<%BANNER%>

Investigation of Physical and Spectral Characteristics of Laser-Induced Plasmas: Applications to Laser-Induced Breakdown...


PAGE 1

INVESTIGATION OF PHYSICAL AND SPECTRAL CHARACTERISTICS OF LASER-INDUCED PLASMAS: APPLICATIONS TO LASER-INDUCED BREAKDOWN SPECTROSCOPY FOR ANALYSIS OF AEROSOLS AND SINGLE PARTICLES By VINCENT PAUL HOHREITER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Vincent Paul Hohreiter

PAGE 3

ACKNOWLEDGMENTS I would like to briefly and generally acknowledge everyonefamily, friends, labmates, colleagueswho helped me make this dissertation happen. Specifically I would like to thank my parents, Vaughn and Christine Hohreiter, whobeyond their constant love and supportalways accepted, fostered, and encouraged my academic progress and misgivings alike. Additionally I owe a debt of gratitude to my advisor, Dr. David Hahn, without whose rigorous academic lead and unwavering personal support I most likely would have abandoned doctoral study years ago. Finally, I would like to thank my wife, Dr. Aline Gubrium, who has inspired me with both love and a tenaciously organized and diligent work ethic. iii

PAGE 4

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iii LIST OF TABLES ............................................................................................................vii LIST OF FIGURES .........................................................................................................viii ABSTRACT .......................................................................................................................xi CHAPTER 1 INTRODUCTION........................................................................................................1 The Laser-Induced Plasma...........................................................................................1 Plasma Formation..................................................................................................2 Spatial and Temporal Evolution............................................................................3 The Laser-Induced Breakdown Spectroscopy (LIBS) Technique................................4 Temporal Properties of Laser-Induced Plasmas for LIBS Analysis.............................6 Stability of Laser-Induced Plasmas..............................................................................8 LIBS and Calibration..................................................................................................10 Spatial Considerations in Plasma Emission................................................................13 2 EXPERIMENTAL METHODS.................................................................................16 The LIBS Experimental Apparatus............................................................................16 Delivery Optics....................................................................................................16 Extraction Optics.................................................................................................17 System Synthesis.................................................................................................18 Plasma Transmission Study........................................................................................19 Transmission Measurements...............................................................................20 Electron Density and Plasma Continuum Measurements...................................21 Laser Shot Stability Study..........................................................................................22 Carbon Calibration Study...........................................................................................24 Solid Phase Carbon Experiments........................................................................25 Gas Phase Carbon Experiments...........................................................................27 LIBS Data Collection..........................................................................................28 Plasma-Aerosol Interaction Study..............................................................................29 The Imaging System............................................................................................30 The Spectroscopy System....................................................................................32 iv

PAGE 5

3 PLASMA TRANSMISSION PROPERTIES.............................................................34 Time-Resolved Plasma Transmission.........................................................................34 Time-Resolved Continuum Measurements................................................................36 Electron Density Measurements.................................................................................38 Plasma Absorption Cross-Section..............................................................................40 Modeling Plasma Transmission..........................................................................40 Plasma Absorption...............................................................................................42 Extrapolation for Small Time Scales...................................................................43 4 LIBS PLASMA STABILITY.....................................................................................46 Effects of Laser Cavity Seeding.................................................................................46 Effects on Shot-to-Shot Precision.......................................................................48 Temporal Effects on the Plasma Formation Process...........................................52 Effects of Aerosol Loading.........................................................................................56 Effects on Shot-to-Shot Precision.......................................................................57 Temporal Effects on the Plasma Formation Process...........................................59 5 LIBS CALIBRATION................................................................................................61 Raw and Baseline-Subtracted Spectra........................................................................61 Calibration Using Atomic Spectra..............................................................................63 Calibration Using Molecular Bands...........................................................................65 Analysis of Matrix Effects..........................................................................................67 Aerosol Loading and Size Considerations..................................................................69 Proposed Mechanism for Particle-Plasma Interaction................................................70 6 SPATIAL CONSIDERATIONS IN PLASMA EMISSION......................................74 Background Emission.................................................................................................74 Spatial Distribution of Atomic Emission....................................................................78 Correlation of Image and Spectral Data.....................................................................81 Rate of Aerosol Dissociation......................................................................................84 Implications for LIBS.................................................................................................90 7 CONCLUSIONS AND FUTURE WORK.................................................................91 Conclusions.................................................................................................................91 Proposal of Future Work............................................................................................92 APPENDIX A SUMMARY OF ATOMIC AND MOLECULAR LINES USED.............................94 B VERIFICATION OF IMAGING SYSTEM...............................................................95 v

PAGE 6

C TABLE OF 1/2 VALUES FOR FREE-ELECTRON DENSITY CALCULATIONS......................................................................................................97 D IMAGES OF NON-HOMOGENEOUS SPATIAL EMISSION................................98 LIST OF REFERENCES.................................................................................................102 BIOGRAPHICAL SKETCH...........................................................................................105 vi

PAGE 7

LIST OF TABLES Table page 4-1 Ratios of the area-integrated spectral peaks (peak-to-base and peak-to-peak) for seeded and unseeded laser operation........................................................................48 5-1 Summary of the carbon calibration response and plasma conditions......................65 A-1 Summary of atomic lines used in spectral analysis..................................................94 A-2 Summary of molecular lines used in spectral analysis.............................................94 C-1 Values of 1/2 for calculation of electron density.....................................................97 vii

PAGE 8

LIST OF FIGURES Figure page 2-1 Optical setup for delivery of laser energy to the sample volume resulting in the creation of a plasma..................................................................................................16 2-2 Optical setup for extraction of plasma emission from sample volume....................17 2-3 Basic LIBS system...................................................................................................18 2-4 LIBS system incorporating second laser for plasma transmission measurements...19 2-5 LIBS system incorporating fast detectors and a digital oscilloscope for additional beam measurements................................................................................22 2-6 Nebulization system coupled with a LIBS system (only Nd:YAG shown) for aerosol analysis........................................................................................................24 2-7 Intensified CCD for imaging combined with a LIBS system for sampling aerosols within the laser-induced plasma.................................................................29 2-8 Plasma image as visualized from a co-axial orientation to the laser beam..............30 3-1 Plasma transmission as a function of time following plasma initiation using 532 and 1064 nm probe laser beams...............................................................................35 3-2 Plasma continuum emission (20 nm bandwidth) for three spectral regions centered at 270, 400 and 530 nm..............................................................................37 3-3 Free electron number density as a function of time following plasma initiation, as measured using the Stark broadened H line width.............................................39 3-4 Measured absorption coefficient for the 532 nm probe laser beam as a function of the corresponding measured free electron density...............................................42 3-5 Predicted electron density values as a function of time following plasma initiation...................................................................................................................44 4-1 Sample spectrum showing atomic emission lines from the plasma spark recorded in ambient air............................................................................................................47 viii

PAGE 9

4-2 N 491.5 /N 493.5 and N 491.5 /C 247.9 peak-to-peak atomic emission ratios for an 800 shot sequence............................................................................................................49 4-3 N 491.5 and C 247.9 peak-to-base atomic emission ratios for an 800 shot sequence....52 4-4 Representative waveforms for the incident laser pulse and the transmitted laser pulse following laser-induced breakdown................................................................53 4-5 Cumulative probability distribution plot for both seeded and unseeded laser operation...................................................................................................................56 4-6 N 491.5 /N 493.5 peak-to-peak atomic emission ratios for an 800 shot sequence...........58 5-1 LIBS spectra in the vicinity of the 247.86 nm carbon atomic emission line for different sources of carbon.......................................................................................62 5-2 Calibration curves based on the 247.86 nm (C I) atomic emission line for the five carbon analyte sources investigated..................................................................64 5-3 LIBS spectra in the vicinity of the CN violet system for the AA-ICP and CO 2 analyte sources.........................................................................................................66 5-4 Schematic detailing the proposed interactions between the rapidly expanding plasma wave and the analyte species.......................................................................71 6-1 Two-dimensional on-axis projection of the laser-induced plasma in an aerosol-free environment......................................................................................................75 6-2 Change in observed diameter of the plasma image as a function of time delay from the end of the laser pulse.................................................................................76 6-3 Center-averaged plasma intensity, normalized to the highest measured value at 2s, as a function of time delay from the end of the laser pulse..............................77 6-4 A spatially localized emission burst from an aerosol in the upper left-hand corner of the plasma image......................................................................................79 6-5 Average diameter (N=30) of emission bursts as a function of plasma delay time...81 6-6 Images (a)-(e) demonstrate the observed growth of emission bursts for time delays of 2, 4, 8, 15, 22, and 30 s respectively.......................................................82 6-7 Three reference spectra taken at 4, 8, and 30us respectively...................................83 6-8 Comparison of measured P/B values for image data and spectral data...................84 6-9 Mass of calcium analyte detected using LIBS on a dilute aerosol matrix of borosilicate glass microspheres................................................................................87 ix

PAGE 10

B-1 Simple grid patterns for optical system evaluation..................................................95 B-2 Image of 1 mm test grid...........................................................................................96 B-3 Image of 2 mm test grid...........................................................................................96 D-1 A strong, central burst is accompanied by a less intense burst in the upper-left corner........................................................................................................................98 D-2 An apparent thermal emission streak.......................................................................99 D-3 A double burst grouped in the central-upper left portion of a relatively weak plasma.....................................................................................................................100 D-4 Two emission bursts beyond the outer edges of a strong background plasma......100 D-5 Four bursts, two very strong, and two fairly weak, within a relatively weak plasma.....................................................................................................................101 x

PAGE 11

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INVESTIGATION OF PHYSICAL AND SPECTRAL CHARACTERISTICS OF LASER-INDUCED PLASMAS: APPLICATIONS TO LASER-INDUCED BREAKDOWN SPECTROSCOPY FOR ANALYSIS OF AEROSOLS AND SINGLE PARTICLES By Vincent Paul Hohreiter December 2005 Chair: David W. Hahn Major Department: Mechanical and Aerospace Engineering The physical and optical characteristics of laser-induced plasmas (LIP) and the precision of measurements using laser-induced breakdown spectroscopy (LIBS) for aerosol and gas phase species are explored. In a series of coherent experiments properties of the LIP and the temporal, spatial, and energetic variations of interactions with entrained particles (aerosols) are evaluated from both a standpoint of basic plasma science and applied atomic spectroscopy. First, the evolution of a LIP is characterized in terms of its temporally-resolved spectral absorptivity, spectral emissivity, and free electron density during the first 500 nanoseconds. Transmission measurements reveal near opacity of the LIP at early times (10-50 ns) and essential transparency at longer times (500 ns). The fundamental change from an absorbing plasma to a non-absorbing plasma during this period is important with respect to radiative energy transfer. xi

PAGE 12

Second, spectral and temporal effects of laser cavity seeding and aerosol presence in the LIP volume are investigated. Improvements in the temporal stability of laser-induced breakdown initiation were observed with laser cavity seeding. Greater shot-to-shot analyte precision, as measured by a nearly 60% reduction in relative standard deviation, was realized in the absence of concomitant aerosols from the analyte sample stream. Third, the effects of analyte phase on the calibration response for LIBS were investigated. Significant differences in the atomic emission signal from carbon were observed when comparing calibration streams of gas-phase and submicron-sized solid-phase species. The resulting calibration curves demonstrated large inter-species variations over a comparable range of atomic carbon concentrations, challenging a widely held assumption that dissociation of constituent species within a LIP results in independence of the analyte atomic emission signal from analyte source. Finally, the LIBS technique is used in conjunction with imaging to measure a rate of dissociation for aerosols entrained in a LIP. Additionally, the time scales of background plasma emission, spatial distribution of atomic emission, and diffusion of atomic species within a LIP are explored. Atomic emission from the plasma volume is demonstrated to occur in spatially localized bursts. The total amount of analyte detected is shown to plateau at times coinciding with significant energy decay of the plasma. xii

PAGE 13

CHAPTER 1 INTRODUCTION The Laser-Induced Plasma The laser-induced plasma is an almost completely ionized gas produced by focused laser irradiance such that the electric field generated at the focal point exceeds the dielctric strength of the medium occupying the focal volume (Weyl 1989). The plasma, also known in the literature as a spark or breakdown, is qualitatively characterized by directionally uniform emission of radiation and is often accompanied by an audible shock wave. The luminous emission, some fraction often falling in the visible portion of the electromagnetic spectrum, is valuable as it corresponds directly and uniquely to the atomic constituents of the sample medium present within the plasma volume. Laser-induced plasmas may be formed on the surface of solids and inside liquid or gaseous media; however, the studies presented herein deal solely with plasmas in a gas phase medium and, hence, further discussion will be limited to these. A region of large energy density corresponding to the laser focal spot is required to create a plasma in a gaseous medium. This generally requires a pulsed laser source and produces temperatures that range from several thousand to tens of thousands of Kelvin in the immediate vacinity of the laser focal volume. Physically, the plasma is created as laser energy breaks all molecular and atomic bonds and strips electrons from atomic nuclei forming a cloud of ions and free electrons. The ions and electrons rapidly recombine to re-form neutral atoms and, in the process, emit photons in narrow, species-specific wavelength bands. Since each known atomic or molecular species contains 1

PAGE 14

2 unique electron configurations and energy transitions, this photonic emission serves as a signature and can be captured and analyzed as such. The following discussion details the various processes which characterize laser-induced plasma formation and evolution. Plasma Formation Laser energy couples into the sample medium through two primary mechanisms, cascade ionization and multi-photon ionization, which explain the plasma formation process. In the cascade ionization process initial free electrons absorb laser energy and collide with neutral atoms and molecules to release more electrons. Writing out the reaction, cascade ionization is described by e + M 2e + M + 1-1 where M represents an atom or molecule. Since two electrons are produced for every one used, the electron population increases exponentially in time. This process is therefore also referred to as avalanche ionization. The second mechanism of plasma formation, multi-photon ionization, is characterized by direct absorption of photons by atoms and molecules and is described by M + n(h) e + M + 1-2 This process does not require an initial population of free electrons and does not tend to cascade as multiple photons are required for the release of one electron. The two ionization processes require large energy densities to proceed, occur simultaneously, and contribute in varying proportions to plasma growth. Since cascade ionization requires an initial population of free electrons, breakdown initiation is thought to generally occur via multi-photon ionization. However, the entire breakdown event is known to be dominated by multi-photon ionization for short wavelengths ( < 1 m) and low pressures (p < 10

PAGE 15

3 Torr) and dominated by cascade ionization for long wavelengths ( > 1 m) and high pressures (p > 100 Torr) (Weyl 1989). Spatial and Temporal Evolution While the laser-induced plasma is highly dependent on many physical parameters such as laser power, laser wavelength, analyte source and concentration, and system optics, the spatial and temporal evolution of most plasmas shows definite trends. Within the first few nanoseconds after the laser pulse reaches the focal spot, the plasma begins a rapid expansion, forms a highly energized cloud of ions and electrons, and is sufficiently hot that radiant emission falls principly in the ultraviolet-visible range (treating the early plasma as a blackbody at 40,000 K, which is within the temperature range reached by many laser induced plasmas, 97% of the emission falls below 400 nmfor less energetic plasmas, the value drops to 86% at 20,000 K and 48% at 10,000 K). This initial period of rapid spatial growth lasts on the order of hundreds of nanoseconds, thereafter the plasma continues to grow for some microseconds, but at a reduced rate. In addition, the early plasma emissionas expectedhas a strong broadband response which tends to dominate the various ionic transitions which are present at small times. After the initial growth phase the plasma begins to decay. It should be noted that growth, in this sense, refers to an increase in volume of the region from which luminous emission emanatesfrom the standpoint of absolute temperature and electron density, the plasma is never more energetic than at the tail end of the initiating laser pulse. At this point electrons and ions recombine yielding neutral atoms and their corresponding spectral lines. The fraction of neutral species present in the focal volume increases steadily during this period, which usually lasts tens of microseconds. The plasma cools significantly from radiative transfer and quenching processes and emission in the visible

PAGE 16

4 spectrum increases, contiuum emission decreases, and distinct spectral lines become prominent. This period, from approximately 1-30 s following the laser pulse, is therefore ideal for capturing and analyzing atomic spectra. From 30 s onward, plasma energy further decays, the plasma volume continues to cool, and atoms begin to combine to form molecules. Molecular spectral bands appear along with fading atomic lines. The plasma continues to emit out to hundreds of microseconds following the initial laser pulse, but the useful signal is increasingly diminished with time. The plasma event ends when no more emission is detectable and the plasma volume is once again occupied by neutral constituents at or near ambient conditions. As a final note on the process of the laser-induced plasma event, the above discussion of clearly demarcated phases of plasma behavior represents a simplification of the phenomena. The occurrence of continuum, ionic, atomic, and molecular emission may well, for any given analyte, overlap significantly depending on the various states and transitions of that analyte. However, it is extremely important and helpful to recognize the nature of temporal variations in plasmassuch recognition plays heavily into plasma spectroscopy applications and is discussed thoroughly in the current work. The Laser-Induced Breakdown Spectroscopy (LIBS) Technique The laser-induced plasma is the fundamental building block for the LIBS technique, the basis of which rests in the ability to accurately extract meaningful spectral data from radiative plasma emission. For LIBS the plasma acts as both the excitation source and the sample volume, and dissociates entrained aerosols (smaller than about two microns as discussed in later sections), molecules, and atomic species resulting in

PAGE 17

5 radiative emission. Spectrally resolving this emission in a useful way is the goal of most LIBS studies. First employed in the 1960s, LIBS has become increasingly popular for qualitative and quantitative spectral analysis because of an inherent advantage over other techniques: very little, if any sample preparation is needed for species under investigation, making in situ measurements possible and straightforward. Numerous literature reviews of the LIBS technique trace its evolution as a means of spectral analysis (Radziemski 1994, Schechter 1997, Smith et al. 2001). Not without its challenges, LIBS requires a significant photon flux to initiate a plasma, especially in a gaseous medium. For a spark to be induced in clean (particulate-free) air, threshold laser fluence has been reported to range from 10-1600 GW/cm 2 with the large variations due to differences in experimental setup (Simeonsson and Mizeolek 1994, Smith et al. 2001). Such large energy densities require powerful pulsed lasers, often large in size and cost. Creating the plasma, however, is just the beginninglaser-grade optics, a diffraction grating, and a collection device (the latter two usually conveniently housed in a modern spectrometer) are required to extract a useful signal from a small plasma volume, which has been shown via absorption measurements to be approximately 1 mm 3 for nominal conditions (Carranza and Hahn 2002c). All studies in this dissertation focus solely on LIBS in a gas phase medium. Gas phase LIBS is a spectral method for characterizing analyte species either in the gaseous state or in aerosol form. An aerosol is a solid or liquid particle suspended in a gas-phase matrix. Aerosols are an ever-present, naturally-occuring component of our atmosphere but can also be easily fabricated with pressurized jets and nebulizers. Much is unknown

PAGE 18

6 about the behavior of ambient aerosol populations; as they represent a constantly changing system in which real-time, in situ analysis does not readily lend itself to conventional means of laboratory study. LIBS, with the ability to dissociate raw samples, presents a technique well-suited to studying aerosols and has been addressed in many studies as a means of aerosol species, concentration, and size detection (Radziemski et al. 1983, Essien et al. 1988, Singh et al. 1997, Neuhauser et al. 1999, Sneddon and Lee 1999, Hahn and Lunden 2000). Temporal Properties of Laser-Induced Plasmas for LIBS Analysis It is known that the LIBS technique uses a highly energetic plasma to vaporize and dissociate matter within the laser-induced plasma, including entrained particles. However, less is known, both theoretically and experimentally, regarding the interactions between the laser-induced plasma and particles. In a recent study, research efforts were focused to quantify these plasma-particle interactions pursuant to quantitative LIBS-based analysis (Carranza and Hahn 2002c). It was found that individual silica particles up to 2 m in diameter were completely vaporized in the laser-induced plasma and yielded a linear analyte response, while larger particles revealed a non-linear response with respect to analyte atomic emission presumably due to incomplete vaporization. While the plasma-particle interaction is not well understood, it appears that the transient nature of the plasma in the early stage of plasma evolution plays an important role in the vaporization process. Hence, rate limitations may be more important than simple plasma and particle heat capacities. Accordingly, as a first step toward understanding the plasma-particle processes, it is important to characterize laser-induced plasmas in the early stages of evolution. To further advance this understanding, this study investigates

PAGE 19

7 the plasma evolution via time-resolved measurements of electron number density, plasma absorptivity, and plasma continuum emission. Laser-induced plasmas have been studied using parameters such as electron temperature, electron density, and emission intensity. Parigger et al. (1995) studied the decay of plasmas produced in hydrogen by a Nd:YAG laser (1064nm) using the hydrogen Balmer series emission lines. They found that at early times, comparable with the plasma-initating laser pulse, electron densities were on the order of 10 19 cm -3 while in the first hundreds of nanoseconds following laser breakdown, the free electron density was about 10 18 cm -3 and the plasma temperature was about 80,000 K (calculated using Boltzmann plots). In a more recent study, Parigger et al. (2003) further explored Stark-broadened emission profiles in gaseous hydrogen plasmas, comparing both H and H measurements in the context of recent treatments of ion dynamics as reported by Oks (2000). For electron densities in excess of 10 17 cm -3 corresponding to delay times less than 500 ns, agreement between inferred electron density values from H and H profiles was very good. However, at longer delay times some discrepancies were noted when using the standard theoretical treatment per Griem (1974, 1997), while good agreement between the H and H results was achieved with the updated treatment of Stark broadening for electron densities approaching 10 16 cm -3 Borghese and Merola (1998) measured a plasma kernel of about 0.02 mm 3 as produced by a fundamental Nd:YAG laser in air during the stage of maximum emission (about 20 ns following plasma initiation). They calculated a corresponding plasma temperature of about 35,000 K at this time using an experimental Planck function, but also calculated higher plasma temperature at earlier times (e.g., 10 5 K at about 10 ns). Using techniques such as

