Studies on the Elemental Measurement of Aerosols Using

Home , Glow discharge, Spectral line shape

A dissertation submitted to the Division of Research and Advanced Studies of the University of Cincinnati

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

in the Department of Biomedical, Chemical, and Environmental Engineering of the College of Engineering and Applied Science

2016

by Lina Zheng

M.S. Environmental Engineering, Peking University, Beijing, China, 2012 B.S. Environmental Engineering, University of Science and Technology Beijing, China, 2009

Committee Dr. Dionysios D. Dionysiou (chair) Dr. George A. Sorial Dr. Pramod S. Kulkarni Dr. Soryong Chae

ABSTRACT

Chemical characterization of aerosols is essential to understand their health effects in environmental and occupational health studies. There has been a great interest in developing low cost, compact, hand-portable and direct-reading instruments for aerosol monitoring. Microplasma spectroscopy methods, such as laser induced breakdown spectroscopy (LIBS), spark emission spectroscopy (SES) and glow discharge optical emission spectroscopy (GD-OES), have been intensively studied and proved as practical technologies for detecting elemental composition of aerosols. The main focus of this dissertation is to develop a low cost and hand-portable methods for near real-time measurement of elemental concentration of aerosol using microplasma spectroscopy.

The dissertation is divided into three parts. In the first part, a corona-based aerosol microconcentrator is designed for efficient concentrating aerosols to a substrate for subsequent analysis using microscale optical spectroscopies. Performance of this corona microconcentrator is determined experimentally by measuring collection efficiency, wall losses, and particle deposition density. An intrinsic spectroscopic sensitivity is determined for the aerosol microconcentrator using SES. Using this intrinsic sensitivity, it is shown that the corona-based microconcentration method provides the best measurement sensitivity compared to alternative particle collection methods, such as filtration, focused impaction using aerodynamic lens, and spot collection using condensational growth. This method has been demonstrated to be very suitable for compact hand- held analytical instrumentation.

The second part of the dissertation focuses on the development of various methods for aerosol measurement using corona microconentration method coupled with microscale optical

spectroscopies. First, a sensitive method has been developed for real-time measurement of carbonaceous aerosol using LIBS or SES. The measurement system is calibrated and detection limits are determined for total atomic carbon using a carbon emission line at 247.856 nm (C I) for various carbonaceous nanoparticles. Measurements of carbon nanotube aerosol at elevated electrode temperature showed improved selectivity to elemental carbon and compared well with the measurements from thermal optical method (NIOSH Method 5040). Second, a multivariate calibration approach has been developed for simultaneous measurement of multiple elements of aerosol using SES. Partial least squares (PLS) regression is performed to construct the relationship between spectra and elemental mass. The PLS model is applied to predict elemental concentration of welding aerosols, showing good measurement accuracy. Third, the corona aerosol microcoentrator is coupled with glow discharge optical emission spectroscopy (GD-OES), forming a low cost method for semi-continuous analysis of aerosol, compared with methods using LIBS and SES. The spectral features and signal stability of this GD-OES aerosol analysis system are investigated. Analytical performance, such as limits of detection and uncertainty, are determined for several elements, and compared with other aerosol analysis methods that employ microplasma spectroscopies.

In the third part, spatial and temporal dynamics of spark microplasma used for aerosol measurement is systemically investigated. The spatial and temporal behavior of carbon emission in spark plasma from carbon black particles is investigated. The plasma parameters, such as excitation temperature and electron density are evaluated using spectroscopic means, and their dependences on interelectrode distance, delay time and pulse energy are presented.

iii

ACKNOWLEDGEMENTS

I would like to express my sincere appreciation to my research advisors, Prof. Dionysios

D. Dionysiou and Dr. Pramod Kulkarni, for their support, encouragement and academic guidance during my study for Ph.D. degree. I would like to thank my committee members, Prof. George A.

Sorial and Prof. Soryong Chae for reviewing my proposal and my dissertation and providing useful comments.

It was a wonderful experience to study in Environmental Science and Engineering Program at University of Cincinnati. I thank the professors whose courses I have taken, Prof. George Sorial,

Prof. Drew McAvoy, Prof. Mingming Lu, and Prof. Tim Keener, etc. I have acquired rich knowledge from their courses. I thank CEAS graduate school for providing me the University

Graduate Scholarship.

It has been a great opportunity for me to work as a worksite student in NIOSH (National

Institute for Occupational Safety and Heath), where I conducted my research projects. I have learned lots of practical skills that are required for a scientist and an engineer. Working with Dr.

Kulkarni was a wonderful experience. He is an outstanding scientist. With his guidance, our research projects went very well. I also appreciate the help from other scientists in NIOSH, Dr.

Eileen M. Birch, Dr. Chaolong Qi, and Gregory Deye, and other Ph.D. students in NIOSH, Huayan

Liang and Shijun Wei.

Special thanks to Dr. Dion’s group members: Xiaodi Duan, Yang He, Bangxing Ren, Jiong

Gao, Ying Huang, Vasileia Vogiazi, Nadeesha Koralegedara, for their help and company when I live alone in Cincinnati.

I am very grateful for all the support from my family. My parents and parents-in-law helped me take care of my little baby in the past year, so that I can focus on my research. Though my

husband and me live in two cities, his love and care around me all the time. Finally, I would like to thank our little boy for making my life more joyful.

DEDICATION

Dedicated to my Collaborators, Dear Friends and Family.

TABLE OF CONTENTS

ABSTRACT ...... i ACKNOWLEDGEMENTS ...... ivv DEDICATION ...... vii TABLE OF CONTENTS ...... vii LIST OF FIGURES ...... x LIST OF TABLES ...... xv LIST OF SYMBOLS AND ABBREVIATIONS ...... xvii

CHAPTER 1 Introduction ...... 1 1.1. Background and motivation ...... 2 1.2 Literature review ...... 7 1.2.1 Aerosol chemical measurement methods ...... 7 1.2.2 Microplasma spectroscopy for aerosol analysis ...... 11 1.3 Dissertation outline ...... 17 1.4 References ...... 18

CHAPTER 2 Aerosol Microconcentrator for Analysis Using Microscale Optical Spectroscopies ...... 25 2.1. Introduction ...... 27 2.2. Methods...... 29 2.2.1. Design of aerosol microconcentrator ...... 29 2.2.2. Experimental setup ...... 30

2.2.3. Analysis of (NH4)2SO4 aerosol ...... 32 2.3. Results and discussion ...... 33 2.3.1. Collection efficiency of the CAM ...... 33 2.3.2. Effects of particle size on collection efficiency ...... 38 2.3.3. Deposition uniformity and side wall losses ...... 40 2.3.4. Effects of spark ablation on collection efficiency ...... 41 2.3.5. Analytical sensitivity ...... 43

vii

2.4. Conclusions ...... 51 2.5. References ...... 51

CHAPTER 3 Near Real-Time Measurement of Carbonaceous Aerosol Using Microplasma Spectroscopy: Application to Measurement of Carbon Nanomaterials ...... 54 3.1. Introduction ...... 56 3.2. Methods...... 58 3.2.1. Experimental setup ...... 58 3.2.2. Calibration method ...... 63 3.3. Results and discussion ...... 64 3.3.1. Sensitivity of different carbon emission lines ...... 64 3.3.2. Plasma characteristics ...... 67 3.3.3. System calibration ...... 69 3.3.4. Limits of Detection ...... 74 3.3.5. CNT Measurement Comparison ...... 75 3.3.6. SES measurements at elevated electrode temperature ...... 77 3.4. Conclusions ...... 84 3.5. References ...... 85

CHAPTER 4 Multivariate Calibration for Measurement of Aerosol Elemental Concentration Using Microplasma Spectroscopy ...... 90 4.1. Introduction ...... 92 4.2. Methods...... 94 4.2.1. Instrumentation ...... 94 4.2.2. Aerosol samples ...... 95 4.2.3. Experimental procedure ...... 96 4.2.4. PLS regression model ...... 96 4.3. Results and discussion ...... 99 4.3.1. Wavelength variables selection ...... 99 4.3.2. Multivariate calibration using PLS ...... 101 4.3.3. Limits of detection ...... 105 4.3.4 Application on welding aerosol measurement ...... 107

viii

4.4. Conclusions ...... 109 4.5. References ...... 109

CHAPTER 5 Method for Rapid Elemental Analysis of Airborne Particles Using Atmospheric Glow Discharge Optical Emission Spectroscopy ...... 112 5.1. Introduction ...... 114 5.2. Methods...... 116 5.2.1. Experimental setup and materials ...... 116 5.2.2. Plasma diagnostics ...... 119 5.2.3. Calibration procedure ...... 121 5.3. Results and discussion ...... 122 5.3.1. Plasma gas temperature and electron density ...... 122 5.3.2. Spatial and Temporal Distribution of Emission Signal ...... 124 5.3.3. Reproducibility of the plasma characteristics ...... 128 5.3.4. Analytical performance ...... 131 5.4. Conclusions ...... 135 5.5. References ...... 135

CHAPTER 6 Spatial and Temporal Dynamics of Pulsed Spark Microplasma Used for Aerosol Analysis ...... 140 6.1. Introduction ...... 142 6.2. Experimental methods ...... 143 6.3. Results and discussion ...... 145 6.3.1. Continuum emission from spark plasma ...... 145 6.3.2. Spatial and temporal evolution of spark emission ...... 147 6.3.3. Plasma imaging ...... 152 6.3.4. Excitation temperature and electron density ...... 155 6.4. Conclusions ...... 161 6.5. References ...... 162

CHAPTER 7 Conclusions and Recommendations for Future Work ...... 165

APPENDIX Supplemental Information for Chapter 2 ...... 170

LIST OF FIGURES

Figure 2.1. Cross-section view of the corona aerosol microconcentrator...... 30

Figure 2.2. Schematic diagram of experimental setup in this study...... 31

Figure 2.3. Collection efficiency as a function of corona current at a volumetric flow rate of 1.5 L min-1 for different collection electrodes. Distance between the corona and the collection electrode was 5 mm. The particle size was 100 nm in diameter. The lines represent the best fit to the experimental data...... 35

Figure 2.4. Variation of (a) collection efficiency, and (b) temporal concentration factor as a function of aerosol flow rate for various electrode diameters. The corona current was 5 µA. The particle size was 100 nm in diameter. The error bars represent standard deviation calculated from three replicate measurements. The lines represent best fits to the experimental data...... 37

Figure 2.5. Comparison of collection efficiency obtained from counting particles by CPC and by determining mass by IC analysis. The corona current was 5 µA. The error bars represent standard deviation calculated from three replicate measurements. The lines represent best fits to the experimental data...... 38

Figure 2.6. Variation of collection efficiency as a function of particle size for different electrode diameters and flow rates at a corona current of 5 µA. The error bars represent standard deviation calculated from three replicate measurements. The lines represent best fits to the experimental data...... 40

Figure 2.7. Variation of collection efficiency as a function of number of sparks (100 nm particles and 1.5 L min-1, 500 µm electrode, 5 µA current) and SEM images of collection electrode surface after (a) 0, (b) 1000, and (c) 2000 sparks. The error bars represent standard deviation calculated from three replicate measurements. The dash line represents the mean of collection efficiency over different number of sparks. The grey shaded region represents area within the 95% confidence interval...... 43

Figure 2.8. (a) Calibration curves for Cr using different electrode diameters (Cr I 360.534 nm emission line was used in calibration); (b) The signal intensity of Cr from spark

emission spectroscopy as a function of particle mass per area. The lines represent linear fits to the experimental data...... 45

Figure 2.9. Measurement sensitivity ( Sc ) of Cr I using CAM at different air flow rates and different electrode diameters. The error bars represent standard deviation. The lines represent best fits to the experimental data...... 48

Figure 2.10. Comparison of the measurement sensitivity ( ) of CAM with alternative concentration or collection methods employing condensational growth droplet concentrator, aerodynamic lens, and filtration...... 50

Figure 3.1. Schematic diagram of the experimental setup used in this work...... 59

Figure 3.2. Schematic diagram of the aerosol preconcentration system (not to scale)...... 61

Figure 3.3. Carbon emission lines identified in this work using spark emission spectroscopy. The spectra correspond to different particulate carbon mass loadings on the collection electrode...... 66

Figure 3.4(a-f). Changes in carbon signal intensity with carbon mass loading using sucrose as analyte, for different emission lines (x-axis represents carbon mass loaded on the collection electrode, ng; y-axis represents carbon signal intensity, arbitrary units [a. u.])...... 66

Figure 3.5. (a) Spatial and temporal signal intensity of carbon emission line at 247.856 nm from spark plasma, (b) Variation of carbon signal intensity as a function of interelectrode distance...... 68

Figure 3.6. (a) Calibration curves for sucrose aerosols with different particle sizes: 50 nm, 100 nm, 150 nm, 200 nm, and 300 nm, (b) Calibration curves for sucrose constructed by changing the collection time and particle number concentration...... 70

Figure 3.7. Calibration curves for sucrose, EDTA, caffeine, sodium carbonate, carbon black, and CNT constructed by SES...... 73

Figure 3.8. Calibration curves for carbon black by LIBS and SES using the collection electrode with a diameter of 500 µm...... 74

Figure 3.9. Comparison of carbon nanotube concentrations measured from SES,

Aethalometer, and NIOSH Method 5040...... 77

Figure 3.10. Change in carbon signal intensity as a function of electrode temperature for various organic (a) and inorganic (b) carbonaceous materials...... 79

Figure 3.11. Comparison of TAC obtained in this work with EC from NIOSH Method 5040. The pie charts show relative compositions of liquid solutions used to generate the test aerosol. The actual liquid concentrations of individual components in these solutions are shown in Table 3.5...... 83

Figure 4.1 A schematic diagram of aerosol spark emission spectroscopy instrumentation...... 94

Figure 4.2. A schematic diagram of the experimental procedure for multivariate calibration...... 97

Figure 4.3. Regression coefficients of PLS-1 model for chromium...... 100

Figure 4.4 Predicted vs measured elemental mass in the collected particle samples for PLS- 1-45 models...... 105

Figure 4.5 Comparison of predicted elemental concentration obtained by ASES coupled with multivariate calibration and referenced elemental concentration obtained by conventional filter method on measurement of welding aerosol...... 108

Figure 5.1. The schematic diagram of the experimental setup...... 119

Figure 5.2. The variation of gas temperature (a) and electron density (b) as a function of interelectrode distance. (The cathode surface is at 0 mm and the anode surface is at 4 mm) ...... 124

Figure 5.3. (a) and (b) Spatially resolved background GD-OES spectra acquired along the axis of the glow discharge plasma in the absence of particle deposition on the collection electrode, (c) and (d) the background spectrum acquired at the collection electrode tip (d = 0 mm) corresponding to (a) and (b) respectively...... 126

Figure 5.5. Space-resolved rf-GD-OES spectra acquired along the axis of the glow xii

discharge plasma in the presence of sucrose particles on the collection electrode...... 128

Figure 5.6. Changes in temperature on the electrode with time when glow discharge was on or off (the glow discharge is on at t=1 s and off at t=120 s)...... 130

Figure 5.7. Variation of Ar I and Pt I signal intensity as a function of time in glow discharge...... 130

Figure 5.9. Calibration curves for C, Cd, Na and Mn obtained using our GD-OES system...... 134

Figure 6.1. Schematic diagram of the experimental setup...... 145

Figure 6.2. Spark emission spectra at a delay time 0, 1, and 5 µs, with a gate width of 1 µs...... 147

Figure 6.3. Space-resolved carbon emission lines from carbon black at delay time of 5 µs with a gate width of 1 µs. (a) C I 247.856 nm and C II 251.206 nm; (b) C II 283.671 nm; (c) C II 426.726 nm. (The cathode is at 0 mm, and the anode is at 5 mm)...... 149

Figure 6.4. The temporal evolution of carbon emission generated by spark discharge: (a) C I 247.856 nm, (b) C II 251.206 nm, (c) C II 283.671 nm, and (d) C II 426.726 nm...... 151

Figure 6.5. ICCD Images of spark plasma generated on carbon black sample at various delay time. (The cathode is at 0 mm, and the anode is at 5 mm) ...... 154

Figure 6.6. Images showing temporal and spatial development of the carbon emission from spark plasma transmitted by a 250 nm bandpass filter. (The cathode is at 0 mm, and the anode is at 5 mm) ...... 154

Figure 6.7. The variation of excitation temperature and electron density as a function of interelectrode distance...... 158

Figure 6.8. The variation of excitation temperature and electron density as a function of delay time...... 159

Figure 6.9. The variation of excitation temperature and electron density as a function of energy (at a distance of 1 mm and a delay time of 5 µs)...... 160

Figure S-1. Corona current as a function of voltage applied between two electrodes for different electrode diameters (The distance between two electrodes is 5 mm)...... 170

xiii

Figure S-2. The estimate particle mass error for various count median diameter and geometric standard deviation by assuming a constant collection efficiency...... 171

xiv

LIST OF TABLES

Table 2.1. Design and operating parameters of alternative particle collection methods for submicrometer size range...... 50

Table 3.1. Description of carbon-containing materials used for calibration ...... 60

Table 3.2. Experimental parameters used in the SES systems ...... 62

Table 3.3. LODs in terms of mass and air concentration obtained by SES ...... 72

Table 3.4. Comparison of different OC/EC thermal analysis methods ...... 81

Table 3.5. Composition and liquid concentration of solutions A, B, C, and D used to produce test aerosols...... 84

Table 4.1. Elemental concentration in the prepared solution...... 95

Table 4.2. Elemental mass in the collected particle samples on the tip of electrode determined by ICP-MS...... 98

Table 4.3. Selected elemental emission lines for each element investigated in this study...... 100

Table 4.4. Comparison of relative RMSECV and R2 from PLS-1 and PLS-2 models with different numbers of variables...... 104

Table 4.5. Limits of detection of aerosol spark emission spectroscopy ...... 107

Table 5.1. Materials used to generate calibration aerosol for elements studies in this work ...... 118

Table 5.2. Comparison of detection limits in this work with other aerosol measurement methods using microplasma spectroscopy ...... 134

Table 6.1. C I and C II spectroscopic data ...... 151

LIST OF SYMBOLS AND ABBREVIATIONS

Abbreviations ASES Aerosol Spark Emission Spectrometer ATOFMS Aerosol Time-of-Flight Mass Spectrometry BC Black Carbon CAM Corona Aerosol Microconcentrator CNF Carbon Nanofibers CNT Carbon Nanotubes CPC Condensation Particle Counter DC Direct Current DI Deionized DMA Differential Mobility Analyzer DSF Dynamic Shape Factor EC Elemental Carbon EDTA Ethylenediaminetetraacetic acid ELCAD Electrolyte cathode discharge ESP Electrostatic Precipitator GC-MS Gas Chromatography Mass Spectroscopy GD-MS Glow Discharge Mass Spectroscopy GD-OES Glow Discharge Optical Emission Spectroscopy HEPA High-Efficiency Particulate Air HV High voltage IC Ion Chromatography ICP-AES Inductively Coupled Plasma Atomic Emission Spectroscopy ICP-MS Inductively Coupled Plasma Mass Spectrometry INAA Instrumental Neutron Activation Analysis LIBS Laser Induced Breakdown Spectroscopy

xvi

LOD Limit of Detection LS-APGD Liquid Sampling-Atmospheric Pressure Glow Discharge LTE Local Thermodynamic Equilibrium MFC Mass Flow Controller MLR Multiple Linear Regression MP Microwave Plasma NIOSH National Institute for Occupational Safety and Health OC Organic Carbon PB/HC-OES Particle Beam/Hollow Cathode-Optical Emission Spectroscopy PEEK Polyether ether ketone PID Proportional Integral Derivative PIXE Proton-induced X-ray Emission PLS Partial Least Squares PTFE Polytetrafluoroethylene PVC Polyvinyl Chloride REL Recommended Exposure Limit RF Radio Frequency RMSECV Root Mean Square Error of Cross Validation RMSEP Root Mean Square Error of Prediction RSD Relative Standard Deviation SCGD Solution-Cathode Glow Discharge SEM Scanning Electron Microscopy SES Spark Emission Spectroscopy SI Supplemental Information SIBS Spark Induced Breakdown Spectroscopy SWNT Single-walled Carbon Nanotubes TAC Total atomic carbon TOR Thermal Optical Reflectance

xvii

TOT Thermal Optical Transmission UV-VIS Ultraviolet-visible XRF X-ray Fluorescence

Symbols

∆휆푆 Stark broadening

∆휆퐷 Doppler broadening

∆휆퐺 Gaussian component

∆휆퐼 Instrument broadening

∆휆퐿 Lorentzian component

∆휆푊 Van der waals broadening

퐴푘𝑖,푍 Transition probability

퐶푐(푑푚) Slip correction factor for the electrical mobility diameter

퐶푐(푑푣푒) Slip correction factor for the volume equivalent diameter

퐶𝑖푛 Incoming aerosol particle mass concentration

퐶푡 Temporal concentration factor

퐸푘,푍 Energy

퐼푍 Line intensity

퐼푡표푡 Total emission signal

푁푒 Electron density 퐻푉 푁표푢푡 Particle number concentration downstream of the CAM with the presence of the electric field across the electrodes

푉=0 푁표푢푡 Particle number concentration downstream of the CAM without the presence of the electric field across the electrodes

푃푍 Partition function of the species in ionization stage Z

푄푓 Aerosol volumetric flow rate 푆∗ Slope of the calibration curve in Figure 2.8(b); intrinsic sensitivity

푆푐 Measurement sensitivity in terms of air concentration

xviii

푆푚 Mass-based sensitivity

푇푒 Electron temperature

푇푔 Gas temperature

푑푒 Electrode diameter

푑푚 Electrical mobility diameter

푑푣푒 Volume equivalent diameter of particle

푔푘,푍 Degeneracy of the upper energy level

푘퐵 Boltzmann constant

푚푒푙푒푐푡푟표푑푒 Particulate mass deposited on the electrode

푚푓𝑖푙푡푒푟 Particulate mass collected on the filter

푚푝 Mass of particles collected on the electrode

푛푒 Electron density

푡푐 Particle collection time

휂푐 Count-based particle collection efficiency

휂푚 Mass-based collection efficiency

휆푘𝑖,푍 Transition line wavelength

𝜌푝 Particle material density ∆퐸 The largest observed transition energy for which the condition holds c Speed of the light h Planck constant I Lower energy level of the species in ionization stage Z K Upper energy level of the species in ionization stage Z L Characteristic length of the plasma M Atomic mass P Gas pressure 퐴 Flat area of the collection electrode 퐼 Signal intensity 퐼(푡) Time-dependent signal intensity xix

푄 Flow rate 푇 Temperature 훿휒 Estimate of the noise level in the spectral data 휂 Collection efficiency 휆 Wavelength 𝜎 Standard deviation of the blank at the selected spectral region 휒 Dynamic shape factor

CHAPTER 1

Introduction

1.1. Background and motivation

Anthropogenic activities, such as combustion of fossil fuels, industrial production, and construction, inevitably lead to release of submicron particles and nanoparticles into the air.

Exposure to such airborne particles could cause respiratory diseases and cancers. Measurement of chemical composition and concentration of aerosol is of great significance for the prevention of exposure and protection of workers’ health.

Most existing methods for determining chemical composition of aerosol are based on filter collection, followed by offline chemical analyses using analytical methods such as Inductively

Coupled Plasma-Mass Spectrometry (ICP-MS) and Atomic Emission Spectroscopy (ICP-AES).

Although filter-based methods are in common use for routine monitoring in both occupational and environmental settings, they have several drawbacks. These filter-based methods are time-and labor- consuming and cannot capture high intensity, short-term exposures that are typically in workplaces. To address these drawbacks, real-time instruments or sensors for chemical analysis of aerosol are needed. Aerosol mass spectrometers have been developed for real-time measurement of chemical composition of aerosols (DeCarlo et al. 2006; Drewnick et al. 2005; Jayne et al. 2000;

Onasch et al. 2012). Though aerosol mass spectrometers have been widely used in atmospheric chemistry studies (Aiken et al. 2008; Canagaratna et al. 2007; Gross et al. 2006), they have not been practical for routine personal exposure measurement in workplace due to their high cost and lack of portability. Currently, there is a growing interest in developing low cost, compact, hand- portable or personal instruments for chemical analysis of aerosol in real-time.

The main focus of this dissertation is to develop a low cost and hand-portable instrument for near real-time measurement of multi-elemental concentration of aerosol using microplasma spectroscopy. This research work can be divided into three parts. The first part deals with the

design, characterization, and optimization of a corona-based aerosol microconcentration method designed for portable analytical instrumentation. The second part focuses on the development of various methods for aerosol measurement using corona microconentration method coupled with microscale optical spectroscopies. The third part deals with the spatial and temporal dynamics of pulsed spark plasma used for aerosol analysis.

Concentration of aerosols, either to a substrate or in a suspended state, is an important step prior to aerosol characterization using instrumentation for improvement of detection sensitivity.

Methods based on collection of aerosols on filters to obtain a time-integrated concentration have been extensively used for routine aerosol chemical measurement through various instrumentations

(Raynor et al. 2011; Spurny 1999). Methods based on virtual impaction have been used to increase the number density (per unit volume) of particles in the suspended state for subsequent instrument measurement (Keskinen et al. 1987; Wu et al. 1989). Methods based on spatial concentration of aerosols, for instance, by creating a focused particle beam, have been well studied and widely used for aerosol analysis by mass spectrometry. The focused particle beam is obtained by expanding the aerosol flow into a low pressure chamber, either through capillary (Murphy and Sears 1964), converging nozzles (Dahneke and Flachsbart 1972) or aerodynamic lenses (Liu et al. 1995). The focused particle beam techniques have also been used for aerosol analysis using microscale spectrometric methods, such as laser-induced break down spectroscopy (LIBS), that probe miniscule samples (Park et al. 2009). However, most of the aerosol concentrating techniques discussed above require relatively large pumping capacity and bulky hardware, and therefore are not suitable for portable and compact instrumentation. Electrostatic collection offers an alternative sampling method that enables use of high flow rates with minimal pressure drop and is suitable for compact field instrumentation. Electrostatic collection has been mainly used for collecting samples

for microscopic analysis (Fierz et al. 2007; Miller et al. 2010). Recently, a concept of aerosol analysis based on electrostatic collection followed by microplasma emission spectroscopy has been proposed (Diwakar and Kulkarni 2012). In order to develop a hand-portable aerosol measurement instrument using the above concept, a corona aerosol microconcentrator is designed and its performance is evaluated in the first part of this dissertation. A one-step corona-induced charging and electrostatic deposition approach has been used for concentrating aerosol on the tip of a microelectrode. The optimized operation condition is obtained for aerosol analysis using spark emission spectroscopy (SES).

While the first part deals with the development of an aerosol microconcentrator (Chapter

2), the second part focuses on exploring various methods for aerosol measurement by coupling this microconcentrator with microplasma-based emission spectroscopies. The second part of this dissertation includes developing a method for measurement of carbonaceous aerosol (Chapter 3), developing a multivariate calibration approach for analysis of multiple elements of aerosol

(Chapter 4), and developing a method for aerosol elemental analysis using glow discharge optical emission spectroscopy (GD-OES) (Chapter 5).

The increasing production and widespread application of carbon nanomaterials, such as carbon nanotubes (CNT) and carbon nanofibers (CNF) could have adverse health effects on workers who are routinely exposed to these materials (NIOSH 2013). Attempts have been made to measure airborne carbon in real time using LIBS (Lee and Yoh 2012; Vors and Salmon 2006). Vors and Salmon (2006) reported a limit of detection (LOD) of 60 µg/m3 of carbon using test aerosols of glucose and NaHCO3 for their laboratory benchtop LIBS system. Lee and Yoh (2012) reported an LOD of 150.4 µg/m3 at a low flow rate and 240.4 µg/m3 at a high sample flow rate in their LIBS system. They used 15 µm test particles and analyzed them in-situ in the aerosol stream flowing

past the LIBS microplasma. Though these and other studies have successfully measured airborne carbon, the detection limits achieved were not adequate for practical aerosol measurement applications, especially for workplace monitoring of CNT aerosols. NIOSH has proposed a recommended exposure limit (REL) for CNT/CNF of 1 µg/m3, as a respirable elemental carbon

(EC), 8-hour time weighted average (NIOSH 2013). Sensitive near real-time techniques that can provide accurate measurements with much lower detection limits are needed. This motivates exploring the applicability of aerosol microconcentration followed by LIBS and SES for near real- time measurement of carbonaceous aerosol. Measurement of total atomic carbon at elevated temperature is used for selectively measurement of elemental carbon (EC), which is a marker of carbon nanomaterials.

For quantitative analysis of aerosol using microplasma spectroscopy, univariate calibration approach has been used to construct calibration curves between a signal intensity and elemental concentration or mass in previous works (Diwakar and Kulkarni 2012; Hunter et al. 2000a; Khalaji et al. 2012; Marcus et al. 1999; Martin et al. 1999). However, in most of spectroscopic methods, quantitatively analysis of their spectra remains challenging due to sample matrix effects (physical and chemical matrix effects) and spectral interferences (Gemperline 2006). Because the spectral emission intensity from one element is affected by the chemical composition of the sample, the conventionally used univariate calibration method would produce large uncertainty when the unknown samples have different chemical matrix from the calibration standards (Laville et al.

2007; Tripathi et al. 2009; Yao et al. 2012). To minimize sample matrix effects, matrix-matched standards or internal standards methods are often used in univariate calibration. However, preparation and certification of matrix-matched standards are tedious for a given analytical requirement. This motivates developing a multivariate calibration approach for simultaneous

analysis of multiple elements in aerosol. Orthogonal experimental samples are designed and partial least squares (PLS) algorithm is employed to build multivariate calibration models.

Approaches based on aerosol preconcentration coupled with LIBS and SES have been developed for measurement of elemental concentration of aerosol in previous works (Diwakar and

Kulkarni 2012). This technique involved deposition of charged particles on the tip of the cathode and excitation of these particles by a laser pulse or a high voltage pulse. Low detection limits (~pg to ~ng) and high time resolution (~minutes) have been achieved by both the two approaches.

However, these methods employ excitation sources that utilize bulky and expensive hardware.

Glow discharge, as an excitation source for the elemental determination, has certain unique advantages such as low cost, low temperature, low operation power, and analytical versatility

(Marcus 2003). Marcus et al. (1999) conducted an elemental analysis of particulate matter (NIST

SRM 1648 urban particulate matter) by a direct injection of particles into a low-pressure glow discharge plasma, and obtained detection limits on the order of tens of nanograms. A lower detection limit may be expected if particles are concentrated on the cathode before GD-OES analysis. In this dissertation, an attempt is made to use glow discharge as an alternative plasma source to analyze the concentrated aerosol sample using the corona microconcentrator designed in the first part (Chapter 2).

Microplasma spectroscopies have been proven to be potential techniques for developing novel analytical instruments. In order to improve their performance in analytical instrumentations, the physics and fundamental aspects of microplasma are needed to be investigated. There has been extensive research on laser-induced plasma diagnostics, such as the plasma size, the emission evolution and distribution, the calculation of excitation temperature and electron density (Aragon and Aguilera 1997; Baig et al. 2012; El Sherbini and Al Aamer 2012; Salik et al. 2013). There are

only several studies on spark plasma diagnostics in literature. Walters has described the process of spark discharge and the emission topography of the spark discharge (Walters 1977; Walters and

Goldstein 1984). A two dimensional image of atomic emission emitted from a spark plasma has been obtained using an imaging spectrograph equipped with a charge-coupled device detector

(Ramli and Wagatsuma 2010). Bye and Scheeline (1993) have calculated the electron temperature

(~15,000 K) and electron density (~2 × 1017 cm-3) of a spark discharge plasma using Saha-

Boltzmann equation. A systematic investigation of spark plasma used in aerosol measurement system has been presented in the third part of this dissertation (Chapter 6), to gain insight into the plasma physical characteristics.

1.2 Literature review

A brief review of the existing methods for chemical analysis of aerosols is present below.

First, the widely used methods, both the off-line methods and the continuous methods, are reviewed. Second, studies on real-time measurement of elemental concentration of aerosol using microplasma spectroscopy (i.e. LIBS, SES, and GD-OES) are presented.

1.2.1 Aerosol chemical measurement methods

Classical methods for quantitative measurement of aerosol chemical composition involve filter collection over several hours, followed by chemical analysis in the laboratory. Different laboratory methods are employed for different analytes, for examples, ion chromatography (IC) for anions and cation species, thermal optical reflectance (TOR) and thermal optical transmission

(TOT) for particulate carbon, gas chromatography mass spectroscopy (GC-MS) for organic aerosol speciation, and inductively coupled plasma mass spectroscopy (ICP-MS) and inductively coupled plasma atomic emission spectroscopy (ICP-AES) for metals. The particulate matter collected on the filters usually requires preprocess before analysis, such as dissolution or digestion. There are 7

also some methods for direct elemental analysis of aerosols on filters, such as X-ray fluorescence analysis (XRF), proton-induced X-ray emission (PIXE), and instrumental neutron activation analysis (INAA). Besides the filter collection method, several other methods for sampling particulate matter are in use. Impactors have been used for collection of size-resolved particulate samples. The diffusion denuder method has been used for sampling semivolatile compounds that can evaporate from the deposited particles during sampling (Possanzini et al. 1983).

Although off-line filter-based measurement methods are in common use for routine aerosol monitoring in both occupational and environmental settings, they have several drawbacks. First, they are time-averaged over several hours or days and cannot capture high-intensity, short-term exposures that are typical in manufacturing workplaces. Second, the turn-around time for analysis is typically several days, hence, immediate interpretations are generally not available for on-site decision making to avoid high concentration hazards exposure. Third, measurement of aerosol composition is influenced by a variety of sampling artifacts. To address these drawbacks, methods for the continuous chemical characterization of aerosols have been developed and applied to ambient and indoor aerosol monitoring over the past two decades. The continuous methods combine field sampling and chemical analysis methods into one unit and report analysis results in the field in near real time (one hour or less). Currently, there have been commercial systems available for continuous analysis of various chemical components of aerosols, such as metals, particulate carbon (e.g., organic carbon (OC), elemental carbon (EC), black carbon (BC), organic

2- - + + + carbon species), anions and cations (e.g., SO4 , NO3 , NH4 , Na , K ).

