Y. Liu, J. Electron Microsc. 54 (2005) 251.

acterization, particularly metal or metal oxide particles, due to

[19] A.R. Petosa, D.P. Jaisi, I.R. Quevedo, M. Elimelech, N. Turenkiji, Environ. Sci.

the sensitivity and selectivity of the ICP-MS detector. However, Technol. 44 (2010) 6532.

[20] N.S. Wigginton, K.L. Haus, M.F. Hochella, J. Environ. Monit. 9 (2007) 1306.

the particles must create a stable aqueous suspension for FFF

[21] R.F. Domingos, M.A. Baalousha, Y. Ju-nam, M.M. Reid, N. Tufenkji, J.R. Lead, G.G.

analysis. The method has been shown to produce comparable

Leppard, K.J. Wilkinson, Environ. Sci. Technol. 43 (2009) 7277.

results to other established sizing methods, such as DLS, for [22] J.R. Lead, K.J. Wilkinson, Environ. Chem. 3 (2006) 159.

determination of particle hydrodynamic diameter. The results [23] C.Y. Wang, et al., Meas. Sci. Technol. 18 (2007) 487.

[24] A. Bootz, V. Vogel, D. Schubert, J. Kreuter, Eur. J. Pharm. Biopharm. 57 (2004)

are also in qualitative agreement with primary particle sizes

369.

determined by microscopy. The FFF–ICP-MS technique is appli- [25] K. Tiede, A.B.A. Boxall, D. Tiede, S.P. Tear, H. David, J. Lewis, J. Anal. At. Spectrom.

cable to environmentally relevant particle concentrations and 24 (2009) 964.

[26] M. Hassellov, J.W. Readman, J.F. Ranville, K. Tiede, Ecotoxicology 17 (2008) 344.

matrices, allowing detection and characterization of nanoparticles

[27] J. Giddings, J. Chem. Phys. 49 (1968) 81.

extracted from biological receptors after exposure. The ability to

[28] J. Giddings, K. Caldwell, Physical Methods of Chemistry, vol. IIIB, Wiley-

discern particle size and composition at ␮g/L concentrations fur- Interscience, New York, 1989, pp. 120–139.

[29] M. Schimpf, Field-ﬂow Fractionation Handbook, Wiley-Interscience, New York,

ther demonstrates the utility of the method for environmental

2000, pp. 71–79.

applications.

[30] S. Lee, P. Rao, S. Prabhakara, M.H. Moon, Anal. Chem. 68 (1996) 1545.

[31] S. Dubascoux, I. Le Hecho, M. Hassellov, F. Von Der Kammer, M. Potin Gautier,

Acknowledgments G. Lespes, J. Anal. At. Spectrom. 25 (2010) 613.

[32] U.S. Environmental Protection Agency, EPA/600/R-99/064, 2000, Ofﬁce of

Water, Washington, DC, 20460.

The use of trade, product, or ﬁrm names in this report is for [33] J.F. Ranville, Environmental Effects of Nanoparticles and Nanomaterials, Clem-

son University, August 21–26, 2010 (Abstract).

descriptive purposes only and does not imply endorsement by

[34] A.J. Bednar, A.R. Poda, A.J. Kennedy, A. Harmon, J.F. Ranville, D.M. Mitrano, J.A.

the U.S. Government. The tests described and the resulting data

Steevens, 31st Annual SETAC North America Meeting, Portland, OR, November

presented herein, unless otherwise noted, were obtained from 11, 2010.

[35] A.J. Kennedy, M.S. Hull, A.J. Bednar, J.D. Goss, J.A. Steevens, Environ. Sci. Technol.

research conducted under the Environmental Quality Technology

44 (2010) 9571.

Program of the United States Army Corps of Engineers by the

179