PAGE 20

8 shadowgraphy and interferometry, Villagran-Muniz et al. (2001) found that the plasma electron density was about 10 18 cm -3 for a pulse energy of 300 mJ, and additionally calculated temperatures behind the shock wave (due to the optical breakdown) of about 10 5 K during the first tens of nanoseconds, when the shock wave is not detached from the plasma. Work by Yalcin et al. (1999) investigated plasma temperature and free electron density, including spatially resolved measurements using Abel-inversion, for delay times as early as 350 ns following plasma initiation. Using Stark-broadened H line widths, they reported peak electron densities of 1.2x10 18 cm -3 and reported little variation (~10%) in this value across the plasma using spatially resolved measurements. All available research shows that the plasma, in its early stage, is a highly energetic system that is characterized by rapid changes, and is not necessarily in thermodynamic equilibrium. For LIBS-based aerosol analysis, these changes need to be assessed in order to identify parameters that could play important roles in the interactions between the plasma and concomitant particles, including free electron densities and plasma absorptivity. In the following study, the evolution of the laser-induced plasma is assessed based on temporally resolved spectral absorptivity and emissivity measurements, along with evaluation of the free electron density via Stark broadening. Stability of Laser-Induced Plasmas Historically, laser-induced breakdown spectroscopy (LIBS) has been applied on a statistical basisspectra from thousands of laser shots are ensemble averaged and results extracted (Radziemski 1994). In recent studies, however, LIBS has been applied using single-shot analysis based on conditional spectral processing to effectively sample ambient aerosol populations by discrete particle analysis (Hahn and Lunden 2000,

PAGE 21

9 Carranza et al. 2001). Because gains in signal-to-noise realized with ensemble averaging are not applicable with single-shot analysis, it is important to maximize the precision on a shot-to-shot basis (Carranza and Hahn 2002b). For example, recent interest in extending the LIBS technique to quantitative analysis of bioaerosols (Boyain-Goitia et al. 2003, Hybl et al. 2003, Morel et al. 2003, Samuals et al. 2003), which entails high precision to distinguish between often-similar elemental signatures, places additional requirements on analyte sensitivity and single-shot precision. Hence, to further establish single-shot LIBS-based techniques as an integrated analytic tool for detection and quantitative analysis of aerosols and other dilute species, the precision must be verified and maximized on a shot-to-shot basis. Therefore, the stability of the plasma-producing laser pulse, the spatial and temporal stability of the plasma itself, and reproducibility of the corresponding spectral emission are all of primary concern. A laser-induced plasma is largely a product of two factorsthe laser irradiance, including both spatial and temporal characteristics, and the sample composition, including homogeneity, in the vicinity of the laser-induced breakdown and subsequent plasma formation. Improving the temporal precision of the plasma formation process and precision of the ensuing plasma spectral emission, thereby improving the precision of the LIBS analyte signal, requires attention to these two factors. In the current work, seeding the laser cavity to improve spatial and temporal quality of the laser beam is explored in concert with the effects of aerosol loading in the laser-induced plasma sample volume. Several recent studies report on the time and spatially dependent absorption of incident radiation by laser-induced plasmas (Aguilera et al. 2003, Bindhu et al. 2003, Bindhu et al. 2004), thereby emphasizing the strong coupling between the plasma

PAGE 22

10 creating laser beam and the consequent laser-induced plasma. The full nature of this laser-plasma coupling is not fully understood and is a focus of on-going work. Hence, an additional goal of the current work is investigating the temporal variation of plasma inception in terms of the effects of aerosol loading and laser cavity seeding. LIBS and Calibration Since the early 1980s, the use of laser-induced breakdown spectroscopy (LIBS) has steadily increased as a quantitative analytical technique, with applications to a wide range of solids, liquids, gases, and aerosols (Rusak et al. 1997, Sneddon and Lee 1999, Lee et al. 2004). In recent studies LIBS has been used in conjunction with single-shot analysis to effectively sample and analyze aerosol populations using discrete particle analysis (Hahn 1998, Hahn and Lunden 2000, Carranza et al. 2003). In other applications, LIBS-based sensing has been successfully implemented for continuous on-line monitoring of air emissions and aerosols (Neuhauser et al. 1999, Zhang et al. 1999, Nunez et al. 2000, Ferioli et al. 2003), and for analysis of ambient air particulate matter (Hahn 1998, Carranza et al. 2001, Lithgow et al. 2004). More recent studies have addressed the feasibility of LIBS for analysis of biological materials and bioaerosols (Boyain-Goitia et al. 2003, Hybl et al. 2003, Morel et al. 2003, Samuals et al. 2003). Important research issues regarding LIBS-based analysis of gaseous sample streams, notably aerosol detection and single particle analysis, include the overall analyte sensitivity, precision, sampling methodology, and calibration schemes. Calibration has been an important issue since Radziemski et al. (1983) first demonstrated the viability of LIBS for detection of atomic species from aerosol samples. In their landmark study, beryllium-rich aerosols were generated by either laser ablation or produced by a nebulizer/heat-chamber system, with the latter used for generation of

PAGE 23

11 calibration curves and subsequent calculation of detection limits. A different nebulizer was used in a continuing study, which produced aerosols estimated to be in the submicron size range (Essien et al. 1988). Calibration curves were generated for three analytes, namely cadmium, lead, and zinc, and were characterized by initial linearity followed by various degrees of saturation at higher concentrations. The saturation effects were attributed to incomplete vaporization of particles. An important finding was the general agreement (within 10%) of lead atomic emission signals of comparable atomic lead concentrations when nebulizing either lead acetate, lead chloride, or lead nitrate. Cadmium revealed a 27% difference in analyte response when comparing nebulized solutions of cadmium nitrate and cadmium chloride. In experiments analogous to the aerosol studies, the relative independence of analyte signals on molecular source for purely gas-phase species was reported in several studies (Dudragne et al. 1998, Tran et al. 2001). Specifically, Dudragne et al. demonstrated that analyte signals for fluorine, chlorine, sulfur and carbon scaled with the number of respective atoms in the constituent molecules for a wide range of compounds, concluding that the parent molecules were fully dissociated in the laser-induced plasma. Tran et al. verified that SF 6 and HF yielded essentially identical fluorine atomic emission signals when the gas composition was adjusted to contain the same atomic fluorine mole fraction. The above comments support a widely used assumption within the LIBS community, namely that complete dissociation of constituent species within the highly energetic laser-induced plasma results in independence of the analyte atomic emission signal on the analyte source. This statement of independence of analyte source must be clearly distinguished from the presence of plasma matrix effects. Plasma matrix effects

PAGE 24

12 are considered changes in a specific analyte emission response due to interactions with or perturbations of the plasma, such as changes in plasma temperature, quenching of atomic emission lines by other species, or loss of an emitting species due to recombination with other atoms. For example, Gleason and Hahn (2001) observed temporal-dependent changes in mercury atomic emission as the overall gas composition of the laser-induced plasma was varied. The distinction between analyte source effects and plasma matrix effects is significant, because when one considers the latter, there is generally an accepted temporal starting point where the atomic concentrations are considered to reflect a matrix-independent baseline. Such a constant temporal starting point was observed with mercury emission as the gas matrix was varied in the study by Gleason and Hahn, noting that the analyte source was held constant. Notwithstanding the apparent de facto acceptance of analyte signal independence on analyte source for laser-induced plasma spectroscopy, to date no studies have systematically examined this issue for LIBS-based analysis of gaseous and aerosol-laden streams. As the LIBS-based analysis of increasingly complex gas-phase and aerosol systems becomes more prevalent, high accuracy and precision become increasingly important. Therefore, precise calibration of the analyte signal response is required, as the number of factors contributing to the emission signal (i.e., laser parameters, optical configuration, spectrometer/detector, and plasma matrix) is sufficiently large to render an a priori calculation of analyte signal response impractical. LIBS calibration schemes are made more difficult when the systems of interest include complex and unknown analyte compositions (i.e., mixtures of gas-phase, solid-phase, and aerosol analyte species), which may exist for ambient air or combustion, exhaust streams. For example, carbon in

PAGE 25

13 ambient air is present as carbon dioxide at several hundred parts-per-million (ppm), but is generally present as fine particulate matter (such as soot), and within biological particles such as pollens and bacterial spores. Similarly, metal species at moderate to high temperatures may exist as vapor-phase species, and also as homogeneously or heterogeneously nucleated particulates. Given the inherent assumption of complete dissociation of analyte species, most calibration schemes to date have focused on a precise and accurate means of analyte introduction, not on the resulting physical state of the analyte. As described above, the use of purely gaseous analyte species and the nebulization of aqueous solutions are the most widely used calibration schemes with LIBS, noting that the former produces a homogenous-phase (i.e. all gaseous) source for laser-induced breakdown, while the latter produces a well-dispersed, high number density mixture of generally micron to submicron-sized aerosols in the gas matrix. The present study examines the phenomena of analyte dissociation and analyte emission for different sources (both gas phase and solid phase) of atomic carbon and the resulting effects on analyte calibration response. Spatial Considerations in Plasma Emission The continued evolution of LIBS as an effective technique for the analysis of aerosol species is dependent on its ability to accurately quantify the presence and concentration of analyte species within individual particles and small groups of aerosols. Primary challenging factors, including (a) that small aerosols may contain only a few femtograms of analyte and (b) that the LIBS technique is known to experience large shot-to-shot fluctuation in signal output, require that more be known about the physics of the plasma-particle interaction on a single-shot basis. Furthermore, identification and

PAGE 26

14 minimization of sources of experimental uncertainty are critical to more accurate detection. While bulk plasma properties and spatially averaged particle-laser interactions can be studied with conventional techniques, methods for measuring laser-induced plasma behavior with respect to discrete aerosols remain limited. Particle position, size, and number density within the plasma volume may have distinct effects on spectral emission. Analysis of these properties for single particles and known discrete populations must be performed in order to examine such connections. While significant progress has been achieved regarding the development of the LIBS technique for use on single particles and aerosols, many fundamental issues remain. Of particular interest is the physical process and time-scale of particle dissociation within a laser-induced plasma and subsequent spatial diffusion of analyte species throughout the plasma. As suggested by recent models of laser-induced plasma behavior, species dissociation and spectral analyte emission within a laser-induced plasma are commonly assumed to occur in a temporally instantaneous and spatially homogeneous way (Ho et al. 1996, Itina et al. 2003,Gornushkin et al. 2004). Although it is recognized that these assumptions are made primarily in order to reduce the mathematical complexity of such models, non-homogeneities in both space and time are for the most part neglected in experimental analysis of laser-induced plasmas as well. An analysis that accounts for spatial and temporal variability considerations could shed greater light on the presence of typically large shot-to-shot fluctuations often present in the data of the vast array of LIBS studies that seek to quantify analyte spectral emission.

PAGE 27

15 Recent studies have proposed and demonstrated that one such spatial parameterthe relative position of a sampled aerosol within the total plasma volumemay play a role among other significant factors in the ability to detect spectral emission as well as the accuracy of detection using LIBS (Lithgow 2005). The current study, using a combination of imaging and spectroscopy, addresses the issue of spatially non-homogeneous atomic emission originating from an aerosol particle within a laser-induced plasma and quantifies a diffusion time scale for both physical aerosol dissociation and analyte-specific emission.

PAGE 28

CHAPTER 2 EXPERIMENTAL METHODS The setup of the LIBS analytical apparatus can be broken down into two sets of optical instruments with two basic functions; delivery of laser irradiance to the sample and extraction of luminous emission from the plasma volume. The LIBS Experimental Apparatus Delivery Optics A schematic of equipment used for delivery of laser energy to the sample is shown below: Condensing Lens Aperature Gallilean Telescope Plasma Nd:YAG Figure 2-1. Optical setup for delivery of laser energy to the sample volume resulting in the creation of a plasma. For simplicity and clarity the telescope and aperature will be omitted from additional schematics of LIBS systems. A pulse of energy, usually 10-20 ns long and 50-300 mJ (for gaseous samples), is emitted by the laser. The energy cross-section for a typical beam (often called the beam-profile) displays a Gaussian curve shape; the beam is more intense at its center than at its edges, and decays exponentially between the two. To obtain a more uniform beam a Galilean telescope consisting of a negative (concave) and positive (convex) lens is used to expand the beam and an aperture removes light at the expanded beams edgeyielding a wider beam with a more uniform profile. In addition to a uniform energy 16

PAGE 29

17 cross-section, a wider beam can be focused more tightly than a narrow one, allowing for a greater spatial energy density. This is advantageous as the next step is to focus the beam to a spot with a condensing lens resulting in a photon flux sufficiently large that the dielectric strength of the gaseous medium is exceeded and a plasma is formed. Extraction Optics As previously detailed, plasma formation results in a burst of electromagnetic radiation centered in the near-ultra-violet and visible portion of the spectrum. Regions of peak intensity, which correlate to atomic and molecular transitions of plasma constituents, develop a few hundred nanoseconds after plasma initiation and must be collected and spectrally resolved. All studies performed for this dissertation collect in a backscatter configuration, which minimizes experimental uncertainty due to spatial variation in plasma formation. igure 2-2. Optical setup for extraction of plasma emission from sample volume. rror then t Plasma Collimation Lens Pierced Mirror From Laser Fiber Optic iCCD Spectrometer Condensing Lens F First a positive lens is used to collimate the spherical plasma emission. A mi urns the collimated beam into a second positive lens which recondenses the plasmaemission onto a fiber optic bundle. The fiber optic feeds a spectrometer which is tuned to a narrow wavelength range of interest, allowing for sub-nanometer resolution of atomic

PAGE 30

18 lines. Spectra are imaged using an intensified charge-coupled device (iCCD) mounted atthe exit slit of the spectrometer. System Synthesis of the laser delivery system with the optics for extraction of plasma emiss igure 2-3. Basic LIBS system. When plasma emission is collected along the same axis as incoming laser radiation, the system is said to be back-collecting. Notice thatadiation the sma emission is the basis of the LIBS technique, auxiliary diagne e and al Combination ionincorporating auspicious properties of bothyields a basic LIBS apparatus. N d:YAG Spectro-meter CCD Pierced Mirror Fiber Optic Plasma F the same lens serves as both the condensing lens to deliver the laser r to the sample and the collimation lens for back-collection of plasma emission. Also the mirror used for extraction must be pierced in the center to allow transmission of the initiating laser pulsethe amount of reflection sacrificed to the hole is negligible on basis of collected light area. While collection of pla ostics are desirable for system characterization within experimental studies; temporal beam sampling and beam energy measurements are most common and arabundantly detailed for experiments in this work. Descriptions of experimental configurations that follow all start with the basic LIBS apparatus described abovmake use of numerous enhancementsadditional optics, secondary laser sources, digit

PAGE 31

19 oscilloscopes, iCCDs for image capture, digital delay generation for temporal sequencing, and optical power metersin analyses other than basic spectral coPlasma Transmission Study llection. The LIBS system used fments of laser-induced plasm igure 2-4. LIBS system incorporating second laser for plasma transmission measurements. A 10ed Nd:YAG laser operating with 275-mJ pulse energy, 10-ns pulse 75-mm or temporally resolved measure a transmission is given below: N d:YAG Spectrometer iCCD Pierced Mirror Fiber Optic Plasma N d:YAG Aperture Calorime ter Probe laserbeam F 64-nm Q-switch width, and 5-Hz pulse repetition rate was used as the plasma source for all experiments. The expanded laser beam (12-mm in diameter) was focused using afocal length UV grade plano-convex lens (CVI, PLCX-50.8-38.6-UV-1064) to create an approximately 20-m diameter focal spot for plasma formation. The plasma emission was collected along the incident beam in a backward direction and separated using a 50-mm diameter elliptical pierced mirror. The collected light was launched into an

PAGE 32

20 optical fiber bundle coupled to a spectrometer (2400-grove/mm grating, 0.12-nm optresolution), and recorded with an intensified charge-coupled device (iCCD) array. Transmission Measurements ical bsorptivity measurements the plasma was formed in the laborahe o the plasma-creating laser beam, and fo of he For the transmission and a tory air. A probe Nd:YAG laser beam was synchronized to the plasma-creating laser by externally triggering both the flashlamp and Q-switch. The temporal jitter between the plasma-initiating laser pulse and the probe laser pulse was about 1 ns. Tprobe laser was operated using either the frequency-doubled 532-nm line (~6 ns fwhm pulse width), or the fundamental 1064-nm line. The pulse energy of the probe laser was maintained at 15 mJ for both the 532 and 1064-nm probe beams. However, the flashlamp pump energy was significantly reduced for the 1064-nm probe beam, which resulted in a broader temporal pulse-width, namely ~30 ns fwhm. The probe laser beam was directed orthogonal t cused at the center of the resulting plasma using a 250-mm focal length lens. Theprobe laser beam alone had insufficient energy (15 mJ) to create a plasma, which was verified by the absence of breakdown with the probe laser alone, as well as the absencebreakdown at long delay times following the 1064-nm laser initiated plasma. The probe laser was translated both horizontally and vertically until the minimum transmission was recorded at a delay time of 20 ns with respect to the plasma-initiating 1064-nm laser pulse. The spatial positioning ensured that the probe laser beam was passed through tmost optically dense region of the plasma. It is noted that in a previous study using a similar probe beam arrangement, the plasma was found to expand to its characteristic volume of 1.4 mm 3 by 20 ns following plasma initiation (Carranza and Hahn 2002c). The transmitted probe laser pulse energy was recorded with a volume-integrating

PAGE 33

21 calorimeter (25-mm detector area), using a 1-minute average at a 5-Hz repetition rThe detector was placed about 0.5 m from the plasma to eliminate any detection of direplasma emission, which was verified by the lack of signal in the absence of the probe laser. The large detector area functioned to mitigate any effects of beam steering on threcorded transmission, although no probe beam displacement was observed during the experiments. The prob ate. ct e e laser beam transmission was determined as the ratio of the average laser pulseues ser pling three locations, each approa ntinuum Measurements ed in an atmospheric pressuream energy transmitted through the plasma divided by the average laser pulse energy recorded in the absence of the laser-induced plasma (i.e. probe laser only). The latter value functioned as the reference pulse energy, and was taken as the average of the valrecorded before and after each plasma measurement. The average relative standard deviation of the reference pulse energy was 1.6% for all experiments; hence probe laenergy drift was not significant. All transmission measurements were recorded a minimum of four times for each reported delay time. Plasma continuum emission was measured by sam ximately 20 nm wide, along the baseline of an ensemble average of 100 spectrtaken from the transmission data. Electron Density and Plasma Co For electron density measurements, the plasma was form re gas cell (Huntington stainless steel 6-way cross) through which a gaseous stof 42 lpm of purified (HEPA filtered), dry air was passed. An additional flow rate of high purity methane was added to the airflow to enhance the hydrogen emission line signal. The overall methane/air flow rate was maintained at 3% methane, below the lower flammability limit (5%) of methane in air, to prevent ignition by the laser-induced

PAGE 34

22 plasma. To obtain the electron density, Stark broadening of the H line was measured. For these measurements, a calibrated blackbody source was used to convert the recordedplasma emission to a relative spectral irradiance scale across the measured spectral bandwidth. All electron density measurements were based on the spectral analysis oensemble average of 100 spectra. Las f an er Shot Stability Study The experimental system stability study is shown schemFigure 2-5. LIBS system incorporating fast detectors and a digital oscilloscope for additional beam measurements. For ac (355 nm) of an injection-seeded Nd:YAG (10 H for the laser-induced plasma atically in Figure 2.5. Fast Detectors To Filtered Air N d:YAG Oscilloscope Spectrometer CCD Pierced Mirror Beam Splitter Fiber Optic Plasma Volume ll experiments, the third harmoni z repetition rate and 280 mJ/pulse) was used to create the plasma using a 50-mm diameter, 75-mm focal length lens. The laser could be operated with or without the

PAGE 35

23 seeder, while maintaining nominally constant pulse energy and timing. The plasma emission was collected on axis with the incident laser beam using a pierced mirror (ROptics, #60.2475) and 75-mm focal length condensing lens. The plasma emission was subsequently fiber-coupled to a 0.3-m spectrometer (2400 groove/mm grating, 0.022 nm/pixel dispersion, ~0.15-nm resolution), where spectral data were recorded using anintensified CCD detector array. For all experiments, LIBS spectra were recorded using signal integration of 10 s width following a delay of 10 s with respect to the incident laser pulse. The las olyn a er pulse waveform was sampled prior to the plasma focal volume and follow fast n a uniforied unt of d as ted within plasma volume). ing transmission through the plasma using a series of optical wedges as beamsamplers, as shown in Figure 2-5. The sampled beams were recorded using matched,photodiode detectors (~1 ns rise time) coupled to a digital oscilloscope (500 MHz bandwidth and 4Gs/s sampling rate). A 355-nm laser line filter was used with bothdetectors to eliminate possible signals originating from the plasma emission. LIBS measurements were recorded either directly in ambient air or withi rm jet of filtered air. For the filtered air measurements, compressed air was dusing an activated alumina desiccant, and passed through a high-efficiency particle arrestor (HEPA) filter to effectively remove all particulates. It is noted that the amowater vapor and carbon dioxide in the compressed air were markedly reduced by the presence of the activated alumina desiccant. The conditioned air stream was expandea free jet with an exit diameter of 1 cm directly into ambient air approximately 1 cm below the laser-induced plasma kernel (L/D = 1 at plasma center, laminar flow expec

PAGE 36

24 These two configurations allowed LIBS data to be recorded under the aerosol loadings of ambient air, a nd under identical experimental conditions but in the absence of partic s ify tudy is shown below: Figure 2-6. Nebulization system coupled with a LIBS system (only Nd:YAG shown) for aerosol analysis. LIBS sample chamber: Six-way cross ulate loadings by using the HEPA-filtered air jet. Independent aerosol size andconcentration data were recorded using a commercial light scattering instrument with sensitivity over a range from 100 nm to 2.5 m particle diameter. This instrument waused to measure the ambient air aerosol loadings in the laboratory air, as well as to verthe absence of any aerosol particles in the HEPA filtered air stream. Carbon Calibration Study The experimental system used for advanced LIBS calibration s Mixing-drying chamber Flow controller (N2) Nebulizer Exhaust Nd-YAG laser Flow controller (CO, CO2, CH4) Condensing Lens