Real-time aerosol chemical measurement methods emerged in 1973 when Davis (1973) developed the real-time signal particle mass spectrometry. Stoffels and Lagergren (1976) developed a direct-inlet mass spectrometry which has a particle inlet for producing a focused

particle beam and transmitting the particles to the ionization region. Later, Prather et al. (1994) developed the aerosol time-of-flight mass spectrometry (ATOFMS) that allows for simultaneous measurement of size and composition of single airborne particles. Portable ATOFMS have been developed and are currently commercially available (Gard et al. 1997). Generally, the principle of aerosol mass spectrometers involves particle beam formation using a series of aerodynamic lens, vaporization and ionization of particles by one or more energy sources, followed by analysis of ions by a mass spectrometer (Wexler and Johnston 2011). Aerosol mass spectrometers can analyze particles that consist of salts, soot, crustal matter, metals, and organic molecules through different methods of vaporization and ionization (Nash et al. 2006). Nowadays, portable aerosol mass spectrometers are widely in use for in-situ chemical analysis of single particles or small groups of particles in aerosol in real time (Aiken et al. 2009; Aiken et al. 2008; Elser et al. 2016; Mohr et al.

2012; Park et al. 2012), and they provide a means for acquiring temporal and spatial information on the origin, reactivity and fate of atmospheric aerosol (Gard et al. 1997; Xu et al. 2015).

Continuous monitoring of particulate metals has also been achieved by a system based on

XRF. Commercially available particulate metal monitors use reel-to-reel filter tape sampling and nondestructive XRF analysis. The particles are collected on a filter tape, and then the deposit is then advanced into the analysis area where the sample is analyzed by XFR for selected metals while the next sample is collected. The Xact 625 Particulate Metals Monitor offered by Cooper

Environmental is capable of measuring multiple metals simultaneously with a time resolution of

15 minutes.

Current methods for measurement of particulate carbon typically measure total carbon

(using various techniques) and attempt to distinguish between EC and organic carbon (OC) by heating a filter sample in inert and oxidizing atmospheres (Birch 1998; Birch and Cary 1996). In

addition, thermal-optical methods monitor the filter transmittance/reflectance to correct for potential positive bias due to ‘char’ formed through sample carbonization. The NIOSH Method

5040, widely used for environmental and occupational monitoring, uses thermal-optical transmittance (TOT) for OC-EC speciation (Birch 1998; Lim et al. 2003). A few laboratory methods for determining particulate carbon have been converted to semi-continuous methods for field application. Turpin et al. (1990) devised the first automatic instrument for in-situ measurement of OC and EC in fine particulate matter with a detection limit of 0.2 µg C/m3 at a sampling interval of two hours. The Rupprecht and Patashnick (R&P) 5400 ambient carbon particulate monitor was reported to have an LOD of 0.25 µg/m3 for a 1-hour collection period

(Rupprecht et al. 1995). Both instruments collect aerosol particles with filters or impactors and determine their carbon content by measuring the amount of CO2 produced through heating the particulate sample to elevated temperatures. At present, a semi-continuous OC-EC analyzer

(Sunset Laboratory Inc., Tigard, OR) is widely used to measure time-resolved particulate organic and elemental carbon concentrations in air quality studies (Batmunkh et al. 2011; Polidori et al.

2006). Normally, this semi-continuous instrument is operated with a 1-2 h sample collection period, depending on the EC concentration, followed by a 20 min sample analysis period (Bae et al. 2004).

OC and EC are determined by thermal analysis approaches, whereas BC is derived from the optical absorption of the dark particles. Portable aethalometer is widely used for continuous measurement of BC concentration of aerosols (Buonanno et al. 2013; Dons et al. 2012; Invernizzi et al. 2011).

The portable aethalometer measures the attenuation of a light beam transmitted through aerosol continuously collected on a filter.

There are two different types of continuous methods for measuring the concentration of anions and cations in aerosol: ion chromatography–based methods and thermal reduction with

detection by gas analyzer (Solomon et al. 2011). The IC based methods involve collecting particles into solution, followed by automatically injecting the solutions into the ion chromatograph

(Solomon and Sioutas 2008). One common approach for collecting particles into solution involves condensational growth followed by virtual impaction (Sioutas et al. 1999). In the methods based

2- - + on thermal reduction, the ion components (i.e. SO4 , NO3 , NH4 ) are converted to their gas-phase species by heating the particulate sample to sufficient temperatures, and the resulting gas species are then measured by appropriate gas analyzers (Solomon and Sioutas 2008). In both methods, for accurate measurements of the particle phase species, a variety of denuders are employed for removing interfering gas-phase species (Al-Horr et al. 2003). Most of the existing continuous methods for aerosol chemical analysis require bulky hardware, and therefore are not suitable for hand-portable and compact instrumentation. There is a need for developing simple and robust methods for continuous monitoring aerosol composition.

1.2.2 Microplasma spectroscopy for aerosol analysis

Microplasma spectroscopy has been demonstrated to be a promising technique for developing low-cost and portable instrumentation for on-line aerosol chemical analysis (Diwakar and Kulkarni 2012). A brief literature review on aerosol analysis using microplasma spectroscopy is presented below.

LIBS for aerosol analysis

The foundation of LIBS as an analytical technique for aerosol analysis was first established by researchers in Los Alamos National Lab in the early 1980s (Radziemski et al. 1983a;

Radziemski et al. 1983b). This work was motivated by the need for a real-time scheme of protecting workers in beryllium contaminated environments. Radziemski et al. (1983a) described the apparatus and the calibration procedure for detecting beryllium in air using LIBS. In addition,

Radziemski et al. (1983b) reported LOD of 0.0006 - 1.2 µg/g for Be, Na, P, As, and Hg in terms of air concentration. Later, they reported the LOD for Cd, Pb and Zn in aerosols were 0019, 0.21 and 0.24 µg/g, respectively, which are well below the threshold limit value of exposure limit

(Essien et al. 1988). These works demonstrated that LIBS is a promising technique for direct detection of aerosols in real time.

In these early works, aerosol analysis using LIBS was accomplished by focusing the laser beam directly into the aerosol flow. It presents a point to point sampling nature when applying the finite-sized laser-induced plasma to detect particles directly, hence, particle concentration must be taken into consideration. Generally, the linear relationship between the LIBS intensity and the particle mass is constructed for calibration. However, saturation was observed at very high particle concentration, due to incomplete vaporization of particles (Essien et al. 1988). For aerosols with low particle concentration, the sampling rate (the ratio of total hits and total laser pulse) is very small, then the spectrochemical information of particle-derived analyte may not be reflected in the average LIBS signal. In order to enhance the LIBS sensitivity, an approach based on random LIBS sampling and conditional analysis of the resulting spectra was proposed by Hahn et al. (1997), where the actual concentration of a given analyte species is determined through multiplying average metal concentration of the hits by sampling rate. In addition, focused particle beam techniques, such as capillary focusing (Cheng 2003) or aerodynamic lens focusing (Park et al.

2009), have also been used for aerosol measurement by LIBS in order to enhance the detection sensitivity.

Besides particle concentration, particle size should also be considered in these direct free stream analysis methods, because the linear relationship between the LIBS intensity and the particle mass is based on complete vaporization of a particle and upper particle size limits exist for

complete particle vaporization. The upper size limits for complete vaporization depend on laser pulse energy, focusing optics, particle types, etc. Carranza and Hahn (2002) investigated the laser- induced plasma vaporization of individual airborne silica particle by plotting the silicon atomic emission signal as a function of particle mass. A linear function was obtained for the particles less than 2.1 µm in diameter, which suggested incomplete particle vaporization occurred for the particles with diameter larger than 2.1 µm. Vors and Salmon (2006) reported an upper size limit of

5 µm for complete vaporization of carbon-rich particles. Gallou et al. (2011) found the upper size limit for complete vaporization of CuSO4 particles was approximately 7 µm. The direct methods were also found to be not effective for detection of elements in ultrafine particles and nanoparticles smaller than 100 nm due to weak intensity of emitted light (Park et al. 2009).

To address the disadvantages of the direct free stream analysis methods and further improve the detection limits, substrate based methods have been developed (Panne et al. 2001; Park et al.

2009). The substrate based methods involve collection of particles on an appropriate substrate followed by LIBS analysis. Filter-based collection offers a convenient way to analyze aerosols using LIBS (Cremers and Radziemski 1985; Gallou et al. 2011; Panne et al. 2001); however, this approach does not allow real-time measurement and suffers from large uncertainties and poor detection limits. Park et al. (2009) developed an LIBS aerosol analysis system with an aerodynamic lens focusing system and a collection substrate. An LOD of 80 ng/m3 for Cu was obtained for a sampling time of 5 minutes, and lower LOD could be achieved by increasing collection time. Such aerodynamic lens focusing system requires relatively large pump capacity and bulky hardware, and therefore not valuable in field application. Diwakar et al. (2012) developed an electrostatic particle preconcentrator for collecting particles on a microneedle tip for subsequent LIBS analysis. This technique presents a promising approach that allows development

of compact, portable near real-time instrumentation for aerosol measurement.

Until now, the commercially available LIBS instruments are mainly designed for solid material analysis. The application of LIBS for aerosol analysis is still in the stage of research. More efforts are needed in order to realize LIBS instrument for aerosol detection.

SES for aerosol analysis

Like LIBS, Spark emission spectroscopy (SES) is a pulsed plasma based atomic emission analytical technique. The laser plasma is generated optically by a high energetic laser pulse, while the spark plasma is generated electrically by a high voltage pulse. An electrical discharge is formed between two electrodes and ablates the cathode material while simultaneously creating a plasma.

The cathode material and sample residing between two electrodes is vaporized, atomized, and excited in the plasma (Falk and Wintjens 1998; Taefi et al. 2010; Walters 1977), resulting atomic emission that provides the elemental composition. SES prevailed in the 1950s and the early 1960s due to its better excitation capability. Meanwhile, in the 1960s, it began to decline in popularity due to development of some other excitation source such as inductively coupled plasma (ICP), microwave plasma (MP), and laser-induce plasma (LIP). Walters (1977) proposed spark discharge as a multielement spectrochemical analysis method and gave a detail on its principle.

SES has been applied for the gas or aerosol samples analysis with no or minimal sample preparation. Hunter et al. (2000a) developed a new real-time technique for monitoring heavy metal aerosols based on spark-induced breakdown spectroscopy (SIBS). Aerosol monitoring was accomplished by detecting elemental emission from spark plasma created between two electrodes where contaminated air sample passed through by a pump. This monitor was able to measure Pb and Cr in real time and in situ with an LOD of 10 µg m-3, and has been successfully applied to monitoring heavy metals in a simulated combustion flue gas. A similar apparatus was employed to

monitor suspended dust by tracing two major elements, calcium and magnesium, and the results showed a good sensitivity to variation of dust caused by normal foot traffic in room (Khalaji et al.

2012). Diwakar and Kulkarni (2012) developed a spark emission spectroscopic system for aerosol elemental measurement, which allowed collecting particles onto the ground electrode tip as well as generating spark plasma. This method provided LODs in the range of 0.44 to 70 ng m-3 for Pb,

Si, Na and Cr. This technique is particularly suitable for implementation in field-portable aerosol instrumentation.

GD-OES for aerosol analysis

The excitation sources discussed above (i.e. laser or spark induced plasma) can be bulky and expensive, making them unsuitable for hand-held, low-cost monitors for aerosol elemental analysis. In this context, the glow-discharge excitation sources offer attractive alternative for development of low-cost aerosol instruments. Glow discharge, as an excitation source for the elemental determination, has certain unique advantages such as low cost, low temperature, low power consumption, and analytical versatility (Marcus 2003). Glow discharge optical emission spectroscopy (GD-OES) and glow discharge mass spectroscopy (GD-MS) have been applied in the bulk elemental analysis of inorganic solid samples (Boumans 1972; Jakubowski et al. 1987) and quantitative depth profile analysis (Jakubowski et al. 2007). In a glow discharge system, the samples work as cathode, and are continuously eroded by bombardment of ions and atoms from the plasma. Then the free atoms ejected from the samples are diffused into the plasma plume, where they are excited through collisions with electrons, metastable gas atoms and ions, and emit element characteristic optical emission (Bengtson 1994; Marcus 1993). In 1993, an electrolyte cathode discharge (ELCAD) was introduced for elemental analysis of solutions (Cserfalvi et al.

1993). Since then, solution-cathode glow discharge (SCGD) and liquid sampling-atmospheric

pressure glow discharge (LS-APGD) have been developed for solution analysis (Doroski et al.

2013; Doroski and Webb 2013; Quarles Jr et al. 2012). These techniques can offer similar detection limits (tens of ppb level) as ICP-AES, but have the advantage of much lower cost and power consumption (Doroski and Webb 2013; Wang et al. 2013). In liquid sampling – glow discharge techniques for solution analysis, such as particle beam/hollow cathode-optical emission spectroscopy (PB/HC-OES) and particle beam-glow discharge mass spectroscopy (PB/GDMS), particle beam technique has been used as a transport-type interface to convert the analytes from liquid phase to gas phase molecules for subsequent analysis (Brewer et al. 2006). The particle beam interface involves the conversion of the liquid effluent into an aerosol, evaporation of the aerosol droplets, and formation of a beam of solute particles (Brewer et al. 2006). Though the

PB/GDOES and PB/GDMS are mainly designed for solution analysis, they have also been applied to aerosol analysis. Marcus et al. conducted elemental analysis of aerosols (NIST SRM 1648 urban particulate matter) by a direct injection of particles into a low-pressure glow discharge plasma through an aerodynamic momentum separator, and obtained limits of detection (LOD) on the order of tens of nanograms (Marcus et al. 1999). However, the aerodynamic momentum method required use of large turbo pumps to create particle beams for direct injection into GD, making it unsuitable for hand-held instrumentation. Compared to direct, in situ analysis methods of Marcus et al., substrate-based collection, followed by GD-OES analysis have proven to be more sensitive. LOD below 1 ppm were obtained by collecting atmospheric particulate matter on a metal plate through a single-orifice impactor stage, followed by analyzing the metal plate in a DC glow discharge mass spectrometer. However, the aerosol sampling time was varied from 3 hours up to 3 days and analysis time was more than 1 hour (Schelles et al. 1996).

There have been no hand-portable instruments available so far for aerosol chemical

analysis. There is a need for improving the microplasma spectroscopy based methods and developing low-cost and portable instruments for aerosol chemical analysis.

1.3 Dissertation outline

Chapter 1 presents the background and motivation of this dissertation and reviews some important literature on aerosol measurement using microplasma spectroscopy.

Chapter 2 presents the design and detailed characterization of a corona-based aerosol collection method. Performance of this method has been determined experimentally by measuring collection efficiency, wall losses, and particle deposition density. The methods was also compared with other microconcentration methods.

Chapter 3, 4 and 5 present microplasma spectroscopy based methods for aerosol measurement, to achieve different objectives. Chapter 3 describes a method for near real-time measurement of carbonaceous aerosols using corona microconcentrator coupled with microplasma emission spectroscopy (LIBS and SES). This method has been successfully applied to measure airborne carbon nanomaterials. Chapter 4 describes a multivariate calibration method for simultaneous measurement of multiple elements of aerosol using SES. PLS regression has been employed to build multivariate calibration models. The prediction capability of PLS models (PLS-

1 and PLS-2) created with different spectral variables is compared. The PLS model has been applied to measurement of welding aerosol. Chapter 5 describes a method for near real-time measurement of aerosols by coupling corona aerosol microconcentrator with glow discharge optical emission spectroscopy.

Chapter 6 presents a systematic diagnostics of spark microplasma used in this aerosol measurement system. The spatial and temporal behavior of neutral and ionic carbon emission from carbon black particles has been investigated using a spectrograph coupled with a gated intensified

charge coupled device (ICCD). The plasma parameters, such as excitation temperature and electron density, have been evaluated using spectroscopic means, and their dependences on interelectrode distance, delay time and pulse energy are presented.

Chapter 7 summarizes the main conclusions and contributions of this dissertation and presents recommendations for future work.

1.4 References

Aiken, A., Salcedo, D., Cubison, M. J., Huffman, J., DeCarlo, P., Ulbrich, I. M., Docherty, K. S., Sueper, D., Kimmel, J., Worsnop, D. R. (2009). Mexico City aerosol analysis during MILAGRO using high resolution aerosol mass spectrometry at the urban supersite (T0)–Part 1: Fine particle composition and organic source apportionment. Atmos. Chem. Phys. 9:6633-6653. Aiken, A. C., Decarlo, P. F., Kroll, J. H., Worsnop, D. R., Huffman, J. A., Docherty, K. S., Ulbrich, I. M., Mohr, C., Kimmel, J. R., Sueper, D. (2008). O/C and OM/OC ratios of primary, secondary, and ambient organic aerosols with high-resolution time-of-flight aerosol mass spectrometry. Environ. Sci. Technol. 42:4478-4485. Al-Horr, R., Samanta, G., Dasgupta, P. K. (2003). A continuous analyzer for soluble anionic constituents and ammonium in atmospheric particulate matter. Environ. Sci. Technol. 37:5711- 5720. Aragon, C. and Aguilera, J. (1997). Two-dimensional spatial distribution of the time-integrated emission from laser-produced plasmas in air at atmospheric pressure. Appl. Spectrosc. 51:1632- 1638. Bae, M.-S., Schauer, J. J., DeMinter, J. T., Turner, Smith, D. (2004). Validation of a semi- continuous instrument for elemental carbon and organic carbon using a thermal-optical method. Atmos. Environ. (1994) 38:2885-2893. Baig, M., Qamar, A., Fareed, M., Anwar-ul-Haq, M., Ali, R. (2012). Spatial diagnostics of the laser induced lithium fluoride plasma. Phys. Plasmas 19:063304. Batmunkh, T., Kim, Y. J., Lee, K. Y., Cayetano, M. G., Jung, J. S., Kim, S. Y., Kim, K. C., Lee, S. J., Kim, J. S., Chang, L. S., An, J. Y. (2011). Time-Resolved Measurements of PM2.5 Carbonaceous Aerosols at Gosan, Korea. J. Air Waste Manage. 61:1174-1182. Bengtson, A. (1994). Quantitative depth profile analysis by glow discharge. Spectrochim. Acta B: At. Spectrosc. 49:411-429. Birch, M. E. (1998). Analysis of carbonaceous aerosols: interlaboratory comparison. Analyst 123:851-857. Birch, M. E. and Cary, R. A. (1996). Elemental carbon-based method for monitoring occupational

exposures to particulate diesel exhaust. Aerosol Sci. Technol. 25:221-241. Boumans, P. (1972). Sputtering in a glow discharge for spectrochemical analysis. Anal. Chem. 44:1219-1228. Brewer, T. M., Castro, J., Marcus, R. K. (2006). Particle beam sample introduction into glow discharge plasmas for speciation analysis. Spectrochim. Acta B: At. Spectrosc. 61:134-149. Buonanno, G., Stabile, L., Morawska, L., Russi, A. (2013). Children exposure assessment to ultrafine particles and black carbon: The role of transport and cooking activities. Atmos. Environ. 79:53-58. Bye, C. A. and Scheeline, A. (1993). Saha-Boltzmann statistics for determination of electron temperature and density in spark discharges using an Echelle/CCD system. Appl. Spectrosc. 47:2022-2030. Canagaratna, M., Jayne, J., Jimenez, J., Allan, J., Alfarra, M., Zhang, Q., Onasch, T., Drewnick, F., Coe, H., Middlebrook, A. (2007). Chemical and microphysical characterization of ambient aerosols with the aerodyne aerosol mass spectrometer. Mass Spectrom. Rev. 26:185-222. Carranza, J. E. and Hahn, D. W. (2002). Assessment of the upper particle size limit for quantitative analysis of aerosols using laser-induced breakdown spectroscopy. Anal. Chem. 74:5450-5454. Cheng (2003). Field measurement comparison of aerosol metals using aerosol beam focused laser- induced plasma spectrometer and reference methods. Talanta 61:127-137. Cremers, D. A. and Radziemski, L. J. (1985). Direct detection of beryllium on filters using the laser spark. Appl. Spectrosc. 39:57-63. Cserfalvi, T., Mezei, P., Apai, P. (1993). Emission studies on a glow-discharge in atmospheric- pressure air using ater as a cathode. J. Phys. D-Appl. Phys. 26:2184-2188. Dahneke, B. and Flachsbart, H. (1972). An aerosol beam spectrometer. J. Aerosol. Sci. 3:345-349. Davis, W. D. (1973). Surface Ionization Mass Spectroscopy of Airborne Particulates. J. Vac. Sci. Technol. 10:278-278. DeCarlo, P. F., Kimmel, J. R., Trimborn, A., Northway, M. J., Jayne, J. T., Aiken, A. C., Gonin, M., Fuhrer, K., Horvath, T., Docherty, K. S. (2006). Field-deployable, high-resolution, time-of-flight aerosol mass spectrometer. Anal. Chem. 78:8281-8289. Diwakar, P., Kulkarni, P., Birch, M. E. (2012). New approach for near-real-time measurement of elemental composition of aerosol using laser-induced breakdown spectroscopy. Aerosol Sci. Technol. 46:316-332. Diwakar, P. K. and Kulkarni, P. (2012). Measurement of elemental concentration of aerosols using spark emission spectroscopy. J. Anal. At. Spectrom. 27:1101-1109. Dons, E., Panis, L. I., Van Poppel, M., Theunis, J., Wets, G. (2012). Personal exposure to black carbon in transport microenvironments. Atmos. Environ. 55:392-398.

Doroski, T. A., King, A. M., Fritz, M. P., Webb, M. R. (2013). Solution-cathode glow discharge - optical emission spectrometry of a new design and using a compact spectrograph. J. Anal. At. Spectrom. 28:1090-1095. Doroski, T. A. and Webb, M. R. (2013). Signal enhancement in solution-cathode glow discharge— optical emission spectrometry via low molecular weight organic compounds. Spectrochim. Acta B: At. Spectrosc. 88:40-45. Drewnick, F., Hings, S. S., DeCarlo, P., Jayne, J. T., Gonin, M., Fuhrer, K., Weimer, S., Jimenez, J. L., Demerjian, K. L., Borrmann, S. (2005). A new time-of-flight aerosol mass spectrometer (TOF-AMS)—Instrument description and first field deployment. Aerosol Sci. Technol. 39:637-658. El Sherbini, A. M. and Al Aamer, A. A. S. (2012). Measurement of plasma parameters in laser- induced breakdown spectroscopy using Si-lines. World J. Nano Sci. Eng. 2:206. Elser, M., Huang, R.-J., Wolf, R., Slowik, J. G., Wang, Q., Canonaco, F., Li, G., Bozzetti, C., Daellenbach, K. R., Huang, Y. (2016). New insights into PM 2.5 chemical composition and sources in two major cities in China during extreme haze events using aerosol mass spectrometry. Atmos. Chem. Phys. 16:3207-3225. Essien, M., Radziemski, L. J., Sneddon, J. (1988). Detection of cadmium, lead and zinc in aerosols by laser-induced breakdown spectrometry. J. Anal. At. Spectrom. 3:985-988. Falk, H. and Wintjens, P. (1998). Statistical evaluation of single sparks. Spectrochim. Acta B: At. Spectrosc. 53:49-62. Fierz, M., Kaegi, R., Burtscher, H. (2007). Theoretical and experimental evaluation of a portable electrostatic TEM sampler. Aerosol Sci. Technol. 41:520-528. Gallou, G., Sirven, J., Dutouquet, C., Bihan, O. L., Frejafon, E. (2011). Aerosols analysis by LIBS for monitoring of air pollution by industrial sources. Aerosol Sci. Technol. 45:918-926. Gard, E., Mayer, J. E., Morrical, B. D., Dienes, T., Fergenson, D. P., Prather, K. A. (1997). Real- time analysis of individual atmospheric aerosol particles: design and performance of a portable ATOFMS. Anal. Chem. 69:4083-4091. Gemperline, P. (2006). Practical guide to chemometrics. CRC press. Gross, D. S., Gälli, M. E., Kalberer, M., Prevot, A. S., Dommen, J., Alfarra, M. R., Duplissy, J., Gaeggeler, K., Gascho, A., Metzger, A. (2006). Real-time measurement of oligomeric species in secondary organic aerosol with the aerosol time-of-flight mass spectrometer. Anal. Chem. 78:2130-2137. Hahn, D. W., Flower, W. L., Hencken, K. R. (1997). Discrete particle detection and metal emissions monitoring using laser-induced breakdown spectroscopy. Appl. Spectrosc. 51:1836-1844. Hunter, A. J. R., Davis, S. J., Piper, L. G., Holtzclaw, K. W., Fraser, M. E. (2000). Spark-induced breakdown spectroscopy: A new technique for monitoring heavy metals. Appl. Spectrosc. 54:575- 582.

Invernizzi, G., Ruprecht, A., Mazza, R., De Marco, C., Močnik, G., Sioutas, C., Westerdahl, D. (2011). Measurement of black carbon concentration as an indicator of air quality benefits of traffic restriction policies within the ecopass zone in Milan, Italy. Atmos. Environ. 45:3522-3527. Jakubowski, N., Dorka, R., Steers, E., Tempez, A. (2007). Trends in glow discharge spectroscopy. J. Anal. At. Spectrom. 22:722-735. Jakubowski, N., Stuewer, D., Vieth, W. (1987). Performance of a glow discharge mass spectrometer for simultaneous multielement analysis of steel. Anal. Chem. 59:1825-1830. Jayne, J. T., Leard, D. C., Zhang, X., Davidovits, P., Smith, K. A., Kolb, C. E., Worsnop, D. R. (2000). Development of an aerosol mass spectrometer for size and composition analysis of submicron particles. Aerosol Sci. Technol. 33:49-70. Keskinen, J., Janka, K., Lehtimäki, M. (1987). Virtual impactor as an accessory to optical particle counters. Aerosol Sci. Technol. 6:79-83. Khalaji, M., Roshanzadeh, B., Mansoori, A., Taefi, N., Tavassoli, S. H. (2012). Continuous dust monitoring and analysis by spark induced breakdown spectroscopy. Opt. Laser Eng. 50:110-113. Laville, S., Sabsabi, M., Doucet, F. R. (2007). Multi-elemental analysis of solidified mineral melt samples by laser-induced breakdown spectroscopy coupled with a linear multivariate calibration. Spectrochim. Acta B: At. Spectrosc. 62:1557-1566. Lee, S. H. and Yoh, J. J. (2012). Spectroscopic detection of carbon particulates from a high speed jet stream with extended plasma visualization. Spectrochim. Acta B: At. Spectrosc. 74-75:144-150. Lim, H. J., Turpin, B. J., Edgerton, E., Hering, S. V., Allen, G., Maring, H., Solomon, P. (2003). Semicontinuous aerosol carbon measurements: Comparison of atlanta supersite measurements. J. Geophys. Res. Atmos. 108. Liu, P., Ziemann, P. J., Kittelson, D. B., McMurry, P. H. (1995). Generating particle beams of controlled dimensions and divergence: I. Theory of particle motion in aerodynamic lenses and nozzle expansions. Aerosol Sci. Technol. 22:293-313. Marcus, R. K. (2003). 5.5 Glow discharge plasmas as tunable sources for elemental speciation. Handbook of Elemental Speciation: Techniques and Methodology 1:334. Marcus, R. K. (1993). Glow discharge spectroscopies. Springer Science & Business Media 17-66. Marcus, R. K., Dempster, M. A., Gibeau, T. E., Reynolds, E. M. (1999). Sampling and analysis of particulate matter by glow discharge atomic emission and mass spectrometries. Anal. Chem. 71:3061-3069. Martin, M. Z., Cheng, M. D., Martin, R. C. (1999). Aerosol measurement by laser-induced plasma technique: A review. Aerosol Sci. Technol. 31:409-421. Miller, A., Frey, G., King, G., Sunderman, C. (2010). A handheld electrostatic precipitator for sampling airborne particles and nanoparticles. Aerosol Sci. Technol. 44:417-427. Mohr, C., DeCarlo, P., Heringa, M., Chirico, R., Slowik, J., Richter, R., Reche, C., Alastuey, A., 21

Querol, X., Seco, R. (2012). Identification and quantification of organic aerosol from cooking and other sources in Barcelona using aerosol mass spectrometer data. Atmos. Chem. Phys. 12:1649- 1665. Murphy, W. and Sears, G. (1964). Production of particulate beams. J. Appl. Phys. 35:1986-1987. Nash, D. G., Baer, T., Johnston, M. V. (2006). Aerosol mass spectrometry: An introductory review. Int. J. Mass Spectrom. 258:2-12. NIOSH (2013). Current Intelligence Bulletin 65: Occupational exposure to carbon nanotubes and nanofibers. DHHS NIOSH Publication No. 2013-145. Onasch, T., Trimborn, A., Fortner, E., Jayne, J., Kok, G., Williams, L., Davidovits, P., Worsnop, D. (2012). Soot particle aerosol mass spectrometer: development, validation, and initial application. Aerosol Sci. Technol. 46:804-817. Panne, U., Neuhauser, R., Theisen, M., Fink, H., Niessner, R. (2001). Analysis of heavy metal aerosols on filters by laser-induced plasma spectroscopy. Spectrochim. Acta B: At. Spectrosc. 56:839-850. Park, K., Cho, G., Kwak, J.-h. (2009). Development of an aerosol focusing-laser induced breakdown spectroscopy (aerosol focusing-LIBS) for determination of fine and ultrafine metal aerosols. Aerosol Sci. Technol. 43:375-386. Park, K., Park, J., Lee, S., Cho, H.-j., Kang, M. (2012). Real time measurement of chemical composition of submicrometer aerosols at urban Gwangju in Korea by aerosol mass spectrometer. Atmos. Environ. 62:281-290. Polidori, A., Turpin, B. J., Lim, H.-J., Cabada, J. C., Subramanian, R., Pandis, S. N., Robinson, A. L. (2006). Local and regional secondary organic aerosol: Insights from a year of semi-continuous carbon measurements at Pittsburgh. Aerosol Sci. Technol. 40:861-872. Possanzini, M., Febo, A., Liberti, A. (1983). New design of a high-performance denuder for the sampling of atmospheric pollutants. Atmos. Environ. (1967) 17:2605-2610. Prather, K. A., Nordmeyer, T., Salt, K. (1994). Real-time characterization of individual aerosol particles using time-of-flight mass spectrometry. Anal. Chem. 66:1403-1407. Quarles Jr, C. D., Gonzalez, J., Choi, I., Ruiz, J., Mao, X., Marcus, R. K., Russo, R. E. (2012). Liquid sampling-atmospheric pressure glow discharge optical emission spectroscopy detection of laser ablation produced particles: A feasibility study. Spectrochim. Acta B: At. Spectrosc. 76:190- 196. Radziemski, L. J., Cremers, D. A., Loree, T. R. (1983a). Detection of beryllium by laser-induced- breakdown spectroscopy. Spectrochim. Acta B: At. Spectrosc. 38:349-355. Radziemski, L. J., Loree, T. R., Cremers, D. A., Hoffman, N. M. (1983b). Time-resolved laser- induced breakdown spectrometry of aerosols. Anal. Chem. 55:1246-1252. Ramli, M. and Wagatsuma, K. (2010). Observation of Atomic Emission Image from Spark

Discharge Plasma by Using Two-dimensional Spectrograph. ISIJ Int. 50:864-867. Raynor, P. C., Leith, D., Lee, K., Mukund, R. (2011). Sampling and analysis using filters. Aerosol measurement: principles, techniques and applications. Rupprecht, G., Patashnick, H., Beeson, D., Green, R., Meyer, M. (1995). A new automated monitor for the measurement of particulate carbon in the atmosphere. Proceedings, Particulate Matter: Health and Regulatory Issues:262-267. Salik, M., Hanif, M., Wang, J., Zhang, X. (2013). Plasma properties of nano-second laser ablated iron target in air. Int. J. Phys. Sci. 8:1738-1745. Schelles, W., Maes, K. J., De Gendt, S., Van Grieken, R. E. (1996). Glow discharge mass spectrometric analysis of atmospheric particulate matter. Anal. Chem. 68:1136-1142. Sioutas, C., Kim, S., Chang, M. (1999). Development and evaluation of a prototype ultrafine particle concentrator. J. Aerosol. Sci. 30:1001-1017. Solomon, P. A., Fraser, M. P., Herckes, P. (2011). Methods for chemical analysis of atmospheric aerosols. Aerosol Measurement: Principles, Techniques, and Applications 3:153-177. Solomon, P. A. and Sioutas, C. (2008). Continuous and semicontinuous monitoring techniques for particulate matter mass and chemical components: a synthesis of findings from EPA’s particulate matter supersites program and related studies. J. Air Waste Manage. 58:164-195. Spurny, K. R. (1999). Analytical Chemistry of Aerosols: Science and Technology. CRC Press. Stoffels, J. J. and Lagergren, C. R. (1976). Direct-inlet mass spectrometer for real-time analysis of particulate contaminants in air. International Mass Spectrometry Conference; Florence, Italy. Taefi, N., Khalaji, M., Tavassoli, S. (2010). Determination of elemental composition of cement powder by spark induced breakdown spectroscopy. Cement Concrete Res. 40:1114-1119. Tripathi, M. M., Eseller, K. E., Yueh, F.-Y., Singh, J. P. (2009). Multivariate calibration of spectra obtained by Laser Induced Breakdown Spectroscopy of plutonium oxide surrogate residues. Spectrochim. Acta B: At. Spectrosc. 64:1212-1218. Turpin, B. J., Cary, R. A., Huntzicker, J. J. (1990). An insitu, time-resolved analyzer for aerosol orgranic and elemental carbon. Aerosol Sci. Technol. 12:161-171. Vors, E. and Salmon, L. (2006). Laser-induced breakdown spectroscopy (LIBS) for carbon single shot analysis of micrometer-sized particles. Anal. Bioanal. Chem. 385:281-286. Walters, J. P. (1977). Spark discharge - application to multielement spectrochemical analysis. Science 198:787-797. Walters, J. P. and Goldstein, S. A. (1984). Emission topography of a stable spark discharge train. Spectrochim. Acta B: At. Spectrosc. 39:693-728. Wang, Z., Schwartz, A. J., Ray, S. J., Hieftje, G. M. (2013). Determination of trace sodium, lithium, magnesium, and potassium impurities in colloidal silica by slurry introduction into an atmospheric-

pressure solution-cathode glow discharge and atomic emission spectrometry. J. Anal. At. Spectrom. 28:234-240. Wexler, A. S. and Johnston, M. V. (2011). Real-time particle analysis by mass spectrometry. Aerosol Measurement: Principles, Techniques, and Applications:233. Wu, J., Cooper, D., Miller, R. (1989). Virtual impactor aerosol concentrator for cleanroom monitoring. The J. Environ. Sci. 32:52-56. Xu, L., Suresh, S., Guo, H., Weber, R. J., Ng, N. L. (2015). Aerosol characterization over the southeastern United States using high-resolution aerosol mass spectrometry: spatial and seasonal variation of aerosol composition and sources with a focus on organic nitrates. Atmos. Chem. Phys. 15:7307-7336. Yao, S., Lu, J., Zheng, J., Dong, M. (2012). Analyzing unburned carbon in fly ash using laser- induced breakdown spectroscopy with multivariate calibration method. J. Analy. At. Spectrom. 27:473-478.