PAGE 37

25 For all experiments, a 1064-nm Q-switched Nd:YAG laser running at a 5-Hz repetition rate, 10-ns pulse width, and 290-mJ/pulse was focused using a 50-mm diameter, 75-mm focal length lens to create the plasma. The plasma emission was collected on-axis with a pierced mirror and fiber-ed to a 0.275ter with a 2400-groove/m2-nm optical resolution). Spectral data were recorded using an iCCD detector array. All carbon emission spectra were recorded using a delay of 10 s with respect to the laser pulse and a signal integration time of 12 s. This detector gating optimized the analyte signal of interest, nameI) line at 247.86 nm. Additional H (656 nm) emission spectra were recorded using a temporal delay of 4 s and a gate width of 4 s, and CN emmhamber, as shown in Figure 2-6, with the goal of providing different physical and chemyte CertiPrep, 10,000 g/ml of carbon as oxalic acid, (COOH)2 ). For the second set of coupl -m spectrome m grating (0.1 incident ly the carbon ( ission spectra were recorded using a te poral delay of 40 s and a gate width of 60s. Five different carbon-containing species were introduced to the LIBS sample c ical states of the carbon carrier species. Calibration stream flow rates and analconcentrations were adjusted to provide a comparable range of atomic carbon mass concentrations for all experiments. Solid Phase Carbon Experiments Two different types of solutions were nebulized to provide a carbon-rich aerosol stream (i.e. solid-phase analyte) to the LIBS sample chamber. For the first set of experiments, carbon solutions were prepared in the range from 500 to 5000 g/ml of atomic carbon by dilution of Atomic Absorption (AA) and Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) grade carbon standard solutions (SPEX

PAGE 38

26 experiments, carbon solutions were prepared in a similar mass concentration range usinga monodisperse particle suspension of 30-nm polyst yrene (C8H8) spheres (Duke Sciene e ff. For all experiments, the carbon mass pensions was adjusted by dilution with ultra-purified deion nce o the e S e rene tific). To avoid foaming within the nebulizer, the polystyrene suspensions werfirst washed with a resin (BIO-RAD, AG 501-X8, 20-50 Mesh, Molecular Biology Grade) to remove the surfactant. Specifically, 50 g of twice-washed resin was added to 300 ml of the polystyrene suspension (1% solids) and stirred for 18 hr, after which thpolystyrene suspension was poured o concentration for the polystyrene sus ized (DI) water, using the manufacturers initial specification (10% solids). The exact mass loadings were then verified by evaporating the water and weighing theremaining dried polystyrene mass corresponding to a known initial suspension volume. The preliminary estimates of carbon concentration based on the manufacturers mass loadings and the measured mass loadings agreed to within 5%, with the small differeattributed to the loss of some particles to the resin beads during surfactant removal. Aerosol flows were generated using a pneumatic nebulizer supplied with 5-liters/min (lpm) of dry nitrogen. A 46-lpm dry nitrogen co-flow was introduced taerosol stream to ensure uniform bulk flow and to facilitate complete desolvation of thnebulizer droplets, resulting in primarily submicron-sized aerosol particles at the LIBsample point. For the given flow scheme approximately 10 3 particles exist in any given plasma volume. The nitrogen supply was HEPA filtered prior to use, and all gases wermetered with precision mass flow controllers. Prior to nebulization, the polystysuspensions were well mixed and then sonified for 5 minutes to minimize any particle agglomeration. The overall aerosol generation system was described in full detail in a

PAGE 39

27 previous publication (Hahn et al. 2001), where nebulization of similar AA-ICP stansolutions provided aerosols with a modal diameter less than 100 nm based on particle sampling and transmission electron microscopy (TEM) analysis. The total mass loading of carbon in the LIBS sample stream was defined by the liquid nebulization rate (nominally 0.12 ml/min), the carbon mass concentration in th dard e liquid). The 0 as e ed mass loadings consiy solution/suspension, and the total gas flow rate (nebulizer flow plus co-flowsolution/suspension concentrations were adjusted to provide a nominal range from 1,00to 10,000 g carbon/m3 through the LIBS sample chamber, which corresponds to about 900 to 9,000 parts per billion (ppb) of carbon on a mass basis. Gas Phase Carbon Experiments Three different carbon-containing gaseous species were used to provide a carbon-rich sample stream (i.e. gas-phase analyte) to the LIBS sample chamber. The three gphase species, carbon dioxide (CO 2 ), carbon monoxide (CO), and methane (CH 4 ), werepurchased as nominally 50-ppm mixtures in a dry nitrogen balance (Spectra Gases). Thactual concentrations ranged from 49.3 to 50.8 ppm, with the true concentrations used forall calculations. The gases were metered with precision mass flow controllers and mixwith the nitrogen co-flow gas stream to provide a range of carbon stent with the values used for the solid-phase experiments. To provide consistencwith the solid-phase experiments, the nebulizer was operated with ultra-purified deionized water, and the total gas flow rate was maintained at the same value for all experiments. In addition, valves on the experimental apparatus allowed the analyte gas streams (CO, CO 2 CH 4 ) to be diverted through the nebulizer for the case of 5 lpm of analyte gas flow rate.

PAGE 40

28 The total mass loading of car bon in the LIBS sample stream was controlled by adjush four and in ). m ) molecular emission electron densities were measured from the Stark broadening of the 65s of ting the relative flow rates of the analyte gas stream and the nitrogen co-flow stream. The experimental conditions were adjusted to provide a nominal range from 1,000 to 7,800 g carbon/m 3 through the LIBS sample chamber. LIBS Data Collection The analyte peak of interest was the carbon (I) atomic emission line at 247.86 nm (21,648 61,982 cm -1 ). The recorded spectral window was approximately 30-nm wide (230 to 260 nm), and was nominally centered on the 247-nm carbon line. For eaccarbon species and carbon mass concentration, a series of 6000 spectra (6 x 1000 shots) were recorded and ensemble-averaged. The experiments were repeated between eight times for each carbon species, with each experiment recorded on different days an attempt to average any fluctuations in the overall system. Spectra were also recorded for the nebulization of DI water only for all experiments to serve as spectral blanks (i.eno carbon emission line In addition to data recorded near the 247-nm carbon line, emission spectra werealso collected for each experimental condition, including DI water only, using 30-nspectral windows centered at 656-nm and 385-nm to capture emission features corresponding to the H atomic emission line and the cyanide (CN band, respectively. Free 6-nm H line using the same procedures reported recently (Carranza and Hahn 2002). Briefly, following the work of Griem, the H line intensities at full-width half-maximum (fwhm) were related to the electron density using the fractional half-widththe reduced Stark profile (Griem 1974).

PAGE 41

29 Plasma-Aerosol Interaction Study Schematic representation of the experimental system for the study of plasma-aerosol interaction is given below: combined with a LIBS system for sampling a. idth a 50mm diameter, 75mm focal length lens to create the plasma inside a sealed, stainless steel six-way cross vacuum section (Huntington) forming a LIBS sample chamber. A nebulized solution of borosilicate glass microspheres (Duke Scientific, 9002) in ultra-pure deionized water was fed into the chamber with a dry HEPA-filtered air co-flow stream. The nebulized solution and co-flow stream, as in previous work, mixed inside a 0.8m drying tube ensuring complete dryout of aerosols upon entry to the LIBS sample chamber (Hahn et al. 2001). The diameter of the microspheres was given by the manufacturer as 2.0.7m and was Nd:YAG (10ns/300mJ/pulse) Spectro meter Pierced Fiber Optic Chamber: 6-way cross iCCD Mirror LIBS Sample iCCD Turning mirror: protects CCD from 1064nm laser emission not absorbed by plasma Narrow-band Line Filter ) (396.2nm/3nm FWHM Aerosol Stream Figure 2-7. Intensified CCD for imagingaerosols within the laser-induced plasm For all experiments, a 1064nm Q-switched Nd:YAG laser with a 10ns pulse wand 300mJ/pulse was focused using

PAGE 42

30 confirmed to be accurate using optical microscopy, although no quantitative analysis was performed. The density of the glass was 2.50g/cc and spheres dry number density was 9.5x1010/g. Calcium, a minor constituent of borosilicate glass present as residual lime (CaO), was the analyte of interest. Typical lime concentrations for borosilicate glass (in which most of the lime present in soda-lime glass has been replaced with boric oxide, B2O3, for modification of material properties) range from ~0.05 to ~1% by mass Figure 2-8. Plasma image as visualized from a co-axial orientation to the laser beamThe Imaging System An imaging system was used to record the position of an entrained particle relatito the plasma kernel, a wide array of experiments can be performed to measure the eof variations in particle position on spectral measurements. To that end, an iCCD synchronized to the Q-switch of the plasma-inducing laser was set up according to the schmatic in Figure 2-7. A 79.6 mm focal length, UV-grade achromatic lens (OptoSigm#027-3015) provided a magnification of 3.6 to the plane of the iCCD with respect to the ve ffect a,

PAGE 43

31 plane containing the plasma. In this configuration the iCCD resolution was found to be 5.3 m/pixel (image targets and derivation of optical resolution is given in Appendix BAdditionally, a high-efficiency 1064-nm dichroic turning mirror (Newport, ). #20QM20HM.15, UV-grade fused silica substrate) was placed in the image path protecting the CCD and intensifier from laser energy at 1064 nm which propagates through the plasma. Plasma emission was forward-collected on the axis of the laser using 1024x1024 pixel iCCD camera (Andor iStar) with the intensifier gate triggered by the Nd:YAG Q-switch. The gate delay of the camera (from the tail end of the laser pulse) and its width were adjusted in time increments ranging from 2.0-30.0 s and from 0.2-3.0 s respectivelythe width maintained at 10% of the delay to allow adequate collection of light by the iCCD as the plasma weakened in time. A 1064 nm mirror was placed in ont of the camera to protect the CCD and intensifier from laser radiation transmission through the plasma which the authors measured to be 10% of total laser power. Additionally10 fr an atomic line filter (CVI, #F03-396.1-4-1.00)3nm full-width at half-maximum centered at 396.2nmcorresponding to a strong Ca II line at 396.85nm, was placed in the beam path so as to collect only calcium emission and background signal over that same narrow spectral range. An achromatic lens with 1064nm anti-reflective coating was used to focus the plasma image onto the CCD. Using a finely spaced and static grid, the magnification and resolution of the imaging system were found to be 3.6 and 5.2m/pixel respectively. The images were sufficiently large that the laser repetition rate which allowed the capture of every laser pulse was limited to 1Hz. For the measurements, one gram of microspheres (9.5x10/g) was diluted with 400cc ultra-pure

PAGE 44

32 DI water to make a stock solution which was fed into the system by a nebulizer wwhen pumped with 5L/min dry HEPA-filtered air, delivered 0.14g/min of solution. resulting concentratio hich, The n of solids, as verified by total dryout weight, was 250g/mL corresally at a or m lied with ICP) standard solution of Calcium diluted to known concentrations of 0.0 (DI water blank), ponding to a number density of 2.4x10 7 /cc, assuming a mean particle diameter of 2m and a density, given by the manufacturer, of 2.5g/cc. The aerosol system was run with a co-flow of 20L/min dry HEPA filtered air. At each gate delay, a series of image frames were collected while running the aerosol system at steady state. Additioneach delay, the system was run with pure DI water only in order to collect a background signal. A measure of the dark current signal was also measured by running the camerwith the lens cap on. The Spectroscopy System Plasma emission was back-collected on-axis with the laser using a pierced mirrand fiber coupled to a 0.275m spectrometer with a 2400 groove/mm grating. Spectral data were recorded over a 30nm window centered at 396nm with an iCCD cameraagain, the intensifier gate triggered by the laser Q-switch. As the spectra could be recorded more rapidly than the images, the laser was run at 5-Hz. A dilute aerosol strea(one aerosol per plasma shot was statistically desired for spectral calibration) was generated by diluting the stock solution, noted above, 10:1 with ultra-purified DI water and the dry air co-flow was increased to 40L/min. The nebulizer was again supp5L/min dry HEPA-filtered air. With the aerosol system running as described, five separate 1000-shot spectral sequences were taken over the same range of gate delays andcorresponding widths given above. In order to calibrate the spectral response of the aerosol stream, the aerosol system was seeded with an inductively-coupled plasma (

PAGE 45

33 2.5, 5.0, and 10.0g-Calcium/mL-solution. Again, five 1000-spectrum sequences were taken over the identical range of intensifier gate delays and corresponding widths as noted above. Previous work demonstrated that mean aerosol diameter from an identical system was approximately 100 nm for a range of ICP liquid analytes (Hahn et al. 2001).

PAGE 46

CHAPTER 3 PLASMA TRANSMISSION PROPERTIES Understanding the parametric changes within a laser-induced plasma with respect to time is important in explaining connections between plasma energy and analyte dissociation. Specifically, the ability of a plasma to absorb energy relates directly to its ability to re-radiate that energy (Kirchoffs Law) both internally and externally. The current supposition is that the plasma is most energetic and therefore most capable of dissociating entrained particles during the times that it is most absorbing and/or energetic. Time-Resolved Plasma Transmission For the following portion of the study, a two-laser system was used to measure temporal changes in transmission (and, hence, absorption) of a plasma produced in air. The time-resolved plasma transmission measurements are presented in Figure 3-1 as a function of delay time between the plasma initiating laser and the probe laser. Data are presented for the 532 and 1064 nm probe beams, along with the temporal profile of the plasma-initiating laser pulse as a reference. Zero delay time corresponds to the peak-to-peak temporal alignment of the two lasers. The transmission profiles reveal the rapid increase in plasma opacity, with the minimum transmission values of 9 and 16% recorded 30 and 40 ns following the plasma initiating pulse for the 532 and 1064 nm probe beams, respectively. It is noted that the 1064 nm probe laser had a considerably wider pulse width, as discussed above; hence the temporal resolution of the 1064 nm data is less than the 532 nm data. These observed time scales for the rapid increase in plasma absorptivity correspond well with the full-width of the plasma initiating laser pulse, 34

PAGE 47

35 204060 80100 532 nm probe Plma Tansmiion asrss(%) 1064 nm probe Reference pulse 0-50050100150200250300350400450500 Delay Time (ns) which r y Figure 3-1. Plasma transmission as a function of time following plasma initiation using532 and 1064 nm probe laser beams. The inset laser pulse profile corresponds to the plasma initiating laser pulse at zero delay time. decays by 99% of its maximum value in 30 ns from the peak intensity. The Figure3-1 data are further analyzed below in the context of electron number density; however, the transmission data alone are evidence that initial plasma processes are closely coupledtemporally to the plasma initiating laser pulse. The abrupt change in slope of the plasma transmission profile at the tail end of the initiating laser pulse may be interpreted as denoting the change from net plasma energy input to net energy dissipation, the latter foexample via radiative transfer and electron recombination (free-bound). An additional interesting feature of the Figure 3-1 data is the return of the plasma transmission to nearly 100% (e.g., 98% and 94% at 532 and 1064 nm, respectively) b

PAGE 48

36 500 ns following plasma initiation. This relatively rapid return to an essentially transparent plasma is significant in the context of radiative transfer and in consideration of new LIBS configurations such as dual pulse LIBS. If a second laser beam is to be coupled to an existing laser-induced plasma, temporal delays beyond the first 0.5 s will most likely result in significantly different laser-plasma interactions than those following the first laser pulse within a few hundreds of nanoseconds. The transmission measurements may also be considered in terms of the plasma expansion. As reported in an eariler paper (Carranza and Hahn 2002a), the plasma expands to its characteristic volume of ~1 mm3 within approximatley 20 ns, corresponding to an initial plasma propagation velocity on the order of 10 km/s. Time-Resolved Continuum Measurements Further insight into the plasma processes are revealed by the time-resolved continuum emission measurements presented in Figure 3-2. The continuummission data correseither 270, 400 or 530 nm. The data are all normalized to their respective maximum temporal emission value. The continuum emission data represent radiative transfer from the plasma, primarily by recombination radiation (free-bound) and Bremsstrahlung emission (free-free). The temporal scale of the continuum emission corresponds markedly with the transmission data discussed above. The continuum emission of all three bands peak between 20 and 40 ns following the peak of the plasma initiating laser pulse, with the 270 nm continuum band peaking ~10 ns before the peak of the 400 and 530 nm bands. A fixed detector gate of 20 ns was used for all measurements; hence temporal resolution is limited to such a value. For all three spectral bands, the continuum e pond to the integrated sum of approximately 20 nm bandwidths centered at

PAGE 49

37 emission was observed to decay by one order of magnitude during the first 200 nrepresents a decrease in effective blackbody temperature of about 70% based on emissivepower per the Planck distribution. Although thermodynamic equilibrium is necessarythe plasma radiative transfer to approach the blackbody function, such a 70% decrease in temperature is quite consistent with reported temperature measurements (see above references) over similar time scales. s, which for 0.0 0.81.0 270nm 400nm 530nma.u.) 0.20.40.60100200300400500 Pasma Contin centered at 270, 400 and 530 nm. All data have been normalized to maximum luum Emission (Time (ns)Figure 3-2. Plasma continuum emission (20 nm bandwidth) for three spectral regions emission intensity for each respective spectral region.

PAGE 50

38 Electron Density Measurements To better characterize the laser-induced plasma over the time scales discussed above, electron number density measurements were also performed. Electron densities were calculated based on the Stark broadening of the H line (656.3 nm), which is well suited to electron density measurements between 10 17 and 10 19 cm -3 (Griem 1997). Following the work of Griem, the Sta rk line intensities at full-width half-maximum (fwhm) are related to the electron density through (Griem 1974) 3/23/22/1)()154)(2(2efwhmne 3-1 where 1/2 is the fractional half-width of the reduced Stark profile (see Appendix C for 1/2 values), e is the electron charge (4.8032x10-10 esu), and ne is the free electron number density. The present data were reduced using the values of 1/2 as reported by Griem (1974), assuming a plasma temperature of 30,000 K. The inversion of electron densities from Stark widths is rather insensitive to temperature, as related to the corresponding reduced line widths. To assess sources of error due to temperature uncertainty in the present data, the measured Stark line widths were also inverted using temperatures of 20,000 and 40,000 K and the corresponding fractional half-widths. Accordingly, the error bars reported with the measured free electron densities represent the maximum error associated with the uncertainty of 10,000 K. The data were also analyzed using the Using the sby 27% at the maximum value (2.6x1018 cm-3) and by 12% at the minimum value (8.9x1017 cm-3), with an average increase of 20% over all measurements. Such differences are assumed within the expected absolute accuracy of the reported electron more recent Stark broadening constants reported by Gigosos and Cardenoso (1996). ame temperature of 30,000 K, the calculated electron densities were increased

PAGE 51

39 density measurements basein the current study. Finallg t in Figure 3-3 as a nction of time following the plasma initiat d on the data analysis as performed y, as noted above, recent theories have addressed the indirect and direct couplinbetween electron and ion broadenings and the impact on Stark broadening (Oks 2000). Over the range of the current Stark line widths, application of this more generalized theory is expected to change the electron densities by 10 to 20 percent based on recenpublished comparisons (Parriger et al. 2003). The measured free electron number densities are presented fuing laser pulse. 5.0 10171.0 10181.5 1018183.0 10Electon Density(cm-3Delay Time (ns)Figure 3-3. Free electron number density as a function of time following plasma initiation, as measured using the Stark broadened H 2.0 102.5 1018180100200300400500600r ) line width.

PAGE 52

40 The hydrogen line widths were not sufficiently resolved from the significant continuum emission for times less than 100 ns. From 100 to 500 ns following plasma initiation, the free electron number density decreases from 2.6x10 18 cm -3 to less than 10 18cm tron number density measured the corres cm-3 rted by Yalcin et al. (1999), as discussed above. It is noted that Yalcin et al. used a similar laser energy and optical set-up, and reported a similar coupling of ~2/3 of the incident laser energy into the plasma, which also agrees well with the value measured with the current LIBS setup (Carranza and Hahn 2002b). Overall, the electron density profile reveals a near linear decay in electron number density between 100 and 400 ns, following by a more gradual decay over the last three data points. This is somewhat in contrast to the non-linear changes in plasma transmission and plasma continuum emission during a similar temporal region, namely between 100 and 400 ns. However, it is noted that the continuum emission is a highly non-linear process (i.e. Planck distribution), while the transmission data must be further reduced to make quantitative comparisons. Plasma Absorption Cross-Section Modeling Plasma Transmission The plasma transmission may be modeled using the Beer-Lambert relationship -3 As observed by the error bars in Figure 3, the uncertainty associated with the assumed plasma temperature is greater in the region of more significant Stark broadening, however, even at the greatest elec ponding relative error is about 5%. The measured electron density of 1.2x10 18at a delay of 350 ns is in exact agreement with the value repo namely: nNLNLNL)()()(exp21 3-2

PAGE 53

41 where ent (NL) i represents the optical depth or turbidity of the i th species, with the absorption cross-section, N the absorbing species number density, and L the absorptionpath length. For the present analysis, it is assumed that free electrons are solely responsible for attenuation of the probe laser beam, which results in the following relation between the measured plasma transmission and the absorption coeffici(defined as K abs = n e) namely: )ln( LKabs. 3-3 While the absorption cross-section is correctly the sum of the absorption and scattering cross-sections, scattering of the probe laser beam by plasma free electrons, Thomson scattering, is neglected due to the state of the plasma frequency in relationthe incident frequencies of the probe laser (Warner and Heiftje 2002). Specifically, fothe first few hundred nanoseconds, the measured electron density is nominally 2x10 to r Debye length and represents the resonance frequency of the free electron oscillations about their equilibrium positions. For the 532 and 1064 nm probe lasers, the corresponding frequencies of 5x10 and 2x10 Hz, respectively, are sufficiently close to the plasma frequency such that significant absorption is expected, thereby dominating any scattering effects. Equation 3-3 enables conversion of the Figure 3-1 transmission data to a more quantitative measure of the plasma properties, namely the plasma optical depth (absorbance) or turbidity. 18 cm -3 which yields a plasma frequency of ~10 13 Hz using the relation 2/13109 (Thorne et al. 1999). The plasma frequency is the reciprocal of the 1414 epnx

PAGE 54

42 Plasma Absorption The reduced plasma turbidity data and the measured electron density data both represent the temporal evolution of the plasma, hence it is interesting to examinecorrelation between these two parameters. To this end, the measured plasma turbidand electron density data were paired at each similar temporal delay over the range100 to 300 ns following plasma in the ity from itiation. The result is presented in Figure 3-4, which revea a linear relatielectron density (R=0.998). lsonship between the 532 nm probe laser turbidity and the free 0.00.20.40.60.81.21.41818181818181818 1.01.61.4 101.6 101.8 102.0 102.2 102.4 102.6 102.8 10 y = -1.3542 + 1.0402e-18x R= 0.99795 asma Absot L)E Plrption (Kexlectron Density (cm-3) Figure 3-4. Measured absorption coefficient for the 532 nm probe laser beam as a function of the corresponding measured free electron density.

PAGE 55

43 An interesting r esult of the Figure 3-4 data is that based on the assumption of free electr ng) long be laser beam. Using this path length, the calculated free-free absorption cross-section is 9.5x10-18 cm2 for 532 nm radiation. This value compares very well with a calculated effective absorption cross section of about 3x10-17 cm2 at 500 nm based on the sum of free-bound and free-free cross sections per Hion (Griem 1997). The absorption coefficients Kabs corresponding to the Figure 3-4 data range from 2 to 12 cm-1. These values agree very well with theoretical calculations reported for 532 nm radiation in a fully ionized hydrogen plasma (Hora and Wilhelm 1970). Extrapolation for Small Time Scales Another interesting result of the measured absorption cross-section is that it enables calculation of electron densities from transmission data at time scales below 100 ns, the lower limit realized in the present study for direct Stark broadening measurements. The extrapolated electron density measurements are presented in Figure 3-5 down to 15 ns following plasma initiation. The free electron density profile peaks at 30 ns (corresponding to the minimum in 532-nm transmission) at a value of 3.6x1018 cm-3 and reveals a value of about 2x1018 cm-3 at 15 ns delay. It is noted that the delay of 30 ns full-width. Accion, the on absorption and the additional approximation of constant optical path over this200 ns time period, the slope of the linear fit yields the free-free (inverse Bremsstrahluabsorption cross-section. In a complementary recent study, the two-laser approach wasused to profile the overall plasma shape under similar experimental conditions (Carranza and Hahn 2002a). It was reported that by about 20 ns following plasma initiation, the plasma had expanded to its characteristic size, with a linear dimension of 1.1 mm athe path of the orthogonal pro corresponds to the 99% decay point of the incident laser pulse, hence essentially thordingly, based on the extrapolation of the measured transmiss e

PAGE 56

44 electr on density profile peaks at the tail end of the incident laser pulse. Overall, the change in electron number density during the first few hundred nanoseconds of plasma evolution is not that great, hence an aerosol particle engulfed by the plasma is subjected to a relatively consistent frequency of electron-particle collisions. This suggests that theplasma temperature (i.e. electron energy ~ temperature) could be more important than absolute free electron density for plasma-particle interactions, namely particle dissociation and vaporization. 1.5 10182.0 10182.5 10 3.0 10183.5 10184.0 1018cm-3) 18050100150200250300ElecTime (ns)Figure 3-5. Predicted electron density values as a function of time following plasma section of 9.5x10 tron De ( initiation. The data are based on the calculated free-free absorption crossnsity -18 cm 2 and the measured 532-nm transmission data.

PAGE 57

45 While temporal regions limited to the first few hundred nanoseconds of plasmaevolution are characterized by significant electron densities, plasma frequencies, and plasma absorptivity, as described above, the longer delay times (500 to 1000 ns) reveal a plasma frequency investigatedtypical of conditions used for LIBS-based gas and aerosol analysisis characterized by significant transient changes in electron density and plasma absorptivity during the first few hundred nanoseconds. The fundamental change from an absorbing (i.e. optically thick) plasma to a non-absorbing plasma during this period is important with respect to the radiative transfer within and from the plasma, and is therefore expected to play a role in the interactions between the laser-induced plasma and incorporated particles. Hence a better understanding of such processes provides a starting point for more complex plasma-particle interactions, as well as furthering the development of plasma modeling capabilities, and advancing LIBS as an analytical technique. p ~10 12 Hz. Such a relatively low plasma frequency is consistent withthe corresponding near transparency of the plasma to the probe laser wavelengths (i.e. incident ), as expected for the condition of p << incident The laser-induced plasma

PAGE 58

CHAPTER 4 LIBS PLASMA STABILITY The shot-to-shot precision of LIBS-based spectral analysis is considered in the context of laser-induced plasma temporal and spectral characteristics, including the effects of laser cavity seeding and the presenAll experiments were recorded using the same spectral window centered at 490nm, which enabled the simultaneous measurement of atomic emission from four pronounced peaks corresponding to three gas phase species, namely the hydrogen beta line at 486.13nm (82,259 102,824cm-1), two neutral nitrogen lines at 491.49nm (86,137 106,478cm-1), and 493.50 nm (86,221 106,478cm-1), and the 247.86nm neutral carbon line (21,648 61,982cm-1), which is observed in second order at 495.71nm. A representative laser-induced plasma spectrum recorded in ambient air for this spectral window is shown in Figure 4-1. Because all three species correspond to gas-phase ambient air species, with the carbon attributed to the approximately 300 ppm of ambient air carbon dioxide and the hydrogen attributed to water vapor, spectral signal variations due to sample non-homogeneity (e.g. as with aerosol-derived analyte species) is eliminated. It is noted that the intensity of the carbon and hydrogen atomic emission lines were significantly reduced in the HEPA filtered air stream due to the reduction in water vapor and carbon dioxide, as discussed above. ce of ambient aerosol particles within the laser-induced plasma volume. Effects of Laser Cavity Seeding 46

PAGE 59

47 500010000Wavelength (nm)Figure 4-1. Sample spectrum showing atomic emission lines from the plasma spline observed in second order. A series of 400 sequential LIBS spectra were recorded in HEPA filtered air wthe laser operating with the injection seeder. Immediately following this set of data, the seeder was turned off, and a second series of 400 spectra were recorded with unseeded laser operation. The full-width, baseline-subtracted peak areas were calculated foof the four atomic emission lines on a shot-to-shot basis. The various peak-to-base (Pand peak-to-peak ratios were calculated for each laser shot, and compared for the seededand unseeded experiments to assess the effects of lase 1500025000480485490495500InituNC ark recorded in ambient air. The 495.7 nm carbon peak is the 247.86 nm carbon I ith r each /B) r cavity seeding on LIBS shot-toH 20000tensy (a..)(2x) N

PAGE 60

48 shot precision. For each emission peak ratio, the average value and standard deviation (N=400 shots) were calculated for both the seeded and unseeded data sets. Effects on Shot-to-Shot Precision The results are summarized in Table 4-1 for all four peak-to-base ratios and all six peak-to-peak combinations, with the latter including both intraspecies and interspecies comparisons. As an example, the N491.5/N493.5 and the N491.5/C247.9 atomic emission ratios are presented in Figure 4-2 for the entire 800 shot sequence. For both cases, Figure 4-2 reveals no apparent differences between the shot-to-shot precision due to seeded or unseeded laser operation. For example, the nitrogen-to-nitrogen emission ratio (N491.5/N493.5) yields an average value of 0.458 with a relative standard deviation (RSD) of 3.39% with laser cavity seeding, while the average value with unseeded operation was 0.457 with a RSD of 3.45%. Table 4-1. Ratios of the area-integrated spectral peaks (peak-to-base and peak-to-peak) for seeded and unseeded laser operation. Each data entry represents thein average and standard deviation based on 400 single-shot spectra recorded HEPA filtered air. The 247.9 nm carbon peak was observed in second order. Seeded Unseeded ) Average Std Dev RSD(%) Average Std Dev RSD(% H(486)/Base 24.5 6.37 26.0 24.5 6.74 27.5 N(491)/Base 13.6 0.535 3.93 13.7 0.537 3.93 N(493)/Base 27.7 0.745 2.69 27.8 0.700 2.52 C(248)/Base 2.63 0.416 15.8 2.58 0.419 16.3 H(486)/N(491) 1.95 0.513 26.3 1.95 0.548 28.1 H(486)/N(493) 0.892 0.233 26.1 0.890 0.248 27.9 H(486)/C(248) 8.94 1.98 22.2 9.13 2.13 23.3 N(491)/N(493) 0.458 0.015 3.39 0.457 0.016 3.45 N(491)/C(248) 4.69 0.785 16.8 4.81 0.796 16.6 N(493)/C(248) 10.2 1.67 16.3 10.5 1.70 16.2

PAGE 61

49 10 1Pak 1 / Peak (a.u 0100200300400500600700800e2.) Seeded LaserUnseeded Laser Shot Count ( N 491.5 / C 247.9 ) ( N 491.5 / N 493.5 )Figure 4-2. It is noted ttio ot aition probability (gA) ine16x10620x10predictensity N491.5 of s inllent agt witxperiml value 8. Hr, eon linea comupper y state 8 cme eed to ber inse to fluons in properties uhe relaioese tws is furcussin the context of the respective peak-to-base ratios. Comparison of the nitrogen-to-carbon emission ratios between the unseeded and seeded laser operation revealed similar N 491.5 /N 493.5 and N 491.5 /C 247.9 peak-to-peak atomic emission ratios for an 800shot sequence. The data were recorded in HEPA filtered air for both seeded operation (shots 1-400) and unseeded operation (shots 401-800). hat the ra f the statistic al weigh nd the trans for these two l s (1.6 s -1 /3.5 6 s -1 ) ts an in ratio /N 493.5 0.459, which i exce reemen h the e enta of 0.45 oweve because the two missi s share mon energ (106,47 -1 ), th intensity ratio is xpect e rath nsitive ctuati plasma such as temperat re. T tive precis n of th o line ther dis ed below

PAGE 62

50 results. These two atomic transitions are characterized by significantly different upper energy levels, namely 106,478 cm-1 and 61,982 cm-1 for nitrogen and carbon, respectively, hence the ratio is expected to better reflect plasma fluctuations. As observed in Figure 4-2, the shot-to-shot fluctuations were observed to vary more than the nitrogen-to-nitrogen peaks, but were not observed to change when comparing seeded and unseeded operation. Specifically, the N491.5/C247.9 emission ratio yielded an average value of 4.69 with a RSD of 16.8% with laser cavity seeding, while the average value with unseeded operation was 4.81 with a RSD of 16.6%, the difference being statistically insignificant. An examination of all peak-to-peak combinations presented in Table 4-1 reveals no significant variation in average values and RSD between the seeded and unseeded laser operation. Of the six peak-to-peak ratios, four showed slight increases in RSD with unseeded operation, while two revealed slight decreases in RSD with unseeded operation. Overall, the average change in RSD was equal to a 3% increase when changing from effects of lion based on both interspecies and intraspecies on if the single atomor seeded to unseeded laser operation. The data support the present conclusion that the aser seeding are negligible with respect to laser shot-to-shot precis atomic emission intensity ratios. The data also support the same conclusiic emission line intensities are explored, specifically using the integrated emission peak to continuum intensity ratios, referred to as the peak-to-base (P/B) ratio. This metric is widely used fLIBS-based spectral analysis as it enhances data precision by normalizing for changes in absolute plasma emission signal levels (Radziemski et al. 1983, Carranza and Hahn 2002a), which can be considerable. Using the P/B ratio as a comparison, the 491.5 and

PAGE 63

51 493.5 nm nitrogen P/B ratios yielded an average value of 13.6 with 3.9% RSD and 27.7with 2.7% RSD, respectively, for the seeded laser operation. Unseeded operatan average of 13.7 with a 3.9% RSD and 27.8 with 2.5% RSD, for the 491.5 and 4nm lines, respectively. As with the peak-to-peak ratios, the differences in the peak-to-base ratios between unseeded and seeded laser operation are statistically insignificant. It is also noted that the average RSD of the two nitrogen peak-to-base ratios is 3.3%, which is in agreement with the RSD recorded for the peak-to-peak ratio of these same two liThe overall agreement in analyte precision between the peak-to-base and peak-tpeak ratios, give ion yielded 93.5 nes. o-n the common upper state of these two nitrogen lines, suggests a limiting precisk-e 4-ese two species in the HEPA-noise is ion dictated by spectral shot noise, which is significant with single-shot intensified CCD spectra. Overall, when comparing seeded and unseeded laser operation, similar results were also obtained with the peak-to-base ratios of the hydrogen and carbon emission peaks, as shown in Table 4-1, although the relative standard deviations of these two species are greater than the values obtained with nitrogen. As an example, the peato-base ratios of the 491.5-nm nitrogen line and the carbon line are presented in Figur3 for the entire 800 shot sequence. It is noted, however, that the increased RSD with carbon and hydrogen reflects the relatively low concentrations of th filtered stream, hence the spectra are characterized by poor analyte signal-toratios overall. In ambient air, for example, the RSD of the hydrogen peak-to-base ratioreduced to 3.4%, which is comparable to the nitrogen values. In aggregate, the above results are evidence of the limited effect of laser cavity seeding on the analyte signal precision, both with peak-to-peak and peak-to-base ratios, and also demonstrate with

PAGE 64

52 nitrogen the high precision obtainable with the LIBS technique under homogenous sample conditions and high analyte signal-to-noise. 02 4680100200300400500600700800Peakto-se RShot Count 10-Baatio (a 12141618.u.) Seeded LaserUnseeded Laser ( N 491.5 ) ( C 247.9 )Figure 4-3. N and Cpeak-to-base atomic emission ratios for an 800 shot operation (shots 1-400) and unseeded operation (shots 401-800). Temporal Effects on the Plasma Formation Process While the effects of laser cavity seeding were not apparent in the resulting LIBS spectra of homogenous gas phase analytes, it is still worthwhile to explore the potential effects of seeding on the overall plasma formation process. For example, the analysis of aerosol species involves the coupling of the discrete particles with the breakdown and 491.5 247.9 sequence. The data were recorded in HEPA filtered air for both seeded

PAGE 65

53 plasma formation processes; hence temporal fluctuations following breakdown initiamay be important. The experimental system was des tion igned to capture individual laser pulse waveforms prior to breakdown (i.e. incident waveform) and following each plasma event (i.e. transmitted waveform). In HEPA filtered air, reference incident waveforms and transmitted waveforms were recorded for each laser pulse for a sequence of 100 shots under seeded operation. The seeder was turned off and a second set of reference and incident waveforms were recorded for unseeded laser operation. A representative pair of incident and transmitted waveforms is shown in Figure 4-4, as averaged over a series of 100 shots. 0.0000.0150.03 0In 0.0450.0600.0750.0900816243240485664tensity (a.u.) Truncation TimeIncident Transmitted Time (ns) Figure 4-4. Representative waveforms for the incident laser pulse and the transmitted laser pulse following laser-induced breakdown. Each waveform represents anaverage of 100 shots.

PAGE 66

54 The plot illustrates the observed time-dependent plasma absorption of a significant fraction of the incident laser irradiance. The appearance of a second local maximum isnoted in the transmitted waveform about two-thirds of the way into the laser pulse. Similar behavior was recorded in a series of recent studies (Bindhu et al. 2003, 2004), and is attributed to a brief saturation or self-regulating condition during the plasma formation process. The temporal onset of significant plasma absorption of the incident waveform may be defined by the location of the first local maximum in the transmitted waveform with respect to th e local maximum of the incident waveform. This temporal truncation is defined in Figure 4-4. Examination of the truncation time on a shot-to-shot basis revealed a degree of variation of the truncation time about the mean truncation time. This variation may be interpreted in terms of the nature of the breakdown process, where the exact spatial and temporal point of breakdown initiation is expected to fluctuate due to variations in the laser pulse field coupled with the statistical nature of multi-photon ionization processes. It is expected that the degree of variation may be influenced by the presence or absence of laser cavity seeding, in that with cavity seeding the temporal laser mode is expected to be more uniform. The temporal variation of breakdown initiation was quantified for 200 laser pulses (100 seeded and 100 unseeded) recorded in the HEPA filtered air stream, with the exact truncation time calculated on a shot-to-shot basis. The mean truncation time was observed to change by slightly less than 2 ns between seeded and unseeded operation. However, it was observed that the laser cavity seeding produced point of bre a slight change in the shape of the overall pulse waveform. Specifically, the average akdown initiation remained relatively constant with respect to the leading

PAGE 67

55 edge, data obability plots of these two data sets were prepared and are shown in Figur ded r ata h all, while the position of the maximum intensity in the incident waveform shifted further into the pulse with seeded operation. Therefore, to provide the more representative measure of variation in the truncation time on a shot-to-shot basis, thewere further processed by calculating the variation in truncation time about the mean truncation time for both seeded and unseeded operation. Cumulative pr e 4-5. Both unseeded and seeded operation produced a set of truncation times that cluster primarily within 1 ns of the mean, however, careful analysis revealed a broadervariation of the breakdown initiation point with unseeded operation. Specifically, theabsolute range in truncation time about the mean was found to be 3.2 ns for the seedata and 3.6 ns for the unseeded data. The standard deviation about the mean was 0.26 ns for seeded operation, and 0.29 ns for unseeded operation. Furthermore, four data points corresponded to truncation times greater than 1.5 ns behind the mean truncation time founseeded operation, compared to only 1 such point for seeded operation, while five dpoints corresponded to truncation times more than 1 ns ahead of the mean time for unseeded operation as compared to only two for seeded operation. Hence the extreme variations in the distribution of truncation times are dominated by unseeded cavity operation data by a factor of three as compared to seeded operation. The conclusion is that the operation of the cavity seeder does produce a more repeatable breakdown processwith respect to the temporal onset of laser-induced breakdown. As discussed above, suceffects may play a role in further understanding of the overall breakdown process, notably in the presence of non-homogeneous samples such as aerosol sampling. Overcavity seeding appears to play a role in enhancing the reliability of the breakdown

PAGE 68

56 process, although as seen above, such an effect does not translate into any significantincrease in precision with LIBS-based spectral analysis. These data suggest tportion of the incident laser pulse, which is coupled strongly to the breakdown process, may be an important factor in the resulting precision of the analyte signal. Therefore, overall laser pulse temporal stability, as observed with cav hat the early ity seeding, may not noticeably affect signal precision because of the relative unimportance of pulse stability following the point of breakdown initiation. -2.0-1.5-1.0-0.50.00.51.01.52.0.0.1151020305070809095999999. 1.999 Seeded UnseededVariationbout Man Trcation me (nsPercent AeunTi) Figure 4-5. Cumulative probability distribution plot for both seeded and unseeded laser operation. Effects of Aerosol Loading To observe the potential effects of aerosol loadings on LIBS precision, it is necessary to compare a constant analyte concentration under changing loadings of concomitant particles. Recall that the HEPA filtered air stream contained significa ntly

PAGE 69

57 less carbon and hydrogen, hence only the nitrogen emission lines were compared to assess the effects of aerosol loading. The removal of carbon and hydrogen only slightly altered the nitrogen signals, with the 400-shot N 491.5 /N 493.5 average equal to 0.44 in ambient air (cavity seeding) and 0.46 in HEPA filtered air (cavity seeding). As befora series of 400 sequential LIBS spectra were recorded with the laser operating directly in ambient air with injection seeding. Immediately following this set of data, the HEPA filtered jet was turned on and a sec e, ond series of 400 spectra were recorded with seeded laser operation. The full-width, baseline-subtracted peak areas were calculated for each spectrum, and were used to calculate the N491.5/N493.5 intensity ratio on a shot-to-shot basis. Effects on Shot-to-Shot Precision The intensity ratios are shown in Figure 4-6 for the entire 800 shot sequence. The nitrogen-to-nitrogen emission ratio (N491.5/N493.5) yielded an average value of 0.440 with a RSD of 5.93% in ambient air with laser cavity seeding, while the average value in the HEPA filtered air stream was 0.458 with a RSD of 3.39% with cavity seeding. Near identical results were obtained with unseeded laser operation, specifically, the average (N491.5/N493.5) intensity ratio was 0.440 with a RSD of 5.81% in ambient air, while the average value in the HEPA filtered air stream was 0.457 with a RSD of 3.45%. The reduced from 5.93% to 3.39%m 5.81% to 3.45% with unseee nce The 491.5-nm emission line produced an average P/B of 11.0 with a RSD of 6.2% in LIBS shot-to-shot precision, as characterized by the relative standard deviation, was with seeded operation, and fro ded operation, by the exclusion of the ambient air particles as realized with thHEPA filtered air stream. Identical results were obtained when comparing the influeof ambient air on the individual nitrogen emission lines, namely the peak-to-base ratios.

PAGE 70

58 ambient air with cavity seeding, while the average P/B in the HEPA filtered air streamwas 13.6 with a RSD of 3.9% also with cavity seeding. With unseeded laser operations, the RSD of this emission line was found to similarly decrease from 6.4% to 3.9%. Overall, both seeded and unseeded laser operation were consistent with realizing a nearly 60% improvement in shot-to-shot precision in the absence of concomitant aerosol particles from the sample stream. Clearly the effects of aerosol loading dominate the effects of laser cavity seeding as manifest in the shot-to-shot precision of both the analytpeak-to-peak and peak-to-base emission ratios. e 0.200.300.40 0.500.600100200300400500600700800491.5 PeakN 495 Pek (a. 0.70N / 3.au.)Shot Count Ambient AirHEPA Filtered AirFigure 4-6. N491.5/N493.5 peak-to-peak atomic emission ratios for an 800 shot sequence. (shots 401-800). For all shots, the laser was operated with the cavity seeder. The data were recorded in ambient air (shots 1-400) and in HEPA filtered air

PAGE 71

59 A commercial light scattering based instrument was used to assess the average ambient air particle loadings in the laboratory air. Typical loadings were approximately 900 particles per cubic centimeter in the size range from 100 nm to 2.5 m, with abo80% of the particles between 100 and 200 nm, 18% between 200 and 300 nm, and the remainder between 300 nm and 2.5 m, with the largest fraction of the remainder weighted toward the smallest size range. In an earlier study, estimates were made for effective laser-focal volume and the laser-induced plasma volume for aerosol interactions (Carranza and Hahn 2002b), namely ~0.01 mm ut the and 1.4 mm3, respectively. The probability of a direct laser beam/aerosol particle interaction is about 1%, while the probability of laser-induced plasma/aerosol particle interaction is about 70%, using Poisson sampling statistics based on these effective sample volumes and aerosol loadings, as described previously (Carranza and Hahn 2002b). Clearly the ambient air particle loadings are such that the probability of interaction with the resulting plasma process is significant. Temporal Effects on the Plasma Formation Process Finally, the effects of aerosol presence on the temporal onset of laser-induced breakdown were assessed in a similar manner as described above. Incident and transmitted waveforms were collected in both ambient air (100 shots) and in HEPA filtered air (100 shots) with the laser operated with cavity seeding for all measurements. For both cases, the range about the mean truncation time was identical, namely 3.2 ns. The relative standard deviation about the mean time was 0.261 ns in the HEPA filtered precision, i 3 air and 0.254 ns in the ambient air. Therefore, in contrast to the effect on plasm spectral t appears that aerosol inclusion does not contribute significantly to temporal a

PAGE 72

60 variat uch ions in the onset of plasma spark inception. This is in agreement with the probability of direct laser beam/particle interaction (~1%) as discussed above. While the1% value represents the average probability of the laser beam encountering a single aerosol, the probability of such an encounter directly within the temporal/spatial regimeof breakdown initiation must be much lower, thereby making the direct impact of sinteractions of second-order importance in the actual breakdown initiation, as reflected in the temporal variation of the breakdown onset.