CHAPTER 2

Aerosol Microconcentrator for Analysis Using Microscale Optical

Spectroscopies

 Submitted to Journal of Aerosol Science 25

Efficient microconcentration of aerosols to a substrate is essential for effectively coupling the collected particles to microscale optical spectroscopies such as laser-induced or spark microplasma, or micro-Raman spectroscopies. In this paper, we present the design, characterization, and optimization of a corona-based aerosol microconcentration method for portable analytical instrumentation. The method involves two coaxial electrodes separated by a few millimeters, one held at a high electrical potential and the other grounded. The particles are collected on the collection (i.e., ground) electrode from a coaxial aerosol flow in a one-step charge- and-collect scheme using corona discharge and electrical precipitation between the two electrodes.

Performance of the corona microconcentration method was determined experimentally by measuring collection efficiency, wall losses, and particle deposition density. An intrinsic spectroscopic sensitivity was experimentally determined for the aerosol microconcentrator. Using this sensitivity, we show that corona-based microconcentration is much superior to alternative methods, including filtration, focused impaction using aerodynamic lens, and spot collection using condensational growth, and offers unique advantages for compact hand-held analytical instrumentation.

2.1. Introduction

Concentrated collection of aerosols has been a topic of great interest in aerosol science.

Microscopic concentration, i.e. collection of aerosol particles over a miniscule area (on the order of 1 mm2 or less) at high number or mass density is particularly important for aerosol characterization using laser spectrometric methods, such as laser-induced breakdown, micro-

Raman and infrared spectroscopies, that probe miniscule samples (Diwakar and Kulkarni 2012).

Depending on the application, concentration of aerosols, either to a substrate or in a suspended state, has been used to achieve different objectives, which range from increasing the number density of suspended particles, increasing the surface density of particles collected on a substrate, and increasing temporal rate of collection. Methods based on collection of aerosols on particulate filters, to obtain a time-integrated concentration, have been extensively used for routine aerosol monitoring (Raynor et al. 2011; Spurny 1999). Methods based on spatial concentration of aerosols, for instance, by creating a focused particle beam, have been well studied and widely used for aerosol analysis by mass spectrometry (Liu et al. 1995; Ziemann et al. 1995). A virtual impactor has been used to increase the number density (per unit volume) of larger particles in the suspended state (Barr et al. 1983), mainly for improvement of instrument detection sensitivity (Keskinen et al. 1987; Wu et al. 1989) and for epidemiological and toxicological studies (Ghio et al. 2000;

Sioutas et al. 1995). A condensational growth technique has also been employed for growing ultrafine particles to super-micron droplets under saturation conditions for subsequent spot collection on an impaction plate (Eiguren Fernandez et al. 2014; Hering et al. 2014) or concentration in a virtual impactor (Sioutas et al. 1999). Focused particle beams also have been employed for efficient collection over microscopic areas. These techniques achieve focusing by expanding the aerosol flow into a low pressure chamber, either through capillary (Murphy and

Sears 1964) converging nozzles (Dahneke and Flachsbart 1972) or aerodynamic lenses (Liu et al.

1995). Such focused particle beam techniques have been applied to various measurements in aerosol systems. Aerodynamic lenses have been used as aerosol inlets in aerosol mass spectrometers for online single particle analysis (Lee et al. 2008; Su et al. 2004), and for aerosol analysis using laser-induced breakdown spectroscopy (Park et al. 2009). Most of the aerosol concentrating techniques discussed above require relatively large pumping capacity and bulky hardware and therefore are not suitable for portable and compact instrumentation. Electrostatic collection offers an alternative sampling method that enables use of high flow rates with minimal pressure drop and is suitable for compact field instrumentation. The point-to-plane type of electrostatic precipitator (ESP) allows one-step charging and collection of particles (Hinds 1999).

Several point-to-plane ESP particle samplers have been designed and evaluated (Cheng et al. 1981;

Miller et al. 2010; Morrow and Mercer 1964). Electrostatic collection has been mainly used for collecting samples for microscopic analysis (Fierz et al. 2007; Miller et al. 2010), and more recently for aerosol analysis using microplasma emission spectroscopy (Diwakar and Kulkarni

2012).

In this study, we present design and detailed characterization of a corona-based aerosol collection method. The system was designed for optimal coupling with optical spectrometric methods (such as laser-induced breakdown, spark discharge, micro-Raman or infrared methods) that require concentrating aerosols to a microscopic area of the deposition substrate for subsequent analysis. Performance of this method was determined experimentally, by measuring collection efficiency, wall losses, and particle deposition density, and compared with other microconcentration methods.

2.2. Methods

2.2.1. Design of aerosol microconcentrator

Figure 2.1 shows the schematic diagram of the corona aerosol microconcentrator (CAM) used in this study. The key components of the CAM include two coaxial electrodes with their tips separated by a distance of 5 mm. Both, the corona and collection (i.e., ground) electrodes, are made of tungsten. The corona electrode has a shaft diameter of 200 µm and converges to a sharp tip of radius 50 µm. The coaxial collection electrode has a relatively flat tip to provide a planar surface for particle deposition. Collection electrodes with four different diameters were investigated in this study: 500, 750, 1000, and 1500 µm. A high positive electric potential (4 – 6.2 kV) was applied on the corona electrode using a DC power supply (Bertan S-230, Spellman Corp., Hauppauge, NY,

USA), whose output voltage was controlled using a PID controller to maintain a stable corona current between the corona and collection electrode. The side walls of the collection electrodes are covered with a dielectric sleeve made of polyether ether ketone (PEEK) (wall thickness varied depending on the electrode size; McMaster-Carr, Princeton, NJ, USA). The external dimensions of the CAM are approximately 3.4 × 3.4 × 6.3 cm. Optical access for optical signal collection is provided by UV-VIS coated fused silica windows on the CAM walls.

Figure 2.1. Cross-section view of the corona aerosol microconcentrator.

2.2.2. Experimental setup

The experimental setup, shown in Figure 2.2, was designed to evaluate collection efficiency of the CAM and detection sensitivity of spark emission spectroscopy. Test aerosols were generated using an atomizer (model 3076, TSI Inc., Shoreview, MN, USA) to aerosolize

(NH4)2SO4 and Cr(NO3)3 solutions. The aerosol was subsequently dried in a diffusion dryer and then passed through a differential mobility analyzer (DMA; model 3081, TSI Inc., Shoreview, MN,

USA), a 210Po neutralizer, followed by an electrostatic precipitator (ESP) to remove all charged particles and obtain a near-monodisperse, uncharged test aerosol. The aerosol flow rate through the DMA was maintained at 0.23 L min-1, and the sheath flow rate in DMA was 2.3 L min-1. The

aerosol exiting the ESP was diluted with HEPA-filtered air at a flow rate of 5 L min-1. The aerosol flow through the CAM was driven by the native pump in the condensation particle counter (CPC; model 3022A, TSI Inc., Shoreview, MN, USA) and a mass flow controller (MFC; model 247 C,

MKS Instruments, Inc., Andover, MA, USA). The MFC was used to adjust the aerosol flow rate through the CAM chamber in a range of 0.6 – 5 L min-1.

Figure 2.2. Schematic diagram of experimental setup in this study.

The particulate mass accumulated on the collection electrode tip of the CAM was measured using spark emission spectroscopy (SES). In the SES system, a high voltage pulse generator

(Cascodium Inc., Andover, MA, USA) was used to produce a spark microplasma in the electrode gap to ablate the collected particulate matter from the collection electrode. The atomic emission signals from the spark microplasma were collected using an optical fiber connected to a spectrograph (IsoPlane SCT320, Princeton Instrument Inc., Trenton, NJ, USA) coupled with a gated ICCD camera (iStar 334T, Andor Technology, South Windsor, CT, USA). A delay time of 5

µs and a gate width of 25 µs were used for spectral acquisition. The atomic emission signal intensity measured during the spark discharge is directly proportional to the abundance of the analyte in the particulate mass deposited on the electrode tip. DMA classified monodisperse

Cr(NO3)3 aerosol with a particle size of 100 nm in diameter was used for studies aimed at establishing collection characteristics of the CAM (with Cr as a target analyte in SES measurement). (NH4)2SO4 test aerosol was used in experiments designed to measure particle loss on the electrode side walls.

2.2.3. Analysis of (NH4)2SO4 aerosol

Ion chromatography (IC) analysis was used to measure the particulate mass of (NH4)2SO4 deposited on the electrode. Concentration of the airborne (NH4)2SO4 aerosol was determined by collecting aerosol on the polytetrafluoroethylene (PTFE) filter, followed by IC analysis. The extraction of (NH4)2SO4 particles (collected on the electrode or filter) was done using the procedures described by Chow and Watson (1999). The collection substrate (electrode or the filter) was placed in an extraction vial and 200 µL of ethanol was added as wetting solution and 3 ml ultrapure DI water as added as a solvent. The vial was then sonicated for 15 minutes in an ultrasonic bath (Model ME 4.6, Mettler Electronics Corp., Anaheim, CA), followed by aging for 24 hours at

o 3 C. This assured complete extraction of the deposited (NH4)2SO4 particles in the solvent. The extracted solution in the vial was analyzed using IC (DionexTM, Thermo Fisher Scientific Inc.,

2- Sunnyvale, CA) for quantitative determination of sulfate ions. The calibration curve for SO4 was constructed using five standard solutions with different concentrations in the range 0.01 – 2 mg L-

2.3. Results and discussion

During the operation of the CAM, a high voltage is applied to the corona electrode to form a nonuniform electrostatic field between the corona electrode and the flat tip of the collection electrode. In the region of sufficiently high electric field strength at the tip of the corona electrode, air undergoes ionization and partial breakdown. Unipolar ions are continuously produced in the corona discharge process surrounding the tip of the corona electrode, which are then transported to the interelectrode space (Chang et al. 1991). The incoming particles acquire electrical charge via diffusion and field charging mechanisms. Subsequently, the charged particles migrate in the electric field and deposit on the collection electrode.

2.3.1. Collection efficiency of the CAM

The collection efficiency of the CAM is governed by the particle diameter, corona current, flow characteristics, electrode diameter, interelectrode distance, and the electrostatic field distribution. It is necessary to choose an optimum set of parameters to maximize the collection

efficiency. Count-based particle collection efficiency (c ) was estimated by measuring the particle number concentration downstream of the CAM using a CPC, with and without the presence of the

HV V 0 electric field across the electrodes ( Nout and Nout ).

V00 HV V c NNN out out out (1)

Figure 2.3 shows the collection efficiency for different electrode sizes (500 µm, 750 µm,

1000 µm, and 1500 µm in diameter, respectively) plotted as functions of the corona current at a flow rate of 1.5 L min-1. As the corona current increases from 0.5 to 6 µA, the collection efficiency for relatively larger electrodes with diameters of 1000 µm and 1500 µm increases approximately by 25 %. These trends are consistent with a study reported earlier (Miller et al. (2010) on a hand- held electrostatic precipitator, with a collection electrode 3 mm diameter. Higher voltage leads to higher corona current and higher charging efficiency, leading to higher collection efficiency. For the smaller electrodes (500 and 750 µm diameter), however, the collection efficiency does not vary much with corona current. Figure S-1 in the Supplemental Information (SI) shows voltage-current characteristics of the CAM for different sizes of collection electrodes. A maximum high voltage exists for mantaining a stable corona current, and higher voltage beyond the maximum leads to frequent sparking or instability. A stable corona current of 5 µA was used for all the subsequent experiments in this study. Figure 2.3 also shows that a larger collection electrode provides a higher collection efficiency, due to a larger deposition surface and a higher electric field strength from a higher voltage under the same corona current (as shown in Figure S-1 in SI).

Figure 2.3. Collection efficiency as a function of corona current at a volumetric flow rate of 1.5

L min-1 for different collection electrodes. Distance between the corona and the collection electrode was 5 mm. The particle size was 100 nm in diameter. The lines represent the best fit to the experimental data.

Figure 2.4(a) shows the collection efficiency plotted as a function of aerosol flow rate for various electrode diameters. As the flow rate increases from 0.6 to 5 L min-1, the collection efficiency decreases for all four electrodes. However, to maximize the sensitivity of the system, it is the total particulate mass collected on the electrode tip that must be maximized. This collected mass is directly proportional to the flow rate ( Q ) and the collection efficiency ( ). Figure 2.4(b) shows the plot of product × as a function of flow rate. This product signifies the time-rate of

microconcentration, and will be denoted as the temporal concentration factor ( Ct ) in this study.

The variation of with for a given collection electrode are described using parabolic best fit

curves and are shown in Figure 2.4(b). The figure shows that is low at very low flow rates even though the collection efficiency is high, then increases with increasing flow rate and reaches a maximum. After the maximum, starts decreasing with further increase of the flow rate because of the low collection efficiency. Maximum can be achieved at a flow rate of approximately 3

L min-1 for 500, 750, and 1000 µm electrodes, and at 5 L min-1 for 1500 µm electrode. The figure also shows that the collection efficiency and the concentration factor also increase with the increasing collection electrode diameter at a given aerosol flow rate.

The collection efficiency presented in Figure 2.4 was based on the measurement of aerosol number concentration at the outlet of the CAM in the presence and the absence of high voltage on

the corona electrode (Eq. (1)). Alternatively, collection efficiency was also obtained using an independent method involving direct measurement of the accumulated mass of particles on the collection electrodes (Zavvos, 2016). In these experiments, the DMA classified (NH4)2SO4 test aerosol was first collected on the collection electrode for a predetermined period of time at a given aerosol flow rate. Simultaneously, to determine the air concentration of the (NH4)2SO4 aerosol entering the CAM, a filter sample was collected at the inlet of the CAM. The particulate mass

deposited on the electrode ( melectrode ) and that collected on the filter ( mfilter ) , over a predetermined period of time, was analyzed for (NH4)2SO4 using IC analysis. The mass-based collection

 efficiency ( m ) was then calculated as,

m mm electrode filter (2)

Figure 2.5 shows the comparison of collection efficiencies obtained using both count- and mass-based measurement methods. The difference in collection efficiency is small and is

approximately within the range of experimental uncertainty of these measurements. Mean uncertainty was obtained using the linear best fit to the experimental data for both the methods.

The mean count-based collection efficiency was approximately 2% greater than the mean mass- based collection efficiency in the particle diameter range 50 – 1000 nm. This demonstrates the equivalency of the two methods. The count-based method was used in the remainder of this study due to its ease of use.

2.3.2. Effects of particle size on collection efficiency

To measure airborne concentration of an analyte in an unknown polydisperse aerosol sample, a key requirement is that the collection efficiency is independent of the particle size. To

examine the size dependence, the collection efficiency of the CAM was measured at different particle diameters in the submicrometer size range using the count-based method and the results are shown in Figure 2.6. The figure shows a slight increase in collection efficiency as a function of particle size at very low or very high flow rates. For each case, a linear best fit to the experimental data is shown. For the 500 µm electrode at 1 L min-1, the collection efficiency remains relatively constant, within the range of experimental uncertainty, over the submicrometer particle size range. When using the 500 µm electrode at 3 L min-1 and the 1000 µm electrode at 3

L min-1 and 5 L min-1, the collection efficiency increases slightly with particle diameter. Similar results were obtained in early studies on collection efficiency of electrostatic precipitators. Cheng et al. (1981) showed that the collection efficiency of a point-to-plane electrostatic precipitator increases with increasing particle size at 0.1 – 1 L min-1, and minimal effects of particle size on collection efficiency were found at the lowest flow rate (0.1 L min-1) in their study. Laskin and

Cowin (2002) also found a nearly constant collection efficiency for 0.1 – 1 µm particles using a point-to-plate electrostatic precipitator operated at a relatively low flow rate of 0.3 L min-1.

However, the collection efficiency of CAM was relatively independent of particle size even at a higher flow rate (1 L min-1) using 500 µm collection electrode, and its increasing trend with particle diameter at 3 L min-1 was less significant than that reported by Cheng et al. (1981), which may be due to different orientation of flow and electric field configuration.

Although the collection efficiency exhibits a slight increasing trend with increasing particle diameter at 3 and 5 L min-1, the size dependence of collection efficiency in Figure 2.6 can be assumed to be represented by a mean value calculated over the entire size range. However, such an approximation may lead to a finite uncertainty, the degree of which will depend on the flow rate, electrode size, and the aerosol size distribution entering the CAM. For a 500 µm electrode, at

3 L min-1, assuming a mean collection efficiency (~10.5%, which is the mean of the measured collection efficiency over the entire particle size range) leads to an error ranging from 0.06% to

36% depending on the nature of the lognormal distribution, as shown in Figure S-2 in SI. For a typical urban aerosol in the accumulation mode which has a count median diameter (dpg) of 100 nm and a geometric standard deviation (σg) of 1.8, the error is estimated to be about 13%.

2.3.3. Deposition uniformity and side wall losses

The variation of number density of particles deposited on the electrode surface was probed using scanning electron microscopy (SEM) (Zavvos, 2016). The spatial distribution of particles was approximately homogeneous over the flat electrode surface.

During particle collection, due to favorable elecrtical field and flow conditions (Figure 2.3), some particles can deposit on the electrode side wall (along its axis) instead of its flat tip. This particulate mass is not ablated by the microplasma. This unablated particulate mass needs to be accounted for to allow accurate retreival of airborne analyte concentration. Two methods were used to quantify the particle loss to the side walls of the electrode: i) counting the particles on side walls using SEM micrographs, and ii) extracting and analyzing the mass of (NH4)2SO4 particles deposited on the side walls by IC chemical analysis (Zavvos, 2016). By counting individual particles per unit area of the micrograph, the fraction of particles deposited on the side wall was estimated to be 7.7%.

In the alternative method using IC analysis, (NH4)2SO4 particles were collected on two electrodes in two separate experiments: a bare electrode and another identical electrode with a mask of thin Teflon™ layer wrapped around its side walls. The Teflon™ mask wrapped around the electrode side walls prevented the particles from depositing directly on the electrode walls.

After each particle collection cycle, the total particulate mass deposited on the bare electrode and masked electrode was extracted in separate vials and analyzed for (NH4)2SO4 using IC. The particle loss determined by IC analysis was in the range of 8 – 17%, slightly higher than the crude estimate of 7.7% obtained from the SEM counting method. Based on these measurements, a constant factor of 12.8%, the mean of the particle loss over different particle diameters, was applied for retrieval of air concentrations to account for wall losses.

2.3.4. Effects of spark ablation on collection efficiency

The coaxial electrode system was used for particle collection as well as for generation of spark discharge in the interelectrode gap for atomic emission measurement. Each analyte measurement task entailed a cyclical sequence of two steps involving collection over a

predetermined time period, followed by multiple ablation events (needed to record emission signal, and subsequently clean the collection surface for the next collection cycle). Depending on the electrode material, it is possible that the repeated spark discharges could alter the morphology of the collection electrode surface, which in turn, may influence the stability of corona or collection characteristics. Experiments were conducted to probe the effect of the number of spark discharges on the electrode collection efficiency. Figure 2.7 shows the variation in particle collection efficiency of the CAM as a function of the number of spark discharges. A continuous train of pulsed spark discharges was produced at a frequency of 10 Hz and particle collection efficiency was experimentally measured (using the procedure described earlier; Eq. (1)) after every 100, 500,

1000, 2000, and 3000 spark discharges. As shown in Figure 2.7, the variation in collection efficiency was within 2% and no significant changes in the collection efficiency was observed even after 3000 sparks. Also shown (Figure 2.7) are SEM images exhibiting the surface morphology of the collection electrode after 0, 1000, and 2000 sparks. No significant change in surface morphology of the electrodes (at the scale shown) were observed even after several thousand ablation events.

95% confidence interval.

2.3.5. Analytical sensitivity

To examine the efficiency of microconcentration, collected particulate mass on the electrode surface was analyzed using spark discharge emission spectroscopy to determine the elemental concentration. Calibration curves were constructed by plotting emission signal intensity as a function of particulate mass. The slope of this calibration curve, denoted here as the mass-

based sensitivity ( Sm ), is an important indicator of the efficacy of microconcentration as well as the overall analytical sensitivity of the method. Figure 2.8(a) shows the calibration curves for Cr constructed for collection electrodes of various diameters (500, 750, 1000, and 1500 µm). DMA classified monodisperse Cr(NO3)3 particles with a diameter of 100 nm were used for calibration.

As shown in Figure 2.8(a), increases with decreasing collection electrode diameter. This suggests efficient coupling of the concentrated particulate mass with the spark microplasma for small diameter electrodes. The data in Figure 2.8(a) can be collapsed into one curve when plotted as a function of surface mass density, defined as the particulate mass per unit area of the electrode

mp surface. Figure 2.8(b) shows signal intensity as a function of surface mass density ( ), A

mp IS * (3) A

* where I is the signal intensity, S is the slope of the calibration curve in Figure 2.8(b), mp is the mass of particles collected on the electrode (ng), and A is the flat area of the collection electrode

(mm2). Here, we defined as the intrinsic spectroscopic sensitivity of the CAM. For a specific spectroscopic system, is a constant for a given analyte. In our system, for Cr was measured to be 107.8 ng-1 mm2. Eq. (3) implies that the signal intensity is a function of surface mass density on the electrode, rather than the absolute total mass itself. This could be explained by the fact that the fraction of the electrode surface sampled by the microplasma during ablation is limited by its finite spatial extent: this fraction is higher when the electrode diameter is equal to or smaller than the diameter of the ablation spot created by the microplasma on the electrode surface. To measure the diameter of the ablation spot, an aluminum disk (25 mm in diameter) was used as a cathode, and a PTFE filter (0.2 mm in thickness) was placed on the disk. After several spark discharges, the ablation area on the PTFE filter was measured using an optical microscope and found to be 510

±130 µm in diameter. This size of the ablation spot is comparable to diameter of 500 µm electrode, which was found to have the highest sensitivity in Figure 2.8(a).

Figure 2.8. (a) Calibration curves for Cr using different electrode diameters (Cr I 360.534 nm emission line was used in calibration); (b) The signal intensity of Cr from spark emission spectroscopy as a function of particle mass per area. The lines represent linear fits to the experimental data.

A larger diameter electrode provides higher particle flux (i.e. time rate of collection) to the

electrode surface, therefore leads to a higher temporal concentration factor ( Ct ). On the other hand, a smaller diameter electrode offers better surface mass density or higher spatial concentration factor. Both the temporal and spatial aspects of microconcentration are important for achieving

superior detection limits in terms of air concentration of the analyte. Using mp C in Qt in Eq.

(3), the emission signal intensity can be expressed as,

S* Qt IC (4) A in where  is the collection efficiency; Q is the air flow rate (L min-1), t is the particle collection

-1 time (min), and Cin is the incoming aerosol particle mass concentration (ng L ). The measurement

sensitivity ( Sc ) of the analyte in terms of air concentration is then given by,

S* Qt S  (5) c A

For most aerosol measurement applications, the objective is to maximize Sc . For our system, this requires an optimum combination of flow rate and electrode diameter. Figure 2.9 shows the detection sensitivity ( ) calculated using Eq. (6) at different flow rates and different electrode diameters for an assumed sampling time of 1 minute. The figure shows that for most electrodes has a maximum; it first increases with increasing flow rate, and then decreases at higher flow rates (the maximum for 1500 µm electrode probably occurs beyond 5 L min-1 and was not measured experimentally). This trend is consistent with those in temporal concentration factor in Figure 2.4(b). Each electrode diameter has an optimum window of flow rate which provides the highest sensitivity. The highest sensitivity (193 µg-1 m3 for Cr I) can be achieved at a flow rate of 46

approximately 3 L min-1 for the 500 µm electrode. The error bars in Figure 2.9 represent the

standard deviation of three experimentally measured sensitivity ( Sc ), which is in the range of 0.6

– 35 µg-1 m3; this uncertainty increases with flow rate. The detection sensitivity can be used to obtain limit of detection (LOD) for a given analyte. The LOD was estimated using 3-𝜎 criteria defined by the International Union of Pure and Applied Chemistry (IUPAC) as,

LOD 3 Sc (6) where  is the standard deviation of the blank at the selected spectral region. The lowest LOD for

Cr was estimated to be 0.1 µg m-3 using 500 µm electrode at 3 L min-1 with a sampling time of 1 minute. As a comparison, it is worth noting that the LOD of a filter-based method using inductively coupled plasma - atomic emission spectrometry (ICP-AES) is approximately 0.23 µg m-3 for an

8-hr sample (at 2 L min-1) (NIOSH 2003). The CAM used in this work offers a factor of 2 improvement in LOD at a sampling time that is over two orders of magnitude smaller. This demonstrates a clear advantage of microconcentration when coupling with microplasma spectroscopies.

The intrinsic spectroscopic sensitivity ( S * ) for an analyte discussed earlier (Fig 10(b)) is a characteristic constant for the specific electrode configuration and plasma characteristics used in our study. It must be noted that this intrinsic sensitivity ( S * ) does not depend on the method employed to collect particles on the collection electrode. Therefore, S * is independent of particle collection method. This allows probing and comparing the efficacy of alternative particle collection methods using filtration, aerodynamic lens, or inertial impaction. However, for the S * to be applicable to other collection methods, it must be ensured that the spark discharge characteristics and the analyte ablation characteristics remain the same when the alternative collection methods are coupled with our spark emission spectroscopy method. This requirement should not be difficult to meet for most alternative particle collection methods. Using the 48

* characteristic S value for Cr I measured for the CAM, the sensitivities (i.e. Sc ) one could achieve using alternative particle collection methods, including filtration, condensational growth followed by focused impaction, and focused impaction using aerodynamic lens, were calculated and compared with that of the CAM. Typical design and operating parameters such as collection efficiency, flow rate, particle deposition area, and sampling time for each collection method were assumed for each collection method and are shown in Table 1. Figure 2.10 shows the calculated measurement sensitivity ( ) obtained for alternative collection methods as a function of the aerosol sampling flow rate. The for CAM (Figure 2.10) is the experimentally measured data

(except at 0.1 L min-1, which was estimated by assuming 100 % collection efficiency). The figure shows that the CAM used in this work provided the highest sensitivity, which varied from 41 to

193 µg-1 m3 for Cr I for flow rates ranging from 0.1 to 5 L min-1. The sensitivity for filter-based method was lower by 1 – 2 orders of magnitude, even when using large sampling flow rates and collection times. This again demonstrates the advantages of spatial microconcentration when coupling to microscale spectroscopies. At 0.1 L min-1, from aerodynamic lens is 54 µg-1 m3, while that from CAM is 41 µg-1 m3; however, 0.1 L min-1 is rather a large flow rate for aerodynamic lens and requires system operation at low pressure requiring very large pumping capacity.

Sensitivity comparable to CAM can be achieved by condensational growth droplet concentrator at relatively high flow rates; for example, an from the condensational growth method of 137 µg-1 m3 (compared to 155 µg-1 m3 for CAM) can be achieved at 1 L min-1. However, the condensational growth method requires bulky hardware and is not conducive to compact portable instrumentation.

Table 2.1. Design and operating parameters of alternative particle collection methods for submicrometer size range.

Particle concentration Collection Flow rate (L Diameter of particle Sampling methods efficiency (%) min-1) deposition area (mm) time (minute) Filtration 100 1 – 20 13c 1 Aerodynamic lensa 96 – 99 0.01 – 0.2 0.5 1 Condensational growth 100 0.1 – 1 1 1 droplet concentrator CAMb 4 – 38 0.1 – 5 0.5 1 a A design tool for aerodynamic lens system (Wang and McMurry 2006a; b) was used to choose the parameters, and 100 nm test particles were used in the calculation. b The collection efficiency of CAM at 0.6, 1, 1.5, 2, 3, 4, and 5 L min-1 was obtained experimentally, and the collection efficiency at 0.1 L min-1 was assumed to be 100%. c 13 mm filter from SKC Inc., is assumed as a conservative estimate, though 25 mm and 37mm filter are most common for personal sampling

Figure 2.10. Comparison of the measurement sensitivity ( Sc ) of CAM with alternative concentration or collection methods employing condensational growth droplet concentrator, aerodynamic lens, and filtration.

2.4. Conclusions

A corona-based particle microconcentrator was designed and evaluated using spark microplasma optical emission spectroscopy. The CAM was designed to achieve aerosol microconcentration on the tip of a microelectrode using one-step corona-induced charging and electrostatic deposition. The collection efficiency was found to be relatively independent of particle size in the submicrometer size range under the operating conditions studied. The optical signal was found to be a function of the surface density of the analyte. An intrinsic spectroscopic sensitivity of CAM for Cr was determined to be 107.8 ng-1mm2 using the spark emission spectroscopy. Using this intrinsic sensitivity, it was shown that the CAM offers the highest spectroscopic sensitivity compared to alternative particle collection methods such as filtration, focused impaction using aerodynamic lens, and spot collection using condensational growth.

Compared to widely-used filter-based atomic emission methods, the CAM offers drastic improvements in LOD and time resolution of measurement. Although, the efficacy of the microconcentration method was demonstrated using spark emission spectroscopy in this work, the method can be readily used for sensitive aerosol analysis using other microscale optical spectroscopies such as laser-induced breakdown, Raman, infrared, and UV-VIS spectroscopies.

The method is particularly well-suited for compact, field-portable or personal instrumentation.

2.5. References

Barr, E. B., Hoover, M. D., Kanapilly, G. M., Yeh, H. C., Rothenberg, S. J. (1983). Aerosol Concentrator Design, Construction, Calibration, and Use. Aerosol Sci. Technol. 2:437-442. Chang, J. S., Lawless, P. A., Yamamoto, T. (1991). Corona discharge processes. IEEE Trans. Plasma Sci. 19:1152-1166. Cheng, Y.-S., Yeh, H.-C., Kanapilly, G. M. (1981). Collection efficiencies of a point-to-plane electrostatic precipitator. Am. Ind. Hyg. Assoc. J. 42:605-610. Chow, J. C. and Watson, J. G. (1999). Ion chromatography in elemental analysis of airborne

particles. Elemental analysis of airborne particles 1:97-137. Dahneke, B. and Flachsbart, H. (1972). An aerosol beam spectrometer. J. Aerosol. Sci. 3:345-349. Diwakar, P., Kulkarni, P., Birch, M. E. (2012). New Approach for Near-Real-Time Measurement of Elemental Composition of Aerosol Using Laser-Induced Breakdown Spectroscopy. Aerosol Sci. Technol. 46:316-332. Diwakar, P. K. and Kulkarni, P. (2012). Measurement of elemental concentration of aerosols using spark emission spectroscopy. J. Anal. At. Spectrom. 27:1101-1109. Eiguren Fernandez, A., Lewis, G. S., Hering, S. V. (2014). Design and Laboratory Evaluation of a Sequential Spot Sampler for Time-Resolved Measurement of Airborne Particle Composition. Aerosol Sci. Technol. 48:655-663. Fierz, M., Kaegi, R., Burtscher, H. (2007). Theoretical and experimental evaluation of a portable electrostatic TEM sampler. Aerosol Sci. Technol. 41:520-528. Ghio, A. J., Kim, C., Devlin, R. B. (2000). Concentrated ambient air particles induce mild pulmonary inflammation in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162:981- 988. Hering, S. V., Spielman, S. R., Lewis, G. S. (2014). Moderated, Water-Based, Condensational Particle Growth in a Laminar Flow. Aerosol Sci. Technol. 48:401-408. Hinds, W. C. (1999). Aerosol Technology: Properties. Behavior, and Measurement of airborne Particles (2nd. Keskinen, J., Janka, K., Lehtimäki, M. (1987). Virtual impactor as an accessory to optical particle counters. Aerosol Sci. Technol. 6:79-83. Laskin, A. and Cowin, J. P. (2002). On deposition efficiency of point-to-plate electrostatic precipitator. J. Aerosol. Sci. 33:405-409. Lee, K.-S., Cho, S.-W., Lee, D. (2008). Development and experimental evaluation of aerodynamic lens as an aerosol inlet of single particle mass spectrometry. J. Aerosol. Sci. 39:287-304. Liu, P., Ziemann, P. J., Kittelson, D. B., McMurry, P. H. (1995). Generating particle beams of controlled dimensions and divergence: I. Theory of particle motion in aerodynamic lenses and nozzle expansions. Aerosol Sci. Technol. 22:293-313. Miller, A., Frey, G., King, G., Sunderman, C. (2010). A handheld electrostatic precipitator for sampling airborne particles and nanoparticles. Aerosol Sci. Technol. 44:417-427. Morrow, P. E. and Mercer, T. T. (1964). A point-to-plane electrostatic precipitator for particle size sampling. Am. Ind. Hyg. Assoc. J. 25:8-14. Murphy, W. and Sears, G. (1964). Production of particulate beams. J. Appl. Phys. 35:1986-1987. NIOSH (2003). ELEMENTS by ICP (Hot Block/HCl/HNO3 Ashing): METHOD 7303. NIOSH Manual of Analytical Methods, Fourth Edition, Issue 1.