PAGE 73

CHAPTER 5 LIBS CALIBRATION Atomic line spectra, measured using the LIBS technique, are expected to increaselinearly with an increase in analyte concentration within the plasma volume. Most studies using LIBS for quantitative measuremen t construct such a calibration curve. A widely used assumption in LIBS research is that the slope of the curve remains unaffected by size or phase differences in analyte species. The following experimental results were compiled for the purpose of analytical examination of this assumption. Experiments required the use of several different analyte sources. For discussion purposes, the nebulized Atomic Absorption (AA) and Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) grade carbon standard solutions will be referred to as the AA-ICP case, the nebulized 30-nm polystyrene suspensions will be referred to as the 30-nm PS case, and the three gas-phase experiments will be referred to as the CO2, CO, and CH4 cases. Raw and Baseline-Subtracted Spectra Representative spectra in the range of the 247.86 nm carbon line are shown in Figure 5-1 for both gas-phase and solid-phase carbon sources. The upper set of four spectra represents the unprocessed plasma emission as recorded, demonstrating the presence of the carbon atomic emission line in combination with significant continuum emission. The lower set of three spectra represents background-subtracted spectra, which are characterized by the presence of the distinct carbon emission line and an essentially zero baseline. The baseline-corrected spectra were obtained by scaling the average DI 61

PAGE 74

62 water only spectrum (also shown in Figure 5-1) to the carbon-containing spectrum of interest for each data set, and then subtracting the scaled background. This process was very successful in removing the continuumflat spectral response surroundingbon emission peak. emission, as evidenced in Figure 5-1 by the the distinct, isolated car 0238240242244246248250252254256258Wavelength (nm)2 different sources of carbon. The upper four spectra represent raw data, the water spectrum (DI) as a blank. All spectra have the same scale and have been shifted vertically for clarity. 100030004000Intity2DIAA-ICP 20005000ens (a.u.)CO30-nm PSCO Figure 5-1. LIBS spectra in the vicinity of the 247.86 nm carbon atomic emission line for lower three curves are background-subtracted spectra using the deionized 60007000AA-ICP30-nm PSC

PAGE 75

63 Calibration Using Atomic Spectra The carbon signal used for this study was the calculated peak-to-base (P/B) ratio. This metric provides a precise measure of laser-induced plasma atomic emission by normalizing out fluctuations in the absolute plasma emission signal (Radziems ki et al. 1983). The reported P/B was calculated using the integrated full peak area of the 247.86 nm carbon emission line divided by the integrated signal of the scaled background spectrum over the same spectral width. Using this procedure, calibration curves were calculated over the range of carbon mass concentrations for each of the carbon analyte sources. The resulting calibration curves are presented in Figure 5-2 for all five analyte sources. All five curves display high linearity, with the slopes and corresponding regression coefficients summarized in Table 5-1. Despite the constant experimental conditions and similar range of atomic carbon mass loadings for all five species, the resulting slopes of the calibration curves reveal marked differences. Most significantly, the calibration slope for the AA-ICP standard was roughly double that of the 30-nm PS curve, and eight times that of the three gas-phase species curves. Consistent with previous findings, the three gas-phase calibration curves group very closely to one another, with the differences in slope being statistically insignificant. In addition, experiments for the three gas-phase species were repeated at iscrete data points with the carbon-containing gas stream fed through the nebulizer. respect to t responsiblesolid-phase carbon species. Clearly, the current results contradict the assumption of d This exercise did not reveal any significant changes in the resulting P/B ratios with he Figure 5-2 data, confirming that nebulization of the analyte species is not for the marked difference in analyte response between the gas-phase and

PAGE 76

64 analyte signal independencsiders both solid-phase and gt, it e on the analyte source when one con aseous-phase analyte sources, or even between varying sized solid-phase analyte sources. Prior to further analysis and discussion of the observed analyte source effecis useful to consider possible changes in the plasma itself (i.e. true matrix effects). 8 CO CO2 012302,0004,0006,0008,00010,000 46 57 CH4 AA-ICP 30-nm PS Signal (P/) CCarbon Concentration (g/m3) Figure 5-2. Calibration curves based on the 247.86 nm (C I) atomic emission line fofive carbon analyte sources investigated. Error bars represent one standard deviation. arbonB r the

PAGE 77

65 Table 5-1. Summary of the carbon calibration response and plasma conditions. Analyte Source Calibration Curve Slope Free Elec. (10 R-value Density Temp. Signal 17 cm -3 ) Plasma (K) CN (P/B) DI Blank N/A N/A 1.06 4.4% 5189 80 N/A AA-ICP 8.24E-04 5.2 % 0.999 1.03 4.6% 4973 10 22.27 30-nm PS 4.25E-04 5.8 % 0.998 1.06 4.4% 4799 100 15.03 CO 1.05E-04 9.6 % 0.997 1.04 4.5% 4952 40 6.98 CO2 1.16E-04 3.2 % 0.997 1.00 4.7% 5114 20 6.07 CH4 9.38E-05 4.2 % 0.998 1.06 4.4% 4994 70 10.18 Calibration Using Molecular Bands In addition to the carbon response as measured with the 247.86 nm atomic emission line, it is also useful to consider additional carbon-related emission bands. Specifically, if the differences in analyte response observed with the 247.86 nm line are the result of differences in the amount of atomic carbon generated within the plasma volume, then one might expect a similar trend with other carbon emission features such as those from CN. easurements of the CN violet system (B2 X2) have been widely observed in laser-induced pla, Ferioli et al. 2003). In the current study, measurements of the CN violet band were performed and analyzed using the 388.3 nm (0,0) and 387.1 nm (1,1) emission lines. M smas and have been used for quantitative analysis (Boyain-Goitia et al. 2003

PAGE 78

66 Representative background-subtracted (using DI water only) CN emission spectr a are sigur hown in F e 5-3. 050010001500 2000 370375400tensit.)2 In y (a.u AA-ICP CO 380385390395 Wavelength (nm) olidFigure 5-3. LIBS spectra in the vicinity of the CN violet system for the AA-ICP and CO 2 analyte sources. The curves have been background-subtracted using the deionized water spectrum as a blank. All spectra have the same scale. The P/B values were calculated for the two lines as described above and then averaged to quantify the CN emission, with the results summarized in Table 5-1 for all five carbon analyte sources. Analysis of the CN emission data reveals a trend that is in agreement with the analyte response observed with the carbon atomic line, namely a significant reduction in analyte signal when comparing the gas-phase species to the sphase species. The primary differences between the atomic carbon and CN emission

PAGE 79

67 results are (i) that the ratios of the various solid-phase to gas-phase P/B values were reduced by about of fa ctor of two, (ii) and that while the CN emission response of the CO2 and CO experiments was consistent, the CN response of the CH4 experiment was about 50% greater than the CO2 and CO result. In summary, the above results clearly support a significant effect of the analyte state (i.e. solid vs. gas phase) on the resulting analyte response, as measured by the corresponding analyte emission peaks. The similar response of the 247.86 nm carbon atomic emission line for all three gas species (which have markedly different C-H-O ratios), coupled with the identical free electron densities for all plasma conditions, lead to the conclusion that the observed differences in analyte response are not the result of differences within the plasma following dissociation (i.e. not emission quenching or analyte recombination effects). In contrast, the differences in analyte response are concluded to arise from differences in the amount of analyte originally dissociated to atomic carbon within the laser-induced plasma volume. To discuss the proposed reasons for such a disparity in analyte dissociation, it is first necessary to consider in more detail the nature of the solid-phase carbon and the overall role of particulates in laser-induced plasmas. Although care was taken to provide a consistent overall matrix for all experiments, including the nebulization of DI water for all cases, it is necessary to examine whether the different analyte species effected the bulk plasma. For example, large variations in plasma properties, such as electron density, resulting from the different analyte species could explain the different analyte response curves. One measure of consistent plasma formation is the continuum plasma emission. Careful examination of the Figure 5-1 data Analysis of Matrix Effects

PAGE 80

68 reveals identical background-subtracted spectra about the carbon emission peak. Thiwas also the case for the remaining analyte sources, namely CO and CH s e ma properties. However, to provide a more -and ed the of stically identical (~1x1017 cm3) for all five analyte sources as well as for the baseline DI water case (see Taental changes in the plasma state wr 4 Based on thconsistent shape of the plasma continuum emission between the DI water case and all five analyte cases, one may conclude similar plas quantitative measure of plasma consistency, the free electron densities, from the measurements of H line Stark broadening, were also evaluated. The fractional halfwidths used to reduce the Stark broadening data were those reported by Griem (1974), assuming a plasma temperature of 10,000 K, which is appropriate for the delay times laser pulse energy used. As discussed previously, the inversion of electron densities from Stark widths is rather insensitive to temperature, as related to the corresponding reducline widths. Nonetheless, to assess sources of error due to temperature uncertainty inpresent data, the measured Stark line widths were also inverted using a temperature20,000 K and the corresponding fractional half-widths. Accordingly, the uncertaintiesreported with the measured free electron densities in Table 5-1 represent the maximum error associated with the uncertainty of 10,000 K and the uncertainty associated with the measured line width (~0.02 nm). This analysis revealed that the free electron densities were stati ble 5-1), indicating no fundam ith the various carbon species. To confirm this, independent measurements of plasma temperature were made using a ratio of N 2 + lines at 375.95 nm and 391.50 nm using the method of Laux et al (2001). The results are summarized in Table 5-1. The temperatures are considerably lower than 10,000 K as the spectra were recorded at longe

PAGE 81

69 delay (40 s) to capture the molecular emission, however the relative uniformity over aspecies suggest a rather invariant plasma over the range of experimental conditions. Aerosol Loading and Size Considerations The nebulizer system used in the current experiments was investigated in detail in previous study (Hahn et al. 2001). In that earlier work, AA ICP-AES grade solutions oiron and titanium were nebulized under essentially identical conditions as the presentstudy. The resulting solid particulates were collected and analyzed using TEM. The particles produced following desolvation (i.e. particles at the LIBS sample point) were found to group about a mean particle size less than 100 nm, and as expected, were found to scale with the mass concentration of analyte in the nebulized solution. Based on this earlier study, the current nebulization of solutions of carbon as oxalic acid over the range of 500 to 5000 g-C/ml is expected to produce carbon-rich aerosols primarily in the sizrange from about 50 to 100 nm. Qualitative light scattering measurements through the LIBS sample chamber revealed the presence of a uniform aerosol cloud of similar scattering intensity for both the AA-ICP and 30-nm PS experiments. This uniform scattering was absent with the three gas-phase carbon species. The chemical nature of the submicron-sized aerosols created from nebulization of the AA-ICP solutions was ndetermined, but it is expected that the particles would be rich in oxalic acid, which takes the form of a transparent, colorless crystal in the solid phase. TEM analysis of the 30 PS suspension particles confirmed a mean size of 28 nm with a relative standard deviation of approximately 16%. Assuming that gas molecules have an effective size on the order of one Angstrom, ll a f e ot nm there was a pronounced range of sizes realized with the three groups of analyte species:

PAGE 82

70 namely ~0.1 nm for the gas phase species, ~30 nm for the polystyrene particles, and ~100nm for the nebulized AA-ICP standards. It is noted that this entire size range falls w ell below a previously est dissociation of silico udy. h a za and at the rapidly expanding plasma can be thought of as a sho ablished 2.1 m upper size limit for complete n oxide particulates in a comparable laser-induced plasma (Carranza and Hahn 2002c); hence the current results support the earlier conclusion that the varied analyte response is not the result of incomplete dissociation within the plasma stemming fromsize effects. Proposed Mechanism for Particle-Plasma Interaction As an alternative to the incomplete dissociation argument, it is proposed that a mechanism is in effect that selectively (i.e. dependent on analyte state) perturbs the analyte species during plasma formation and growth, thereby affecting the resulting amount of analyte present within the subsequent laser-induced plasma. Such a mechanism must selectively remove molecular species from within the resulting plasma volume to account for the analyte response data observed in this stIn addition, one would expect the depletion of molecular species to be combined witsecondary size effect, such that solid-phase species preferentially remain within the plasma volume in a manner that scales with particle size or mass. Previous measurements of the plasma volume using a temporal probe laser revealed that the laser-induced plasma grew from its initial breakdown volume (i.e. laser focal volume) to an effective plasma volume of about 1 mm 3 over a time-scale of about 20 ns (CarranHahn 2002b). One may consider th ckwave-like phenomenon, in which the highly energetic pressure and electron wave rapidly expands from the plasma kernel. As the plasma wave expands, molecular and

PAGE 83

71 particulate species are pushed toward the edge of the plasma volume. However, due to the many orders of magnitude difference in the mass of gas-phase species (molecules) and solid-phase species (particulates), it is expected that the efficiency at which a givspecies is carried by the plasma wave will scale inversely with particle mass. In othewords, an effective analyte slip factor will exist as the plasma wave expands. A schematic of this proposed effect is presented in Figure 5-4. en r g plasma wave and the analyte species. Large particulates or aerosols (upper whereas nano-scale analyte species (lower image pair) will tend to be depleted and electron densities are significantly lower. Analyte species at the particulate scale Figure 5-4. Schematic detailing the proposed interactions between the rapidly expandinimage pair) are shown to resist outward radial transport by the plasma wave, from the plasma core. Within such a model, analyte species at the molecular level would tend to be selectively swept out of the plasma core toward the plasma edges, where temperatures

PAGE 84

72 would be less affected by these drag forces due to inertial effects, and would therefore selectively remain in the plasma core volume, leading to more complete dissociation antherefore a greater effective analyte concentration and emission response. Some preliminary LIF imaging of LIBS plasmas using gold aerosols and particulates support this model (Nakata and Okada 1999). Based on such a working model, it is expected that a fin d ite particle size (i.e. mass) exists where the efficiency of remaining within the plasma core reaches a plateau, plasmit for the estimvolumzed is Particles bedisplay significant size effects with respect to the analyte response. that essentially the point in which all such particulates effectively remain within the core a volume. While determination of such a particle size requires further validation of the hypothesis itself, as well as additional experimental work and modeling, it is useful to consider some limiting values based on the available data. A linear analyte response was observed for silicon oxide particles over the size range 1-2.1 m, as reported previously (Carranza and Hahn 2002c). Hence one micron may be considered an upper limated particle size necessary to minimize analyte depletion within the plasma e. Based on the current study, particle size effects are concluded to exist over the 30 to 100 nm size range. Accordingly, one may reasonably conclude that the critical size for an analyte species such that preferential removal from the plasma core is minimiin the range of 0.1-1 m. Particles above this range are expected to display a linear analyte response (i.e. no particle size effect) up to the limit for complete vaporization. low this range, including down to the molecular regime, are expected to Overall, the range of data reported in the current study supports the conclusionthere is preferential accumulation and dissociation of solid-phase analyte species within

PAGE 85

73 the plasma center resulting in an enhanced analyte response, as measured by atomic andmolecular emission, as compared to gas-phase analytes. It is proposed that such an effect is realized by the interaction of the analyte species (solid or gas phase) with the expanding plasma, with relative inertial effects playing a key role. To date, published studies have explored the interaction o f aerosols in laser-induced plasmas (Lushnikov and Neginins an 1993, Schoolcraft et al. 2000, Gornushkin et al. 2004); however, the coupling of such interactions, including particle dynamics and plasma emission processes, remaongoing issue. An important outcome of this study is demonstration of the need to produce calibration schemes for use with LIBS that reflect as much as possible the physical state of the analyte species of interest.

PAGE 86

CHAPTER 6 SPATIAL CONSIDERATIONS IN PLASMA EMISSION The physical process and time-scale of particle dissociation within a laser-induced plasma and the subsequent spatial diffusion of analyte atomic species released from the particle throughout the plasma has not been well understood to date. Even recent theoretical models of laser-induced plasma behavior treat species dissociation and spectral analyte emission within a laser-induced plasma as temporally instantaneousspatially homogeneous phenomena. In an effort and to further explore these issues of basic plasm science which underlie LIBS methodologyknowledge of particle vaporization and dissociation and the presence and significance of spatial inhomogeneities in analyte emission; namely, experimental data and analysis that considers and accounts for spatial and temporal variability in atomic emission from aerosol sources within a laser-induced plasma is presented. It should be noted that the term dissociation is used herein as a general description to denote a variety of complex processes including particle melting and vaporization, fragmentation, sublimation, molecular ionization, atomic ionization, as well as concomitant heat and mass transfer. Background Emission A characteristic image of the laser induced plasma for the experimental blank is shown in Figure 6-1, below. The figure represents the on-axis projected image of the three-dimensional plasma, and gives a measure of both the optical size and intensity of the plasma. The plasma images were recorded for a series of delays and detector gate widths, as presented in Chapter 2. To first gain an understanding of the plasma a 74

PAGE 87

75 characteristics in the absence of aerosols, images were recorded in the presence of only HEPA filtered air and nebulized DI water, as described previously. 1 mm Figure 6-1. Two-dimensional on-axis projection of the laser-induced plasma in an aerosol-free environment. A series of 20 individual images were averaged together for each delay time. When compared over time, the averaged images revealed that the plasma size grew essentially linearly with time, while the plasma emission intensity, as measured by ICCD counts, decayed exponentially in time. It beyond the longest reported delayal e sis t should also be noted tha time in the study, 30s, the plasma shape was observed to deviate from a sphericstructure toward a more annular intensity profile. This atypical structure was observedbetween delay times from about 30 to 50s, after which the measured intensity profile once again took on a spherical structure. The overall intensity of the image in the rangof 30 to 100s was characterized by a marked decrease, making qualitative analy

PAGE 88

76 difficult. The primary region of analysis was therefore confined to delay times up to 30s, leaving the latter delay times and related plasma behavior as a subject for future ork. By directly measuring the cross-sectional area of the plasma image, the growth of the plasma was quantified over the range of delays, with the results shown in Figure 6-2 below. As observed in the figure, the plasma growth in size is basically linear with time, other than a brief deviation at a delay of 8s. w 2.6Intensifier Gate Delay [s] Figure 6-2. Change in observed diameter of the plasma image as a function of time delay from the end of the laser pulse. Error bars represent an estimated 0.2mm uncertainty in locating the plasma edge above intensity fluctuations. The overall plasma size in the temporal region of interest, ~3.5mm diameter, is somewhat larger than the emission-based plasma diameter of 1.7mm reported previously 2.833.23.43.63.8405101520253035Diametlasma Image [mm] P er of

PAGE 89

77 by Carranza and Hahn (2002b). However, in the earlier study, the reported volume wan effective plasma volume based on a spectral-equivalent concentration and an a priordeterm as i ined analyte mass released from the aerosol, hence the absolute agreement betwe igure 6-3. en the two different measurements is not necessarily expected. In addition to measuring the plasma size, the images were also used to calculate theemission intensity profile. A 400-pixel intensity average about the center of each plasma image was taken and the results are shown in F 0.000.200.400.600.801.001.2005101520253035Normalized plasma center intensity Delay (us) asma intensity, normalized to the highest measured value Figure 6-3. Center-averaged plrepresent standard deviation. at 2s, as a function of time delay from the end of the laser pulse. Error bars

PAGE 90

78 The plasma emission intensity, which corresponds to continuum emission, is observed to decay exponentially with time, which is consistent with the intensity profiles reported in Chapter 3. Clearly the plasma images are consistent with the spectral dproduced by collecting and dispersing the spatially integrated plasma e ata mission, although more sing the plasma imaging system, deionized water was replaced in the nebulizer with the stock suspension of nominally 2-m borosilicate microspheres described in Chapter 2. The particle suspension was adjusted to promote the creation of an appropriate particle number density at the sample point such that the LIBS plasma was expected to sample an aerosol particle with a reasonable probability, namely with a ~10-20% sample rate. Introduction of the nebulized aerosol stream to the LIBS sample chamber led to the appearance of localized, luminous bursts or hot spots within the plasma imagethe position of which were observed to be spatially random throughout the plasma. An example of such an image is shown below in Figure 6-4. That this emission was due to the spatially non-homogeneous presence of calcium atoms is strongly supported by the insertion of the narrow line filter (396.8nm) in the image collection path that corresponds to the ionized calcium atomic emission line at 396.85nm (0-25,192 cm-1). Hence the localized, luminous feature is concluded to result from the atoms released by a single borosilicate particle that is sampled within the laser-indud plasma. It is noted that such hot-spots (i.e. particlonly background structure (Figure 6-1) was visible in the absence of the aerosol stream. detailed comparisons are made below. Spatial Distribution of Atomic Emission For investigation of the plasma-particle interactions u ce e hits) were only observed with the nebulization of the particle suspension;

PAGE 91

79 corner of the plasma image. The delay time for this image was 2sanalyte emission. The image was recorded using a 3nm interference filter The diameter of each luminous burst was averaged for each time delay for a series of 30 such bursts to provide a quantitative measure of the diffusion length-scale of the calcium atoms. A plot of the measured emission burst diameter as a function of delay time is presented in Figure 6-5. Atomic diffusion coefficients were calculated from the 222 Figure 6-4. A spatially localized emission burst from an aerosol in the upper left-hand sufficiently early that background emission is comparable in strength to centered at a wavelength of 396.2nm. slope of this curve, yielding values of 0.063m/s at 2s and 0.017 m/s at 30s, with an average value of 0.04 m/s over the temporal range of 2-30s. These values suggest a diffusion time-scale, whereby individual aerosols are dissociated with time and subsequently act as expanding (via diffusion) localized atomic sources in which the 1 mm

PAGE 92

80 corresponding atomic emission is readily distinguished from the background plasma emission. Heiftje et al. (1987) reported such an approach for droplet desolvation and solute-particle vaporization processes within an inductively coupled plasma. The current diffusion rate, 0.04 m2/s, is consistent with the rates used by Heiftje; assuming a plasma temperature (T) of 15,000K and scaling by T3/2. It should be noted that microsecond-scale diffusion rates stand in stark contrast long-standing theoretical models and empirical interpretations of aerosol-derived atomic emission, in which the analyte species are assumed to uniformly diffuse through the plasma volume on a rapid, nanosecond timeRather, the current observations suggest that the rates of analyte diffusion within a ser-induced plasma have much in common with diffusion mechanisms at work in much lower tempfor temperaturlaser-induced plasmr a-center Figure 6-3, and supports established theory regarding temporal decay of LIP emission. e. -scal la erature plasmas and even flames, given appropriate scaling to account e differences; yet another implication of energy-driven dissociation within as. The plasma image sequence shown in Figure 6-6, demonstrates the diffusion of localized atomic emission with time. Steady spatial growth of the clearly defined emission burst occurs over the 2-30s range of delay times. Note also that the burst does not fill the entire plasma even at the longest time, 30s, which stands in contrast to the widely used assumption in LIBS practice that analyte emission occurs ovethe entire plasma volume for a significant portion of the temporal emission process. Another important observation is the clear decrease in burst intensity and plasmintensity for the series of images in Figure 6-6, which confirms the measurement of

PAGE 93

81 1.50 0.0005101520253035 Figure 6-5. Average diameter (N=30) of emission bursts as a function of plasma delayspectral metrics. Here, the peak-to-base (P/B) ratio is taken as the signal of interest and compared between the spectral image sets and traditional spectral data sets. Representative spectra with salient atomic and molecular line structure present within the range 380-410nm are shown below in Figure 6-7. The corresponding P/B trends over all delay times are shown below in Figure 6-8. 0.251.00ion Bsm]Delay [us] time. Error bars represent standard deviation. Correlation of Image and Spectral Data While the imaging data provide a qualitative measure of the physical emission process, it is important, in terms of the overall role of LIBS as an analytical technique, to substantiate the validity of the image data through correlation with traditional, established 0.500.751.25Emssiurt Diameter [m

PAGE 94

82 (a) (b) ( c) (d) (e) (f) delays of 2, 4, 8, 15, 22, and 30 Figure 6-6. Images (a)-(e) demonstrate the observed growth of emission bursts for time s respectively. Both spatial growth and intensity decay are observed with increasing time delay. All images reflect the same spatial scale, shown in (a). 1 mm 4 s 2 s 8 s 15 s 22 s30 s

PAGE 95

83 0100030004000380385390395400405410Spec Intensity [a.u.]Wavelength (nm)Ca II 396.85Ca II 393.37O I 394.7CN Violet &N2+ Bands30s8s4s 2000tral evolution of the primary calcium line at 396.85nm in addition to structure belonging to other ions and molecules present in the plasma. It is important to realize that the image data represents a local P/B ratio, such that normalization to the continuum emission is performed in the immediate vicinity of the emission burst. This is in contrast to the traditional spectral data, which is spatially averaged over the entire projected plasma area. Given equal apparatus response, one might assume that the localized P/B would be substantially increased as compared with mall delay timsintermedia Figure 6-7. Three reference spectra taken at 4, 8, and 30us respectively demonstrate the the spatially averaged P/B. The good agreement between the two at s e and as much as 100% greater relative P/B response from the spectral data at the te delayshowever, suggests that the spatially-resolved and spectrally

PAGE 96

84 resolved plasma signals are characterized by slightly different temporal profiles. The more general and important observation, though, is that the two trends correlate well and suggest the use of image data to selectively analyze spatial regions could be utilized to improve detection limits using LIBS. 0.000.200.400.600.801.001.2005101520253035 Normalized Spectra Normalized ImageP/B [Normalized to Value at 30us]Delay (us) Figure 6-8. Comparison of measured P/B values for image data and spectral data. Errorespective P/B values at a delay of 30 s. Rate of Aerosol Dissociation To analyze the rate at which the experimental aerosols were dissociated within the laser-induced plasma, the Ca II line at 396.85 nm was selected as the analyte peainterest for LIBS analysis. The calcium signal used for spectral analysis was the r bars represent one standard deviation. The data are normalized to the k of

PAGE 97

85 calculated peak-to-base (P/B) ratio. This metric provides a precise measure of laser-induced plasma atomic emission by normalizing out fluctuations in the absolute plasma emission signal (Carranza and Hahn 2002b). The reported P/B was calculated using theintegrated full peak-area of the 396.8n m calcium emission line divided by the integrated signal of a scaled background spectrum over the same spectral width. Using the above described calibration procedure for ICP-standard solutions of calcium at 0, 2.5, 5 and 10g/mL, a calibration curve was constructed at each delay time. The P/B values for the experimental borosilicate aerosol streams were then measured and compared with the calibration curve at each intensifier gate delay (2-30s)providing a measure of the equivalent concentration of calcium at each of the six delays. In order to proceed with calculating an experimental value for absolute mass of calcium detected for a given aerosol particle, a clear relationship between the spectrally-derived equivalent calcium concentration, the effective plasma volume, and absolute mass of calcium in a single aerosol is needed. Following the analysis of Hahn and Lunden (2000), an equivalent mass-concentration per plasma event corresponding to a single aerosol hit is related to the average mass of a single particle through the relation M f = M N 6-1 where Meq e resulting LIBS spectrum per a traditional spectral calibration curve response function, f is the frequency of particle hits per laser shot, ass of an average aerosol particle, and eq avg D is the equivalent analyte mass concentration (mass/volume) based on th M avg is the analyte m N D (particles/volume) is the number density of aerosols in the sample matrix. For dilute aerosol streams (~1 particle/plasma volume or fewer, such as those

PAGE 98

86 used in the current work) the application of Poisson statistics to estimate frequency isvalid and leads to the relation f = 1 exp(-) 6-2 where is the average number of particles per plasma volume. It is readily shown that = V 6-3 whereingle aerosass The r e nt work was taken as the sphere of rotation formed by a plasma using the directly measured P N D V P is a measure of the characteristic plasma volume. Again assuming a low particle hit-rate corresponding to dilute aerosol streams (i.e. values of much less than unity), the exp(-) term of Equation 6-2 can be approximated by 1 (to order 2 ). Combination of this approximation with Equations 6-1, 6-2, and 6-3 leads to useful relationship for calculating an experimentally determined analyte mass for a s ol particle, namely M single = M eq V P 6-4 Equation 6-4 is significant, in that it allows calculation of the absolute analyte mby multiplying the equivalent analyte concentration by the effective plasma volumeequivalent analyte concentration is readily calculated by using the resulting spectrum (oaverage particle hit spectrum) in combination with a traditional calibration curve. For th current study, calibration curves were constructed by nebulizing known concentrations of calcium standards (SPEX ICP-AES solutions) over a range of resulting calcium concentrations. The equivalent calcium concentration values where then calculated using the ensemble-averaged spectrum corresponding to a number (~200) of particle hits recorded for the dilute aerosol stream of the borosilicate particles. The equivalent concentrations were then multiplied by the effective plasma volume, which in the curre

PAGE 99

87 optical diameters as given in Figure 6-2. The resulting product provides a direct measuof the absolute calcium mass re re leased as a function of time, which is presented in Figure -9. 6 01002 600 00l M 300400500ss of Calcm etected [f 05101520253035Delay Time [s] Figure 6-9. Mass of calcium analyte detected using LIBS on a dilute aerosol matrix ofincrease with timethe subsequent plateau corresponds to a marked decrease The asymptote in the curve of Figure 6-9, taken to represent the maximum num TotaaiuDg] borosilicate glass microspheres. The first four data points show a logarithmic in plasma energy. The error bars represent standard deviation. ber of calcium atoms within the plasma volume, indicates that approximately 400fg of calcium are present in each aerosol. Using an average aerosol diameter of 2m, 0.7 m standard deviation, and density of 2.5g/cc (as reported by the manufacturer), ~400fg

PAGE 100

88 translates to 2% calcium by mass. No independent measurements of calcium concentration in the aerosols were performed, but the 2% value is in the range reported for bo rosilicate glasses industry-wide. The plateau in Figure 6-9 suggests that the plasma has fully dissociated the calcium aerosol at approximately 15s from the end of the laser pulse, and only partial dissociation has occurred before this time. Hence, the concept of a rate-limit with regards to particle dissociation within a laser-induced plasma, as first put forth by Carranza and Hahn (2002c), is strongly supported. In that work, the authors observe that, while several orders of magnitude more energy exists in the focused laser pulse than exists in the bond energy of a given solid particulate, the measurable spectral signal (directly proportional to atomically dissociated analyte mass in the plasma volume) experiences a plateau as particle size reaches a critical value of about 2m. It is proposed in the current study that dissociation processes may plateau at some critical value of plasma energy or electron density (see Chapter 3 for full treatment of this relationship), which corresponds directly to a critical time since the plasma decays at a known, measurable rate. Note in Figure 6-3 that the plasma emission at 15s has decayed to less than 10% of its value at 2sthe Figure 6-9 occurs at nearly the same delay r 3 (see Figuradiative transfer, hence m thermal transport (via convective/conductive processes) and free electron interaction change in concavity for both Figure 6-3 and time. This is consistent with the rapid decrease in electron density as reported in Chaptere 3-3). Such behavior indicates that the plasma has lost significant energy via ay no longer be energetic enough to dissociate additional mass (i.e. atoms) at the appreciable rate necessary for complete dissociation at these timescales. This observation could give rise to an interesting theoretical model whereby both

PAGE 101

89 (analogous to e-beam sputtering) effects are considered in tandem with regaparticle dissociation within a laser-induced plasma. One might consider the resulting ratof energy transfer via these two mechan rd to e isms, which might explain the time scales obsery er ved with the current experimental data. Such calculations are recommended for future work. Summarizing the observations of Carranza and Hahn (2002c), if the aerosols entrained in the plasma volume are sufficiently small to be fully dissociated within the period of high reactivity, then a linear analyte emission response will be realized, independent of the aerosol number density. If, however, the aerosols are too large, full dissociation will not occur, again irrespective of aerosol number density, indicating a limit to the mass or particle size that can be fully dissociated and ionized by a single plasma event. The emission response will then level off at a much lessened rate of releasean effective saturation with respect to molecular and atomic dissociation of analyte atoms from the bulk aerosol particle. The current study suggests a temporal analog to the findings of the prior work and interpretation of the results leads to a corollary for the earlier hypothesis which is as follows: aerosols entrained in the plasma volume that are not sufficiently small to be fulldissociated in the time that the plasma volume is sufficiently energetic shown in this study to correspond approximately to the time interval 0-15s following the laser pulseexperience an asymptotic flattening (or plateau) of analyte emission response after that time has elapsed, indicating that the dissociation event has given way to a much slowprocess (possibly a thermal process). Establishing this temporal threshold across variations in LIBS system parameters is a subject of future study.

PAGE 102

90 Implications for LIBS The results above give a strong indication that atomic emission from aerosols within the laser-induced plasma is, for the most part, a spatially non-uniform eventis an important consideration in the design and analysis of experiments employing LIBfor aerosol di This S agnostics. Furthermore, while optically measurable diffusion of atomic specie of s. s with r ity and temperature) are significant enough to pro ions s from aerosols occurs within the laser-induced plasma during the time periodspectral emission, sources of atomic analyte seemingly fail to occupy the entire plasma volume over the range of intensifier gate delay times typically used in LIBS measurements. This is an important result as it implies an ever-present spatial factor influencing the ability to precisely collect analyte emission from laser-induced plasmaThe current study also suggests a discernable time scale for the dissociation of aerosols within a laser induced plasma. A plateau of analyte concentration coincideasymptotic energy decay within the plasma yielding a characteristic time window ovewhich the plasma energetics (i.e. plasma dens mote rapid dissociation of entrained mass. Rate-limited particle dissociation shouldbe further explored as the implications exist for plasmas from all sources and interactwith all forms of matter.

PAGE 103

CHAPTER 7 obser The fundamental change from an absorbing, optically thick plasma to a non-absorbing plasma during the first few hundred nanoseconds is important with respect to the radiative transfer within and from the plasma, and is therefore expected to play a role in the interactions between the laser-induced plasma and incorporated particles. The shot-to-shot precision of LIBS-based spectral analysis was shown, in the context of laser-induced plasma temporal and spectral characteristics, to be adversely affected by the presence of aerosols in the plasma volumean important consideration for researchers using spectroscopy for gas-phase and aerosol-phase samples. CONCLUSIONS AND FUTURE WORK The preceding chapters bring together a vast quantity of data pertaining to the properties of laser-induced plasmas and their subsequent interactions with concomitant aerosols. The following represents an attempt to bring the primary salient conclusions together along with ideas for future work generated by this dissertation. Conclusions Below, in chapter order, are the primary conclusions drawn from the data and vations of the studies contained herein, along with brief comments as to the perceived importance of these conclusions to plasma science generally and the LIBS technique specifically: Temporally resolved intensity measurements have shown that the laser-induced plasma is characterized by significant transient changes in electron density and plasma absorptivity during the first few hundred nanoseconds, yielding a free-free absorption cross-section of 9.5x10 -18 cm 2 for 532nm radiation. Resolution of these temporal gradients sheds light on the relevant time-scales of observable, physical plasma behavior including interaction with other sources of radiation and with entrained matter. 91

PAGE 104

92 Seeding of the pulsed laser cavity to produce a single-mode structure was observed to offer a statistical advantage with regard to the precision of temporal onset of breakdown of the laser induced plasma, but not on the analyte signal precision. Calibration experiments using a standard carefully controlled to yield comparable minvestigatedrevse aerosol species s. s compared to gas-phase analytes. It is proposed that such an effect is realized by the interaction of the analyte species (solid or g the expanding plasma, with relative rom -ion eak-to-base measurements as compared with spatially integrated ne. S ass. Proposal of Future Work resultfollow from the present study: t l transition in the process governing plasma absorptivity at early times from free-free absorption to photoionization, which would explain some degree of wavelength functionality. LIBS systemloading and ba olar analyte ckground matrix across all species ealed a clear trend in which small, solid-pha exhibited a stronger calibration response per mole of analyte than did gas phasespecies. This result challenges a widely used assumption that the slope of the calibration curve remains unaffected by size or phase differences in analyte specie Differences in the calibration response of analyte species based on size and phase support the conclusion that there is preferential accumulation and dissociation of solid-phase analyte species within the plasma center resulting in an enhanced analyte response, as measured by atomic and molecular emission, a as phase) with inertial effects playing a key role. Imaging and spectral analysis of plasma events reveal that atomic emission faerosols within the laser-induced plasma is, for the most part, a spatially nonuniform event. This observation is of significance in the improvement of detectlimits for LIBS, as spatially selective spectral analysis would theoretically offerimprovements in pmeasurements alo The combination of imaging and spectroscopy applied in the current work to LIBfor aerosol analysis also suggests a discernable time scale for the dissociation of aerosols within a laser induced plasma, yielding a diffusion coefficient of 0.04m 2 /s.A plateau of analyte concentration coincides with asymptotic energy decay within the plasma, a characteristic corresponding to the time window over which the plasma is energetic enough to produce rapid dissociation of entrained aerosol m As with any academic study, the results of this work and implications of these s suggest additional study. Below are some recommendations for future work that Further exploration of the observed wavelength dependency of the free-free absorption cross-section, which is not predicted theoretically for the inverse Bremsstrahlung process alone, is warranted. Current thought and modeling suggesa tempora

PAGE 105

93 Comprehensive theoretical modeling for the dissociation and subsequent diffusion of an aerosol particle as a result of interaction with a laser-induced plasma woulhelp to substantiate the data and physical model put forth in the LIBS calibration d study. Additional experimentation, using different analyte species, analyte sources, d s to plasma. Based on the current results, there is an indication that electron ound to be characteristic of spectral emission from aerosols. This would clearly be the optimal system with which to study al tensifier gate delay times (30-50s) over which continuum emission is readily detectable (~0-100s). At the very least, researchers working dissociation process suggested by the current study. It is anticipated from the present observations thathich coincides with earlier ce whereas larger aerosols and particulates, which do not fully dissociate within the m aerosols (~15s for the present study). and laser configurations would also aid in solidifying the knowledge of phase ansize-based variations in calibrating LIBS systems. A second theoretical model to capture the simultaneous and combined effects of conductive and convective heat transfer and electron bombardment with regardenergy transfer considerations during particle dissociation within the laser-induced bombardment may be important with respect to the particle dissociation processes,as temporal gradients in electron density are less steep than those in temperature. Successfully build a system to repeatably execute the optical trapping of a single microscopic aerosol. Such a system would enable further investigation, on a singleparticle basis, of what the results in this study suggest on a statistical basis regarding spatial inhomogeneities f spatial variations in aerosol emission from laser-induced plasmas, including evaluation of spatial trends and differences in emission from ions versus neutratomic species. Further investigation of the annular structure of continuum emission observed for the middle-range of in with LIBS in this temporal range should be aware of the phenomenon, whether or not it is shown to be a universal effect (and not just an artifact of the system used)or have any bearing on signal response. Explore a range of aerosol particle sizes, both smaller and larger than the 2m diameter aerosols used at present, in order to evaluate the nature of the rate-limited aerosols smaller than 2m (w published data on the upper size limit for full particle dissociation) will experienfull dissociation (plateau in mass removal rate) at proportionally earlier times, laser-induced plasma, will experience little temporal deviation from the mass-removal plateau of the 2

PAGE 106

APPENDIX A SUMMARY OF ATOMIC AND MOLECUL AR LINES USED ble L Ta A-1. Summary of atomic lines used in spectral analysis. ine Wavelength (nm)Upper State (cm -1 )Lower State (cm -1 ) g k A ki (106 s-1) C I 247.86 61,982 21,648 102 N I 491.49 106,478 86,137 1.62 N I 493.50 106,478 86,221 3.52 H () 656.29 97,492 82,259 388 H () 486.13 102,824 82,259 124 Si I 288.16 40,992 6,299 567 C a II 393.37 25,414 0 588 C a II 396.85 25,192 0 280 Table A-2. Summary of molecular lines used in spectral analysis. Line Wavelength (nm) N 2 + 375.95 N 2 + 391.50 CN 388.30 CN 387.10 94

PAGE 107

APPENDIX B VERIFICATION OF IMAGING SYSTEM In order to provide a measure of optical resolution two gri ds were constructed with 1 mm2 and mm spacing. Figure B-1. Simple grid patterns for optical system evaluation; (a) represents 1 mm spacing and (b) 2 mmcing. Pae to-scale but have been enlarged for clarity. The test grids above were d and ao microscope slides, placed in the focal plane, back-illuminated, and imaged we iCCD within the particle trapping experimental setup. The known cell spacing was measured in pixels and a resolution of 5.3 microns/pixel was found for both grids. Note that neither image displays distortion or bending of lines and the response appears linear over the entire CCD array. spa tterns ar printe ttached t ith th 95

PAGE 108

96 Figure B-2. Image of 1 mm test grid. F igure B-3. Image of 2 mm test grid.

PAGE 109

APPENDIX C TABLE OF 1/2 VALUES FOR FREE-ELECTRON DENSITY CALCULATIONS The following table was adapted from Griem (1974). Each given temperature increment and value of free-electron number density corresponds to an 1/2 value per equation 3-1. The units of 1/2 are Angstroms per c.g.s. field strength. Table C-1. Values of 1/2 for calculation of electron density. T = 10,000 K Ne (cm-3) (A/Field) 1.00E+15 0.00777 1.00E+16 0.0134 1.00E+17 0.0186 1.00E+18 0.0215 T = 20,000 K Ne (cm-3) (A/Field) 1.00E+15 0.00601 1.00E+16 0.0114 1.00E+17 0.0175 1.00E+18 0.0226 1.00E+19 0.0235 T = 30,000 K Ne (cm-3) (A/Field) 1.00E+15 0.00498 1.00E+16 0.01 1.00E+17 0.0166 1.00E+18 0.0225 1.00E+19 0.0257 T = 40,000 K Ne (cm-3) (A/Field) 1.00E+15 0.0045 1.00E+16 0.00922 1.00E+17 0.0158 1.00E+18 0.0223 1.00E+19 0.0269 97

PAGE 110

APPENDIX D IMAGES OF NON-HOMOGENEOUS SPATIAL EMISSION The following images were taken during the course of data collection and are presented here to give an indication of the locus of possible effects when imaging emission from aerosol-derived analyte species. When aerosol loadings are increased, and were observed over the ran of delay tim reported. Additionally, certain aerosols appeared to leave a the or trad them; resembling a comet and possibly suggesting that a measurable component of bulk aerosol motion, in addition to diffusion, is present due to the a formation. multiple emission bursts within the same plasma are inevitable ge es rmal streak ce behin force s wrought by plasm Figure D-1. A strong, central burst is accompanied by a less intense burst in the upper-t corne lef r. 98

PAGE 111

99 Also, due to the spatially random distribution of aerosol location, bursts near the plasma center were no more common than those near the plasma edges, however a couple are shown, such as the one above, as they did ocshould also b be higher than t than cur but are not represented in the text. It e noted that, because temperatures near the plasma-center tend to hose near the plasma edges, the P/B response of central bursts tend to be lowerbursts at or near the plasma edge. Figure D-2. An apparent thermal emission streak, giving the impression of bulk aerosol motion from the center towards the edge of the plasma.

PAGE 112

100 Figure D-3. A double burst grouped in the central-upper left portion of a relatively weak plasma. Figure D-4. Two emission bursts beyond the outer edges of a strong background plasma.