Park, K., Cho, G., Kwak, J.-h. (2009). Development of an aerosol focusing-laser induced breakdown spectroscopy (aerosol focusing-LIBS) for determination of fine and ultrafine metal aerosols. Aerosol Sci. Technol. 43:375-386. Raynor, P. C., Leith, D., Lee, K., Mukund, R. (2011). Sampling and analysis using filters. Aerosol measurement: principles, techniques and applications. Sioutas, C., Kim, S., Chang, M. (1999). Development and evaluation of a prototype ultrafine particle concentrator. J. Aerosol. Sci. 30:1001-1017. Sioutas, C., Koutrakis, P., Burton, R. M. (1995). A technique to expose animals to concentrated fine ambient aerosols. Environ. Health Persp. 103:172. Spurny, K. R. (1999). Analytical Chemistry of Aerosols: Science and Technology. CRC Press. Su, Y., Sipin, M. F., Furutani, H., Prather, K. A. (2004). Development and Characterization of an Aerosol Time-of-Flight Mass Spectrometer with Increased Detection Efficiency. Anal. Chem. 76:712-719. Wang, X. and McMurry, P. H. (2006a). A design tool for aerodynamic lens systems. Aerosol Sci. Technol. 40:320-334. Wang, X. and McMurry, P. H. (2006b). Instruction manual for the aerodynamic lens calculator. Aerosol Sci. Technol. 40:1-10. Wu, J., Cooper, D., Miller, R. (1989). Virtual impactor aerosol concentrator for cleanroom monitoring. The J. Environ. Sci. 32:52-56. Zavvos, K. (2016). Collection and Pre-Concentration of Aerosol for Optical Spectroscopies, University of Cincinnati. Ziemann, P. J., Liu, P., Rao, N. P., Kittelson, D. B., McMurry, P. H. (1995). Particle beam mass spectrometry of submicron particles charged to saturation in an electron beam. J. Aerosol. Sci. 26:745-756.

CHAPTER 3

Near Real-Time Measurement of Carbonaceous Aerosol Using Microplasma

Spectroscopy: Application to Measurement of Carbon Nanomaterials

 Published on Aerosol Science and Technology 54

A sensitive, field-portable microplasma spectroscopy method has been developed for real- time measurement of carbon nanomaterials. The method involves microconcentration of aerosol on a microelectrode tip for subsequent analysis for atomic carbon using laser-induced breakdown spectroscopy (LIBS) or spark emission spectroscopy (SES). The spark-induced microplasma was characterized by measuring the excitation temperature (15,000 – 35,000 K), electron density (1.0

× 1017 – 2.2 × 1017 cm-3), and spectral responses as functions of time and interelectrode distance.

The system was calibrated and detection limits were determined for total atomic carbon (TAC) using a carbon emission line at 247.856 nm (C I) for various carbonaceous materials including sucrose, EDTA, caffeine, sodium carbonate, carbon black, and carbon nanotubes. The limit of detection for total atomic carbon was 1.61 ng, equivalent to 238 ng m-3 when sampling at 1.5 L min-1 for 5 min. To improve the selectivity for carbon nanomaterials, which consist of elemental carbon (EC), the cathode was heated to 300 °C to reduce the contribution of organic carbon to the total atomic carbon. Measurements of carbon nanotube aerosol at elevated electrode temperature showed improved selectivity to elemental carbon and compared well with the measurements from thermal optical method (NIOSH Method 5040). The study shows that the SES method to be an excellent candidate for development as a low-cost, hand-portable, real-time instrument for measurement of carbonaceous aerosols and nanomaterials.

3.1. Introduction

Carbon nanomaterials, such as carbon nanotubes (CNT) and carbon nanofibers (CNF) have promising technological applications. However, the increasing production and widespread application of nanomaterials could have adverse health effects on workers who are routinely exposed to these materials (NIOSH 2013). Measurement of airborne nanomaterial concentration and composition is of great significance for the prevention of exposure and protection of workers’ health. Most existing methods for measurement of carbonaceous particles involve filter collection over several hours, followed by off-line analysis (Birch and Cary 1996; Chow et al. 1993).

Although filter-based methods are in common use for routine monitoring in both occupational and environmental settings, they have several drawbacks. In particular, they are time-integrated methods that are labor and time-intensive, with typical analysis turnaround times of several days to weeks. Field-portable, near real-time instruments can meaningfully augment filter-based methods to provide more timely as well as accurate characterization of aerosols to which workers may be exposed.

There are several real-time atomic and molecular spectrometric methods for carbonaceous aerosol analysis. Mass spectrometry based, real-time instruments have been widely used for both elemental and molecular analyses in atmospheric chemistry studies (DeCarlo et al. 2006; Gross et al. 2006), but they have not been practical for workplace aerosol measurements due to their high cost and lack of portability. Optical spectroscopies, particularly microplasma-based emission spectroscopies, are attractive alternatives for elemental speciation measurements (Broekaert 2002).

These techniques involve identifying and quantifying elements based on their characteristic atomic emission, which occurs at particular wavelengths when the sample is placed in a high-temperature excitation medium such as a flame, plasma, arc, or spark (Broekaert 2002; Diwakar and Kulkarni

2012; Hunter et al. 2000b; Martin et al. 1999; You et al. 1996). These plasma spectroscopy techniques have been applied to measure the carbon content of some solid samples, such as soil, coal, and fly ash (Dong et al. 2012; Glumac et al. 2010; Yao et al. 2012). Vors and Salmon (2006) reported a limit of detection (LOD) of 60 µg/m3 of carbon using test aerosols of glucose and

NaHCO3 for their laboratory benchtop laser-induced breakdown spectroscopy (LIBS) system. Lee and Yoh (2012) reported an LOD of 150.4 µg/m3 at a low flow rate and 240.4 µg/m3 at a high sample flow rate in their LIBS system. They used 15 µm test particles and analyzed them in-situ in the aerosol stream flowing past the LIBS microplasma. Though these and other studies have successfully measured airborne carbon, the detection limits achieved were not adequate for practical aerosol measurement applications, especially for workplace monitoring of CNT aerosols.

NIOSH has proposed a recommended exposure limit (REL) for CNT/CNF of 1 µg/m3, as a respirable elemental carbon (EC), 8-hour time weighted average (NIOSH 2013). Sensitive near real-time techniques that can provide accurate measurements with much lower detection limits are needed. The objective of this study was to investigate the effectiveness of LIBS and spark emission spectroscopy (SES) for near real-time measurement of carbonaceous aerosol using preconcentration techniques developed earlier (Diwakar and Kulkarni 2012).

Current methods for measurement of carbonaceous aerosols typically measure total carbon

(using various techniques) and attempt to distinguish between EC and organic carbon (OC) by heating a filter sample in inert and oxidizing atmospheres (Birch 1998; Birch and Cary 1996). The thermal-optical analyzer on which Method 5040 is based has been successfully applied to workplace monitoring of diesel particulate matter (Birch and Cary 1996) and carbon nanomaterials

(Birch et al. 2011a; Dahm et al. 2012; NIOSH 2013). At present, a semi-continuous OC-EC analyzer (Sunset Laboratory Inc., Tigard, OR), is widely used to measure time-resolved particulate

organic and elemental carbon concentrations in air quality studies (Batmunkh et al. 2011; Polidori et al. 2006). Normally, this semi-continuous instrument is operated with a 1-2 h sample collection period, depending on the EC concentration, followed by a 20 min sample analysis period (Bae et al. 2004).

In this paper, we describe a near real-time, plasma-based method for analysis of submicrometer airborne carbonaceous particles. The method collects particles on an electrode tip using electrostatic principles (Diwakar and Kulkarni 2012), following which emission spectrometric analysis is conducted. Various carbonaceous materials including organic and inorganic carbon (carbonates and carbon allotropes) were used to calibrate the instrument and establish the LODs. LODs, uncertainty, accuracy, and applications to nanomaterial measurements in air are presented and discussed.

3.2. Methods

3.2.1. Experimental setup

A schematic diagram of the experimental setup is shown in Figure 3.1. The major components include: (i) an aerosol generation system, (ii) an aerosol collection system, and (iii) an SES system.

Aerosol generation

Test aerosols were generated using a pneumatic atomizer (model 3076, TSI Inc., Shoreview,

MN, USA) to aerosolize solutions or suspensions prepared by mixing carbon-containing materials in ultra-filtered DI water (Fisher Scientific, Pittsburgh, PA, USA). The aerosol was passed through a diffusion dryer to remove associated water, and the resulting dry aerosol was passed through a differential mobility analyzer (DMA; model 3080, TSI Inc., Shoreview, MN, USA) to obtain a near-monodisperse test aerosol (with geometric standard deviation less than about 1.1 nm) in the 58

size range of 10 to 300 nm. An uncharged and near-monodisperse test aerosol was then produced by passing the aerosol through a neutralizer (a Po-210 source) and an electrostatic precipitator

(ESP). Typical number concentrations of the test aerosols generated in this study were in the range of 1×103 – 2×104 cm-3, depending on the carbon materials and concentrations of the solutions or suspensions.

Figure 3.1. Schematic diagram of the experimental setup used in this work.

Various carbon materials are used for calibration in this study, including sucrose, EDTA, caffeine, sodium carbonate (Na2CO3), carbon black, and CNT. Detailed description of these materials is shown in Table 3.1. Powder samples of 99.9% pure sucrose, EDTA, caffeine, and

Na2CO3 (Fisher Chemical, Pittsburgh, PA; Sigma-Aldrich, St. Louis, MO) were dissolved in ultra- filtered DI water to obtain solutions for aerosol generation. A carbon black suspension was

prepared by adding carbon black (REGAL® 400R, Cabot Corporation, Billerica, MA; >99% elemental carbon) to ultra-filtered DI water, and a CNT suspension was obtained by mixing hydrophilic single-walled carbon nanotubes (P7-SWNT, Carbon Solutions, Inc., Riverside, CA) into ultra-filtered DI water. The P7-SWNT nanotubes are surface functionalized with polyethyleneglycol and possess carbonaceous purity greater than 90%.

A constant flow rate of 1.5 L min-1 was maintained through the aerosol collection system and was driven by the native pump in the condensation particle counter (CPC; model 3022A, TSI

Inc., Shoreview, MN, USA). The overall flow scheme was controlled using a mass flow controller

(MFC; model 247 C, MKS Instruments, Inc., Andover, MA, USA).

Table 3.1. Description of carbon-containing materials used for calibration

Material Manufacturer Chemical Density, Carbon fraction composition g/cm3 by mass, %

Sucrose Fisher Chemical C12H22O11 1.58 42

EDTA Aldrich C10H16N2O8 0.86 41

Caffeine Aldrich C8H10N4O2 1.23 49

Sodium Carbonate Fisher Chemical Na2CO3 2.54 11 Carbon black Cabot C 1.8 > 99 CNT (P7-SWNT) Carbon Solutions C 2.1 > 90

Aerosol collection

An electrostatic aerosol collection system designed in an earlier study (Diwakar and

Kulkarni 2012) was used, with some modifications needed for monitoring atomic emission from carbon. As shown in Figure 3.2, the system consisted of two coaxial electrodes with a separation distance of 5 mm. A high positive potential (~5 kV) was applied on the corona electrode through a DC power supply (Bertan S-230, Spellman Corp., Hauppauge, NY, USA). The sidewalls of the electrodes were covered with a high dielectric strength sheath [polyether ether ketone (PEEK);

McMaster-Carr, Princeton, NJ, USA] to avoid particle deposition on the sidewalls. The corona electrode, made of tungsten, was 500 µm in diameter. The electrode has a sharp tip, with an approximate radius of 100 m, and was used to create a corona around its tip. The ground electrode was 1500 µm in diameter and had a relatively flat tip to provide a planar surface for particle deposition. Platinum ground electrodes were used to minimize spectral interference in carbon detection at 247.856 nm. The ground electrode was attached to a miniature resistive heating element to allow heating of the ground electrode up to 300 ºC. The ground electrode used for particle deposition was also used as the cathode for producing spark discharge.

Figure 3.2. Schematic diagram of the aerosol preconcentration system (not to scale).

SES setup 61

In the SES system, a high voltage pulse generator (Cascodium Inc., Andover, MA), was used to produce a spark microplasma in the interelectrode gap to ablate the particulate matter collected on the cathode. The pulse generator was designed for spark spectroscopy, with an output pulse energy up to 800 mJ. The size of the microplasma was measured to be about 500 µm in diameter. This diameter was measured by ablating a thin film of ink coated on the ground electrode surface and subsequently measuring the diameter of ablated spot using an optical microscope. The atomic emission signals from the microplasma were collected using a fiber optic cable connected to a broadband spectrometer with a wavelength range of 200 – 980 nm and a resolution of about

0.1 nm (LIBS2500 Plus, Ocean Optics Inc., Dunedin, FL, USA). Data acquisition and triggering of the spectrometers and laser and high voltage pulse generator were accomplished through

OOILIBS software (Version 4.5.07, Ocean Optics, Dunedin, FL, USA). A delay time of 5 µs was used in the SES system (Diwakar and Kulkarni 2012). A summary for the experimental parameters used in the SES systems is shown in Table 3.2. To probe the evolution of the carbon plasma generated by spark discharge, the temporal profiles of the C I transition (2s22p2 1S to 2s22p3s 1Po) at 247.856 nm were recorded using an intensified charge coupled device (ICCD; iStar 334T, Andor

Technology, South Windsor, CT, USA) at different distances from the collection electrode (cathode) surface and at different delay times.

Table 3.2. Experimental parameters used in the SES systems

Spark energy 200 mJ Delay time 5 µs Integration time 1 ms Spectrometer wavelength range 200–980 nm Aerosol flow rate 1.5 L min-1 Operating voltage for corona 5 kV Distance between the electrodes 5 mm Diameter of corona electrode 500 µm Diameter of collection electrode 1500 µm

3.2.2. Calibration method

Six different carbon materials were used to obtain calibration curves in this work: sucrose,

EDTA, caffeine, sodium carbonate (Na2CO3), carbon black, and single-walled carbon nanotubes

(SWCNT). A solution or suspension for each material was prepared for aerosol generation. The aerosol was collected on the platinum collection electrode (ground) for a predetermined time period ranging from a few seconds to a few minutes. The particle mass deposited on the electrode for a given particle diameter is given by (DeCarlo et al. 2004):

3 휋 퐶푐(푑푣푒) 푚푝 = 휂퐶𝑖푛푄푓 푡푐 𝜌푝 ( 푑푚) (1) 6 휒퐶푐(푑푚) where 휂 is the capture efficiency of particles, 퐶𝑖푛 is the particle concentration flowing into the chamber, 푄푓 is the aerosol volumetric flow rate, 푡푐 is the particle collection time, 𝜌푝 is the particle material density, 푑푣푒 is the volume equivalent diameter of particle, 푑푚 is the electrical mobility diameter, 퐶푐(푑푣푒) is slip correction factor for the volume equivalent diameter, 퐶푐(푑푚) is slip correction factor for the electrical mobility diameter, and 휒 is the dynamic shape factor (DSF) in the transition regime. DSF is defined as the ratio of the drag force on a nonspherical particle to the drag force on a spherical particle which has the same volume equivalent diameter and travels at the same relative velocity (Kulkarni et al. 2011). The DSF is 1 for spherical particles. For nonspherical particles, DSF is greater than 1 and needs to be determined experimentally for each

퐶푐(푑푣푒) particle type. For particles with 푑푚 > 100 nm and 휒 < 2.3, is less than 1.1. By assuming 퐶푐(푑푚)

퐶 (푑 ) 푐 푣푒 = 1, Equation (1) can be simplified as follows, 퐶푐(푑푚)

휋 푑 3 푚 = 휂퐶 푄 푡 𝜌 ( 푚) (2) 푝 𝑖푛 푓 푐 푝 6 휒

Particle capture efficiency was calculated by measuring the particle number concentration downstream of the collection unit using a CPC, with and without the presence of the electric field

퐻푉 푉=0 across the electrodes ( 푁표푢푡 and 푁표푢푡 ).

푉=0 퐻푉 푁표푢푡 − 푁표푢푡 휂 = 푉=0 (3) 푁표푢푡

Mass loading on the collection electrode was changed by varying the collection time or the particle concentration at the inlet. For a given aerosol generated, and for a given mass, three samples were collected over the same collection time. The final calibration curve was constructed by averaging over these three replicate measurements for each mass level.

The carbon material deposited on the electrode was then ablated by a laser-induced spark

(in the LIBS setup) or the high voltage pulsed spark (in the SES setup). Usually, a single pulse was sufficient to ablate the entire particulate mass; higher loadings required an additional one or two pulses. The atomic emission from each ablation was collected using the spectrometer. The signal intensity was subsequently converted to a peak area, defined as the integrated atomic emission line intensity below the peak after subtracting the baseline area. The baseline spectrum was obtained using an identical measurement, except with the absence of analyte on the electrode. For those measurements that required multiple pulses, the total signal intensity was calculated as the sum of the carbon signal intensity from each pulse. The calibration curve was constructed by plotting signal intensity as a function of mass of atomic carbon in the particulate matter deposited on the collection electrode. The LOD was estimated using 3-𝜎 criteria defined by the International Union of Pure and Applied Chemistry (IUPAC; Boumans 1994).

3.3. Results and discussion

3.3.1. Sensitivity of different carbon emission lines

Figure 3.3 shows the spectra obtained from spark discharge excitation of sucrose particles collected in the preconcentrator. Six carbon emission lines, including neutral emission line C I and

ionic emission line C II, were identified using the National Institute of Standards and Technology

(NIST) atomic spectra database. As shown in Figure 3.3, atomic emission peaks for carbon appear at wavelengths of 247.856 nm, 426.726 nm, 657.805 nm, 723.132 nm, 723.642 nm, and 940.573 nm when analyzing sucrose by SES. For a given carbon mass loading, the strongest signal was observed at 247.856 nm. Comparing the four spectra obtained from sucrose particles at different carbon loadings (Figure 3.3), we found that only the 247.856 nm emission line was detected at loadings below 15.9 ng. All carbon emission peaks appeared when the carbon loadings were above

31.8 ng. These measurements show that the sensitivity (S) for carbon measurement is different for different emission lines, with the 247.856 nm line exhibiting the highest sensitivity. This finding is consistent with other studies on emission lines for atomic carbon (Bricklemyer et al. 2011; Dong et al. 2012; Glumac et al. 2010; Vors and Salmon 2006). Figure 3.4(a-f) shows the intensity of all the carbon emission lines from sucrose plotted as a function of carbon mass. Based on the sucrose data, we assumed that sensitivity will be highest at 247.856 nm for the other carbonaceous aerosols, without consideration of effects of allotrope or molecular structure on carbon signal. Therefore, this emission line was used for all materials studied in this work.

Figure 3.3. Carbon emission lines identified in this work using spark emission spectroscopy. The spectra correspond to different particulate carbon mass loadings on the collection electrode.

3.3.2. Plasma characteristics

Figure 3.5(a) shows the contour map of the spatial and temporal variation of the carbon emission signal (the spatial resolution is 0.4 mm and time resolution is 1 µs). After initiation of the plasma, carbon emission only appeared in the region close to the cathode surface, at 1 µs, and the carbon emission signal was weak. With further evolution of the spark plasma, the carbon plume expands towards the anode. At 11 µs, the emission signal was highest at 1.4 mm away from the cathode surface (Figure 3.5(b)), and the carbon plume extended upward towards the anode surface.

After 11 µs, the carbon emission signal intensity decreased rapidly with time. The lifetime of the plasma was estimated to be about 15 µs. The spatial and temporal trend in spark plasma emission observed in our system was consistent with those described in earlier studies on spark discharge formation and evolution (Walters 1972; Walters and Goldstein 1984). As Walters (1977) noted, when a high voltage pulse is applied between the electrodes, a conducting ion channel is first established to induce dielectric breakdown of air and is directed towards the cathode. This conducting channel further spatially expands due to energy deposition in the interelectrode space, leading to a pulsed spark discharge. After the energy deposition initiates, the plasma plume at the cathode surface expands rapidly due to space-charge effects, while the central conducting channel rapidly shrinks as the energy deposition in the interelectrode gap ceases. Our measurements are consistent with these mechanisms of spark discharge formation and propagation. From these measurements (Figure 3.5(a)), we estimated the transport time of the carbon species between the cathode and anode to be about 8 µs. The axial velocity of carbon species, defined as the ratio of inter-electrode distance to transport time between the cathode and anode, was estimated to be 0.6 mm/µs. Axial velocity reported by Walters and Goldstein (1984) was approximately 1.0 mm/µs for copper species and 0.8 mm/µs for silver in a similar pulsed spark plasma. The variations in the

axial velocity of emission fronts are perhaps due to differences in spark characteristics, electrode characteristics, electrode configuration, and the chemical species.

Figure 3.5. (a) Spatial and temporal signal intensity of carbon emission line at 247.856 nm from spark plasma, (b) Variation of carbon signal intensity as a function of interelectrode distance.

Using a Boltzmann plot, the estimated plasma excitation temperature in our system, based on C II emission, varied in the range of 15,000 – 35,000 K depending on the interelectrode distance and delay time. The electron density was estimated by examining Stark broadening of the Hα line

(656.28 nm) and was found to be in the range 1.0 × 1017 – 2.2 × 1017 cm-3. The laser-induced plasma in our system was not characterized in this study; however, Harilal et al. (1997) have reported the excitation temperature in a similar LIBS carbon plasma to be in the range of 1.6 – 2.4 eV (equivalent to 18,500 – 27, 800 K), and an electron density in the range of 1.1 × 1017 – 2.1 ×

1017 cm-3. Both plasma types, i.e., laser and spark-induced, exhibited similar ranges for excitation temperature and electron density.

3.3.3. System calibration

The effect of particle size on the carbon signal was probed before constructing calibration curves for polydisperse carbonaceous aerosol. Calibration curves for sucrose particles of diameter

50 nm, 100 nm, 150 nm, 200 nm and 300 nm were obtained and compared using the SES system.

Figure 3.6 shows calibration curves for various particle diameters over the particulate mass range of 4 – 140 ng. To construct these calibration curves, the mass of particles collected on the electrodes was varied and the corresponding emission signal was recorded for each mass loading. Three methods were used to vary the particulate mass loading on the electrode. First, the mass loading was changed by changing the particle diameter of the DMA-classified aerosol (Figure 3.6(a)). The second approach involved changing the mass loading by changing the collection time over which particles accumulate on the electrode, for a given particle size (100 nm) and aerosol number concentration (Figure 3.6(b), filled symbols). The third method involved changing the inlet concentration of the aerosol to change the particulate mass loading (while keeping the size and collection times the same). Figure 3.6(a)-(b) show that

within the range of experimental uncertainty, the sucrose calibration curves agree well. The method sensitivity (S), based on the calibration curves, agreed to within 15% for these three methods. The data in Figure 3.6(a) also imply that the emission signal was independent, within the range of experimental uncertainty, of the particle size deposited on the collection electrode. This ensures that calibration curves constructed using monodisperse particles were applicable when analyzing polydisperse aerosol in the submicrometer size range investigated.

Calibration curves for organic and inorganic materials, including EDTA, caffeine, Na2CO3, carbon black, and CNT were also obtained using the SES setup. The slope of each linear calibration curve for different carbon materials and the correlation coefficients are shown in Table 3.3. The sensitivity (S) for carbon measurement was highest for CNT, followed by sucrose, carbon black, caffeine, Na2CO3, and EDTA. The reasons for difference in sensitivity across the different materials are unknown, but part of the variability could relate to varying particle morphology

(across different carbon materials), which indirectly affects the mass loading estimation (which assumed spherical particle shape). For nonspherical particles with χ>1, particulate mass calculated using spherical mobility diameter would lead to an overestimation of calculated mass (Kulkarni et al. 2011), which can range from approximately 33% at χ = 1.1 to 237% at nonspherical shapes with χ = 1.5. This error in mass estimation may partially explain the difference in sensitivity across particle types, especially for carbon black and CNT, which were relatively more nonspherical in shape. Dynamic shape factors of carbon black and CNT particles were not measured in this study.

However, DSFs of approximately 1.5 for diesel exhaust particles ((Park et al. 2004)) and soot particles (푑푚 = 100 nm; Slowik et al. (2004)) have been reported for aerosols generated by the combustion of diesel or gaseous fuel, of which elemental or black carbon was the main component.

Ku and Kulkarni (2015) reported a DSF of 1.8 for CNT particles (푑푚 = 100 nm) generated by

pneumatically atomizing CNT suspension (using the same system used in this work). Based on these values, a DSF of 1.5 was assumed for carbon black particles and 1.8 for CNT particles in this study. Since sucrose, EDTA, caffeine and Na2CO3 aerosol were obtained from dissolved solids, their DSF was assumed to be 1.

Table 3.3. LODs in terms of mass and air concentration obtained by SES

LODs* Sensitivity RSD% Material Air R2 (S) Mass, ng concentration, range ng/m3 Sucrose 28.3 1.63 241 0.96 7 - 12 EDTA 22.6 1.94 287 0.99 3 - 7 Caffeine 25.9 1.70 252 0.96 2 - 9 Na2CO3 24.2 1.82 269 0.97 2 - 14 Carbon black 25.6 1.76 261 0.95 3 - 11 CNTs 34.8 1.26 187 0.96 3 - 12 Ensemble 27.9 1.61 238 0.97 2 - 14 *LOD in terms of air concentration calculated by assuming a collection time of 5 min, flow rate of 1.5 L min-1, and collection efficiency of 90%.

Figure 3.7 shows a calibration curve for the pooled data from all carbonaceous materials used in this study. A single calibration curve was obtained for the entire data set. The sensitivity of the ‘ensemble’ calibration curve was 27.9 a.u. ng-1, which was about 23% higher than the lowest sensitivity measured (for EDTA) and 20% lower than the highest sensitivity measured (for CNT) in this study. The ensemble calibration curve in Figure 3.7 is useful for measurement of carbonaceous aerosol where the nature or structure of the carbonaceous matter is not known a priori. The pooled, ensemble calibration curve in Figure 3.7 captures uncertainties in atomic emission signal due to different molecular forms of carbon allotropes, different non-carbon matrices (i.e. matric effects), and uncertainties associated with the particle collection (i.e. deposited particulate mass on electrodes). 72

Using the ensemble calibration curve may lead to an increased error ranging from 1.4% for sucrose to 23% for EDTA. This error is comparable to the uncertainty involved in the measurement of the calibration curves. The relative standard deviation (RSD) calculated by three replicate measurements varied in a range of 2–14%, as shown in Table 3.3, representing the reproducibility of a spectral response when measuring airborne particles using SES.

Figure 3.7. Calibration curves for sucrose, EDTA, caffeine, sodium carbonate, carbon black, and

CNT constructed by SES.

The applicability of LIBS on carbonaceous aerosol measurement in the same aerosol collection system was also studied. Figure 3.8 shows calibration curves obtained from both microplasma systems, LIBS and SES, for carbon black particles using the 500 µm electrode. As mentioned earlier, the laser beam was oriented perpendicular to the longitudinal axis of the electrode. As discussed elsewhere (Diwakar et al., 2012), this orientation required a correction to account for mass loading in the shadow region. The correction factor was approximately 14% of

the total mass collected on the tip (Diwakar and Kulkarni 2012) in our setup. In the SES system, no such correction was necessary due to the symmetrical nature of the ablation (about the longitudinal axis of the electrode). Comparable sensitivity for carbon black was obtained from both LIBS and SES microplasma systems.

Figure 3.8. Calibration curves for carbon black by LIBS and SES using the collection electrode with a diameter of 500 µm.

3.3.4. Limits of Detection

As listed in Table 3.3, the mass LODs range from 1.26 ng to 1.94 ng. By assuming a collection time of 5 min and a flow rate of 1.5 L min-1, the LOD in terms of air concentration is in the range 186 – 287 ng/m3. Much lower LODs can be achieved by increasing the collection time and/or flow rate. Obviously, there is a conflict between LOD and time resolution. However, thermal analysis methods have higher EC LODs (e.g., typically about 1 µg/m3) and require filter sampling over relatively long periods, and/or collection at high flow rates (air volumes), to achieve

comparable LODs (Bae et al. 2004; Birch and Cary 1996; Rupprecht et al. 1995; Schauer et al.

2003; Turpin et al. 1990). Other studies on measurement of carbonaceous particulate matter by plasma spectroscopy reported carbon LODs of 60 µg/m3 or higher (Lee and Yoh 2012; Vors and

Salmon 2006), much higher than the LODs found in our study.

3.3.5. CNT Measurement Comparison

SES measurements of unknown CNT aerosol samples were compared with those from the other commonly used methods for measurement of carbonaceous aerosol. Four suspensions with different CNT concentrations were prepared and used to generate the test aerosols. The airborne concentration of carbon in the test aerosol was determined using our SES system. The ensemble calibration curve (Figure 3.7) was used to obtain the carbon particulate mass deposited on the electrode, which was then inverted to an air concentration. Parallel and independent measurements of the test aerosol were obtained using two separate methods: i) NIOSH Method 5040, and ii) a portable aethalometer (microAeth®Model AE51, AethLabs, San Francisco, CA, USA) for ‘black carbon’ (BC).

As discussed, NIOSH Method 5040 is a thermal-optical technique that measures elemental and organic carbon. In addition to diesel particulate matter (as EC), Method 5040 has been applied to workplace monitoring of CNT/CNF (Birch et al. 2011b). Because CNT/CNF are composed of

EC, EC is a quantitative measure of airborne CNT/CNF (Birch et al. 2011b). A manual OC-EC split is assigned in the 5040 analysis of CNT/CNF as the relatively large particle size (μm-scale as opposed to nanoscale DPM) and agglomerate structure make the auto-split unreliable (NIOSH

2013). The portable aethalometer measures optical absorption of BC (i.e., light absorbing) aerosol collected on a filter, at 880 nm wavelength. However, aethalometer measurements can result in large bias if the size distribution and refractive index of the aerosol are vastly different from those

of the aerosol used for instrument calibration. All three methods (SES, NIOSH Method 5040, and aethalometer) use different measurement principles. As such, the agreement between them will depend on the aerosol properties, including particle size distribution, refractive index, relative abundance of organic and inorganic carbon, sample matrix components, particle morphology and structure. Figure 3.9 shows a comparison of CNT aerosol measurements by SES and the aethalometer, with EC determined by NIOSH Method 5040. SES measurements were, on an average, 13.8% lower than the EC results determined by NIOSH Method 5040 (with manual split), while the average aethalometer measurements were higher by 12.0%. We note that the three methods used in this study measure different physical or chemical properties of the aerosol, which, in turn serve as a surrogate measures of particular species or form of particulate carbon of interest.

All three methods should agree closely, if: i) the same calibration aerosol is used calibration of all methods, ii) all methods are calibrated with respect to the same reference measurement method, and iii) particle size/shape and sample matrices of the unknown sample and calibration aerosol are the same. However, this was not the case in our study; default calibrations for both aethalometer and the NIOSH methods were used. As a result there was a large difference in the size distribution and chemical composition of the calibration aerosol as well as the calibration methods used for the three methods. These factors may explain the difference in carbon measurements reported in

Figure 3.9.

Figure 3.9. Comparison of carbon nanotube concentrations measured from SES, Aethalometer, and NIOSH Method 5040.

3.3.6. SES measurements at elevated electrode temperature

Our SES method measures total atomic carbon (TAC), including the EC and OC fractions in the particulate sample. To reduce the potential interference of OC (if present), we examined the effectiveness of SES measurements at elevated electrode temperature. Specifically, the particle collection electrode was heated to a temperature of up to 300 °C (higher temperatures could not be maintained in this study, though they could possibly be achieved with redesign of the heating elements). The OC fraction can vary, but it will likely be reduced with increasing temperature, through various processes such as evaporation, oxidation, and thermal decomposition. The degree of OC interference depends on the nature of the organic matter. Under the thermal conditions of the electrode surface in our SES system, held at or below 300 °C, vaporization is the most likely mechanism for removing organic compounds from the particulate matrix on the electrode

(Lapuerta et al. 2007). However, some nonvolatile organic compounds (e.g., sucrose) could decompose to form volatile components (Šimkovic et al. 2003), and any residue or other nonvolatile organic components could preferentially be oxidized from the electrode. In our system, carbonization of carbonaceous particulate matter in an inert atmosphere does not occur as oxygen is present continuously during the analysis.

In our tests, particles were first collected at room temperature over a predetermined period, following which the heater was activated to increase the electrode temperature, which was then held at a set value for two minutes. Subsequently, the spark microplasma was introduced and the atomic emission signal was recorded using the procedure described earlier. During the sample heating, a constant air flow rate of 1.5 L min-1 was maintained through the chamber, which ensured quick removal of products of evaporation away from the interelectrode space. Figure 3.10 shows the atomic emission signals for a given initial mass, as a function of electrode temperature, for the organic (a) and inorganic (b) materials examined in this study. All organic materials, with the exception of humic acid sodium salt, show a clear reduction in signal with increasing temperatures, above approximately 200 °C (Figure 3.10(a)). There was no detectable particulate carbon on the electrode beyond 300 °C (except for humic acid sodium salt). This behavior is consistent with the fact that these organic materials have boiling point or decomposition temperatures below 300 °C

(Šimkovic et al. 2003; Wendlandt 1960). However, the signal from humic acid sodium salt was unchanged, even when the electrode was heated to 300 °C, likely due to the higher temperature required to remove this material (manufacturer specifies a melting point greater than 300 °C). We conducted thermal optical analysis of humic acid sodium salt using the Sunset Laboratory EC/OC analyzer. The thermal program was modified to conduct the analysis in the oxidative mode (5%

O2/He mixture) to approximately mimic the oxidative conditions in SES analysis. No peak was

observed below 300 °C, implying that very little mass would be lost from the electrode below

300 °C.

Figure 3.10. Change in carbon signal intensity as a function of electrode temperature for various organic (a) and inorganic (b) carbonaceous materials.

None of the inorganic carbon materials tested showed a change in the emission signal with increasing electrode temperature (Figure 3.10(b)), but electrode temperatures higher than 300 °C could not be achieved due to design constraints. Figure 3.10 indicates that the OC contribution to the TAC could be further reduced if SES measurements are conducted after heating the electrode beyond 300 °C.