PAGE 113

101 Figure D-5. Four bursts, two very strong, and two fairly weak, within a relatively weak plasma.

PAGE 114

LIST OF REFERENCES Aguilera J.A., Aragon C., Bengoechea J., Applied Optics, 42 (2003) 5938-5946. Bindhu C.V., Harilal S.S., Tillack M.S., Najmabadi F., Gaeris A.C., Applied Spectroscopy, 58 (6) (2004), 719-726. Bindhu C.V., Harilal S.S., Tillack M.S., Najmabadi F., Gaeris A.C., Journal of Applied Physics, 94 (12) (2003), 7204-7207. Borghese A. and Merola S.S., Applied Optics, 37 (1998) 3977-3983. Boyain-Goitia A.R., Beddows D.C.S., Griffiths B.C., Telle H.H., Applied Optics, 42 (2003), 6119-6132. Carranza J.E., Iida K., and Hahn D.W., Applied Optics 42 (2003) 6022-6028. Carranza J.E., Hahn D.W., Spectrochimica Acta Part B 57 (2002a) 779-790. -1539Carranza J.E., Hahn D.W., Analytical Chemistry 74 (2002c) 5450-5454. Carranza J.E., Fisher B.T., Yoder G.D., and Hahn D.W., Spectrochimica Acta Part B, 56 (2001) 851-864. Dudragne L., Adam Ph., Amouroux J., Applied Spectroscopy, 52 (1998) 1321-1327. Essien M., Radziemski L.J., and Sneddon J., Journal of Analytical Atomic Spectrometry, 3 (1988) 985-988. Ferioli F., Puzinauskas P.V., Buckley S.G., Applied Spectroscopy 57 (2003) 1183-1189. Gleason R.L. and Hahn D.W., Spectrochimica Acta Part B,56, (2001), 419-430. Gornushkin I.B., Kazakov A.Y., Omenetto N., Smith B.W.,Winefordner J.D., Spectrochimica Acta Part B 59 (2004) 401-418. Griem H.R., Principles of Plasma Spectroscopy, Cambridge University Press, UK 1997, Chapter 5. Griem H.R., Spectral Line Broadening by Plasmas, Academic Press, New York 1974. Carranza J.E., Hahn D.W., Journal of Analytical Atomic Spectroscopy, 17 (2002b) 1534. 102

PAGE 115

103 Hahn D.W., Applied Physics Letters, 72 (1998) 2960-2962. Hahn D.W., Carranza J.E., Arsenault G.R., Johnsen H.A., Henken K.R., Review of Scientific Instruments, 72 (2001) 3706-3713. Hahn D.W. and Lunden M.M., Aerosol Science and Technology, 33 (2000) 30-48. Heiftje G.M., Miller R.M., Pak Y., Wittig E.P., Analytical Chemistry, 59 (1987) 2861-9 (1996) Itina T.E., Hermann J., Delaporte Ph., Sentis M., Applied Surfactant Science, 208-209 Quantitative Spectroscopy & Radiative Transfer 68 (2001) 473-482. Lee W.B.,Wu J.Y., Sneddon J., Applied Spectroscopy Reviews 39 (2004) 27-97. Lithgow G.A. and Buckley S.G., Applied Physics Letters, 87 (2005) 1-3. 38 (2004) 3319-3328. Lushnikov A.A., Negin A.E., Journal of Aerosol Science 24 (1993) 707-735. Neuhauser R.E., Panne U., Niessner R., and Wilbring P., Fresenius Journal of Analytical (2000) 1805-1816. Oks Eansfer, 65:405-414 (2000). Parigger C., Lewis J.W.L., Plemmons D.H., Journal of Quantum Spectroscopy and 2872 Ho J.R., Grigoropoulous C.P., Humphrey J.A.C., Journal of Applied Physics 77205-7215. Hybl J.D., Lithgow G.A., Buckley S.G., Applied Spectroscopy, 57 (2003) 1207-1215. (2003) 27-32. Kerker M., The Scattering of Light and Other Electromagnetic Radiation, Academic, New York (1969). Laux C.O., Gessman R.J., Kruger C.H., Roux F., Michaud F., Davis S.P., J. of Lithgow G.A., Robinson A.L., Buckley S.G., Atmospheric Environment, Morel S., Leone N., Adam P., Amouroux J., Applied Optics 42 (2003) 6184-6191. Nakata Y., Okada T., Appl. Phys. A, 69 (Suppl.) (1999) S275-S278. Chemistry, 364 (1999) 720-726. Nunez M.H., Cavalli P., Petrucci G., and Omenetto N., Applied Spectroscopy. 54 ., Journal of Quantum Spectroscopy Radiative Tr Parigger C.G., Plemmons D.H., Oks E., Applied Optics, 42 (2003) 5992-6000. Radiative Transfer, 53 (1995) 249-255.

PAGE 116

104 Radziemski L.J., Microchemical Journal, 50 (1994) 218-234. 5 Analytical Chemistry, 27 (1997) 257-290. Samu., DeLucia F.C. McNesby K.L., Miziolek A.W., Applied Optics, 42 (2003) 6205-6209. Schechter I., Reviews in Analytical Chemistry, 16 (1997) 173-298. (2000) 5143-5150. Simeonsson J.D. and Miziolek A.W., Applied Physics B, 59 (1994) 1-9. Powder y, 55, (2001)1455-1461. Villagniz M., Sobral H., Camps E., IEEE Transactions Plasma Science, 29 (2001) 613-616. Weyl G.M. Physics of Laser-Induced Breakdown: An Update, in Laser-Induced Plasmas l Zhang H., Yueh F.Y., and Singh J.P., Applied Optics, 38 (1999) 1459-1466. Radziemski L.J., Loree T.R., Cremers D.A., Hoffman N.M., Analytical Chemistry, 5(1983) 1246-1252. Rusak D.A., Castle B.C., Smith B.W., and Winefordner J.D., Critical Reviews in als A.C Schoolcraft T.A., Constable G.S., Zhigilei L.V., Garrison B.J., Analytical Chemistry 72 Singh J.P., Yueh F.Y., Zhang H.S., and Cook R.L., Process and Control Quality, 10 (1997) 247-258. Smith B.W., Hahn D.W., Gibb E., Gornushkin I., and Winefordner J.D., KONAand Particle N 0 19 (2001) 25-33. Sneddon J. and Lee Y.-I., Analytical Letters, 32 (1999) 2143-2162. Tran M., Smith B.W., Hahn D.W., Winefordner J.D., Applied Spectroscop ran-Mu and Applications, edited by L.J. Radziemski and D.A. Cremers (1989), Marce Dekker, New York, Chapter 1. Yalcin S., Crosley D.R., Smith G.P., and Faris G.W., Applied Physics B, 68 (1999) 121-130.

PAGE 117

BIOGRAPHICAL SKETCH West Palm Beach where he completed grade school. He received his Bachelor of Science degree (1999) and Master of Science degree (2001), both in mechanical enginity of Florida. Vincent Hohreiter was born in Fort Lauderdale, FL, and grew up in neighboring eering, from the Univers 105


Permanent Link: http://ufdc.ufl.edu/UFE0012403/00001

Material Information

Title: Investigation of Physical and Spectral Characteristics of Laser-Induced Plasmas: Applications to Laser-Induced Breakdown Spectroscopy for Analysis of Aerosols and Single Particles
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0012403:00001

Permanent Link: http://ufdc.ufl.edu/UFE0012403/00001

Material Information

Title: Investigation of Physical and Spectral Characteristics of Laser-Induced Plasmas: Applications to Laser-Induced Breakdown Spectroscopy for Analysis of Aerosols and Single Particles
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0012403:00001


This item has the following downloads:


Full Text












INVESTIGATION OF PHYSICAL AND SPECTRAL CHARACTERISTICS OF
LASER-INDUCED PLASMAS: APPLICATIONS TO LASER-INDUCED
BREAKDOWN SPECTROSCOPY FOR ANALYSIS OF AEROSOLS AND SINGLE
PARTICLES















By

VINCENT PAUL HOHREITER


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Vincent Paul Hohreiter















ACKNOWLEDGMENTS

I would like to briefly and generally acknowledge everyone-family, friends,

labmates, colleagues-who helped me make this dissertation happen. Specifically I

would like to thank my parents, Vaughn and Christine Hohreiter, who-beyond their

constant love and support-always accepted, fostered, and encouraged my academic

progress and misgivings alike. Additionally I owe a debt of gratitude to my advisor, Dr.

David Hahn, without whose rigorous academic lead and unwavering personal support I

most likely would have abandoned doctoral study years ago. Finally, I would like to

thank my wife, Dr. Aline Gubrium, who has inspired me with both love and a tenaciously

organized and diligent work ethic.
















TABLE OF CONTENTS



A C K N O W L E D G M E N T S ......... .................................................................................... iii

LIST OF TABLES ............. ..... ........................ ......................... vii

LIST OF FIGURES ............. ................... ............ .......... ............... .. viii

ABSTRACT .............. .................. .......... .............. xi

CHAPTER

1 IN TR OD U CTION ............................................... .. ......................... ..

T he L aser-Induced Plasm a ........................................ ....................................... 1
Plasm a Form action .............................................. .... .... ................ .2
Spatial and Temporal Evolution............................ ..... .. .................3
The Laser-Induced Breakdown Spectroscopy (LIBS) Technique.............................4
Temporal Properties of Laser-Induced Plasmas for LIBS Analysis............................6
Stability of Laser-Induced Plasm as ....................... ................. ........................... 8
L IB S and C alibration .............. .... .................... .. ........... ............. .. ............ .. 10
Spatial Considerations in Plasma Emission.............................................. ..........13

2 EXPERIM ENTAL M ETHODS ........................................ ........................... 16

The LIBS Experimental Apparatus ........... .. .................. ......... ............... 16
D eliv ery O ptics............................................... ................ 16
Extraction Optics ................. .. ........... ............ .. ............. 17
Sy stem Sy nth esis ............................................. ................. 18
Plasma Transmission Study ................................. .... ............... 19
Transmission M easurem ents .............. ... .. .......................................20
Electron Density and Plasma Continuum Measurements ..................................21
L aser Shot Stability Stu dy ............................................................... .....................22
C arbon C alibration Study .................................................. .............. ............... 24
Solid Phase Carbon Experim ents ............................................. ............... 25
Gas Phase Carbon Experim ents...................................... ......................... 27
LIB S D ata Collection ........ ............ ...... .......... ................ ......... 28
Plasm a-A erosol Interaction Study ........................................ ......................... 29
The Im aging System .................. ........................ .... ........ .. ........ .... 30
The Spectroscopy System ........................................................ ................ 32









3 PLASMA TRANSMISSION PROPERTIES .................................. .................34

Tim e-R resolved Plasm a Transm ission..................................... ..............................34
Time-Resolved Continuum M easurements ..................................... .................36
Electron D ensity M easurem ents ...................................................... .... ........... 38
Plasm a A bsorption Cross-Section ........................................ ......................... 40
M odeling Plasm a Transm mission ........................................ ....... ............... 40
Plasm a A bsorption............... .............................. .. .......... ............. 42
Extrapolation for Sm all Tim e Scales........................................ ............... 43

4 LIB S PLA SM A STA BILITY .......................................................... ............... 46

E effects of L aser C avity Seeding ..................................................................... .. .... 46
Effects on Shot-to-Shot Precision ............................................ ............... 48
Temporal Effects on the Plasma Formation Process................ ..................52
Effects of A erosol Loading................................................................. ....................56
Effects on Shot-to-Shot Precision ............................................ ............... 57
Temporal Effects on the Plasma Formation Process............................... ....59

5 L S C A L IB R A T IO N ............................................... ........................... ................6 1

Raw and Baseline-Subtracted Spectra................................................................ 61
Calibration U sing Atom ic Spectra................................................... ...... ......... 63
Calibration U sing M olecular B ands ........................................ ....................... 65
A analysis of M atrix Effects............. ................... ........................... ............... 67
Aerosol Loading and Size Considerations..................... ... ....................... 69
Proposed Mechanism for Particle-Plasma Interaction....................................70

6 SPATIAL CONSIDERATIONS IN PLASMA EMISSION................ ................. 74

B background E m mission ....................................................................... .............. 74
Spatial Distribution of Atomic Emission .......................................... 78
Correlation of Image and Spectral Data .............. .......... .................. .............. .81
R ate of A erosol D association ............................ ........... ................................... 84
Im plications for L IB S .............................................. .................... 90

7 CONCLUSIONS AND FUTURE WORK ............... ......................................91

C o n clu sio n s ....................................................... ................ 9 1
Proposal of Future W ork .......................................................................... 92

APPENDIX

A SUMMARY OF ATOMIC AND MOLECULAR LINES USED ...........................94

B VERIFICATION OF IMAGING SYSTEM.................................. ...............95





v









C TABLE OF 01/2 VALUES FOR FREE-ELECTRON DENSITY
C A L C U L A T IO N S ........................................................................... .....................9 7

D IMAGES OF NON-HOMOGENEOUS SPATIAL EMISSION.............................98

L IST O F R E F E R E N C E S ...................................................................... ..................... 102

BIOGRAPHICAL SKETCH ............................................................. ...............105
















LIST OF TABLES


Tablege

4-1 Ratios of the area-integrated spectral peaks (peak-to-base and peak-to-peak) for
seeded and unseeded laser operation ................................................ ............... 48

5-1 Summary of the carbon calibration response and plasma conditions. ...................65

A-i Summary of atomic lines used in spectral analysis ..........................................94

A-2 Summary of molecular lines used in spectral analysis ................................. 94

C-l Values of al/2 for calculation of electron density ..................................................97
















LIST OF FIGURES


Figure page

2-1 Optical setup for delivery of laser energy to the sample volume resulting in the
creation of a plasm a........... ..................................... ..... .... .. ............ 16

2-2 Optical setup for extraction of plasma emission from sample volume ................. 17

2-3 B asic L IB S system .......................... .................. ................... .. ..... 18

2-4 LIBS system incorporating second laser for plasma transmission measurements...19

2-5 LIBS system incorporating fast detectors and a digital oscilloscope for
additional beam m easurem ents. ........................................ ......................... 22

2-6 Nebulization system coupled with a LIBS system (only Nd:YAG shown) for
aerosol analy sis. .................................................... ................. 24

2-7 Intensified CCD for imaging combined with a LIBS system for sampling
aerosols w within the laser-induced plasm a ...................................... ..................29

2-8 Plasma image as visualized from a co-axial orientation to the laser beam ............30

3-1 Plasma transmission as a function of time following plasma initiation using 532
and 1064 nm probe laser beam s. ........................................................ ........... 35

3-2 Plasma continuum emission (20 nm bandwidth) for three spectral regions
centered at 270, 400 and 530 nm ...................................... .......................... 37

3-3 Free electron number density as a function of time following plasma initiation,
as measured using the Stark broadened Ha line width...................... ..............39

3-4 Measured absorption coefficient for the 532 nm probe laser beam as a function
of the corresponding measured free electron density............... .... .............. 42

3-5 Predicted electron density values as a function of time following plasma
initiation ............................................................................44

4-1 Sample spectrum showing atomic emission lines from the plasma spark recorded
in am bient air........................................................................ ... ...... ...... 47









4-2 N491.5/N493.5 and N491.5/C247.9 peak-to-peak atomic emission ratios for an 800
sh ot sequ en ce............................. .................................................... ............... 4 9

4-3 N491.5 and C247.9 peak-to-base atomic emission ratios for an 800 shot sequence....52

4-4 Representative waveforms for the incident laser pulse and the transmitted laser
pulse following laser-induced breakdown.........................................................53

4-5 Cumulative probability distribution plot for both seeded and unseeded laser
operation ............................................................................................. .56

4-6 N491.5/N493.5 peak-to-peak atomic emission ratios for an 800 shot sequence. ..........58

5-1 LIBS spectra in the vicinity of the 247.86 nm carbon atomic emission line for
different sources of carbon. ........................................ .......................................62

5-2 Calibration curves based on the 247.86 nm (C I) atomic emission line for the
five carbon analyte sources investigated ....................................... ............... 64

5-3 LIBS spectra in the vicinity of the CN violet system for the AA-ICP and CO2
analyte sources. .......................................................................66

5-4 Schematic detailing the proposed interactions between the rapidly expanding
plasma wave and the analyte species. ........................................... ............... 71

6-1 Two-dimensional on-axis projection of the laser-induced plasma in an aerosol-
free environm ent. ............................................. ................... .... ...... 75

6-2 Change in observed diameter of the plasma image as a function of time delay
from the end of the laser pulse. ............................................................................ 76

6-3 Center-averaged plasma intensity, normalized to the highest measured value at
2[ts, as a function of time delay from the end of the laser pulse............................77

6-4 A spatially localized emission burst from an aerosol in the upper left-hand
corner of the plasma image. ................. ......... .................... 79

6-5 Average diameter (N=30) of emission bursts as a function of plasma delay time...81

6-6 Images (a)-(e) demonstrate the observed growth of emission bursts for time
delays of 2, 4, 8, 15, 22, and 30 ts respectively ............. ......................................82

6-7 Three reference spectra taken at 4, 8, and 30us respectively .................................83

6-8 Comparison of measured P/B values for image data and spectral data. ..................84

6-9 Mass of calcium analyte detected using LIBS on a dilute aerosol matrix of
borosilicate glass m icrospheres ........................................ .......................... 87









B-l Simple grid patterns for optical system evaluation...............................................95

B -2 Im age of 1 m m test grid. ........................................ ............................................96

B -3 Im age of 2 m m test grid. ........................................ ............................................96

D-1 A strong, central burst is accompanied by a less intense burst in the upper-left
c o rn e r ...................................... .................................... ................ 9 8

D-2 An apparent therm al emission streak ............................................ ............... 99

D-3 A double burst grouped in the central-upper left portion of a relatively weak
plasm a........................................................ ................... ... .... ... ... 100

D-4 Two emission bursts beyond the outer edges of a strong background plasma. .....100

D-5 Four bursts, two very strong, and two fairly weak, within a relatively weak
plasm a ......... ......... ......... ..................................... ........................... 10 1















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

INVESTIGATION OF PHYSICAL AND SPECTRAL CHARACTERISTICS OF
LASER-INDUCED PLASMAS: APPLICATIONS TO LASER-INDUCED
BREAKDOWN SPECTROSCOPY FOR ANALYSIS OF AEROSOLS AND SINGLE
PARTICLES

By

Vincent Paul Hohreiter

December 2005

Chair: David W. Hahn
Major Department: Mechanical and Aerospace Engineering

The physical and optical characteristics of laser-induced plasmas (LIP) and the

precision of measurements using laser-induced breakdown spectroscopy (LIBS) for

aerosol and gas phase species are explored. In a series of coherent experiments

properties of the LIP and the temporal, spatial, and energetic variations of interactions

with entrained particles (aerosols) are evaluated from both a standpoint of basic plasma

science and applied atomic spectroscopy.

First, the evolution of a LIP is characterized in terms of its temporally-resolved

spectral absorptivity, spectral emissivity, and free electron density during the first 500

nanoseconds. Transmission measurements reveal near opacity of the LIP at early times

(10-50 ns) and essential transparency at longer times (500 ns). The fundamental change

from an absorbing plasma to a non-absorbing plasma during this period is important with

respect to radiative energy transfer.









Second, spectral and temporal effects of laser cavity seeding and aerosol presence

in the LIP volume are investigated. Improvements in the temporal stability of laser-

induced breakdown initiation were observed with laser cavity seeding. Greater shot-to-

shot analyte precision, as measured by a nearly 60% reduction in relative standard

deviation, was realized in the absence of concomitant aerosols from the analyte sample

stream.

Third, the effects of analyte phase on the calibration response for LIBS were

investigated. Significant differences in the atomic emission signal from carbon were

observed when comparing calibration streams of gas-phase and submicron-sized solid-

phase species. The resulting calibration curves demonstrated large inter-species

variations over a comparable range of atomic carbon concentrations, challenging a widely

held assumption that dissociation of constituent species within a LIP results in

independence of the analyte atomic emission signal from analyte source.

Finally, the LIBS technique is used in conjunction with imaging to measure a rate

of dissociation for aerosols entrained in a LIP. Additionally, the time scales of

background plasma emission, spatial distribution of atomic emission, and diffusion of

atomic species within a LIP are explored. Atomic emission from the plasma volume is

demonstrated to occur in spatially localized bursts. The total amount of analyte detected

is shown to plateau at times coinciding with significant energy decay of the plasma.














CHAPTER 1
INTRODUCTION

The Laser-Induced Plasma

The laser-induced plasma is an almost completely ionized gas produced by focused

laser irradiance such that the electric field generated at the focal point exceeds the

dielctric strength of the medium occupying the focal volume (Weyl 1989). The plasma,

also known in the literature as a spark or breakdown, is qualitatively characterized by

directionally uniform emission of radiation and is often accompanied by an audible shock

wave. The luminous emission, some fraction often falling in the visible portion of the

electromagnetic spectrum, is valuable as it corresponds directly and uniquely to the

atomic constituents of the sample medium present within the plasma volume. Laser-

induced plasmas may be formed on the surface of solids and inside liquid or gaseous

media; however, the studies presented herein deal solely with plasmas in a gas phase

medium and, hence, further discussion will be limited to these.

A region of large energy density corresponding to the laser focal spot is required to

create a plasma in a gaseous medium. This generally requires a pulsed laser source and

produces temperatures that range from several thousand to tens of thousands of Kelvin in

the immediate vacinity of the laser focal volume. Physically, the plasma is created as

laser energy breaks all molecular and atomic bonds and strips electrons from atomic

nuclei forming a cloud of ions and free electrons. The ions and electrons rapidly

recombine to re-form neutral atoms and, in the process, emit photons in narrow, species-

specific wavelength bands. Since each known atomic or molecular species contains









unique electron configurations and energy transitions, this photonic emission serves as a

signature and can be captured and analyzed as such. The following discussion details the

various processes which characterize laser-induced plasma formation and evolution.

Plasma Formation

Laser energy couples into the sample medium through two primary mechanisms,

cascade ionization and multi-photon ionization, which explain the plasma formation

process. In the cascade ionization process initial free electrons absorb laser energy and

collide with neutral atoms and molecules to release more electrons. Writing out the

reaction, cascade ionization is described by

e- +M- 2e- +A 1-1

where M represents an atom or molecule. Since two electrons are produced for every one

used, the electron population increases exponentially in time. This process is therefore

also referred to as avalanche ionization.

The second mechanism of plasma formation, multi-photon ionization, is

characterized by direct absorption of photons by atoms and molecules and is described by

M n(hv) -) e + M 1-2

This process does not require an initial population of free electrons and does not

tend to cascade as multiple photons are required for the release of one electron. The two

ionization processes require large energy densities to proceed, occur simultaneously, and

contribute in varying proportions to plasma growth. Since cascade ionization requires an

initial population of free electrons, breakdown initiation is thought to generally occur via

multi-photon ionization. However, the entire breakdown event is known to be dominated

by multi-photon ionization for short wavelengths (X < 1 tm) and low pressures (p < 10









Torr) and dominated by cascade ionization for long wavelengths (X > 1 [tm) and high

pressures (p > 100 Torr) (Weyl 1989).

Spatial and Temporal Evolution

While the laser-induced plasma is highly dependent on many physical parameters

such as laser power, laser wavelength, analyte source and concentration, and system

optics, the spatial and temporal evolution of most plasmas shows definite trends. Within

the first few nanoseconds after the laser pulse reaches the focal spot, the plasma begins a

rapid expansion, forms a highly energized cloud of ions and electrons, and is sufficiently

hot that radiant emission falls principle in the ultraviolet-visible range (treating the early

plasma as a blackbody at 40,000 K, which is within the temperature range reached by

many laser induced plasmas, 97% of the emission falls below 400 nm-for less energetic

plasmas, the value drops to 86% at 20,000 K and 48% at 10,000 K). This initial period of

rapid spatial growth lasts on the order of hundreds of nanoseconds, thereafter the plasma

continues to grow for some microseconds, but at a reduced rate. In addition, the early

plasma emission-as expected-has a strong broadband response which tends to

dominate the various ionic transitions which are present at small times.