It is well known that OC-EC measurement by thermal analysis techniques is method dependent. Among other factors, speciation depends on the sample composition, particle morphology, atmosphere surrounding the sample, sample substrate and matrix, thermal program, instrument design (including char correction method, if used), and nature of the thermal decomposition process (pyrolysis/combustion/volatilization). Thermal techniques used for OC-EC speciation in airborne particulate matter have mainly included thermal-optical transmittance (TOT;

Birch and Cary 1996; Birch 1998; Peterson and Richards 2002; Chow et al. 2005; Bauer et al.

2009; Pavlovic et al. 2014), thermal/optical reflectance (TOR; Han et al. 2007; Chow et al. 2007;

Pavlovic et al. 2014), thermogravimetric analysis (TGA; Iwatsuki et al. 1998; Stratakis and

Stamatelos 2003; Lapuerta et al. 2007); and thermal manganese dioxide oxidation (TOM; Fung

1990; Park et al. 2005). Current methods are mainly based on NIOSH Method 5040 and similar protocols, which use a Sunset Laboratory analyzer, or the IMPROVE protocol, based on a different analyzer design (Desert Research Institute [DRI], Reno, NV). Some of the OC-EC methods that have been applied, including direct-reading methods, and their operating parameters are summarized in Table 3.4. Different sample types can exhibit varying degrees of carbonization. A major cause of disagreement between the samples is the extent of carbonization and correction for it which, in turn, depends on the thermal program and instrument design (Birch 1998). In inert atmosphere, a maximum of 550 °C (used in the IMPROVE protocol) gives higher EC results than

the Speciation Trends Network (STN) and NIOSH-based protocols, which typically use 850 °C or higher (Birch and Cary 1996; Birch 1998).

Table 3.4. Comparison of different OC/EC thermal analysis methods

Method OC Analysis EC Analysis Carrier gas Temperatur Residence Carrier gas Temperatur Residence e plateaus time (s) e plateaus time (s) (°C) (°C) Our method N/A N/A N/A Ambient air 300 120 NIOSH Method He 250, 500, 60,60,60, 90 2% O2, 98% 650, 750, 30, 30, 60, 5040_TOT† 600, 850 He 850, 940 120 (Birch 1998) IMPROVE_TO He 120, 250, 150 – 580 2% O2, 98% 550, 700, 150 – 580 R (Chow et al. 450, 550, He 800 2007) STN_TOR/TO He 310, 480, 60, 60, 60, 2% O2, 98% 600, 675, 45, 45, 45, T (Chow et al. 615, 900 90 He 750, 815, 45, 120 2005) 920 HKUST- He 250, 550, 150, 150, 1% O2, 98% 650, 750, 150, 150, 3_TOT (Chow 650, 850 150, 110 He 850, 890 150, 150 et al. 2005) TMO (Fung et He 120, 525 180, 300 2.5% O2, 750 180 al. 2002) 97.5% He TGA (Lapuerta N2 450 Heating Ambient air 500 1800 et al. 2007) rate: 3 °C per min TGA (Iwatsuki Ambient air 430 3600 Ambient air Continuous Heating et al. 1998) (430 to 800) rate: 10 °C per min ACE- He 340, 500, 60, 60, 60, 2% O2, 98% 550, 625, 45, 45, 45, Asia_TOT 615, 870, 90, 45 He 700, 775, 45, 45, 120 (Schauer et al. oven off 850, 900 2003) EUSAAR_TO He 200, 300, 120, 150, 2% O2, 98% 500, 550, 120, 120, T (Cavalli et al. 450, 650 180, 180 He 700, 850 70, 80 2010) Coulometric N2 200, 400, 120, 120, O2 800 270 method 550 240 (ZH1/120.44 1995) R&P 5400 Ambient air 340 600 Ambient air 600 750 (continuous analyzer) (Rupprecht et al. 1995) †Temperature steps and times may vary but give comparable OC-EC results (as single fractions). Maximum in helium: ≥ 650 °C, with 850 °C typical. Maximum in O2/He: 920 °C, but samples can oxidize at higher or lower 81

temperatures. A manual OC-EC split may be required with some applications (NIOSH 2016).

To further probe the degree of selectivity for EC, measurements were conducted on four test aerosols, which were atomized from four liquid solutions (A–D) of organic and elemental/inorganic materials mixed in varying proportions. Composition and liquid concentration of each component in these solutions are shown in Table 3.5. As our focus was on the measurement of EC, only one EC and three OC fractions were used. These mixtures resulted in EC/OC ratios in the range 1-13 as shown in Table 3.5. The carbon particulate mass deposited on the electrode was obtained based on the ensemble calibration curve (Figure 3.7), and then inverted to an air concentration. Total atomic carbon results from SES measurements of the test aerosols generated from these four samples (A–D) were compared with the corresponding EC measurements by

NIOSH Method 5040. Results of the comparison are shown in Figure 3.11, at electrode temperatures of 25 °C and 300 °C. The pie charts in Figure 3.11 show the composition of the mixtures as relative fractions of CNT, sucrose, EDTA, and caffeine in each sample: A, B, C, and

D. The ensemble calibration curve was used to obtain TAC in SES measurements. The lowest EC concentration was limited to 12 µg/m3 in these tests. At this air concentration, the detection limits of NIOSH Method 5040 required a minimum of 2 hours of sample collection. The test aerosol was relatively stable during this 2-h period. SES measurements (with collection time of 2 minutes) were obtained every 30 minutes during this 2-h period (for each sample mixture). Each data point in Figure 3.11 is the average of four SES measurements. Figure 3.11 shows linear fits to data obtained at two electrode temperatures of 25 and 300 °C. A slope of 1 for these linear fits would indicate excellent agreement between TAC and EC. The Figure 3.11 shows that, for the carbon concentration range studied, SES measurements overestimate EC by 6%, 14%, 28% and 29% for sample A, B, C and D, respectively at 25 °C; whereas they underestimate the EC about 13%at an

electrode temperature of 300 °C. The TAC at 25 °C was obtained before the separation of OC and

EC, such that it was higher than the EC results from NIOSH Method 5040. In contrast, the TAC at

300 °C agreed well with EC results determined by NIOSH Method 5040, being within the range of measurement uncertainty. Thus, our study indicates that TAC measurement at elevated electrode temperature can improve the selectivity for EC.

Table 3.5. Composition and liquid concentration of solutions A, B, C, and D used to produce test aerosols.

Carbon concentration in the solutions, µg/mL Samples Total atomic carbon EC/OC CNT Sucrose EDTA Caffeine concentration, µg/mL A 40 1 1 1 43 13.3 B 80 5 5 5 95 5.3 C 120 15 15 15 165 2.7 D 160 50 50 50 310 1.1

The time resolution of our SES carbon measurement will depend on aerosol sampling flow rate, carbonaceous aerosol concentration, and the limits of quantification (LOQ) of the method.

Assuming a flow rate of 2 L min-1, the calculated particle collection time for achieving LOD in terms of mass was in the range of 1 – 15 minutes at an air concentration of carbonaceous aerosol in the range 0.2 – 3 µg/m3. Each measurement cycle consists of three following steps: i) particle collection (~minutes), ii) wait time (< 30s), and iii) repeated ablations of the particulate sample and atomic emission measurement (< 30s). The apparent time resolution (i.e. duty cycle), corresponding to the above particle collection times, therefore would be in the range of approximately 2 – 16 minutes. The commercially available, semi continuous OC-EC aerosol analyzer from Sunset Laboratory Inc. has time resolution of 30 minutes to 8 hours, and the R&P

5400 Ambient Particulate Carbon Monitor has time resolution of 1 hour (Lim et al. 2003). The

SES method is capable of providing more sensitive detection at higher time resolution.

3.4. Conclusions

Microelectrode-based preconcentration was successfully coupled with microplasma emission spectroscopy, using laser-induced and spark plasmas as excitation sources, for near real- time measurement of carbonaceous aerosols. Calibration curves were obtained for various pure organic and inorganic materials including, sucrose, EDTA, caffeine, Na2CO3, carbon black, and 84

CNT. The measurement sensitivity was found to vary across types of carbonaceous material for a given excitation source; the difference between the lowest and the highest measured sensitivity was about 54%. A single calibration curve could be obtained by pooling together the calibration data for all the organic and inorganic carbonaceous materials. The method intrinsically measures

TAC. It was shown that selectivity to inorganic carbon can be improved by conducting measurements at elevated electrode temperatures of up to 300 oC. Measurements at electrode temperatures beyond 300 oC should allow further improvement in measurement selectivity to engineered carbonaceous nanomaterials. LOD (in terms of TAC) was found to be in the range of

1.26 – 1.94 ng, which corresponds to 187 – 287 ng m-3, in terms of air concentration at a sample flow rate of 1.5 L min-1 and a sampling time of 5 min. The reproducibility of spectral response for laboratory generated aerosol was in the range of 2 – 14%. Measurement of carbon concentration of test aerosols, generated by using complex mixtures of organic, inorganic, and CNT material, using our SES method agreed well with those from the NIOSH method 5040. The SES method presented here has higher sensitivity and time resolution (in the range of 2 – 16 min) compared to current commercial methods. The method is particularly suited for development of low-cost portable hand-held instruments for personal or mobile aerosol measurement applications.

3.5. References

Bae, M.-S., Schauer, J. J., DeMinter, J. T., Turner, Smith, D. (2004). Validation of a semi- continuous instrument for elemental carbon and organic carbon using a thermal-optical method. Atmos. Environ. (1994) 38:2885-2893. Batmunkh, T., Kim, Y. J., Lee, K. Y., Cayetano, M. G., Jung, J. S., Kim, S. Y., Kim, K. C., Lee, S. J., Kim, J. S., Chang, L. S., An, J. Y. (2011). Time-Resolved Measurements of PM2.5 Carbonaceous Aerosols at Gosan, Korea. J. Air Waste Manage. 61:1174-1182. Bauer, J. J., Yu, X.-Y., Cary, R., Laulainen, N., Berkowitz, C. (2009). Characterization of the sunset semi-continuous carbon aerosol analyzer. J. Air Waste Manage. 59:826-833. Birch, M. E. (1998). Analysis of carbonaceous aerosols: interlaboratory comparison. Analyst

123:851-857. Birch, M. E. and Cary, R. A. (1996). Elemental carbon-based method for monitoring occupational exposures to particulate diesel exhaust. Aerosol Sci. Technol. 25:221-241. Birch, M. E., Ku, B.-K., Evans, D. E., Ruda-Eberenz, T. A. (2011a). Exposure and Emissions Monitoring during Carbon Nanofiber Production-Part I: Elemental Carbon and Iron-Soot Aerosols. Ann. Occup. Hyg. 55:1016-1036. Birch, M. E., Ku, B.-K., Evans, D. E., Ruda-Eberenz, T. A. (2011b). Exposure and emissions monitoring during carbon nanofiber production—Part I: elemental carbon and iron–soot aerosols. Ann. Occup. Hyg. 55:1016-1036. Boumans, P. (1994). Detection limits and spectral interferences in atomic emission spectrometry. Anal. Chem. 66:459A-467A. Bricklemyer, R. S., Brown, D. J., Barefield, J. E., Clegg, S. M. (2011). Intact soil core total, inorganic, and organic carbon measurement using laser-induced breakdown spectroscopy. Soil Sci. Soc. Am. J. 75:1006-1018. Broekaert, J. A. C. (2002). The development of microplasmas for spectrochemical analysis. Anal. Bioanal. Chem. 374:182-187. Cavalli, F., Viana, M., Yttri, K. E., Genberg, J., Putaud, J. P. (2010). Toward a standardised thermal- optical protocol for measuring atmospheric organic and elemental carbon: the EUSAAR protocol. Atmos. Meas. Tech. 3:79-89. Chow, J., Watson, J., Chen, L.-W., Paredes-Miranda, G., Chang, M.-C., Trimble, D., Fung, K., Zhang, H., Zhen Yu, J. (2005). Refining temperature measures in thermal/optical carbon analysis. Atmos. Chem. Phys. 5:2961-2972. Chow, J. C., Watson, J. G., Chen, L.-W. A., Chang, M. O., Robinson, N. F., Trimble, D., Kohl, S. (2007). The IMPROVE_A temperature protocol for thermal/optical carbon analysis: maintaining consistency with a long-term database. J. Air Waste Manage. 57:1014-1023. Chow, J. C., Watson, J. G., Pritchett, L. C., Pierson, W. R., Frazier, C. A., Purcell, R. G. (1993). The DRI thermal/optical reflectance carbon analysis system: description, evaluation and applications in US air quality studies. Atmos. Environ. A General Topics 27:1185-1201. Dahm, M. M., Evans, D. E., Schubauer-Berigan, M. K., Birch, M. E., Deddens, J. A. (2012). Occupational exposure assessment in carbon nanotube and nanofiber primary and secondary manufacturers: mobile direct-reading sampling. Ann. Occup. Hyg. 57(3):328-344. DeCarlo, P. F., Kimmel, J. R., Trimborn, A., Northway, M. J., Jayne, J. T., Aiken, A. C., Gonin, M., Fuhrer, K., Horvath, T., Docherty, K. S. (2006). Field-deployable, high-resolution, time-of-flight aerosol mass spectrometer. Anal. Chem. 78:8281-8289. DeCarlo, P. F., Slowik, J. G., Worsnop, D. R., Davidovits, P., Jimenez, J. L. (2004). Particle morphology and density characterization by combined mobility and aerodynamic diameter measurements. Part 1: Theory. Aerosol Sci. Technol. 38:1185-1205.

Diwakar, P., Kulkarni, P., Birch, M. E. (2012). New Approach for Near-Real-Time Measurement of Elemental Composition of Aerosol Using Laser-Induced Breakdown Spectroscopy. Aerosol Sci. Technol. 46:316-332. Diwakar, P. K. and Kulkarni, P. (2012). Measurement of elemental concentration of aerosols using spark emission spectroscopy. J. Anal. At. Spectrom. 27:1101-1109. Dong, M., Mao, X., Gonzalez, J. J., Lu, J., Russo, R. E. (2012). Time-resolved LIBS of atomic and molecular carbon from coal in air, argon and helium. J. Anal. At. Spectrom. 27:2066-2075. Fung, K. (1990). Particulate carbon speciation by MnO2 oxidation. Aerosol Sci. Technol. 12:122- 127. Fung, K., Chow, J. C., Watson, J. G. (2002). Evaluation of OC/EC speciation by thermal manganese dioxide oxidation and the IMPROVE method. J. Air Waste Manage. 52:1333-1341. Glumac, N. G., Dong, W. K., Jarrell, W. M. (2010). Quantitative Analysis of Soil Organic Carbon Using Laser-Induced Breakdown Spectroscopy: An Improved Method. Soil Sci. Soc. Am. J. 74:1922-1928. Gross, D. S., Gälli, M. E., Kalberer, M., Prevot, A. S., Dommen, J., Alfarra, M. R., Duplissy, J., Gaeggeler, K., Gascho, A., Metzger, A. (2006). Real-time measurement of oligomeric species in secondary organic aerosol with the aerosol time-of-flight mass spectrometer. Anal. Chem. 78:2130-2137. Han, Y., Cao, J., Chow, J. C., Watson, J. G., An, Z., Jin, Z., Fung, K., Liu, S. (2007). Evaluation of the thermal/optical reflectance method for discrimination between char-and soot-EC. Chemosphere 69:569-574. Harilal, S. S., Bindhu, C. V., Issac, R. C., Nampoori, V. P. N., Vallabhan, C. P. G. (1997). Electron density and temperature measurements in a laser produced carbon plasma. J. Appl. Phys. 82:2140- 2146. Hunter, A. J. R., Morency, J. R., Senior, C. L., Davis, S. J., Fraser, M. E. (2000). Continuous emissions monitoring using spark-induced breakdown spectroscopy. J. Air Waste Manage. 50:111- 117. Iwatsuki, M., Kyotani, T., Matsubara, K. (1998). Fractional Determination of Elemental Carbon and Total Soluble and Insoluble Organic Compounds in Airborne Particulate Matter by Thermal Analysis Combined with Extraction and Heavy Liquid Separation. Anal. Sci. 14:321-326. Ku, B. K. and Kulkarni, P. (2015). Measurement of transport properties of aerosolized nanomaterials. J. Aerosol Sci. 90:169-181. Kulkarni, P., Baron, P. A., Willeke, K. (2011). Nonspherical Particle Measurement: Shape Factor, Fractals, and Fibers. John Wiley & Sons, Inc, 507-547. Lapuerta, M., Ballesteros, R., Rodríguez-Fernández, J. (2007). Thermogravimetric analysis of diesel particulate matter. Meas. Sci. Technol. 18:650.

Lee, S. H. and Yoh, J. J. (2012). Spectroscopic detection of carbon particulates from a high speed jet stream with extended plasma visualization. Spectrochim. Acta. B At. Spectrosc. 74-75:144-150. Lim, H. J., Turpin, B. J., Edgerton, E., Hering, S. V., Allen, G., Maring, H., Solomon, P. (2003). Semicontinuous aerosol carbon measurements: Comparison of Atlanta Supersite measurements. J. Geophys. Res. Atmos. 108(D7):8419. Martin, M. Z., Cheng, M. D., Martin, R. C. (1999). Aerosol measurement by laser-induced plasma technique: A review. Aerosol Sci. Technol. 31:409-421. NIOSH (2013). Current Intelligence Bulletin 65: Occupational Exposure to Carbon Nanotubes and Nanofibers. DHHS NIOSH Publication No. 2013-145, U.S. Department of Health and Human Services, Washington DC. Park, K., Kittelson, D. B., McMurry, P. H. (2004). Structural properties of diesel exhaust particles measured by transmission electron microscopy (TEM): Relationships to particle mass and mobility. Aerosol Sci. Technol. 38:881-889. Park, S. S., Bae, M. S., Schauer, J. J., Ryu, S. Y., Kim, Y. J., Cho, S. Y., Kim, S. J. (2005). Evaluation of the TMO and TOT methods for OC and EC measurements and their characteristics in PM 2.5 at an urban site of Korea during ACE-Asia. Atmos. Environ. 39:5101-5112. Pavlovic, J., Kinsey, J., Hays, M. (2014). The influence of temperature calibration on the OC–EC results from a dual-optics thermal carbon analyzer. Atmos. Meas.Tech. 7:2829-2838. Peterson, M. R. and Richards, M. H. (2002). Thermal-optical-transmittance analysis for organic, elemental, carbonate, total carbon, and OCX2 in PM2. 5 by the EPA/NIOSH method, in Proceedings, Symposium on Air Quality Measurement Methods and Technology-2002, Air & Waste Management Association, Pittsburgh, PA, pp. 83. Polidori, A., Turpin, B. J., Lim, H.-J., Cabada, J. C., Subramanian, R., Pandis, S. N., Robinson, A. L. (2006). Local and regional secondary organic aerosol: Insights from a year of semi-continuous carbon measurements at Pittsburgh. Aerosol Sci. Technol. 40:861-872. Rupprecht, G., Patashnick, H., Beeson, D., Green, R., Meyer, M. (1995). A new automated monitor for the measurement of particulate carbon in the atmosphere, in Proceedings, Particulate Matter: Health and Regulatory Issues, pp. 262-267. Schauer, J. J., Mader, B., Deminter, J., Heidemann, G., Bae, M., Seinfeld, J. H., Flagan, R., Cary, R., Smith, D., Huebert, B. (2003). ACE-Asia intercomparison of a thermal-optical method for the determination of particle-phase organic and elemental carbon. Environ. Sci. Technol. 37:993-1001. Šimkovic, I., Šurina, I., Vričan, M. (2003). Primary reactions of sucrose thermal degradation. Journal. Anal. Appl. Pyrolysis 70:493-504. Slowik, J. G., Stainken, K., Davidovits, P., Williams, L., Jayne, J., Kolb, C., Worsnop, D. R., Rudich, Y., DeCarlo, P. F., Jimenez, J. L. (2004). Particle morphology and density characterization by combined mobility and aerodynamic diameter measurements. Part 2: Application to combustion-generated soot aerosols as a function of fuel equivalence ratio. Aerosol Sci. Technol. 38:1206-1222. 88

Stratakis, G. and Stamatelos, A. (2003). Thermogravimetric analysis of soot emitted by a modern diesel engine run on catalyst-doped fuel. Combust. Flame 132:157-169. Turpin, B. J., Cary, R. A., Huntzicker, J. J. (1990). AN INSITU, TIME-RESOLVED ANALYZER FOR AEROSOL ORGANIC AND ELEMENTAL CARBON. Aerosol Sci. Technol. 12:161-171. Vors, E. and Salmon, L. (2006). Laser-induced breakdown spectroscopy (LIBS) for carbon single shot analysis of micrometer-sized particles. Anal. Bioanal. Chem. 385:281-286. Walters, J. P. (1972). Formation and growth of a stabilized spark discharge. Appl. Spectrosc. 26:323-353. Walters, J. P. (1977). Spark discharge - Application to multielement spectrochemical analysis. Science 198:787-797. Walters, J. P. and Goldstein, S. A. (1984). Emission topography of a stable spark discharge train. Spectrochim. Acta. B-At. Spectrosc. 39:693-728. Wendlandt, W. (1960). Thermogravimetric and Differential Thermal Analysis of (Ethylenedinitrilo) tetraacetic Acid and Its Derivatives. Anal. Chem. 32:848-849. Yao, S., Lu, J., Zheng, J., Dong, M. (2012). Analyzing unburned carbon in fly ash using laser- induced breakdown spectroscopy with multivariate calibration method. J. Anal. At. Spectrom. 27:473-478. You, J., Depalma, P. A., Marcus, R. K. (1996). Nebulization and analysis characteristics of a particle beam-hollow cathode glow discharge atomic emission spectrometry system. J. Anal. At. Spectrom. 11:483-490. ZH1/120.44 (1995). Von der Berufsgenossenschaften anerkannte Analysenverfahren zur Feststellung der Konzentrationen krebserzeugender Arbeitstoffe in der Luft in Arbeitsbereichen. Method No. 44: Diesel Engine Emission Carl Heymanns Verlag, Köln.

CHAPTER 4

Multivariate Calibration for Measurement of Aerosol Elemental

Concentration Using Microplasma Spectroscopy

A multivariate calibration approach has been developed for measurement of aerosol elemental concentration, using spark emission spectroscopy (SES) and partial least squares (PLS) regression. A training set consisting of 25 orthogonal aerosol samples with 9 factors (elements:

Cr, Mn, Fe, Ni, Cu, Zn, Cd, Pb, Ti) and 5 levels (elemental concentrations) was designed. Spectral information was obtained for each aerosol sample using aerosol spark emission spectroscopy

(ASES) at a time resolution of 1 minute. Simultaneous filter samples were collected for 30 minutes for determination of elemental concentration using a standard laboratory method of digestion followed by inductively coupled plasma mass spectrometry (ICP-MS). PLS-1 and PLS-2 regression were performed to construct the relationship between spectra and mass of analytes. Comparison of PLS models constructed with different input wavelength variables shows that selection of the relevant variables could improve the prediction ability of PLS method. In this particular application, multivariate calibration with selected variables presents slightly better prediction capability than the univariate calibration. The detection limit for the nine elements was in the range of 0.03 – 1.63 g/m3, determined from the multivariate calibration model. The performance of the multivariate calibration model was tested for prediction of elemental concentration of welding aerosols generated by aerosolizing a suspension with welding fume material. The relative root mean square error of prediction (RMSEP) is 12% for Cr, 36% for Fe, 18% for Mn and 12% for Ni, demonstrating the good accuracy of the developed SES method using multivariate calibration.

4.1. Introduction

Exposure to airborne particles containing toxic metals could have adverse health effects on workers, thereby prompting the need for reliable measurement methods for exposure characterization. The most widely used methods for determining concentration of particulate toxic metals involve filter collection, followed by off-line analyses. These methods are time-consuming, and workers cannot obtain instant feedback on their exposure to inhalable hazards. Developing real-time methods for measuring aerosol chemical composition is of great significance to protect workers’ health.

Microplasma spectroscopy, such as laser-induced breakdown spectroscopy (LIBS), spark emission spectroscopy (SES) and glow discharge-optical emission spectroscopy (GD-OES), has been considered as an effective technique for real-time and in-situ measurement of elemental species in aerosol phases (Diwakar and Kulkarni 2012; Hunter et al. 2000a; Khalaji et al. 2012;

Marcus et al. 1999; Martin et al. 1999). In this technique, aerosol samples can be analyzed directly in a free or focused aerosol stream, or collected on a substrate followed by spectroscopic analysis.

For quantitative analysis of aerosols using microplasma spectroscopy, univariate calibration approach has been used to construct calibration curves between signal intensity and elemental concentration or mass in previous works (Diwakar and Kulkarni 2012; Hunter et al. 2000a; Khalaji et al. 2012; Marcus et al. 1999; Martin et al. 1999). However, in most of spectroscopic methods, quantitative spectral analyses remains challenging due to sample matrix effects (physical and chemical matrix effects) and spectral interferences (Gemperline 2006). Because the spectral emission intensity from one element is affected by the chemical composition of the sample, the conventionally used univariate calibration method would produce large uncertainty when unknown samples have different chemical matrix from the calibration standards (Laville et al.

Multivariate calibration has been employed in the spectroscopic analysis of mixtures, such as steels, alloys, coal, and soil (Doucet et al. 2007; Feng et al. 2013; Golloch and Wilke 1997;

Gonzaga and Pasquini 2012; Martin et al. 2010). Yao et al. (2012) analyzed unburned carbon in fly ash using LIBS and multivariate calibration method, showing that the multivariate calibration had a better performance than univariate calibration as matrix effects on spectroscopic signals caused by other components in fly ash can be taken into account. Zaytsev et al. (2014) compared single and multivariate calibration for determination of Si, Mn, Cr and Ni in high-alloyed stainless steels using LIBS. He found that multivariate analysis of spectra data was more effective and accurate in the case of overlapping analytical lines due to the complex composition. The application of multivariate calibration in spectroscopic analysis could provide an advantage by accounting for matrix effects as well as elimination of spectral interference, especially for spectrograph with poor resolution.

In previous work, we have developed a near real-time aerosol analysis system based on corona preconcentration and SES (Diwakar and Kulkarni 2012). This method has shown excellent accuracy, precision, detection limits and time resolution through univariate calibration for single element. In this study, we provide an alternative calibration method for simultaneous analysis of multiple elements in aerosol using SES. Partial least squares (PLS) regression was employed to build multivariate calibration models. The prediction capability of PLS models (PLS-1 and PLS-

2) created with different spectral variables was compared. Application of the PLS model to welding

aerosol measurements was presented.

4.2. Methods

4.2.1. Instrumentation

Spark emission spectroscopy described in a previous study was used for analytical measurement of aerosol samples (Diwakar and Kulkarni 2012). Figure 4.1 shows the schematic diagram of aerosol spark emission spectroscopy (ASES) instrumentation. This method involves collection of particles onto a small electrode tip (500 µm in diameter) with a corona microconcentrator, followed by ablation of particles using a high voltage (HV) pulse generator

(200 mJ per pulse; ARC-2, Cascodium Inc., Andover, MA,). The optical emission from excited atomic and ionic species in the spark-induced plasma was collected by a broadband spectrometer

(200 – 900 nm wavelength range, 0.1 nm resolution; LIBS 2500 Plus; Ocean Optics Inc.; Dunedin,

FL) for spectrochemical analysis. A delay time of 5 µs and a gate width of 1 ms were used. The spectra data was used to identify elements and determine their mass in the collected particle samples.

Figure 4.1 A schematic diagram of aerosol spark emission spectroscopy instrumentation.

4.2.2. Aerosol samples

Aerosol samples containing multiple elements were generated through aerosolization of target solutions. These solutions were prepared with metal nitrates. A multilevel multifactor training set was designed for multivariate calibration (Brereton 1997). It consisted of 25 samples with 9 mutually orthogonal factors (elements: Cr, Mn, Fe, Ni, Cu, Zn, Cd, Pb, Ti) and 5 levels

(elemental concentrations), as shown in Table 4.1.

Table 4.1. Elemental concentration in the prepared solution.

Sample Elemental concentration in the solution, mg/L No. Cr Mn Fe Ni Cu Zn Cd Pb Ti 1 20 20 20 20 20 20 20 20 20 2 20 5 10 5 40 40 20 10 40 3 5 10 5 40 40 20 10 40 10 4 10 5 40 40 20 10 40 10 30 5 5 40 40 20 10 40 10 30 30 6 40 40 20 10 40 10 30 30 20 7 40 20 10 40 10 30 30 20 40 8 20 10 40 10 30 30 20 40 30 9 10 40 10 30 30 20 40 30 40 10 40 10 30 30 20 40 30 40 5 11 10 30 30 20 40 30 40 5 5 12 30 30 20 40 30 40 5 5 20 13 30 20 40 30 40 5 5 20 30 14 20 40 30 40 5 5 20 30 5 15 40 30 40 5 5 20 30 5 30 16 30 40 5 5 20 30 5 30 10 17 40 5 5 20 30 5 30 10 10 18 5 5 20 30 5 30 10 10 20 19 5 20 30 5 30 10 10 20 5 20 20 30 5 30 10 10 20 5 10 21 30 5 30 10 10 20 5 10 5 22 5 30 10 10 20 5 10 5 40 23 30 10 10 20 5 10 5 40 40 24 10 10 20 5 10 5 40 40 20 25 10 20 5 10 5 40 40 20 10

4.2.3. Experimental procedure

Figure 4.2 shows the schematic diagram of the experimental procedure for multivariate calibration. Test aerosol was generated using a pneumatic atomizer (Model 3080, TSI Inc.,

Shoreview, MN), which was then passed through a diffusion dryer. The dry aerosol was introduced into the corona microconcentrator and a PVC (polyvinyl chloride) filter simultaneously. For SES analysis, the aerosol was collected on the ground electrode tip in the corona microconcentrator for

1 minute at a flow rate of 2 L min-1, followed by ablation of particulate matter by spark discharge.

For complete ablation of the collected particles, four high voltage pulses were employed, and the sum of the four spectra was used as the spectrum of one measurement. Each aerosol sample was analyzed in five replicates, and their average gave one spectrum representing each sample. To determine elemental concentration in aerosol, test aerosol was collected on a PVC filter for 30 minutes at a flow rate of 2 L min-1. The filter sample was digested following NIOSH Method 7303 and then analyzed using inductively coupled plasma mass spectrometry (ICP-MS). The mass of each element in the collected particulate samples with ASES were calculated using equation: 푚 =

휂퐶𝑖푛푄푓푡푐, where 푚 is elemental mass; 휂 is collection efficiency; 퐶𝑖푛 is elemental concentration;

푄푓 is aerosol flow rate; and 푡푐 is collection time. The results are shown in Table 4.2. The correlation coefficients for each two factors are close to 0, ensuring that all the element factors in the aerosol samples are mutually orthogonal. PLS regressions were performed to construct models for relating spectra and elemental mass. These PLS models were validated with full cross validation and external test samples.

4.2.4. PLS regression model

PLS regression is a multivariate calibration technique that provides a model for relating two data matrices, a set of predictor variables X (n observations, m variables) and a set of response

variables Y (n observations, p variables), by a linear multivariate model. In our case, n observations are the 25 samples in the training set, m variables are the SES spectral data at different wavelengths and p response variables are elemental mass of each sample. Unlike traditional multiple linear regression (MLR), PLS models the structure of X and Y, such that it can analyze data with numerous, strongly collinear, noisy X variables and simultaneously model several response Y variables (Wold et al. 2001). The principle of PLS is that it finds a few new X variables (called X- scores), as linear combinations of the old X, and thereafter uses the X-scores as predictors of Y.

Figure 4.2. A schematic diagram of the experimental procedure for multivariate calibration.

Table 4.2. Elemental mass in the collected particle samples on the tip of electrode determined by

ICP-MS.

Sample Elemental mass collected on the tip of electrode, ng No. Cr Mn Fe Ni Cu Zn Cd Pb Ti 1 21.0 21.5 19.5 22.5 25.5 19.0 20.0 20.5 17.5 2 18.5 5.0 8.8 4.3 37.0 34.0 18.0 8.5 32.0 3 6.0 9.5 4.4 40.0 36.0 15.5 8.5 35.5 8.0 4 10.5 5.0 39.5 43.0 21.0 8.5 39.0 12.0 27.0 5 6.0 41.0 45.5 22.5 10.5 36.0 9.5 30.0 28.0 6 37.5 38.5 18.6 9.5 37.0 8.5 28.5 29.5 18.0 7 38.5 20.5 9.5 41.5 10.0 23.5 29.0 20.0 34.5 8 19.5 10.5 32.5 10.0 28.0 23.5 19.5 38.5 26.5 9 10.0 38.0 9.2 30.0 28.0 16.0 38.0 29.0 32.0 10 38.5 10.5 24.5 31.5 20.0 35.0 29.5 39.0 2.6 11 11.5 30.0 29.0 22.0 38.5 25.0 39.0 6.0 5.0 12 31.5 32.0 19.0 44.0 30.0 36.0 6.0 5.5 18.5 13 30.5 21.0 36.0 33.0 40.0 5.0 5.0 19.5 26.0 14 18.0 36.5 26.5 39.5 5.0 4.5 18.5 27.5 3.9 15 39.5 30.0 41.5 5.5 5.0 16.5 29.5 5.0 28.0 16 27.5 37.5 4.3 4.0 17.5 23.5 4.7 26.5 8.5 17 34.0 4.6 4.1 18.5 25.0 3.1 25.5 9.0 7.5 18 5.5 4.7 22.0 28.5 4.7 23.0 8.5 8.0 11.0 19 5.5 20.0 23.5 4.8 27.5 8.5 9.0 19.0 4.3 20 17.5 27.0 4.4 29.0 8.5 8.0 18.0 4.8 7.5 21 27.5 4.9 26.0 8.5 8.5 16.0 4.5 8.5 4.5 22 6.0 29.5 9.2 9.0 19.0 3.9 9.0 5.0 36.5 23 29.0 10.0 9.6 20.0 6.0 8.0 4.8 36.0 35.0 24 9.5 8.5 19.0 3.7 7.5 2.8 32.0 34.0 16.0 25 10.0 20.0 4.9 9.5 5.5 35.5 40.5 21.5 9.0

Both PLS-1 and PLS-2 multivariate calibration models with different selected spectral variables were constructed for each analyte. In PLS-1 model, only one element was modeled each time, while all the nine elements were calibrated in one PLS-2 model (Brereton 2000). The number of latent variables for each model was chosen at which the relative Root Mean Square Error of

Cross Validation (RMSECV) was the lowest (Brereton 2007). The performance of the calibration models were compared by examining their relative RMSECV and correlation coefficient R2 of cross validation. The PLS analysis was carried out using the software The Unscrambler 10.0 X

(CAMO Software Inc., Woodbridge, NJ).

4.3. Results and discussion

4.3.1. Wavelength variables selection

In multivariate calibration using PLS for spectroscopic analyses, selection of input variables has an effect on the prediction ability of models. In order to obtain the PLS regression model with better precision of the predictions, wavelengths (variables) that carry most relative information or exhibit good sensitivity and linearity for the interested analytes should be selected as input variables of PLS models. In this study, the selection of wavelength variables was carried out by examining the regression coefficients of PLS analysis for each element using the selected full spectra with a wavelength range of 300 – 540 nm. The wavelengths which have larger regression coefficients represent the emission lines that are better correlated to the mass of analyte and more important for the models. Figure 4.3 shows the regression coefficients of the PLS models for Cr constructed with the full spectra ranging from 300 to 540 nm. From Figure 4.3 (a), the maximum of regression coefficients was observed around 520 nm. The zoomed-in figure (Figure

4.3 (iii)) for the wavelength region around 520 nm presents three overlapped peaks. These three peaks are in good agreement with chromium emission lines listed in NIST atomic spectral database, which are Cr I 520.45 nm, Cr I 520.60 nm, and Cr I 520.84 nm. Besides these three overlapped peaks, there are three consecutive peaks observed in a wavelength region around 360 nm (Figure

4.3 (i)) and 425 nm (Figure 4.3 (ii)), respectively. They correspond to chromium emission lines Cr

I 357.87 nm, Cr I 359.35 nm, Cr I 360.53 nm, Cr I 425.43 nm, Cr I 427.48 nm, and Cr I 428.97 nm through identification using NIST atomic spectral database. Similarly, we selected the relevant wavelengths for all the other elements of interest in this study by analyzing their PLS regression coefficients, and the results are shown in Table 4.4. The highlighted emission line represents the 99

one with the highest regression coefficient for each element. Five variables for each emission peak

(center wavelength, left two and right two pixels) were used, and 350 variables in total were selected as input variables for PLS models.

Figure 4.3. Regression coefficients of PLS-1 model for chromium.

Table 4.3. Selected elemental emission lines for each element investigated in this study.

Elements Emission lines/nm 357.8682, 359.3481, 360.5320, 425.4331, 427.4806, 428.9733, 520.4505, Cr 520.6021, 520.8415 Mn 344.1985, 346.0314, 347.4038, 348.2904, 403.076, 403.307, 403.449, 482.352 371.9935, 373.4864, 373.7132, 374.5561, 375.8233, 382.0425, 404.5813, Fe 406.3594, 407.1738 341.476, 342.371, 345.289, 345.847, 346.165, 349.296, 351.034, 351.505, Ni 352.454, 356.637, 361.939 Cu 324.754, 327.396, 510.554, 515.324, 521.820 Zn 328.233, 330.258, 468.014, 472. 215, 481.053, 491.162, 492.403 Cd 346.620, 361.051, 467.815, 479.991, 508.582, 533.748, 537.813 Pb 363.957, 368.346, 405.781 334.941, 338.028, 338.376, 368.521, 375.769, 376.132, 453.324, 453.478, Ti 498.173, 499.107, 501.419 *The highlighted emission line represents the one with the highest regression coefficient for each element.

100

4.3.2. Multivariate calibration using PLS

To understand the effect of input variables on the prediction ability of PLS regression models, we constructed PLS regression models using three X matrices containing different numbers of variables. These three X matrices include: (i) variables containing the whole spectra ranging from 300 to 540 nm (4312 variables), (ii) variables containing all the emission peaks related to the analytes (all the emission peaks shown in Table 4.3, 350 variables), and (iii) a subset of variables containing only one emission peak for each element (i.e. the one with the highest regression coefficient, shaded in Table 4.3, 45 variables in total for all the nine elements). Both

PLS-1 and PLS-2 models were performed to construct the relationship between spectra and elemental mass. To simplify the model description, these models were denoted as PLS-1-4312,

PLS-1-350, PLS-1-45, PLS-2-4312, PLS-2-350, and PLS-2-45, respectively. Due to the limited numbers of samples in the training set, full cross validation was employed to estimate how reliable these models might be when used for prediction of unknown samples. It should be noted that the use of cross validation here was mainly for comparison of the performance of different PLS models.

The prediction ability of PLS models needs to be further evaluated through validation using unknown test samples, which will be discussed in Section 4.3.4.

Table 4.4 shows the relative RMSECV and R2 obtained from cross validation for different models. Comparison of the PLS models constructed with different numbers of variables shows that the models with selected variables (PLS-1-350, PLS-1-45, PLS-2-350, and PLS-2-45) were significantly superior to the models with the whole spectra (PLS-1-4312 and PLS-2-4312). For

PLS-1 models, as the variable number is reduced from 4312 to 45, the average relative RMSECV for the nine elements is reduced from 36% to 25% and the average R2 is improved from 0.65 to

0.83. This result demonstrates that variable selection plays an important role in multivariate

101

calibration on the spectra data of SES. Although PLS model has the ability to deal with numerous variables, the prediction ability is reduced because the majority of data consist of noise if the raw spectral data are applied in the calibration model. Therefore, the use of selected variables that are most relevant to the underlying problem as input variables is essential to construct the optimal PLS model. For PLS-1 models with the selected wavelength variables, the prediction ability of PLS-1-

350 and PLS-1-45 are quite similar. This is because the signal intensities of multiple emission peaks from one element are related. Selection of one emission peak for each element as input variables of the PLS model is sufficient for constructing accurate multivariate calibration models.

Table 4.4 also shows that the relative RMSECVs of PLS-1 models are 2 – 4% smaller than

PLS-2 models. This suggests that PLS-1 models provide better prediction ability than PLS-2 models. However, several PLS-1 models have to be performed separately for each Y variable, while one PLS-2 model could model multiple Y variables simultaneously.

Using the 25 samples in the training set, univariate calibration model was also constructed for each element by plotting the signal intensity of an emission line as a function of the elemental mass. For each element, the emission line with the highest regression coefficient was used. The univariate calibration models were described using linear regression curves. The R2 and RMSE of the univariate regression are also shown in Table 4.5. For most elements, the univariate calibration gives similar R2 and RMSE as those of PLS models with selective variables (i.e. PLS-2-350, PLS-

2-45, PLS-1-350, and PLS-1-45). Some studies on sample analysis using LIBS suggest that multivariate calibration provides improved accuracy and precision compared with univariate calibration (Tripathi et al. 2009; Yao et al. 2012). However, the performance of univariate approach depends on both the spectral overlap and the design of the calibration experiments (Demir and

Brereton 1998). A comparison of univariate and multivariate calibration for determination of

102

elements in stainless steels using LIBS, showed that the univariate calibration provided the best prediction if appropriate reference lines could be found and analytical lines were not overlapped

(Zaytsev et al. 2014). When field samples or poor designed mixed samples are used for calibration, the conclusions about the effectiveness of different calibration approaches may be unreliable. In this study, we designed orthogonal experimental samples and selected the non-overlapping emission peaks. We found both the univariate and multivariate calibration present similar performance.

Figure 4.5 shows the regression results (open circle) and the cross-validation results (solid triangle) of PLS-1-45 models that give the correlation between the referenced and predicted elemental mass for the 25 particulate samples. These models were constructed with 2, 5, 8, 5, 4, 5,

2, 6, 5 latent variables for elements Cr, Mn, Fe, Ni, Cu, Zn, Cd, Pb, and Ti, respectively. The PLS regression results (open circle) show a strong correlation between the measured and predicted elemental mass, with R2 values in the range of 0.81 – 0.96 for different elements. The cross validation results (solid triangle) also show a strong correlation with an average R2 values of 0.83, indicating good prediction ability of this model. It was noted that a low R2 (0.67) and high

RMESCV (37%) for Fe was observed, which is probably due to the low detection sensitivity for

Fe in our system (the detection limit is shown in Table 4.5). A lower RMSECV might be obtained if samples with larger Fe mass were used in the calibration model.

103

Table 4.4. Comparison of relative RMSECV and R2 from PLS-1 and PLS-2 models with different numbers of variables.

R2 Element PLS-2-4312 PLS-2-350 PLS-2-45 PLS-1-4312 PLS-1-350 PLS-1-45 Univariate Cr 0.73 0.73 0.88 0.78 0.79 0.79 0.81 Mn 0.66 0.68 0.71 0.73 0.79 0.81 0.75 Fe 0.20 0.31 0.71 0.26 0.55 0.67 0.57 Ni 0.58 0.75 0.75 0.66 0.83 0.82 0.71 Cu 0.69 0.92 0.89 0.80 0.91 0.94 0.89 Zn 0.60 0.83 0.79 0.68 0.81 0.85 0.83 Cd 0.87 0.90 0.93 0.90 0.92 0.90 0.92 Pb 0.15 0.63 0.73 0.20 0.70 0.84 0.84 Ti 0.67 0.85 0.88 0.82 0.85 0.88 0.84 Average 0.57 0.73 0.81 0.65 0.79 0.83 0.80 Relative RMSECV, % Element PLS-2-4312 PLS-2-350 PLS-2-45 PLS-1-4312 PLS-1-350 PLS-1-45 Univariate

Cr 32 32 21 29 28 28 28 Mn 38 37 35 34 29 28 34 Fe 58 54 35 56 44 37 52 Ni 46 36 35 44 29 30 42 Cu 35 18 21 27 19 15 21 Zn 37 24 27 33 25 23 24 Cd 23 20 17 20 19 21 18 Pb 56 37 32 55 34 24 25 Ti 35 23 21 25 23 21 24 Average 40 31 27 36 28 25 30

104

Figure 4.4 Predicted vs measured elemental mass in the collected particle samples for PLS-1-45 models.

4.3.3. Limits of detection

According to 3- 𝜎 criteria defined by the International Union of Pure and Applied

Chemistry (IUPAC), the limit of detection (LOD) is expressed as,

퐿푂퐷 = 3𝜎/푆 (1) where 𝜎 is the standard deviation of the blank at the selected emission line and 푆 is the sensitivity given by the slope of the calibration curve of univariate calibration. Blank measurements were

105

taken without analyte collected on the electrode, and 𝜎 was obtained by averaging over 20 replicate blank measurements.

LOD in multivariate calibration was determined analogously to univariate calibration

(Braga et al. 2010; Valderrama et al. 2007):

1 퐿푂퐷 = 3.3 훿휒 = 3.3 훿휒‖퐛‖ (2) 푆퐸푁 where 훿휒 is an estimate of the noise level in the data and 푆퐸푁 is the sensitivity of the PLS model, that gives the fraction of analytical signal due to the increase of the concentration of a particular analyte at unit concentration. 훿휒 was obtained by measuring the variation of the noise in the selected regions for each analyte. 푆퐸푁 was estimated as the inverse of the Euclidian norm of the regression coefficients vector. Table 4.5 shows the LOD obtained from both univariate calibration and multivariate calibration (PLS-1-45 model) for elemental measurement using SES. The LOD in terms of mass for multivariate calibration is in the range of 0.15 – 4.04 ng depending on elements, with the lowest LOD for Cr and the highest for Fe. The LOD for multivariate calibration is in the range of 0.03 – 1.63 g/m3 in terms of air concentration, calculated by assuming a flow rate of 2

L/min, collection time of 5 min, and collection efficiency of 0.5 for our system. By comparing the

LOD from univariate and multivariate calibration, the largest difference was found for Fe, where the LOD from univariate calibration is twice that from multivariate calibration. In general, both methods did not present any significant difference for most analytes. This result is consistent with that found by Braga et al. (2010) in a study on determination of micronutrients in plant materials using LIBS.

106

Table 4.5. Limits of detection of aerosol spark emission spectroscopy

Limits of detection

In terms of air concentration*, µg Elements In terms of mass, ng m-3

Multivariate Univariate Multivariate Univariate calibration calibration calibration calibration

Cr 0.15 0.20 0.03 0.04

Mn 1.37 1.83 0.27 0.37

Fe 4.04 8.13 0.81 1.63

Ni 2.01 3.52 0.40 0.70

Cu 1.06 1.30 0.21 0.26

Zn 1.78 2.12 0.36 0.42

Cd 0.79 1.32 0.16 0.26

Pb 2.42 3.99 0.48 0.80

Ti 1.11 1.12 0.22 0.22 *Assuming a flow rate of 2 L min-1 and a collection time of 5 min.

4.3.4 Application on welding aerosol measurement

We applied the multivariate calibration model PLS-1-45 on measurement of heavy metal concentration of welding aerosol. Test welding aerosols were generated by aerosolizing suspension of stainless steel welding fume reference material (HSL SSWF-01, Health & Safety Laboratory,

Buxton, UK), and aerosols with five different concentration levels were obtained through dilution with clean air. Test aerosols were collected for 2 minutes using ASES system at 2 L/min. The elemental mass in the collected sample was predicted using PLS-1-45 models, and then inverted to an air concentration. Three replicated measurements were performed by ASES for each test aerosol sample. Meanwhile, the elemental concentrations of test welding aerosols were determined by filter collection (for 30 minutes) and ICP-MS analysis. The test aerosol was relatively stable during these 30 minutes. Figure 4.5 shows the comparison of predicted and referenced elemental concentration. Four elements including Cr, Fe, Mn, and Ni were detected in this welding fume, which agreed with the elements reported in the certification of SSWF-01 material. The figure

107

shows that a good agreement was obtained between the predicted elemental concentration by

ASES and the measured concentration by conventional filter method. Relative root mean square error of prediction (RMSEP) was used as a parameter for estimating the accuracy of our multivariate calibration model. The relative RMSEP was 12% for Cr, 36% for Fe, 18% for Mn and

12% for Ni. The error bars in Figure 4.5 show the standard deviation of three replicated measurements. The relative standard deviation varied in the range of 7 – 19%, demonstrating good precision of our aerosol measurement system.

108

4.4. Conclusions

This study provides a multivariate calibration approach for measurement of aerosol elemental concentration using microplasma spectroscopy. The employed spark emission spectroscopy in combination with suitable multivariate method allows one to make full use of the emission spectra without loss of selectivity and multielement capability. PLS regression coefficients can be used for the selection of appropriate wavelengths. Selection of variables including only emission peaks from the analyte of interested results in improvements in the prediction ability of multivariate calibration model. Developing specific model for each analyte

(PLS-1 model) demonstrates improved prediction ability compared with PLS-2 models. In this application of SES, the comparison of multivariate and univariate calibration showed similar performance in prediction ability. The PLS models were applied to measurement of welding aerosol, showing results comparable to those obtained by conventional filter collection followed by ICP-MS method.

4.5. References

Braga, J. W. B., Trevizan, L. C., Nunes, L. C., Rufini, I. A., Santos, D., Krug, F. J. (2010). Comparison of univariate and multivariate calibration for the determination of micronutrients in pellets of plant materials by laser induced breakdown spectrometry. Spectrochim. Acta B At. Spectrosc. 65:66-74. Brereton, R. G. (1997). Multilevel multifactor designs for multivariatecalibration. Analyst 122:1521-1529. Brereton, R. G. (2000). Introduction to multivariate calibration in analytical chemistry. Analyst 125:2125-2154. Brereton, R. G. (2007). Applied chemometrics for scientists. John Wiley & Sons. Demir, C. and Brereton, R. G. (1998). Multivariate calibration on designed mixtures of four pharmaceuticals. Analyst 123:181-189. Diwakar, P., Kulkarni, P., Birch, M. E. (2012). New approach for near-real-time measurement of elemental composition of aerosol using laser-induced breakdown spectroscopy. Aerosol Sci. Technol. 46:316-332.

109

Diwakar, P. K. and Kulkarni, P. (2012). Measurement of elemental concentration of aerosols using spark emission spectroscopy. J. Anal. At. Spectrom. 27:1101-1109. Doucet, F. R., Belliveau, T. F., Fortier, J.-L., Hubert, J. (2007). Use of chemometrics and laser- induced breakdown spectroscopy for quantitative analysis of major and minor elements in aluminium alloys. Appl. Spectrosc. 61:327-332. Feng, J., Wang, Z., Li, L., Li, Z., Ni, W. (2013). A nonlinearized multivariate dominant factor- based partial least squares (PLS) model for coal analysis by using laser-induced breakdown spectroscopy. Appl. Spectrosc. 67:291-300. Fink, H., Panne, U., Niessner, R. (2002). Process analysis of recycled thermoplasts from consumer electronics by laser-induced plasma spectroscopy. Anal. Chem. 74:4334-4342. Gemperline, P. (2006). Practical guide to chemometrics. CRC press. Golloch, A. and Wilke, K. (1997). Fast survey analysis of gold alloys by means of a spark emission spectrometer and multivariate calibration. J. Anal. At. Spectrom. 12:1225-1230. Gonzaga, F. B. and Pasquini, C. (2012). A compact and low cost laser induced breakdown spectroscopic system: Application for simultaneous determination of chromium and nickel in steel using multivariate calibration. Spectrochim. Acta B At. Spectrosc. 69:20-24. Hunter, A. J. R., Davis, S. J., Piper, L. G., Holtzclaw, K. W., Fraser, M. E. (2000). Spark-induced breakdown spectroscopy: A new technique for monitoring heavy metals. Appl. Spectrosc. 54:575- 582. Khalaji, M., Roshanzadeh, B., Mansoori, A., Taefi, N., Tavassoli, S. H. (2012). Continuous dust monitoring and analysis by spark induced breakdown spectroscopy. Opt. Laser Eng. 50:110-113. Laville, S., Sabsabi, M., Doucet, F. R. (2007). Multi-elemental analysis of solidified mineral melt samples by laser-induced breakdown spectroscopy coupled with a linear multivariate calibration. Spectrochim. Acta B At. Spectrosc. 62:1557-1566. Marcus, R. K., Dempster, M. A., Gibeau, T. E., Reynolds, E. M. (1999). Sampling and analysis of particulate matter by glow discharge atomic emission and mass spectrometries. Anal. Chem. 71:3061-3069. Martin, M. Z., Cheng, M. D., Martin, R. C. (1999). Aerosol measurement by laser-induced plasma technique: A review. Aerosol Sci. Technol. 31:409-421. Martin, M. Z., Labbé, N., André, N., Wullschleger, S. D., Harris, R. D., Ebinger, M. H. (2010). Novel multivariate analysis for soil carbon measurements using laser-induced breakdown spectroscopy. Soil Sci. Soc. Am. J. 74:87-93. Tripathi, M. M., Eseller, K. E., Yueh, F.-Y., Singh, J. P. (2009). Multivariate calibration of spectra obtained by Laser Induced Breakdown Spectroscopy of plutonium oxide surrogate residues. Spectrochim. Acta B At. Spectrosc. 64:1212-1218. Valderrama, P., Braga, J. W. B., Poppi, R. J. (2007). Variable selection, outlier detection, and

110

figures of merit estimation in a partial least-squares regression multivariate calibration model. A case study for the determination of quality parameters in the alcohol industry by near-infrared spectroscopy. J. Agric. Food Chem. 55:8331-8338. Wold, S., Sjöström, M., Eriksson, L. (2001). PLS-regression: a basic tool of chemometrics. Chemometr. Intell. Lab. 58:109-130. Yao, S., Lu, J., Zheng, J., Dong, M. (2012). Analyzing unburned carbon in fly ash using laser- induced breakdown spectroscopy with multivariate calibration method. J. Anal. At. Spectrom. 27:473-478. Zaytsev, S. M., Popov, A. M., Chernykh, E. V., Voronina, R. D., Zorov, N. B., Labutin, T. A. (2014). Comparison of single-and multivariate calibration for determination of Si, Mn, Cr and Ni in high- alloyed stainless steels by laser-induced breakdown spectrometry. J. Anal. At. Spectrom. 29:1417- 1424.

111

CHAPTER 5

Method for Rapid Elemental Analysis of Airborne Particles Using

Atmospheric Glow Discharge Optical Emission Spectroscopy

 To be submitted to Analytical Chemistry 112

A new, low-cost approach based on application of atmospheric radio frequency glow discharge (rf-GD) optical emission spectroscopy (OES) has been developed for near real-time measurement of elemental concentrations in airborne particulate matter. This method involves deposition of aerosol particles on the tip of a ground electrode of a coaxial microelectrode system, followed by atomization and excitation of the particulate matter using the rf-GD. The resulting atomic emissions were recorded using a spectrometer for elemental identification and quantification. The glow discharge plasma was characterized by measuring spatially resolved gas temperatures (378 – 1438 K) and electron densities (2 – 5 × 1014 cm-3). Spatial analysis of the spectral features showed that the collision and excitation of the collected particles occurred in the region near the collection electrode. The temporal analysis of spectral features in the rf-GD showed that the collected particles were continuously ablated; the time for complete ablation of

193 ng of sucrose particles was found to be approximately 2 s. The system was calibrated using

100 nm particles containing C, Cd, Mn, and Na, respectively. Our method provides limits of detection in the range of 0.055 – 1.0 ng in terms of absolute elemental mass, and a measurement reproducibility of 5 – 28%. This study demonstrates that the rf-GD can be an excellent excitation source for the development of low-cost hand-held sensors for elemental measurement of aerosols.

113

5.1. Introduction

Airborne particles have great effect on global climate, air quality, and human health (Pöschl

2005). In particular, long term inhalation of toxic particulate matter could pose a significant health risk to those who are routinely exposed to airborne particles, such as in workplaces. Measurement of exposure to metals is essential to environmental and occupational health studies. Existing elemental analysis methods, such as atomic absorption spectroscopy (AAS), X-ray fluorescence spectroscopy (XRFS), inductively coupled plasma atomic emission spectroscopy (ICP-AES) and mass spectroscopy (ICP-MS), etc., are effective for chemical analysis of particulate matter and widely used for routine monitoring personal exposure because of their high sensitivity and accuracy (Brown et al. 2010; Manalis et al. 2005; Voutsa and Samara 2002). Normally, these methods require particle collection on filters over several hours, followed by subsequent laboratory analysis. These methods are labor- and time-intensive. Low-cost, field portable, near real-time instruments for chemical analysis of aerosol are needed to address these needs.

Several plasma-based techniques have been used for elemental analysis of aerosols, which have employed excitation sources ranging from spark microplasma (Diwakar and Kulkarni 2012;

Hunter et al. 2000; Khalaji et al. 2012), laser-induced breakdown (Diwakar et al. 2012; Park et al.

2009; Radziemski et al. 1983; Vors and Salmon 2006), microwave induced plasma (Duan et al.

2000; Kulkarni and Efthimion 2015). However, the excitation sources used in these methods can be bulky and expensive, making them unsuitable for hand-held, low-cost monitors for aerosol elemental analysis. In this context, the glow-discharge excitation sources offer attractive alternative for development of low-cost aerosol instruments. Glow discharge, as an excitation source for the elemental determination, has certain unique advantages such as low cost, low temperature, low power consumption, and analytical versatility (Marcus 2003). Glow discharge

114

optical emission spectroscopy (GD-OES) and glow discharge mass spectroscopy (GD-MS) have been applied in the bulk elemental analysis of inorganic solid samples (Boumans 1972;

Jakubowski et al. 1987) and quantitative depth profile analysis (Jakubowski et al. 2007). In a glow discharge system, the samples work as cathode, and is continuously eroded by bombardments of ions and neutrals from the plasma. Then the free atoms ejected from the samples are diffused into the plasma plume, where they are excited through collisions with electrons, metastable gas atoms and ions, and emit element characteristic optical emission (Bengtson 1994; Marcus 2013). In 1993, an electrolyte cathode discharge (ELCAD) was introduced for elemental analysis of solutions

(Cserfalvi et al. 1993). Since then, solution-cathode glow discharge (SCGD) and liquid sampling- atmospheric pressure glow discharge (LS-APGD) have been developed (Doroski et al. 2013;

Doroski and Webb 2013; Quarles Jr et al. 2012). These techniques can offer similar detection limits

(tens of ppbs) as ICP-AES, but have the advantage of much lower cost and power consumption

(Doroski and Webb 2013; Wang et al. 2013). In liquid sampling – glow discharge techniques for solution analysis, such as particle beam/hollow cathode-optical emission spectroscopy (PB/HC-

OES) and particle beam-glow discharge mass spectroscopy (PB/GDMS), particle beam technique has been used as a transport-type interface to convert the analytes from liquid phase to gas phase molecules for subsequent analysis (Brewer et al. 2006). The particle beam interface involves the conversion of the liquid effluent into an aerosol, evaporation of the aerosol droplets, and formation of a beam of solute particles (Brewer et al. 2006). Though the PB/GDOES and PB/GDMS are mainly designed for solution analysis, they have also been applied to aerosol analysis. Marcus et al. (1999) conducted elemental analysis of aerosols (NIST SRM 1648 urban particulate matter) by a direct injection of particles into a low-pressure glow discharge plasma through an aerodynamic momentum separator, and obtained limits of detection (LOD) on the order of tens of nanograms.

115

However, the aerodynamic momentum method required use of large turbo pumps to create particle beams for direct injection into GD, making it unsuitable for hand-held instrumentation. Compared to direct, in situ analysis methods of Marcus et al., substrate-based collection, followed by GD-

OES analysis have proven to be more sensitive. LOD below 1 ppm were obtained by collecting atmospheric particulate matter on a metal plate through a single-orifice impactor stage, followed by analyzing the metal plate in a DC glow discharge mass spectrometer, but the aerosol sampling time was varied from 3 hours up to 3 days and analysis time was more than 1 hour (Schelles et al.

1996).

The objective of this study is to develop a near real-time method for aerosol elemental analysis using a low-cost radio frequency atmospheric glow discharge excitation source. A corona- based microconcentration method (Zheng et al. 2016) is used for microscopic collection of airborne particles, followed by elemental analysis using glow discharge emission spectroscopy. A methodology was developed for automated and semi-continuous analysis of aerosol. The spectral features and signal stability of this GD-OES aerosol analysis system were investigated. The glow discharge plasma was characterized by measuring its gas temperature and electron density using spectroscopic method. Analytical performance, such as calibration curves, uncertainty, and limits of detection, were determined for select elements of interest (C, Cd, Mn, and Na), and compared with other microplasma spectroscopy methods.

5.2. Methods

5.2.1. Experimental setup and materials

A schematic diagram of the experimental setup is shown in Figure 5.1. The major components included: (i) an aerosol generation system, (ii) an aerosol collection system, and (iii) a GD-OES system. 116

The aerosol generation and collection systems were similar to those described in previously published studies.(Diwakar et al. 2012; Diwakar and Kulkarni 2012) Test aerosols were generated using a pneumatic atomizer (Model 3080, TSI Inc., Shoreview, MN, USA) to atomize solutions containing analytes and then passed through a diffusion dryer. Through a differential mobility analyzer (DMA; model 3080, TSI Inc., Shoreview, MN, USA), a neutralizer and an electrostatic precipitator, near-monodisperse uncharged particles were obtained. In this study, particles of 100 nm in diameter were used for all the measurements. Table 5.1 shows materials containing C, Na,

Cd, and Mn used for calibration. Stock standard solutions were diluted using ultrafiltered DI water to obtain the target solutions for aerosol atomization, ranging from 100 to 1000 µg mL-1 depending on the analyte.

The test aerosol particles were then introduced into a corona aerosol microconcentrator

(CAM). The CAM consisted of two coaxial electrodes with a distance of 4 mm. A high positive potential (~5 kV) was applied on the corona electrode through a DC power supply (Bertran S-230,

Spellman Corp., Hauppauge, NY, USA). The corona electrode, made of tungsten, was 200 µm in diameter. This electrode has a sharp tip with an approximate radius of 50 m, and was used to create corona around its tip. The ground electrode, made of platinum, was 500 µm in diameter, and has a relatively flat tip to provide a planar surface for particle deposition. The aerosol particles entering the CAM were collected on the tip of a grounded electrode. The sidewalls of the ground electrode were covered with a high dielectric strength sheath [polyether ether ketone (PEEK), 1.58 mm outer diameter and 0.40 mm wall thickness; McMaster-Carr, Princeton, NJ, USA]. The flat tip of the ground electrode was bare to allow aerosol sample collection. This same electrode system was also used to produce radio-frequency glow discharge at the tip of the collection electrode. A constant flow rate of 1.5 L min-1 was maintained through the aerosol collection system and was

117

driven by the native pump in the condensation particle counter (CPC; model 3022A, TSI Inc.,

Shoreview, MN, USA). The overall flow scheme was controlled using a mass flow controller

(MFC; model 247 C, MKS Instruments, Inc., Andover, MA, USA).

Table 5.1. Materials used to generate calibration aerosol for elements studies in this work

Element Source Chemical Density, g/cm3 Mass fraction, % composition C Sucrose solution C12H22O11 1.58 42 (Fisher Chemical) Cd Elemental standard solution Cd(NO3)3 3.6 47 (inorganic ventures) Mn Elemental standard solution Mn(NO3)2 1.54 30 (inorganic ventures) Na Sodium carbonate solution Na2CO3 2.54 11 (Fisher Chemical)

Glow discharge was generated in argon atmosphere using a radio frequency power supply

(PVM500, with a maximum output voltage of 1.6 kV and a frequency of 27.6 kHz). After the particle collection, pre-purified argon gas was introduced into the chamber at a constant flow rate of 0.9 L min-1 at atmospheric pressure. Once the glow discharge was initiated, the collected particulate matter on the ground electrode surface was ablated over few seconds (the time required for complete ablation of the sample depends on the particle mass). The glow discharge in the interelectrode gap was imaged onto the spectrograph (IsoPlane SCT320, Princeton Instrument Inc.,

Trenton, NJ, USA) using an UV-grade plano-convex lens (f=50 mm). The spectrograph was coupled with a gated intensified charge-coupled device (iStar 334T, Andor Technology, South

Windsor, CT, USA), which allowed recording of space- and time-resolved emission spectra from the glow discharge during the particulate sample ablation. The multi-track mode of the ICCD was used to record the space-resolved spectra, and the kinetic mode was used to record the time- resolved spectra. The wavelength calibration was achieved using an Hg-Ar lamp (Ocean Optics

118

Inc., Dunedin, FL, USA). Triggering of the spectrograph, RF power supply and data acquisition were accomplished through the built-in digital delay generator in the ICCD.

Figure 5.1. The schematic diagram of the experimental setup.

5.2.2. Plasma diagnostics

Gas temperature was determined from van der Waals broadening of argon emission line at

603.2 nm (Christova et al. 2004; Munoz et al. 2009; Sismanoglu et al. 2009). The profile of the spectral line Ar I 603.212 nm can be fitted to a Voigt function, which is the convolution of a

Gaussian function and a Lorentzian function (Munoz et al. 2009). Of various broadening mechanisms, instrumental broadening (∆휆퐼 ) and Doppler broadening (∆휆퐷 ) contribute to the

Gaussian component (∆휆퐺), while Stark broadening ( ∆휆푆) and Van der waals broadening (∆휆푊)

119

contribute to the Lorentzian component (∆휆퐿), as the following equations (Christova et al. 2004),

2 2 2 ∆휆퐺 = ∆휆퐼 + ∆휆퐷 (1)

∆휆퐿 = ∆휆푆 + ∆휆푊 (2)

Gaussian and Lorentzian components were separated from the deconvolution of fitted

Voigt profile based on the Leverberg-Marquardt non-linear algorithm for least squares. Fixing

Gaussian component has been shown to increase the accuracy of deconvolution (Konjevic et al.

2012). In a glow discharge plasma with low temperature, Doppler broadening was estimated to be

0.003 nm by the following equation (Djurović and Konjević 2009) (assuming T=2000K), which is negligible.

푇 ∆휆 = 7.16 × 10−7 휆( )1/2 (3) 퐷 푀 where 휆 is the wavelength (nm), 푇 is the temperature (K), and M is the Ar atomic mass (a.m.u.).

Therefore, Gaussian width was mainly caused by instrumental broadening, which was 0.05 nm for our instrument. Lorentzian width of was obtained from the deconvolution of fitted Voigt profile by fixing the Gaussian component at 0.05 nm. As the contribution of Stark broadening to Lorentzian width was negligible for 603.2 nm emission line (Munoz et al. 2009), van der Waals broadening was determined as the Lorentzian width. A simplified equation between the gas temperature

(푇푔) and the van der Waals broadening of a given spectral line was given by (Hofmann et al. 2012)

퐶푊 ∆휆푊 = 0.7 (4) 푇푔 where 퐶푊 is a coefficient that depends on the transition and the nature of the interacting atoms considered and 퐶푊 = 4.217 nm for Ar I (603.21 nm) (Yubero et al. 2007).

Stark broadening of Hβ (486.133 nm) was used for electron density calculation. Stark broadening was obtained by subtracting van der Waals broadening from the Lorentzian width. The van der Waals broadening of Hβ is given by (Belostotskiy et al. 2010) 120

−3 푃 ∆휆푊 = 6.8 × 10 0.7 (5) 푇푔 where P is the gas pressure (Torr), 푇푔 is the gas temperature (K). The stark broadening is given by

(Belostotskiy et al. 2010)

−10 2⁄3 Δ휆푆 = 2.5 × 10 훼(푛푒, 푇푒)푛푒 (6)

-3 where 푛푒 is the electron density (cm ) and 훼(푛푒, 푇푒) is the reduced wavelength separation for the selected transition and is a function of both of the electron density and electron temperature.

Generally, 훼(푛푒, 푇푒) was assumed to be a constant (0.077) in a glow discharge plasma that has

13 14 -3 푇푒 = 1–10 eV and 푛푒 = 10 –10 cm (Belostotskiy et al. 2010).

5.2.3. Calibration procedure

The calibration procedure consisted of the following steps: (i) generation of test aerosols,

(ii) collection of particles on the flat tip of the ground electrode for predetermined amount of time,

(iii) ablation of the collected particles by glow discharge, (iv) recording of the time-resolved emission spectra during glow discharge, (v) calculation of the emission signal for analyte of interest for each spectrum, (vi) calculation of the time-integrated signal intensity for an analyte of interest, (vii) construction of the calibration curve by plotting the integrated signal intensity as a function of analyte mass. Particulate mass deposited on the electrode for the known diameter of particles is given by,

휋 푚 = 휂퐶 푄 푡 𝜌 푑 3 (7) 푝 𝑖푛 푓 푐 푝 6 푣푒 where 휂 is the capture efficiency of particles, 퐶𝑖푛 is the particle concentration flowing into the chamber, 푄푓 is the aerosol volumetric flow rate, 푡푐 is the particle collection time, 𝜌푝 is the particle material density, 푑푣푒 is the volume equivalent diameter of particle. Assuming the particles are spherical, the volume equivalent diameter is equal to the electrical mobility diameter. Particle

121

capture efficiency was calculated by measuring the particle number concentration downstream of the collection unit using a CPC, with or without the presence of the electric field across the

퐻푉 푉=0 electrodes ( 푁표푢푡 and 푁표푢푡 ).

푉=0 퐻푉 푁표푢푡 − 푁표푢푡 휂 = 푉=0 (8) 푁표푢푡

The particulate mass loadings on the ground electrode of 1 to 100 ng were achieved by varying the collection time. For each mass loading, three replications were performed and the final calibration curve was constructed by averaging over three independent sets of measurements. The time for complete ablation of particles deposited on the electrode using glow discharge was measured, typically several seconds. The atomic emission from glow discharge was recorded kinetically with a gate width of 500 ms over a total cumulative period of 10 seconds for individual measurement. The total emission signal from the target analyte with the known mass was obtained by the time-dependent signal over the life of the glow discharge. The calibration curve was constructed by plotting the total signal intensity as a function of mass loaded on the collection electrode.

5.3. Results and discussion

5.3.1. Plasma gas temperature and electron density

Figure 5.2 (a) shows the gas temperature plotted as a function of interelectrode distance.

Higher gas temperatures were observed in the region close to the electrode than in the interelectrode gap. This is consistent with earlier studies on DC glow discharge, which show that the gas temperature is higher at the cathode surface and decreases with distance from cathode

(Obradovic and Kuraica 2006; Revel et al. 2000; Winter et al. 2008). This is possibly due to the increasing collisions between gas molecules and electrons emitted from the cathode surface. In our

122

RF glow discharge system, the two electrodes alternated between cathode and anode, leading to a higher temperature in the region close to both electrodes. The gas temperature of glow discharge in our system is in the range of 378 – 1438 K, which is similar to those reported in the literature

(Munoz et al. 2009; Yubero et al. 2013). Different approaches have been applied for measuring the gas temperature of glow discharge, such as filtered Rayleigh scattering (Yalin et al. 2002), Doppler broadening (Obradovic and Kuraica 2006), heat transport equilibrium (Revel et al. 2000) and van der Waals broadening (Munoz et al. 2009). The gas temperature depends on the type of glow discharge, carrier/working gas composition, gas flow rate, and the electrical characteristics of the discharge. The gas temperature of DC glow discharge was 700–900 K reported by Obradovic and

Kuraica (2006), 300–800 K by Yalin et al. (2002), and 300–360 K by Revel et al. (2000). The gas temperature of high frequency AC glow discharge was reported by Munoz et al. (2009) to be 120–

2000 K and Yubero et al. (2013) to be 800–1900 K. The temperature of dielectric barrier discharge was 315–460 K reported by Ionascut-Nedelcescu et al. (2008).

The electron density was deduced from the Stark broadening of the hydrogen Balmer (Hβ) line. Figure 5.2 (b) shows the electron density plotted as a function of interelectrode distance. The electron density reaches a maximum close to the cathode surface (negative glow region), and then decrease with the distance from the cathode. This trend agrees well with the prediction of one dimensional model of an argon microdischarge (Wang 2006). The electron density of our GD-OES system is in the order of 1014 cm-3, which is consistent with earlier results in literature (Belostotskiy et al. 2010; Qian et al. 2010; Zhou et al. 2013). This is lower by 3 to 5 orders of magnitude compared to that in pulsed spark discharge or laser-induced spark (Diwakar and Kulkarni 2012;

Dong et al. 2012; Harilal et al. 1998).

123

Figure 5.2. The variation of gas temperature (a) and electron density (b) as a function of interelectrode distance. (The cathode surface is at 0 mm and the anode surface is at 4 mm)

5.3.2. Spatial and Temporal Distribution of Emission Signal

Understanding spatial-temporal dynamics of the GD are important to optimize the signal- to-noise ratio of our rf-GD system. We probed spatial and temporal dynamics of emission signal in the interelectrode space during rf-GD. Figure 5.3 (a) and (b) shows key emission spectra, acquired at different locations along the longitudinal axis of the two electrodes in the interelectrode space in the absence of any analyte on the ground electrode. Figure 5.3 (c) and (d) shows the spectrum obtained at the collection electrode tip (at 0 mm). Several platinum and argon emissions were observed for our rf-GD-OES system. The platinum emission signal from the collection/ground electrode occurs mainly within 1 mm of the ground electrode surface, with the highest signal appearing at the electrode tip. As expected, the argon emission signal appears across the entire interelectrode gap. These measurements suggest that the excitation of atoms ejected from

124

the collection electrode, through collisions with ions, electrons, or other atoms in the glow discharge plasma, mainly occurs near the cathode electrode surface due to the high density for both negatively and positively charged particles in this region. This observation is consistent with those reported for the solution cathode GD-OES and a DC glow discharges at atmospheric pressure

(Davis and Marcus 2001; Staack et al. 2005). The region where platinum emission was observed matches the ‘negative glow’ (NG) region in a typical structure of low-pressure glow discharge.

The NG region is the source of light used in GD-OES and allows acquiring most analytical information. It is noted that both the platinum and argon emissions were observed below the cathode tip in the GD-OES system. This is consistent with the observation of glow discharge in an earlier study by Jiang, in which he found that parts of the plasma could run out of the discharge space between the electrodes and surround the cathode side walls (Jiang et al. 2014). Figure 5.3

(c) and (d) also shows that most emission lines are from neutrals, most likely due to the relatively low temperature of the glow discharge. Ionic emissions can be observed in GD for some elements with low ionization energies, but they are relatively weak (Doroski et al. 2013). GD-OES provides fewer emission lines compared to LIBS and spark microplasma emission spectroscopy. In addition, the linewidths are smaller and molecular band emissions are limited. These factors can potentially lower the possibility of spectral interferences (Broekaert 2003).

125

The temporal characteristics of analyte signal were investigated by acquiring time-resolved spectra with sucrose particles deposited on the collection electrode. Because glow discharge is a continuous plasma, during which the analyte is ablated layer by layer, the analyte signal is a function of time. Figure 5.4(a) shows the color contour plot for the time-resolved spectra obtained after the glow discharge was initiated (at t=0, with a gate width of 0.5 s). Figure 5.4 (b) and (c) shows the spectra obtained at t=0 s and t= 4 s, respectively. The carbon emission signal (C I 247.9 nm) is highest at t=0 s and then decreases with time, whereas the platinum

126

Figure 5.4. Time-resolved rf-GD-OES spectra acquired between 0 and 4 seconds after the glow discharge was initiated in the presence of sucrose particles on the collection electrode tip, (b) the spectrum acquired at t=0 s, and (c) the spectrum acquired at t=4 s. emission signals (Pt I 262.8 nm and Pt I 265.9 nm) appear at 0.5 s and then increase with time. At t=2 s, the carbon emission signal disappears, and the platinum emission signal reaches a maximum and remains unchanged after that. The decreasing carbon signal indicates that the sucrose particles were gradually ablated by the glow discharge. The amount of particulate carbon was predetermined

127

to be 81 ng using Eq. (1). It takes approximately two seconds for complete ablation of the particulate sucrose (81 ng carbon). Figure 5.5 shows that the spatially resolved spectra acquired in the presence of sucrose particle deposited on the cathode tip. It is seen that the atomic emission from the ablated particles also occurs in the region near the collection electrode. These spatial and temporal characteristics of our rf-GD system were used to optimize the signal-to-noise ratio and operating characteristics.

Figure 5.5. Space-resolved rf-GD-OES spectra acquired along the axis of the glow discharge plasma in the presence of sucrose particles on the collection electrode.

5.3.3. Reproducibility of the plasma characteristics

Continuous operation of the glow discharge can lead to the localized heating of the collection electrode and/or changes in electrode surface morphology (due to plasma etching), which in turn may affect particle collection characteristics, subsequently affecting the accuracy and precision of analyte signal measurement. To probe these artefacts, we measured the

128

temperature changes on the cathode using a thermocouple probe. Figure 5.6 shows variation of the temperature as a function of time after the glow discharge started and stopped. After the glow discharge was initiated, the cathode temperature rapidly increases to approximately 220 °C. After about t=20 s, the temperature approaches the equilibrium value. The increasing electrode temperature is resulted from the energy transfer from the reactive particles (ions, electrons, metastables, photons, etc.) in plasma to the electrode surface (Thomann et al. 2006). The time dependence of the electrode temperature and its value are consistent with that reported by Kristya

(2012) on an atmospheric pressure glow discharge in helium. The figure also shows that once the glow discharge is turned off, the electrode temperature drops to room temperature after approximately 30 s. The rapid heating and cooling of the electrode assures short collection cycles.

We investigated the repeatability of Ar I and Pt I signal during a single glow discharge event. Figure 5.7 shows variation of Ar I and Pt I signals as a function of time. A continuous glow discharge was produced for two minutes during which its optical emission spectra were recorded every 0.5 s. As shown in Figure 5.7, no significant variation was observed for Ar I (the relative standard deviation was 1.2% for Ar I). However, the signal intensity of Pt I increases with time.

This could probably be attributed to the increased cathodic sputtering yield resulting from the thermal effects during the glow discharge (Marcus 2013; Nelson 1965).

129

Figure 5.6. Changes in temperature on the electrode with time when glow discharge was on or off (the glow discharge is on at t=1 s and off at t=120 s).

Figure 5.7. Variation of Ar I and Pt I signal intensity as a function of time in glow discharge.

130

5.3.4. Analytical performance

Calibration curves for different analytes were constructed by depositing a known particulate mass on the collection electrode, following by measurement of emission signal as a function of time as described earlier. Figure 5.8 shows change of cumulative carbon emission signal (C I 247.9 nm) as a function of time for different particulate loadings. The particulate mass on the electrode tip was varied by changing the collection time, which varied from 1 to 5 minutes.

The cumulative carbon emission signal increases with time (and reaches maximum and remains unchanged after several seconds), indicating that the particles collected on the electrode tip were gradually ablated by glow discharge. The collected particulate matter is ablated completely at the moment when the cumulative carbon emission signal reaches maximum and remains unchanged.

From Figure 5.8, the time duration required for complete ablation of collected particulate carbon, ranging from 16.2 ng to 64.9 ng, was approximately 2 s. Larger mass loadings above 80 ng, takes slightly longer time for complete ablation (2.5 s).

The time-dependent signal intensity (퐼(푡) ) of the analyte from the glow discharge was

푇 integrated to obtain the total emission signal (퐼 ), such that 퐼 = 퐼(푡) 푑푡 , where 푇 is the 푡표푡 푡표푡 ∫0 period for which the glow discharge was turned on. Therefore 퐼푡표푡 is the cumulative signal at T.

Using the data in Figure 5.8, calibration curve was constructed by plotting the integrated signal intensity 퐼푡표푡 as a function of elemental mass (mp) deposited on the electrode tip. Figure 5.9 shows representative calibration curves for C, Cd, Mn, and Na. The selected analytical emission line for each element was C I 247.8 nm, Cd I 508.6 nm, Mn I 403.1 nm, and Na I 589.0 nm. Calibration curves were described using a linear fit. Three sets of measurements were performed for each mass loading. Each data point on the calibration curve represents the average over three replicates. The error bar represents the standard deviation around the mean, and the relative standard deviation,

131

which indicates the reproducibility of spectral response for a given analyte mass, varied in the range of 5 – 28% for all the elements. The overall repeatability of our system was comparable to reported values for LIBS aerosol techniques that use the similar particle preconcentration method

(Diwakar et al. 2012). The uncertainty in the calibration process is from three factors: i)

Repeatability of spatial and temporal characteristics of the plasma in the interelectrode space , ii) uncertainty in optical emission measurement, and iii) uncertainty associated with the estimation of an analyte mass deposited on the electrode tip. As has been noted before, other key sources, such as variation in particle shape, particle number concentration and collection efficiency fluctuation, are important for accurate mass estimation (Diwakar et al. 2012).

The limit of detection (LOD) was estimated using 3-𝜎 criteria defined by the International

Union of Pure and Applied Chemistry (IUPAC) as (Boumans 1994),

퐿푂퐷 = 3𝜎⁄푆 (9) where 𝜎 is the standard deviation of the blank at the selected spectral region and S is the sensitivity given by the slope of the calibration curve. Blank measurement was obtained in the absence of any particulate mass on the electrode using identical signal collection and spectral processing algorithm. The standard deviation of the blank was obtained by averaging over 15 replicate blank measurements for the spectral region corresponding to each elemental emission line. The mass

LOD was in the range of 0.55 – 1.0 ng depending on elements analyzed, as listed in Table 5.1. The

LOD in terms of air concentration was 7 to 134 ng m-3 at a flow rate of 1.5 l min-1 for a sampling time of 5 minutes. A much lower LOD can be achieved by either increasing sampling time or flow rate. Table 5.2 shows the comparison of LODs resulted from different aerosol measurement methods using microplasma spectroscopy, such as GD-OES, LIBS, and SES. The LODs of GD-

OES method coupled with corona particle preconcentration were significantly better than those

132

from particle beam interface GD-OES reported earlier (Marcus et al. 1999). Also LODs in this study were comparable to those for the LIBS and spark microplasma system coupled with our preconcentration system (Diwakar et al. 2012; Diwakar and Kulkarni 2012). It should be noted that the detector used in our study is ICCD, which provides tenfold increase in signal sensitivity compared to CCD spectrometer in our earlier studies. Thus glow discharge would give higher

LODs than laser or spark induced plasma assuming the same spectroscopic detection system is used.

Figure 5.8. Change of signal response (C I 247.9 nm) for different particulate carbon mass with glow discharge time.

133

Figure 5.9. Calibration curves for C, Cd, Na and Mn obtained using our GD-OES system.

Table 5.2. Comparison of detection limits in this work with other aerosol measurement methods using microplasma spectroscopy

GD-OES SES GD-OES LIBS particle beam Corona Corona corona Elements interface microconcentration microconcentration microconcentration (Marcus et al. (Diwakar and Kulkarni (Diwakar et al. 2012) (this work) 1999) 2012) C 0.49 (65)* ------Cd 1 (134) -- 5.03 (670) -- Cr -- -- 0.035 (47) 0.011 (1.5) Cu -- -- 0.138 (18) -- Fe -- 200 (N/A) -- -- Mn 0.28 (36) -- 0.155 (20) -- Na 0.055 (7) 30 (N/A) 0.018 (2.4) 0.028 (3.7) Pb ------1.75 (233) Si ------0.8 (107) Ti -- -- 0.44 (59) -- V -- 22 (N/A) -- -- *0.49 is the LOD in terms of mass (ng), and 65 is the LOD in terms of air concentration (ng m-3) assuming a collection time of 5 minutes and a flow rate of 1.5 l min-1

134

5.4. Conclusions

The CAM method has been effectively coupled with atmospheric glow discharge optical emission spectroscopy to allow near real-time measurements of elemental concentration of aerosols. The coaxial electrode system in CAM is inherently well-suited for continued particle collection and rf-GD generation. The spatio-temporal dynamics of the rf-GD in CAM were probed to optimize the signal-to-noise ratio for elemental measurement. Artefacts from electrode heating and electrode surface sputtering were negligible and did not interfere with elemental measurement.

Calibration curves were obtained by measuring the cumulative emission signal as a function of time for a given elemental mass deposited on the electrode tip. For elements studied in this work

(C, Cd, Mn and Na), our method provides detection limits in the range of 7 – 134 ng m-3 at a flow rate of 1.5 l min-1 for a 5 minute sampling period. The reproducibility of spectral response was in the range of 5 – 28%. Several other elements, though not studied in this work, can be measured using this system. The near real-time approach developed in this work offers an excellent alternative to development of low-cost, hand-held spectrometers for elemental analysis of aerosols for environmental and occupational health measurement applications.

5.5. References

Belostotskiy, S. G., Ouk, T., Donnelly, V. M., Economou, D. J., Sadeghi, N. (2010). Gas temperature and electron density profiles in an argon dc microdischarge measured by optical emission spectroscopy. J. Appl. Phys. 107. Bengtson, A. (1994). Quantitative depth profile analysis by glow discharge. Spectrochimi. Acta B: At. Spectrosc. 49:411-429. Boumans, P. (1972). Sputtering in a glow discharge for spectrochemical analysis. Anal. Chem. 44:1219-1228. Boumans, P. (1994). Detection limits and spectral interferences in atomic-emission spectrometry. Anal. Chem. 66:A459-A467.

135

Brewer, T. M., Castro, J., Marcus, R. K. (2006). Particle beam sample introduction into glow discharge plasmas for speciation analysis. Spectrochimi. Acta B: At. Spectrosc. 61:134-149. Broekaert, J. (2003). Optical emission spectrometry with glow discharges. Glow Discharge Plasmas in Analytical Spectroscopy:15. Brown, R. J. C., Jarvis, K. E., Disch, B. A., Goddard, S. L., Adriaenssens, E., Claeys, N. (2010). Comparison of ED-XRF and LA-ICP-MS with the European reference method of acid digestion- ICP-MS for the measurement of metals in ambient particulate matter. Accredit. Qual. Assur. 15:493-502. Christova, M., Castanos-Martinez, E., Calzada, M. D., Kabouzi, Y., Luque, J. M., Moisan, M. (2004). Electron density and gas temperature from line broadening in an argon surface-wave- sustained discharge at atmospheric pressure. Appl. Spectrosc. 58:1032-1037. Cserfalvi, T., Mezei, P., Apai, P. (1993). Emission studies on a glow-discharge in atmospheric- pressure air using water as a cathode. J. Phys. D-Appl. Phys. 26:2184-2188. Davis, W. C. and Marcus, R. K. (2001). An atmospheric pressure glow discharge optical emission source for the direct sampling of liquid media. J. Anal. At. Spectrom. 16:931-937. Diwakar, P., Kulkarni, P., Birch, M. E. (2012). New Approach for Near-Real-Time Measurement of Elemental Composition of Aerosol Using Laser-Induced Breakdown Spectroscopy. Aerosol Sci. Technol. 46:316-332. Diwakar, P. K. and Kulkarni, P. (2012). Measurement of elemental concentration of aerosols using spark emission spectroscopy. J. Anal. At. Spectrom. 27:1101-1109. Djurović, S. and Konjević, N. (2009). On the use of non-hydrogenic spectral lines for low electron density and high pressure plasma diagnostics. Plasma Sources Sci. Technol. 18:035011. Dong, M., Mao, X., Gonzalez, J. J., Lu, J., Russo, R. E. (2012). Time-resolved LIBS of atomic and molecular carbon from coal in air, argon and helium. J. Anal. At. Spectrom. 27:2066-2075. Doroski, T. A., King, A. M., Fritz, M. P., Webb, M. R. (2013). Solution-cathode glow discharge - optical emission spectrometry of a new design and using a compact spectrograph. J. Anal. At. Spectrom. 28:1090-1095. Doroski, T. A. and Webb, M. R. (2013). Signal enhancement in solution-cathode glow discharge— optical emission spectrometry via low molecular weight organic compounds. Spectrochimi. Acta B: At. Spectrosc. 88:40-45. Duan, Y., Su, Y., Jin, Z., Abeln, S. P. (2000). A field-portable plasma source monitor for real-time air particulate monitoring. Anal. Chem. 72:1672-1679. Harilal, S. S., Bindhu, C. V., Nampoori, V. P. N., Vallabhan, C. P. G. (1998). Temporal and spatial behavior of electron density and temperature in a laser-produced plasma from YBa2Cu3O7. Appl. Spectrosc. 52:449-455. Hofmann, S., van Gessel, A. F. H., Verreycken, T., Bruggeman, P. (2012). Power dissipation, gas

136

temperatures and electron densities of cold atmospheric pressure helium and argon RF plasma jets (vol 20, 065010, 2011). Plasma Sources Sci. Technol. 21. Hunter, A. J. R., Morency, J. R., Senior, C. L., Davis, S. J., Fraser, M. E. (2000). Continuous emissions monitoring using spark-induced breakdown spectroscopy. J. Air & Waste Mgt. Assoc. 50:111-117. Ionascut-Nedelcescu, A., Carlone, C., Kogelschatz, U., Gravelle, D. V., Boulos, M. I. (2008). Calculation of the gas temperature in a throughflow atmospheric pressure dielectric barrier discharge torch by spectral line shape analysis. J. Appl. Phys. 103. Jakubowski, N., Dorka, R., Steers, E., Tempez, A. (2007). Trends in glow discharge spectroscopy. J. Anal. At. Spectrom. 22:722-735. Jakubowski, N., Stuewer, D., Vieth, W. (1987). Performance of a glow discharge mass spectrometer for simultaneous multielement analysis of steel. Anal. Chem. 59:1825-1830. Jiang, W., Tang, J., Wang, Y., Zhao, W., Duan, Y. (2014). Characterization of argon direct-current glow discharge with a longitudinal electric field applied at ambient air. Sci. Rep. 4. Khalaji, M., Roshanzadeh, B., Mansoori, A., Taefi, N., Tavassoli, S. H. (2012). Continuous dust monitoring and analysis by spark induced breakdown spectroscopy. Opt. Laser Eng. 50:110-113. Konjevic, N., Ivkovic, M., Sakan, N. (2012). Hydrogen Balmer lines for low electron number density plasma diagnostics. Spectrochimi. Acta B: At. Spectrosc. 76:16-26. Kristya, V. I. (2012). Calculation of time dependence of the electrode surface temperature in the cathode spot of atmospheric pressure normal glow discharge. J. Surf. Ingestig.-X-Ray Synchro. 6:255-258. Kulkarni, P. and Efthimion, P. (2015). Electrode-assisted microwave-induced plasma spectroscopy, US Patents, US 9091597 B2. Manalis, N., Grivas, G., Protonotarios, V., Moutsatsou, A., Samara, C., Chaloulakou, A. (2005). Toxic metal content of particulate matter (PM10), within the Greater Area of Athens. Chemosphere 60:557-566. Marcus, R. K. (2003). 5.5 Glow Discharge Plasmas as Tunable Sources for Elemental Speciation. Handbook of Elemental Speciation: Techniques and Methodology 1:334. Marcus, R. K. (2013). Glow discharge spectroscopies. Springer Science & Business Media. Marcus, R. K., Dempster, M. A., Gibeau, T. E., Reynolds, E. M. (1999). Sampling and analysis of particulate matter by glow discharge atomic emission and mass spectrometries. Anal. Chem. 71:3061-3069. Munoz, J., Dimitrijevic, M. S., Yubero, C., Calzada, M. D. (2009). Using the van der Waals broadening of spectral atomic lines to measure the gas temperature of an argon-helium microwave plasma at atmospheric pressure. Spectrochimi. Acta B: At. Spectrosc. 64:167-172. Nelson, R. S. (1965). An investigation of thermal spikes by studying the high energy sputtering of 137

metals at elevated temperatures. Philos. Mag. 11:291-302. Obradovic, B. M. and Kuraica, M. M. (2006). Optogalvanic effect and measurement of gas temperature in an abnormal glow discharge. Appl. Phys. Lett. 89. Pöschl, U. (2005). Atmospheric aerosols: Composition, transformation, climate and health effects. Angew. Chem. Int. Ed. 44:7520-7540. Park, K., Cho, G., Kwak, J.-H. (2009). Development of an Aerosol Focusing-Laser Induced Breakdown Spectroscopy (Aerosol Focusing-LIBS) for Determination of Fine and Ultrafine Metal Aerosols. Aerosol Sci. Technol. 43:375-386. Qian, M., Ren, C., Wang, D., Zhang, J., Wei, G. (2010). Stark broadening measurement of the electron density in an atmospheric pressure argon plasma jet with double-power electrodes. J. Appl. Phys. 107. Quarles Jr, C. D., Gonzalez, J., Choi, I., Ruiz, J., Mao, X., Marcus, R. K., Russo, R. E. (2012). Liquid sampling-atmospheric pressure glow discharge optical emission spectroscopy detection of laser ablation produced particles: A feasibility study. Spectrochimi. Acta B: At. Spectrosc. 76:190- 196. Radziemski, L. J., Loree, T. R., Cremers, D. A., Hoffman, N. M. (1983). Time-resolved laser- induced breakdown spectrometry of aerosols. Anal. Chem. 55:1246-1252. Revel, I., Pitchford, L. C., Boeuf, J. P. (2000). Calculated gas temperature profiles in argon glow discharges. J. Appl. Phys. 88:2234-2239. Schelles, W., Maes, K. J., De Gendt, S., Van Grieken, R. E. (1996). Glow discharge mass spectrometric analysis of atmospheric particulate matter. Anal. Chem. 68:1136-1142. Sismanoglu, B. N., Amorim, J., Souza-Correa, J. A., Oliveira, C., Gomes, M. P. (2009). Optical emission spectroscopy diagnostics of an atmospheric pressure direct current microplasma jet. Spectrochimi. Acta B: At. Spectrosc. 64:1287-1293. Staack, D., Farouk, B., Gutsol, A., Fridman, A. (2005). Characterization of a dc atmospheric pressure normal glow discharge. Plasma Sources Sci. Technol. 14:700. Thomann, A. L., Semmar, N., Dussart, R., Mathias, J., Lang, V. (2006). Diagnostic system for plasma/surface energy transfer characterization. Rev. Sci. Instrum. 77:6. Vors, E. and Salmon, L. (2006). Laser-induced breakdown spectroscopy (LIBS) for carbon single shot analysis of micrometer-sized particles. Anal. Bioanal. Chem. 385:281-286. Voutsa, D. and Samara, C. (2002). Labile and bioaccessible fractions of heavy metals in the airborne particulate matter from urban and industrial areas. Atmos. Environ. 36:3583-3590. Wang, Q. (2006). Plasma diagnostics and modeling of direct current microplasma discharges at atmospheric pressure. Thesis (Ph.D.)-University of Huston. Wang, Z., Schwartz, A. J., Ray, S. J., Hieftje, G. M. (2013). Determination of trace sodium, lithium, magnesium, and potassium impurities in colloidal silica by slurry introduction into an atmospheric- 138

pressure solution-cathode glow discharge and atomic emission spectrometry. J. Anal. At. Spectrom. 28:234-240. Winter, J., Lange, H., Golubovskii, Y. B. (2008). Gas temperature in the cathode region of a dc glow discharge with a thermionic cathode. J. Phys. D-Appl. Phys. 41. Yalin, A. P., Ionikh, Y. Z., Miles, R. B. (2002). Gas temperature measurements in weakly ionized glow discharges with filtered Rayleigh scattering. Appl. Opt. 41:3753-3762. Yubero, C., Dimitrijevic, M. S., Garcia, M. C., Calzada, M. D. (2007). Using the van der Waals broadening of the spectral atomic lines to measure the gas temperature of an argon microwave plasma at atmospheric pressure. Spectrochimi. Acta B: At. Spectrosc. 62:169-176. Yubero, C., Garcia, M. C., Varo, M., Martinez, P. (2013). Gas temperature determination in microwave discharges at atmospheric pressure by using different Optical Emission Spectroscopy techniques. Spectrochimi. Acta B: At. Spectrosc. 90:61-67. Zheng, L., Kulkarni, P., Zavvos, K., Liang, H., Birch, M. E., Dionysiou, D. D. (2016). Aerosol Microconcentrator for Analysis using Microscale Optical Spectroscopie. J. Aerosol Sci. in review Zhou, Y.-J., Yuan, Q.-H., Li, F., Wang, X.-M., Yin, G.-Q., Dong, C.-Z. (2013). Nonequilibrium atmospheric pressure plasma jet using a combination of 50 kHz/2 MHz dual-frequency power sources. Phys. Plasmas 20.

139

CHAPTER 6

Spatial and Temporal Dynamics of Pulsed Spark Microplasma Used for

Aerosol Analysis

140

The spatial and temporal dynamics of spark plasma used for spectrochemical analysis of aerosol was investigated. The spark plasma was generated by applying a high voltage pulse on two coaxial electrodes, with carbon black particles deposited on the cathode. Spectroscopic techniques are used to probe spatial and temporal evolution of the spark microplasma to gain insights into its dynamics. The time and spatial resolved emission spectrum from plasma and its images were recorded using a spectrograph coupled with an intensified charge-coupled device.

The C II emission lines (251.206, 283.671, and 426.726 nm) were used to construct Boltzmann plots for determination of excitation temperature, and Stark broadening of Hα emission line

(656.28 nm) was used for electron density calculation. Both the plasma imaging and spectroscopic analysis show that the life time of spark plasma is approximately 15 µs. The excitation temperature of spark plasma is in the range of 15,000 – 35,000 K, and the electron density is in the range of

(1.0– 2.2) × 1017 cm-3. Variation of excitation temperature and electron density as a function of interelectrode distance and delay time were studied. The influence of high voltage pulse energy on the excitation temperature and electron density was also probed. Our findings provide physical insight into the spark plasma and enhance the performance of spark emission spectroscopy in elemental analysis.

141

6.1. Introduction

Spark emission spectroscopy (SES) has been applied to elemental analysis of a broad range of samples, such as alloys, cement powder, soil, fuel, and dust (Golloch and Wilke 1997; Kawahara et al. 2009; Khalaji et al. 2012; Schmidt et al. 2012; Taefi et al. 2010). SES can analyze multi- elements with no or minimal sample preparation, and it provides good detection limits and reproducibility (Diwakar and Kulkarni 2012; Walters 1977). In this technique, a high voltage pulse is applied on two coaxial electrodes to generate a high energetic electrical spark, which heats the sample to a high temperature to excite the atoms within it and produce optical emission (Falk and

Wintjens 1998; Taefi et al. 2010; Walters 1977). The solid samples are usually analyzed by being embedded in an electric conductive material. Gas or aerosol samples have also been in situ analyzed by creating spark plasma between two electrodes where the sample passed through

(Hunter et al. 2000b; Khalaji et al. 2012). Developing novel instruments based on SES for elemental analysis of various samples has been of great interest (Dittmar et al. ; Golloch and

Siegmund 1997). In order to improve the performance of SES in analytical instrumentations, the physics and fundamental aspects of spark microplasma are needed to be investigated.

The process of spark discharge was first described by Walters (1977). The events composing spark discharge fall into three classes: i) formation and expansion of a conductive ion channel, ii) excitation of electrode material, transportation of excited species, and radiation, iii) development of postdischarge environment. Later, Walters and Goldstein (1984) obtained the

“emission topography” of the spark discharge using time and space resolved spectroscopy and characterized the transportation of excited species in the interelectrode gap. A two dimensional image of atomic emission emitted from a spark plasma has been obtained using an imaging spectrograph equipped with a charge-coupled device detector (Ramli and Wagatsuma 2010).

142

Studies on spatial distribution and temporal evolution of emission from spark discharge provide detailed information on the excitation process in spark plasma, and are also useful for determining the optimum conditions for the analytical application.

In microplasma spectroscopy, some important plasma parameters such as excitation temperature and electron density are required in order to produce reliable spectral features. There has been extensive research on laser-induced plasma diagnostics, such as the excitation temperature, the electron density and their dependence on delay time, distance from the target surface and laser irradiance (Baig et al. 2012; El Sherbini and Al Aamer 2012; Salik et al. 2013).

However, relatively little information is available on spark plasma diagnostics. Bye and Scheeline

(1993) have calculated the electron temperature (~15,000 K) and electron density (~2 × 1017 cm-3) of a spark discharge plasma using Saha-Boltzmann equation. Until now, a systematic investigation of spark plasma characteristics has not been reported.

In our previous study, Diwakar and Kulkarni (2012) proposed a method for near real-time measurement of elemental concentration in aerosol, which involves preconcentration of particles onto the cathode tip, followed by ablation of the particles by a spark discharge and optical emission spectroscopic analysis. The objective of this study is to investigate spatial and temporal dynamics of the spark microplasma used in this aerosol measurement system. The spatial and temporal behavior of neutral and ionic carbon emission from carbon black particles was investigated using a spectrograph coupled with a gated ICCD. The excitation temperature and electron density were evaluated using spectroscopic means, and their dependences on interelectrode distance, delay time and pulse energy were presented.

6.2. Experimental methods

A schematic diagram of the experimental setup is shown in Figure 6.1. It consists of three

143

components: i) an aerosol generation system, ii) an aerosol collection system, and iii) an SES system. An aerosol was generated using a pneumatic atomizer (model 3076, TSI Inc., Shoreview,

MN, USA) to aerosolize the target suspension prepared by mixing carbon black (REGAL® 400R,

Cabot Corporation, Billerica, MA) into ultra-filtered DI water. The aerosol was passed through a diffusion dryer to remove associated water, and the resulting dry particles were deposited on the ground electrode tip in a corona aerosol microconcentrator (CAM) developed in an earlier study

(Diwakar et al. 2012). The CAM system consists of two coaxial electrodes with a separation distance of 5 mm. Both the two electrodes were 500 µm in diameter. The corona electrode, made of tungsten, has a sharp tip and was used to create a corona around its tip. The ground electrode, made of platinum, has a relatively flat tip to provide a planar surface for particle deposition. A high positive potential (~5 kV) was applied on the corona electrode through a DC power supply (Bertran

S-230, Spellman Corp., Hauppauge, NY, USA). For each measurement, the same amount of particulate carbon black was deposited on the ground electrode by using the same aerosol flow rate and collection time.

In the SES system, a high voltage pulse generator (Cascodium Inc., Andover, MA), was used to produce a spark microplasma in the interelectrode gap to ablate the particulate carbon black collected on the cathode. The pulse generator has an output energy in the range of 50 – 800 mJ per pulse. The emission from atomic and ionic carbon species in the spark plasma was collected and imaged onto the slit of a spectrograph (IsoPlane SCT320, Princeton Instrument Inc., Trenton, NJ,

USA) using an ultraviolet-grade plano-convex lens with a focal length of 50 mm and a diameter of 25 mm. The spectrograph was coupled with a gated intensified charge coupled device (ICCD) detector (iStar 334T, Andor Technology, South Windsor, CT, USA), which allows to record space and time resolved spectra. A slit width of 10 µm was used for spectra collection, while a slit width

144

of 3 mm was used for image collection. The space-resolved spectra was recorded using multi-track mode. The time-resolved spectra were obtained by changing delay time while keeping the same gate width (1 µs). All the spectra were obtained with background correction. The plasma images were recorded in the image mode using a gate width of 250 ns. Three gratings are available, i.e.

1200/300 (1200 l/mm groove density and 300 nm blaze wavelength), 1200/500, and 1800/500.

The selection of grating depended on the demanded resolution and interested wavelength region.

Triggering of spectrograph, high power pulse and data acquisition were accomplished through the built-in DDG in the ICCD. The wavelength calibration was achieved using an Hg-Ar lamp (Ocean

Optics Inc., Dunedin, FL, USA).

Figure 6.1. Schematic diagram of the experimental setup.

6.3. Results and discussion

6.3.1. Continuum emission from spark plasma

In the initial stage of plasma, the emission is dominated by the continuum emission due to 145

radiative recombination and Bremsstrahlung effects (Anabitarte et al. 2012). The continuum emission contains little spectroscopic information and tends to overshadow the emission lines from the excited species present in the plasma. For this reason, the spectra acquisition should be delayed in order to avoid this continuum. Figure 6.2 shows the spectra obtained from spark induced carbon plasma at different decay time of 0 µs, 1 µs and 5 µs, using a gate width of 1 µs. The wavelength window in Figure 6.2 is 268 – 298 nm, in which C II emission lines at 274.649 nm and 283.671 nm are located. The spectrum obtained at the beginning of plasma evolution (0 delay) shows the continuum emission. After that, the ionic carbon emission line at 283.671 nm was observed at 1

µs. This suggests that the continuum emission occurred within 1 µs after the ignition of spark plasma. This is consistent with the earlier observation by Waters in a spark discharge system

(Walters and Goldstein 1984). This result is also comparable to nanosecond pulsed laser induced plasma, in which continuum emission appears during the laser pulse and lasts for several hundred nanoseconds to several microseconds depending on laser wavelength (Singh and Thakur 2007).

The time gate of decay of this continuum radiation also changes with experimental parameters, such as pulse duration, ambient pressure, or sample features (Anabitarte et al. 2012). In our experimental condition, the SES spectra acquisition must be delayed by 1 µs to improve the ratio of signal with respect to the continuum background for the purpose of elemental analysis.

146

Figure 6.2. Spark emission spectra at a delay time 0, 1, and 5 µs, with a gate width of 1 µs.

6.3.2. Spatial and temporal evolution of spark emission

Due to the transient nature of pulsed spark plasma, the populations of the various species present in the plume rapidly evolve with time and position. The spatial and temporal behavior characterization was carried out with spatially and temporally resolved spectra in air at atmospheric pressure. Three spectral windows that contain one atomic carbon emission line (C I

247.856 nm) and three ionic carbon emission lines (C II 251.206 nm, C II 283.671 nm, and C II

426.726 nm) were used. As shown in Section 3.4, these ionic carbon emission lines can be used for excitation temperature calculation.

Figure 6.3 shows the spatially resolved spectra of spark emission from carbon black at a delay time of 5 µs. These spectra were taken axially from the inter-electrode gap with a spatial resolution of 0.4 mm. At 5 µs, all the four carbon emission lines have their maximum signals at

0.6 mm above the cathode surface, and decrease with distance from that position on both directions.

147

The extension of emission is approximately 6 mm above and 2 mm below the cathode surface.

This suggests that neutral and ionized carbon species that were generated from excitation of carbon black particles on the cathode by spark discharge, expanded toward two opposite directions away from the cathode surface. The expansion of spark induced plasma in this study is in the same order as laser induced plasma. Baig et al. (2012) showed that the extension of neutral lithium emission along laser direction was 2 mm, with the strongest emission near the target surface, using a 1064 nm laser at the irradiance of 46 GW/cm2. The extension of emission and the location of the maximum emission signal varied depending on laser irradiance, with larger extension at higher laser irradiance. Aragon et al. showed that the extension of copper, iron, and magnesium emission along laser direction varied in the range of 3 – 7 mm, and the distance of the maximum of emission intensity to the target surface varied in the range of 2 – 4 mm using laser power density ranging

80 – 900 GW/cm2 (Aragon and Aguilera 1997). The emission observed below the cathode surface in the spark discharge system was likely due to the size of cathode tip in the similar micro scale as the spark plasma size, leading to the expansion of species below the cathode surface. We also observed a weak increasing trend of the carbon signals when it is approaching to the anode surface, which may be caused by a small amount of carbon particles attached on the anode surface while they were introduced into the preconcentrator.

148

Figure 6.3. Space-resolved carbon emission lines from carbon black at delay time of 5 µs with a gate width of 1 µs. (a) C I 247.856 nm and C II 251.206 nm; (b) C II 283.671 nm; (c) C II

426.726 nm. (The cathode is at 0 mm, and the anode is at 5 mm)

149

To probe the evolution of carbon emission generated by spark discharge, the spatial resolved spectra profiles were also recorded at different delay time. The signal intensity was calculated as the peak height with base line subtraction. Figure 6.4 shows the temporal evolution of carbon emission at 247.856 nm, 251.206 nm, 283.671 nm, and 426.726 nm. After the initiation of the plasma, the carbon emission only appeared in the region close to the cathode surface at 1, and the carbon emission signal was weak. With further evolution of the spark plasma, the carbon plume expands towards the anode, and the emissions become more intense. C I and C II emission exhibit different expansion velocities and life time. It is seen that C I emission at 247.856 nm extends up to the anode surface at ~ 6 µs (Figure 6.4 (a)), while C II emission at 251.206 nm,

283.671 nm, and 426.726 nm extends up to the anode at 4 µs (Figure 6.4 (b-c)). From these measurements, axial expansion velocity of carbon species, defined as the ratio of inter-electrode distance to transport time between cathode and anode, was estimated to be 0.8 mm/µs for C I emission and 1.25 mm/µs for C II emission. Axial velocity reported by Walters and Goldstein

(1984) was approximately 1.0 mm/µs for the copper species and 0.8 mm/µs for silver species in a similar pulsed spark plasma. The variations in the axial velocity of emission fronts were possibly due to difference in spark characteristics, electrode characteristics, electrode configuration, and chemical species (Walters and Goldstein 1984). From Figure 6.4, the carbon signals achieve their highest value at 11 µs for C I and at 6 µs for C II, and then decrease rapidly with time. The life time of C I emission is estimated to be approximately 15 µs, while it is 10 µs for C II emission.

This indicates that neutral carbon emission dominated after the spark plasma had expanded and cooled, which is consistent with the observation in laser induced plasma(Le Drogoff et al. 2001).

Figure 6.4 also shows that the emission at 426.726 nm is confined to a region closer to the cathode surface in comparison to the spatial distribution of the emissions at 247.856, 251.206, and

150

283.671 nm. This can be explained by the difference in upper energy levels of the transitions.

During the plasma plume expansion, the spectral lines from highly ionized species are observed close to the target whereas those from lower ionization and neutral species are observed in the plume away from the target (Aguilera and Aragón 2004; Aragon and Aguilera 1997; Singh and

Thakur 2007). As shown in Table 6.1, it has the highest upper energy level for the C II transition

2s23d 2D – 2s24f 2Fo at 426.726 nm, resulting in emission observed much closer to the target surface.

Figure 6.4. The temporal evolution of carbon emission generated by spark discharge: (a) C I

247.856 nm, (b) C II 251.206 nm, (c) C II 283.671 nm, and (d) C II 426.726 nm.

Table 6.1. C I and C II spectroscopic data

-1 Wavelength, nm 푬풌, eV 품풌 푨풌풊, s Transition 151

247.856 7.68 2.80× 107 2s22p2 1S – 2s22p3s 1Po 251.206 18.65 3.37 × 108 2s2p2 2P – 2p3 2Do 283.671 16.33 1.32 × 108 2s2p2 2S – 2s23p 2Po 426.726 20.95 9.54 × 107 2s23d 2D – 2s24f 2Fo

6.3.3. Plasma imaging

The behavior of plasma produced by high voltage pulse was also studied by recording temporally resolved images using a gated ICCD camera system. Figure 6.5 shows the temporal evolution of plasma by recording total emission from carbon black at delay time of 0 to 15 µs.

Each image in this Figure was recorded during an independent spark discharge event. The plasma image at 0 µs shows continuum emission with high signal intensity, which is consistent with the spectrum obtained at 0 µs shown in Figure 6.2. As the continuum emissions decays, the plasma containing ionic species was formed in the inter-electrode gap and expanded to the surrounding nearly symmetrically to the axis of electrodes, with higher density in the center. At 6 µs, both the plasma size and density reached their maximum levels. After that, the plasma could still keep the similar size, but weaken gradually. The spatial and temporal trend in spark plasma observed in our system was consistent with those described in the earlier studies on spark discharge formation and evolution (Walters 1972; Walters and Goldstein 1984). As Walters (1977) noted, when a high voltage pulse is applied between the electrodes, a conducting ion channel is first established to induce dielectric breakdown of air and is directed towards the cathode. This conducting channel further spatially expands due to energy deposition in the interlelectrode space, leading to a pulsed spark discharge. After the energy deposition initiates, the plasma plume at the cathode surface expands rapidly due to space-charge effects, while the central conducting channel rapidly shrinks as the energy deposition in the interelectrode gap further ceases. Our measurements of plasma evolution are consistent with these mechanisms of sparks discharge formation and propagation.

152

The images in Figure 6.5 show the total emission, including the emission from excitation of carbon black particles collected on the cathode as well as molecules in air. To probe the evolution of carbon species in plasma, a bandpass filter was used to produce narrow-wavelength- band images of the plasma. With a proper choice of peak wavelength and spectral bandwidth, the bandpass filter could isolate and transmit a single transition, providing a means of measuring the temporal and spatial dependence of the wavelength-associated specie distribution in plasma. In this study, a bandpass filter with peak wavelength of 250 nm and FWHM of 11 nm was used to selectively detect carbon emission, which permitted radiation from C I transition at 247.856 nm and C II transition at 251.206 nm. Figure 6.6 shows images of carbon emission from spark plasma transmitted by a 250 nm bandpass filter. It is seen that carbon emission appeared close to the cathode surface at an early stage of plasma, and then expanded toward the anode. Both the expansion and signal intensity reach maximums at 6 µs after the initiation of spark plasma, and then shrunk. These observations by ICCD imaging of carbon emission are consistent with those distributions of spatially and temporally resolved spectra at C I 247.856 nm and C II 251.206 nm in Figure 6.4(a, b). From Figure 6.6, we also note that the lateral extension of carbon emission is approximately 3 mm.

153

Figure 6.5. ICCD Images of spark plasma generated on carbon black sample at various delay time. (The cathode is at 0 mm, and the anode is at 5 mm)

Figure 6.6. Images showing temporal and spatial development of the carbon emission from spark plasma transmitted by a 250 nm bandpass filter. (The cathode is at 0 mm, and the anode is at 5 mm)

154

6.3.4. Excitation temperature and electron density

The excitation temperature was determined using Boltzmann plot method. In this method, the atomic emission spectral line intensity 퐼푍 is a measurement of the population of the corresponding energy level of the element in the plasma. If the plasma is in local thermodynamic equilibrium (LTE) and optically thin, the intensity 퐼푍 of a spectral line occurring between the upper energy level k and the lower energy level i of the species in ionization stage Z is given by

Boltzmann law,

ℎ푐 푁0 퐸푘,푍 퐼푍 = 퐿 푔푘,푍퐴푘𝑖,푍exp (− ) (1) 4휋휆푘푖,푍 푃푍 푘퐵푇 where 퐼푍 is the line intensity, 휆푘𝑖,푍 is the transition line wavelength, 푔푘,푍 is degeneracy of the upper energy level k, 퐴푘𝑖,푍 is the transition probability, 푘퐵 is the Boltzmann constant, T is the plasma temperature, 퐸푘,푍 is the energy, h is the Planck constant, c is the speed of the light, L is the characteristic length of the plasma, 푁0 is the total number density and 푃푍 is the partition function of the species in ionization stage Z. By taking the natural logarithm, Eq. (1) can be rewritten as,

퐼푍휆푘푖,푍 1 ℎ푐퐿푁0 ln ( ) = − 퐸푘,푍 + ln ( ) (2) 푔푘,푍 퐴푘푖,푍 푘퐵푇 4휋푃푍

Eq. (2) yields a linear plot and the value of T is deduced from the slope of the Boltzmann plot. The carbon emission lines used to estimate the plasma temperature are 251.205, 283.671 and 426.726 nm. These transitions were selected because they have the great difference between their corresponding upper energy levels to determine the temperature more accurately. The spectroscopic data of these transitions were taken from NIST atomic spectra database and given in

Table 6.1.

The electron density of the plasma was determined by measurement of the stark broaden line of Hα at 656.28 nm. Of various broadening mechanisms (i.e. Stark, Doppler, and resonance

155

broadening), the Stark effects have the most important contribution to line broadening in spark plasma. Using Stark broadening of Hα line, the electron density can be calculated as (Ashkenazy et al. 1991)

∆휆 푛 = 8.02 × 1012( 푠)3/2 푐푚−3 (3) 푒 훼 where ∆휆푆 is FWHM of the Stark broaden line in Å and 훼 is half-widths of reduced Stark profiles.

훼 is a function of the temperature and electron density as provided in Griem's table (Kepple and

Griem 1968). The Stark line width was extracted from the measured line width ∆휆표푏푠푒푟푣푒푑 by subtracting instrument broadening ∆휆𝑖푛푠푡푟푢푚푒푛푡 and Doppler line broadening ∆휆퐷표푝푝푙푒푟

(Camacho et al. 2011). ∆휆푂푏푠푒푟푣푒푑 was obtained from Lorentzian fitting operated in OriginLab 8.5.

Instrumental broadening depends on the device. In our experiments, ∆휆퐼푛푠푡푟푢푚푒푛푡 was 0.05 nm measured by an Hg-Ar lamp (HG-1, Ocean Optics Inc., Dunedin, FL, USA). Doppler broadening

−7 1/2 can be obtained as: ∆휆퐷표푝푝푙푒푟 = 7.16 × 10 휆(푇⁄푀) , where 휆 is the wavelength in nm, 푇 is the temperature in K, and 푀is the atomic mass in amu.

Spatial dependence. The estimation of electron density and excitation temperature of the spark induced carbon plasma were carried out for different interelectrode distance in a time resolved manner using spark energy of 200 mJ and gate width of 1 µs. To probe the space dependence of plasma parameters, time averaged excitation temperature and electron density were plotted as a function of interelectrode distance (Figure 6.7). The excitation temperature of spark plasma ranges from 15,000 to 35,000 K, with highest excitation temperature at 0.6 mm above the cathode surface, and decreases with distance to 0.6 mm in the region close to cathode (-1 to 4 mm).

As the plasma expands toward the anode, the thermal energy is converted into the kinetic energy, causing the temperature to drop off rapidly. The spatial dependence of spark plasma excitation temperature is similar with those observed in laser induced plasmas (Baig et al. 2012; Harilal et al.

156

1997; Harilal et al. 1998). Harilal et al. (1997) evaluated the laser induced carbon plasma excitation temperature with respect to the distance from the target surface and found that the temperature decreased with distance perpendicular to the target surface. The electron density estimated by Stark

17 17 -3 broadening of Hα line in our study was in the range 1.0 × 10 –2.2 × 10 cm , with relatively higher electron density in the regions close to the cathode as well as the anode. The spatial dependence of spark induced plasma electron density was different from that observed in laser induced plasmas, which have decreasing electron density with distance to the target surface

(Harilal et al. 1997; Harilal et al. 1998). The higher electron density observed near the anode surface can be explained by the formation of a conducing ion channel directed towards the cathode when a high voltage pulse is applied between the electrodes. Diwakar and Kulkarni (2012) noted that the plasma in the interelectrode space can be described using two overlapping regions as the expansion of both the central ion channel and the spark plasma plume, and both of them contribute the electrons. Therefore, two regions with higher electron density were observed, with one close to cathode and the other close to anode. Both the excitation temperature and the electron density of spark induced carbon plasma in our study are comparable to those attained in a laser produced carbon plasma using similar energy input (Harilal et al. 1997).

157

Figure 6.7. The variation of excitation temperature and electron density as a function of interelectrode distance.

Temporal evolution. Temporal variation of excitation temperature and electron density was studied in a space averaged manner with a time resolution of 1 µs. Figure 6.8 shows the variation of space averaged excitation temperature and electron density plotted as a function of delay time. It was found that both the plasma excitation temperature and electron density increases in the early stages of plasma, and reach the maximum after 4 µs delay, and then decrease linearly with delay time. This temporal evolution trend of excitation temperature and electron density observed in spark plasma is different from that of laser plasma. In the early stages of laser plasma evolution, both the excitation temperature and electron density are high and decrease rapidly with time, but afterwards get stabilized for a long period (Dong et al. 2012; Harilal et al. 1997; Harilal et al. 1998). At the initiation of spark plasma, the temperature is low. At this period, a conducting ion channel is established and directed towards the cathode. The channel expands spatially and the

158

energy is deposited in the interelectrode gap, leading to a controlled spark discharge. The spark plasma expands isothermally within the time of the duration of ion channel, thus the space averaged temperature increased rapidly. Afterward, a rapid decrease in excitation temperature occurred at 5 µs which coincided with the time when the conducting channel ceased but the plasma continued to expand adiabatically. In our study, the correlation between the electron density and temperature demonstrates that the raise in the plasma temperature by spark discharge promotes plasma ionization and, therefore, increases the electron density.

Figure 6.8. The variation of excitation temperature and electron density as a function of delay time.

Effect of spark energy. Figure 6.9 shows variation of the excitation temperature and electron density of spark produced carbon black plasma with respect to spark energy at a distance

1 mm from cathode surface and at delay time 5 µs. As pulse energy increases from 80 to 200 mJ, the excitation temperature increases from 13,000 to 20,000 K, and saturates at higher energy level,

159

meanwhile the electron density increases from 1.1 × 1017 cm-3 to 2.0 × 1017 cm-3 and then saturates.

The variation of excitation temperature and electron density with respect to pulse energy could be described using exponent functions, as shown in Figure 6.9. These results suggest the nature and characteristics of spark produced plasma strongly depend on high voltage pulse energy. The saturation in excitation temperature and electron density at higher pulse energy is likely to due to plasma shielding or the formation of a self-regulating regime at higher irradiances (Harilal et al.

1997). Because of this saturation phenomenon, the pulse energy is recommended to be 200 mJ in spark induced plasma spectroscopy.

Figure 6.9. The variation of excitation temperature and electron density as a function of energy

(at a distance of 1 mm and a delay time of 5 µs).

Local thermodynamic equilibrium (LTE). The calculation of excitation temperature was carried out by assuming that the plasma is in LTE. It is noted that the LTE is approached when

160

electron collisions are the major processes of de-excitation. A McWhirter criterion must be satisfied by the plasma to be in LTE (Anabitarte et al. 2012):

12 3 1⁄2 푁푒 ≥ 1.6 × 10 ∆퐸 푇푒 (4) where ∆퐸 (eV) is the largest observed transition energy for which the condition holds, and 푇푒 is the excitation temperature (K). If we consider the largest energy, ΔE=5.0 eV, and the highest temperature, T=35,000 K, the minimal electron density for LTE is 3.7×1016 cm-3. From Figure 6.7, the experimentally determined electron density of spark plasma is significantly higher than this criterion value. Therefore, it is proper to calculate the excitation temperature using the Boltzmann plot.

6.4. Conclusions

The spatial and temporal dynamics of spark microplasma used for spectrochemical analysis of aerosol were investigated. The spatial and temporal behavior of the transient emission from carbon was investigated using imaging and spectroscopic techniques. Carbon emission appeared close to the cathode surface at an early stage of plasma, and then expanded toward the anode. The neutral and ionic carbon emissions exhibit different spatial and temporal behavior. The C I emission lasts longer than C II emission, indicating that neutral carbon emission dominated after the spark plasma had expanded and cooled. During the plasma plume expansion, the spectral lines from highly ionized species are observed closer to the target compared with those from lower ionization and neutral species. The spark plasma excitation temperature was evaluated using the

Boltzmann plot method, and the electron density was calculated using Stark broadened profile of

Hα emission line. In comparison to the laser induced carbon plasma, the spark induced carbon plasma has comparatively similar excitation temperature (15,000 to 35,000 K) and electron density

(1.0 × 1017–2.2 × 1017 cm-3). In spark plasma, the excitation temperature was highest at 0.6 mm

161

above the cathode surface. Higher electron density was observed in the regions close to the cathode as well as the anode due to the contributions of both the central ion channel and spark plasma plume. The time dependence of excitation temperature and electron density was consistent with each other. Both the excitation temperature and electron density of the spark plasma increase as the ion channel forms and expands at the early stages of plasma evolution, and decrease as the plasma decays. By investigating the influence of pulse energy on plasma parameters, it was found that both the excitation temperature and electron density increases with pulse energy and saturates at higher energy levels. This study provides detailed information on the dynamics of spark plume generation and expansion, and these results are proven to improve the analytical figures of spark emission spectroscopy and implement SES techniques in instrumentation development.

6.5. References

Aguilera, J. A. and Aragón, C. (2004). Characterization of a laser-induced plasma by spatially resolved spectroscopy of neutral atom and ion emissions.: Comparison of local and spatially integrated measurements. Spectrochim. Acta B At. Spectrosc. 59:1861-1876. Anabitarte, F., Cobo, A., Lopez-Higuera, J. M. (2012). Laser-induced breakdown spectroscopy: fundamentals, applications, and challenges. ISRN Spectrosc. 2012. Aragon, C. and Aguilera, J. (1997). Two-dimensional spatial distribution of the time-integrated emission from laser-produced plasmas in air at atmospheric pressure. Appl. Spectrosc. 51:1632- 1638. Ashkenazy, J., Kipper, R., Caner, M. (1991). Spectroscopic measurement of electron-density of capillary plasma based on stark-broadening of hydrogen lines. Phys. Rev. A 43:5568-5574. Baig, M., Qamar, A., Fareed, M., Anwar-ul-Haq, M., Ali, R. (2012). Spatial diagnostics of the laser induced lithium fluoride plasma. Phys. Plasmas 19:063304. Bye, C. A. and Scheeline, A. (1993). Saha-Boltzmann statistics for determination of electron temperature and density in spark discharges using an Echelle/CCD system. Appl. Spectrosc. 47:2022-2030. Camacho, J. J., Diaz, L., Santos, M., Poyato, J. M. L. (2011). Time-resolved optical emission spectroscopic measurements of He plasma induced by a high-power CO2 pulsed laser. Spectrochim. Acta B At. Spectrosc. 66:57-66. Dittmar, A., Klein, G., Laemmerzahl, U., Lammerzahl, U. Spark emission spectroscopy of 162

anorganic substances in solution, Faru Tec Ges Umwelttechnik Mbh. Diwakar, P., Kulkarni, P., Birch, M. E. (2012). New Approach for Near-Real-Time Measurement of Elemental Composition of Aerosol Using Laser-Induced Breakdown Spectroscopy. Aerosol Sci. Technol. 46:316-332. Diwakar, P. K. and Kulkarni, P. (2012). Measurement of elemental concentration of aerosols using spark emission spectroscopy. J. Anal. At. Spectrom. 27:1101-1109. Dong, M., Mao, X., Gonzalez, J. J., Lu, J., Russo, R. E. (2012). Time-resolved LIBS of atomic and molecular carbon from coal in air, argon and helium. J. Anal. At. Spectrom. 27:2066-2075. El Sherbini, A. M. and Al Aamer, A. A. S. (2012). Measurement of Plasma Parameters in Laser- Induced Breakdown Spectroscopy Using Si-Lines. World J. Nano Sci. Eng. 2:206. Falk, H. and Wintjens, P. (1998). Statistical evaluation of single sparks. Spectrochim. Acta B At. Spectrosc. 53:49-62. Golloch, A. and Siegmund, D. (1997). Sliding spark spectroscopy - rapid survey analysis of flame retardants and other additives in polymers. Fresenius J. Anal. Chem. 358:804-811. Golloch, A. and Wilke, K. (1997). Fast survey analysis of gold alloys by means of a spark emission spectrometer and multivariate calibration. J. Anal. At. Spectrom. 12:1225-1230. Harilal, S. S., Bindhu, C. V., Issac, R. C., Nampoori, V. P. N., Vallabhan, C. P. G. (1997). Electron density and temperature measurements in a laser produced carbon plasma. J. Appl. Phys. 82:2140- 2146. Harilal, S. S., Bindhu, C. V., Nampoori, V. P. N., Vallabhan, C. P. G. (1998). Temporal and spatial behavior of electron density and temperature in a laser-produced plasma from YBa2Cu3O7. Appl. Spectrosc. 52:449-455. Hunter, A. J. R., Morency, J. R., Senior, C. L., Davis, S. J., Fraser, M. E. (2000). Continuous emissions monitoring using spark-induced breakdown spectroscopy. J. Air Waste Manage. 50:111- 117. Kawahara, N., Tomita, E., Takemoto, S., Ikeda, Y. (2009). Fuel concentration measurement of premixed mixture using spark-induced breakdown spectroscopy. Spectrochim. Acta B At. Spectrosc. 64:1085-1092. Kepple, P. and Griem, H. R. (1968). Improved Stark profile calculations for the hydrogen lines H α, H β, H γ, and H δ. Phys. Rev. 173:317. Khalaji, M., Roshanzadeh, B., Mansoori, A., Taefi, N., Tavassoli, S. H. (2012). Continuous dust monitoring and analysis by spark induced breakdown spectroscopy. Opt. Laser Eng. 50:110-113. Le Drogoff, B., Margot, J., Chaker, M., Sabsabi, M., Barthelemy, O., Johnston, T., Laville, S., Vidal, F., Von Kaenel, Y. (2001). Temporal characterization of femtosecond laser pulses induced plasma for spectrochemical analysis of aluminum alloys. Spectrochim. Acta B At. Spectrosc. 56:987-1002.

163

Ramli, M. and Wagatsuma, K. (2010). Observation of Atomic Emission Image from Spark Discharge Plasma by Using Two-dimensional Spectrograph. ISIJ Int. 50:864-867. Salik, M., Hanif, M., Wang, J., Zhang, X. (2013). Plasma properties of nano-second laser ablated iron target in air. Int. J. Phys. Sci. 8:1738-1745. Schmidt, M. S., Sorauf, K. J., Miller, K. E., Sonnenfroh, D., Wainner, R., Bauer, A. J. (2012). Spark-induced breakdown spectroscopy and multivariate analysis applied to the measurement of total carbon in soil. Appl. Opt. 51:B176-B182. Singh, J. P. and Thakur, S. N. (2007). Laser-induced breakdown spectroscopy. Elsevier. Taefi, N., Khalaji, M., Tavassoli, S. (2010). Determination of elemental composition of cement powder by spark induced breakdown spectroscopy. Cement Concr. Res. 40:1114-1119. Walters, J. P. (1972). Formation and growth of a stabilized spark discharge. Appl. Spectrosc. 26:323-. Walters, J. P. (1977). Spark discharge - application to multielement spectrochemical analysis. Science 198:787-797. Walters, J. P. and Goldstein, S. A. (1984). Emission topography of a stable spark discharge train. Spectrochim. Acta B At. Spectrosc. 39:693-728.

164

CHAPTER 7

Conclusions and Recommendations for Future Work

165

This dissertation provides methodologies for real-time aerosol chemical analysis using microplasma spectroscopy. The findings of this work are meaningful to the development of low- cost and hand-portable instruments for aerosol chemical analysis. The main conclusions and significant contributions of this dissertation are summarized in this chapter, and recommendations for future work are also outlined.

In Chapter 2, a corona-based aerosol microconcentration method was developed for efficient concentration of aerosols to a microscopic area of a substrate for subsequent analysis by microscale optical spectroscopy. It was shown that the corona aerosol microconcentrator (CAM) offers the highest spectroscopic sensitivity compared to alternative particle collection methods such as filtration, focused impaction using aerodynamic lens, and spot collection using condensational growth. This method is particularly well-suited for compact, field-portable or personal instrumentation.

The corona-based aerosol microconcentration method involves collection of particles on the collection electrode from a coaxial aerosol flow in a one-step charge-and-collect scheme using corona discharge and electrical precipitation. Simulation of particle transport and deposition in the

CAM was conducted and showed that corona leads to higher flow velocities directed towards the collection electrode, which helps improve particle deposition. The collection efficiency of the

CAM was experimentally examined as a function of particle size, flow rate and electrode size. The collection efficiency was found to be relatively independent of particle size in the submicrometer size range under the operating conditions studied. The collection efficiency increased with electrode diameter and decreased with flow rate. Spark discharge emission spectroscopy was used to evaluate the detection sensitivity of the CAM at different flow rates and electrode diameters. It was found that each electrode diameter has an optimum window of flow rate which provides the

166

highest sensitivity. The highest sensitivity was achieved at a flow rate of approximately 3 L min-1 for the 500 µm electrode. The results provided some guidance on how to design and optimize a corona-based aerosol microconcentrator that could be coupled with microscale spectroscopy for aerosol chemical analysis. Since this work mainly deals with coupling CAM with microplasma spectroscopy (i.e., LIBS, SES and GD-OES) for elemental analysis of aerosol, future work should be aimed at coupling the CAM with other microscale spectroscopic methods, such as micro-Raman and near infrared spectroscopy, to achieve molecular and structure analysis of aerosol.

In Chapter 3, real-time measurement of carbonaceous aerosols was performed using the corona-based aerosol microconcentration method and spark emission spectroscopy. The system was calibrated for total atomic carbon (TAC) using a carbon emission line at 247.856 nm (C I) for various carbonaceous materials including sucrose, EDTA, caffeine, sodium carbonate, carbon black, and carbon nanotubes. It was shown that a single calibration curve could be obtained for all the organic and inorganic carbonaceous materials and the LOD for TAC was 1.61 ng, equivalent to 238 ng m-3 when sampling at 1.5 L min-1 for 5 min. Measurements at elevated electrode temperatures of up to 300 oC could reduce the contribution of organic carbon to the TAC, and thereby improved the selectivity to elemental carbon. Using the calibration curve, SES measurements of unknown CNT aerosol samples at elevated temperature compared well with the measurements from thermal optical method. The SES method showed advantages of low detection limits and high time resolution. The SES method can be an excellent candidate for development as a low-cost and hand-portable instrument for monitoring airborne carbon nanomaterials.

In Chapter 4, simultaneous measurement of multiple elements in aerosols was performed using the SES method and a multivariate calibration approach. Partial least squares (PLS) regression is used to construct the relationship between spectra and mass of analyses. Results show

167

that selection of the relevant variables as model inputs could improve the prediction ability of PLS models. The study in Chapter 4 demonstrates that PLS regression was a useful tool for measuring multiple elements in complex aerosol samples. Here, experimental designed aerosol samples were used for building calibration models. Future work could be focused on developing multivariate calibration models for various specific measurements using field samples that are sufficiently representative of all the future samples.

A low-cost semi-continuous aerosol analysis method based on GD-OES was developed in

Chapter 5. The corona aerosol concentrator designed in Chapter 1 was used for collection of particles for subsequent analysis by GD-OES. The signal intensity from the analyte was found to be decreased with glow discharge time, indicating that the collected particles are gradually ablated by the glow discharge. The cumulative signal intensity was proportional to the mass of analyte in the particulate sample. This method provides good detection limits for metals and good measurement reproducibility. Although the general analytical performance of the GD-OES method is not as good as LIBS or SES method, the glow discharge can be an excellent low-cost excitation source for development of low-cost hand-held sensors for aerosol measurement.

The physics and fundamental aspects of glow discharge and spark discharge plasma were investigated in Chapters 5 and 6. The plasma parameters, such as temperature and electron density were evaluated using spectroscopic means. The glow discharge plasma has a gas temperature in the range of 378 – 1438 K and electron density of 2 – 5 × 1014 cm-3, whereas the spark plasma has an excitation temperature in the range of 15,000 – 35,000 K and the electron density in the range of (1.0– 2.2) × 1017 cm-3. Spatial analysis of the glow discharge spectral feature showed that the collision and excitation of the collected particles occurred in the region near the collection electrode. The excited species by spark discharge appeared close to the cathode surface at an early

168

stage of the spark plasma, and then expanded toward the anode. The emissions became more intense with the evolution of the spark plasma, reached a maximum and then decreased rapidly with time. Both the plasma imaging and spectroscopic analysis showed that the life time of spark plasma was approximately 15 µs. The studies in Chapters 5 and 6 provided results on diagnostics of microplasma, which have significance on developing analytical instrumentations utilizing microplasma spectroscopy.

This dissertation provides methodologies for real-time aerosol chemical analysis using microplasma spectroscopy. All the results of this work were obtained from bench-top experimental setup. Field-portable instrumentations for aerosol chemical measurement could be implemented by integrating all the key instruments used in the experiments. Further, low-cost and compact aerosol analysis instruments or sensors could be possibly designed by miniaturizing each component of the experimental setup. In the future, field studies on real-time monitoring workers’ exposure to particulate metals can be conducted using the developed hand-portable instruments.

The aerosol chemical analysis methods proposed in this dissertation allows determination of multiple elements with excellent detection limits. In future work, size-resolved elemental analysis of aerosols should be explored by adding a differential mobility analyzer at the inlet of the aerosol chemical analysis instrument.

169

APPENDIX Supplemental Information for Chapter 2

Figure S-1. Corona current as a function of voltage applied between two electrodes for different electrode diameters (The distance between two electrodes is 5 mm).

170

Figure S-2. The estimate particle mass error for various count median diameter and geometric standard deviation by assuming a constant collection efficiency.

171