After the initial growth phase the plasma begins to decay. It should be noted that

growth, in this sense, refers to an increase in volume of the region from which luminous

emission emanates-from the standpoint of absolute temperature and electron density,

the plasma is never more energetic than at the tail end of the initiating laser pulse. At this

point electrons and ions recombine yielding neutral atoms and their corresponding

spectral lines. The fraction of neutral species present in the focal volume increases

steadily during this period, which usually lasts tens of microseconds. The plasma cools

significantly from radiative transfer and quenching processes and emission in the visible









spectrum increases, contiuum emission decreases, and distinct spectral lines become

prominent. This period, from approximately 1-30 [ts following the laser pulse, is

therefore ideal for capturing and analyzing atomic spectra.

From 30 [ts onward, plasma energy further decays, the plasma volume continues to

cool, and atoms begin to combine to form molecules. Molecular spectral bands appear

along with fading atomic lines. The plasma continues to emit out to hundreds of

microseconds following the initial laser pulse, but the useful signal is increasingly

diminished with time. The plasma event ends when no more emission is detectable and

the plasma volume is once again occupied by neutral constituents at or near ambient

conditions.

As a final note on the process of the laser-induced plasma event, the above

discussion of clearly demarcated phases of plasma behavior represents a simplification of

the phenomena. The occurrence of continuum, ionic, atomic, and molecular emission

may well, for any given analyte, overlap significantly depending on the various states and

transitions of that analyte. However, it is extremely important and helpful to recognize

the nature of temporal variations in plasmas-such recognition plays heavily into plasma

spectroscopy applications and is discussed thoroughly in the current work.

The Laser-Induced Breakdown Spectroscopy (LIBS) Technique

The laser-induced plasma is the fundamental building block for the LIBS

technique, the basis of which rests in the ability to accurately extract meaningful spectral

data from radiative plasma emission. For LIBS the plasma acts as both the excitation

source and the sample volume, and dissociates entrained aerosols (smaller than about two

microns as discussed in later sections), molecules, and atomic species resulting in









radiative emission. Spectrally resolving this emission in a useful way is the goal of most

LIBS studies.

First employed in the 1960s, LIBS has become increasingly popular for qualitative

and quantitative spectral analysis because of an inherent advantage over other techniques:

very little, if any sample preparation is needed for species under investigation, making in

situ measurements possible and straightforward. Numerous literature reviews of the

LIBS technique trace its evolution as a means of spectral analysis (Radziemski 1994,

Schechter 1997, Smith et al. 2001).

Not without its challenges, LIBS requires a significant photon flux to initiate a

plasma, especially in a gaseous medium. For a spark to be induced in clean (particulate-

free) air, threshold laser fluence has been reported to range from 10-1600 GW/cm2 with

the large variations due to differences in experimental setup (Simeonsson and Mizeolek

1994, Smith et al. 2001). Such large energy densities require powerful pulsed lasers,

often large in size and cost. Creating the plasma, however, is just the beginning-laser-

grade optics, a diffraction grating, and a collection device (the latter two usually

conveniently housed in a modern spectrometer) are required to extract a useful signal

from a small plasma volume, which has been shown via absorption measurements to be

approximately 1 mm3 for nominal conditions (Carranza and Hahn 2002c).

All studies in this dissertation focus solely on LIBS in a gas phase medium. Gas

phase LIBS is a spectral method for characterizing analyte species either in the gaseous

state or in aerosol form. An aerosol is a solid or liquid particle suspended in a gas-phase

matrix. Aerosols are an ever-present, naturally-occuring component of our atmosphere

but can also be easily fabricated with pressurized jets and nebulizers. Much is unknown









about the behavior of ambient aerosol populations; as they represent a constantly

changing system in which real-time, in situ analysis does not readily lend itself to

conventional means of laboratory study. LIBS, with the ability to dissociate raw samples,

presents a technique well-suited to studying aerosols and has been addressed in many

studies as a means of aerosol species, concentration, and size detection (Radziemski et al.

1983, Essien et al. 1988, Singh et al. 1997, Neuhauser et al. 1999, Sneddon and Lee 1999,

Hahn and Lunden 2000).

Temporal Properties of Laser-Induced Plasmas for LIBS Analysis

It is known that the LIBS technique uses a highly energetic plasma to vaporize and

dissociate matter within the laser-induced plasma, including entrained particles.

However, less is known, both theoretically and experimentally, regarding the interactions

between the laser-induced plasma and particles. In a recent study, research efforts were

focused to quantify these plasma-particle interactions pursuant to quantitative LIBS-

based analysis (Carranza and Hahn 2002c). It was found that individual silica particles

up to 2 |tm in diameter were completely vaporized in the laser-induced plasma and

yielded a linear analyte response, while larger particles revealed a non-linear response

with respect to analyte atomic emission presumably due to incomplete vaporization.

While the plasma-particle interaction is not well understood, it appears that the transient

nature of the plasma in the early stage of plasma evolution plays an important role in the

vaporization process. Hence, rate limitations may be more important than simple plasma

and particle heat capacities. Accordingly, as a first step toward understanding the

plasma-particle processes, it is important to characterize laser-induced plasmas in the

early stages of evolution. To further advance this understanding, this study investigates









the plasma evolution via time-resolved measurements of electron number density, plasma

absorptivity, and plasma continuum emission.

Laser-induced plasmas have been studied using parameters such as electron

temperature, electron density, and emission intensity. Parigger et al. (1995) studied the

decay of plasmas produced in hydrogen by a Nd:YAG laser (1064nm) using the

hydrogen Balmer series emission lines. They found that at early times, comparable with

the plasma-initating laser pulse, electron densities were on the order of 1019 cm-3 while in

the first hundreds of nanoseconds following laser breakdown, the free electron density

was about 1018 cm-3, and the plasma temperature was about 80,000 K (calculated using

Boltzmann plots). In a more recent study, Parigger et al. (2003) further explored Stark-

broadened emission profiles in gaseous hydrogen plasmas, comparing both Ha and Hp

measurements in the context of recent treatments of ion dynamics as reported by Oks

(2000). For electron densities in excess of 1017 cm-3, corresponding to delay times less

than 500 ns, agreement between inferred electron density values from H, and Hp profiles

was very good. However, at longer delay times some discrepancies were noted when

using the standard theoretical treatment per Griem (1974, 1997), while good agreement

between the Ha and Hp results was achieved with the updated treatment of Stark

broadening for electron densities approaching 1016 cm-3. Borghese and Merola (1998)

measured a plasma kernel of about 0.02 mm3 as produced by a fundamental Nd:YAG

laser in air during the stage of maximum emission (about 20 ns following plasma

initiation). They calculated a corresponding plasma temperature of about 35,000 K at

this time using an experimental Planck function, but also calculated higher plasma

temperature at earlier times (e.g., 105 K at about 10 ns). Using techniques such as









shadowgraphy and interferometry, Villagran-Muniz et al. (2001) found that the plasma

electron density was about 1018 cm-3 for a pulse energy of 300 mJ, and additionally

calculated temperatures behind the shock wave (due to the optical breakdown) of about

105 K during the first tens of nanoseconds, when the shock wave is not detached from the

plasma. Work by Yalcin et al. (1999) investigated plasma temperature and free electron

density, including spatially resolved measurements using Abel-inversion, for delay times

as early as 350 ns following plasma initiation. Using Stark-broadened H, line widths,

they reported peak electron densities of 1.2xl018 cm-3, and reported little variation

(-10%) in this value across the plasma using spatially resolved measurements.

All available research shows that the plasma, in its early stage, is a highly energetic

system that is characterized by rapid changes, and is not necessarily in thermodynamic

equilibrium. For LIBS-based aerosol analysis, these changes need to be assessed in order

to identify parameters that could play important roles in the interactions between the

plasma and concomitant particles, including free electron densities and plasma

absorptivity. In the following study, the evolution of the laser-induced plasma is assessed

based on temporally resolved spectral absorptivity and emissivity measurements, along

with evaluation of the free electron density via Stark broadening.

Stability of Laser-Induced Plasmas

Historically, laser-induced breakdown spectroscopy (LIBS) has been applied on a

statistical basis-spectra from thousands of laser shots are ensemble averaged and results

extracted (Radziemski 1994). In recent studies, however, LIBS has been applied using

single-shot analysis based on conditional spectral processing to effectively sample

ambient aerosol populations by discrete particle analysis (Hahn and Lunden 2000,









Carranza et al. 2001). Because gains in signal-to-noise realized with ensemble averaging

are not applicable with single-shot analysis, it is important to maximize the precision on a

shot-to-shot basis (Carranza and Hahn 2002b). For example, recent interest in extending

the LIBS technique to quantitative analysis of bioaerosols (Boyain-Goitia et al. 2003,

Hybl et al. 2003, Morel et al. 2003, Samuals et al. 2003), which entails high precision to

distinguish between often-similar elemental signatures, places additional requirements on

analyte sensitivity and single-shot precision. Hence, to further establish single-shot

LIBS-based techniques as an integrated analytic tool for detection and quantitative

analysis of aerosols and other dilute species, the precision must be verified and

maximized on a shot-to-shot basis. Therefore, the stability of the plasma-producing laser

pulse, the spatial and temporal stability of the plasma itself, and reproducibility of the

corresponding spectral emission are all of primary concern.

A laser-induced plasma is largely a product of two factors-the laser irradiance,

including both spatial and temporal characteristics, and the sample composition,

including homogeneity, in the vicinity of the laser-induced breakdown and subsequent

plasma formation. Improving the temporal precision of the plasma formation process and

precision of the ensuing plasma spectral emission, thereby improving the precision of the

LIBS analyte signal, requires attention to these two factors. In the current work, seeding

the laser cavity to improve spatial and temporal quality of the laser beam is explored in

concert with the effects of aerosol loading in the laser-induced plasma sample volume.

Several recent studies report on the time and spatially dependent absorption of

incident radiation by laser-induced plasmas (Aguilera et al. 2003, Bindhu et al. 2003,

Bindhu et al. 2004), thereby emphasizing the strong coupling between the plasma-









creating laser beam and the consequent laser-induced plasma. The full nature of this

laser-plasma coupling is not fully understood and is a focus of on-going work. Hence, an

additional goal of the current work is investigating the temporal variation of plasma

inception in terms of the effects of aerosol loading and laser cavity seeding.

LIBS and Calibration

Since the early 1980s, the use of laser-induced breakdown spectroscopy (LIBS) has

steadily increased as a quantitative analytical technique, with applications to a wide range

of solids, liquids, gases, and aerosols (Rusak et al. 1997, Sneddon and Lee 1999, Lee et

al. 2004). In recent studies LIBS has been used in conjunction with single-shot analysis

to effectively sample and analyze aerosol populations using discrete particle analysis

(Hahn 1998, Hahn and Lunden 2000, Carranza et al. 2003). In other applications, LIBS-

based sensing has been successfully implemented for continuous on-line monitoring of

air emissions and aerosols (Neuhauser et al. 1999, Zhang et al. 1999, Nunez et al. 2000,

Ferioli et al. 2003), and for analysis of ambient air particulate matter (Hahn 1998,

Carranza et al. 2001, Lithgow et al. 2004). More recent studies have addressed the

feasibility of LIBS for analysis of biological materials and bioaerosols (Boyain-Goitia et

al. 2003, Hybl et al. 2003, Morel et al. 2003, Samuals et al. 2003). Important research

issues regarding LIBS-based analysis of gaseous sample streams, notably aerosol

detection and single particle analysis, include the overall analyte sensitivity, precision,

sampling methodology, and calibration schemes.

Calibration has been an important issue since Radziemski et al. (1983) first

demonstrated the viability of LIBS for detection of atomic species from aerosol samples.

In their landmark study, beryllium-rich aerosols were generated by either laser ablation or

produced by a nebulizer/heat-chamber system, with the latter used for generation of









calibration curves and subsequent calculation of detection limits. A different nebulizer

was used in a continuing study, which produced aerosols estimated to be in the

submicron size range (Essien et al. 1988). Calibration curves were generated for three

analytes, namely cadmium, lead, and zinc, and were characterized by initial linearity

followed by various degrees of saturation at higher concentrations. The saturation effects

were attributed to incomplete vaporization of particles. An important finding was the

general agreement (within 10%) of lead atomic emission signals of comparable atomic

lead concentrations when nebulizing either lead acetate, lead chloride, or lead nitrate.

Cadmium revealed a 27% difference in analyte response when comparing nebulized

solutions of cadmium nitrate and cadmium chloride. In experiments analogous to the

aerosol studies, the relative independence of analyte signals on molecular source for

purely gas-phase species was reported in several studies (Dudragne et al. 1998, Tran et al.

2001). Specifically, Dudragne et al. demonstrated that analyte signals for fluorine,

chlorine, sulfur and carbon scaled with the number of respective atoms in the constituent

molecules for a wide range of compounds, concluding that the parent molecules were

fully dissociated in the laser-induced plasma. Tran et al. verified that SF6 and HF yielded

essentially identical fluorine atomic emission signals when the gas composition was

adjusted to contain the same atomic fluorine mole fraction.

The above comments support a widely used assumption within the LIBS

community, namely that complete dissociation of constituent species within the highly

energetic laser-induced plasma results in independence of the analyte atomic emission

signal on the analyte source. This statement of independence of analyte source must be

clearly distinguished from the presence of plasma matrix effects. Plasma matrix effects









are considered changes in a specific analyte emission response due to interactions with or

perturbations of the plasma, such as changes in plasma temperature, quenching of atomic

emission lines by other species, or loss of an emitting species due to recombination with

other atoms. For example, Gleason and Hahn (2001) observed temporal-dependent

changes in mercury atomic emission as the overall gas composition of the laser-induced

plasma was varied. The distinction between analyte source effects and plasma matrix

effects is significant, because when one considers the latter, there is generally an accepted

temporal starting point where the atomic concentrations are considered to reflect a

matrix-independent baseline. Such a constant temporal starting point was observed with

mercury emission as the gas matrix was varied in the study by Gleason and Hahn, noting

that the analyte source was held constant. Notwithstanding the apparent defacto

acceptance of analyte signal independence on analyte source for laser-induced plasma

spectroscopy, to date no studies have systematically examined this issue for LIBS-based

analysis of gaseous and aerosol-laden streams.

As the LIBS-based analysis of increasingly complex gas-phase and aerosol systems

becomes more prevalent, high accuracy and precision become increasingly important.

Therefore, precise calibration of the analyte signal response is required, as the number of

factors contributing to the emission signal (i.e., laser parameters, optical configuration,

spectrometer/detector, and plasma matrix) is sufficiently large to render an apriori

calculation of analyte signal response impractical. LIBS calibration schemes are made

more difficult when the systems of interest include complex and unknown analyte

compositions (i.e., mixtures of gas-phase, solid-phase, and aerosol analyte species),

which may exist for ambient air or combustion, exhaust streams. For example, carbon in









ambient air is present as carbon dioxide at several hundred parts-per-million (ppm), but is

generally present as fine particulate matter (such as soot), and within biological particles

such as pollens and bacterial spores. Similarly, metal species at moderate to high

temperatures may exist as vapor-phase species, and also as homogeneously or

heterogeneously nucleated particulates. Given the inherent assumption of complete

dissociation of analyte species, most calibration schemes to date have focused on a

precise and accurate means of analyte introduction, not on the resulting physical state of

the analyte. As described above, the use of purely gaseous analyte species and the

nebulization of aqueous solutions are the most widely used calibration schemes with

LIBS, noting that the former produces a homogenous-phase (i.e. all gaseous) source for

laser-induced breakdown, while the latter produces a well-dispersed, high number density

mixture of generally micron to submicron-sized aerosols in the gas matrix. The present

study examines the phenomena of analyte dissociation and analyte emission for different

sources (both gas phase and solid phase) of atomic carbon and the resulting effects on

analyte calibration response.

Spatial Considerations in Plasma Emission

The continued evolution of LIBS as an effective technique for the analysis of

aerosol species is dependent on its ability to accurately quantify the presence and

concentration of analyte species within individual particles and small groups of aerosols.

Primary challenging factors, including (a) that small aerosols may contain only a few

femtograms of analyte and (b) that the LIBS technique is known to experience large shot-

to-shot fluctuation in signal output, require that more be known about the physics of the

plasma-particle interaction on a single-shot basis. Furthermore, identification and









minimization of sources of experimental uncertainty are critical to more accurate

detection.

While bulk plasma properties and spatially averaged particle-laser interactions can

be studied with conventional techniques, methods for measuring laser-induced plasma

behavior with respect to discrete aerosols remain limited. Particle position, size, and

number density within the plasma volume may have distinct effects on spectral emission.

Analysis of these properties for single particles and known discrete populations must be

performed in order to examine such connections.

While significant progress has been achieved regarding the development of the

LIBS technique for use on single particles and aerosols, many fundamental issues remain.

Of particular interest is the physical process and time-scale of particle dissociation within

a laser-induced plasma and subsequent spatial diffusion of analyte species throughout the

plasma. As suggested by recent models of laser-induced plasma behavior, species

dissociation and spectral analyte emission within a laser-induced plasma are commonly

assumed to occur in a temporally instantaneous and spatially homogeneous way (Ho et al.

1996, Itina et al. 2003,Gornushkin et al. 2004). Although it is recognized that these

assumptions are made primarily in order to reduce the mathematical complexity of such

models, non-homogeneities in both space and time are for the most part neglected in

experimental analysis of laser-induced plasmas as well. An analysis that accounts for

spatial and temporal variability considerations could shed greater light on the presence of

typically large shot-to-shot fluctuations often present in the data of the vast array of LIBS

studies that seek to quantify analyte spectral emission.









Recent studies have proposed and demonstrated that one such spatial parameter-

the relative position of a sampled aerosol within the total plasma volume-may play a

role among other significant factors in the ability to detect spectral emission as well as

the accuracy of detection using LIBS (Lithgow 2005). The current study, using a

combination of imaging and spectroscopy, addresses the issue of spatially non-

homogeneous atomic emission originating from an aerosol particle within a laser-induced

plasma and quantifies a diffusion time scale for both physical aerosol dissociation and

analyte-specific emission.














CHAPTER 2
EXPERIMENTAL METHODS

The setup of the LIBS analytical apparatus can be broken down into two sets of

optical instruments with two basic functions; delivery of laser irradiance to the sample

and extraction of luminous emission from the plasma volume.

The LIBS Experimental Apparatus

Delivery Optics

A schematic of equipment used for delivery of laser energy to the sample is shown

below:




Nd:YAG
Plasma
Gallilean Aperature Condensing
Telescope Lens



Figure 2-1. Optical setup for delivery of laser energy to the sample volume resulting in
the creation of a plasma. For simplicity and clarity the telescope and
aperature will be omitted from additional schematics of LIBS systems.

A pulse of energy, usually 10-20 ns long and 50-300 mJ (for gaseous samples), is

emitted by the laser. The energy cross-section for a typical beam (often called the beam-

profile) displays a Gaussian curve shape; the beam is more intense at its center than at its

edges, and decays exponentially between the two. To obtain a more uniform beam a

Galilean telescope consisting of a negative (concave) and positive (convex) lens is used

to expand the beam and an aperture removes light at the expanded beam's edge-

yielding a wider beam with a more uniform profile. In addition to a uniform energy









cross-section, a wider beam can be focused more tightly than a narrow one, allowing for

a greater spatial energy density. This is advantageous as the next step is to focus the

beam to a spot with a condensing lens resulting in a photon flux sufficiently large that the

dielectric strength of the gaseous medium is exceeded and a plasma is formed.

Extraction Optics

As previously detailed, plasma formation results in a burst of electromagnetic

radiation centered in the near-ultra-violet and visible portion of the spectrum. Regions of

peak intensity, which correlate to atomic and molecular transitions of plasma

constituents, develop a few hundred nanoseconds after plasma initiation and must be

collected and spectrally resolved. All studies performed for this dissertation collect in a

backscatter configuration, which minimizes experimental uncertainty due to spatial

variation in plasma formation.

Collimation
Pierced Lens
From Laser Mirror

tPlasma
Condensing C:L
Lens

GiCCD
ob Spectrometer C
Fiber
Optic



Figure 2-2. Optical setup for extraction of plasma emission from sample volume.

First a positive lens is used to collimate the spherical plasma emission. A mirror

then turns the collimated beam into a second positive lens which recondenses the plasma

emission onto a fiber optic bundle. The fiber optic feeds a spectrometer which is tuned to

a narrow wavelength range of interest, allowing for sub-nanometer resolution of atomic









lines. Spectra are imaged using an intensified charge-coupled device (iCCD) mounted at

the exit slit of the spectrometer.

System Synthesis

Combination of the laser delivery system with the optics for extraction of plasma

emission-incorporating auspicious properties of both-yields a basic LIBS apparatus.

Pierced
Mirror

Nd:YAG Mirsm
Plasma




SSpectro- CCD
Fiber meter
Optic



Figure 2-3. Basic LIBS system. When plasma emission is collected along the same axis
as incoming laser radiation, the system is said to be back-collecting.

Notice that the same lens serves as both the condensing lens to deliver the laser radiation

to the sample and the collimation lens for back-collection of plasma emission. Also the

mirror used for extraction must be pierced in the center to allow transmission of the

initiating laser pulse-the amount of reflection sacrificed to the hole is negligible on the

basis of collected light area.

While collection of plasma emission is the basis of the LIBS technique, auxiliary

diagnostics are desirable for system characterization within experimental studies;

temporal beam sampling and beam energy measurements are most common and are

abundantly detailed for experiments in this work. Descriptions of experimental

configurations that follow all start with the basic LIBS apparatus described above and

make use of numerous enhancements-additional optics, secondary laser sources, digital









oscilloscopes, iCCDs for image capture, digital delay generation for temporal

sequencing, and optical power meters-in analyses other than basic spectral collection.

Plasma Transmission Study

The LIBS system used for temporally resolved measurements of laser-induced

plasma transmission is given below:

Nd:G Probe laser beam







Nd:YAG
Plasma


Aperture l -

iCCD Fiber
Spectrometer Opc Calorimeter






Figure 2-4. LIBS system incorporating second laser for plasma transmission
measurements.

A 1064-nm Q-switched Nd:YAG laser operating with 275-mJ pulse energy, 10-ns

pulse width, and 5-Hz pulse repetition rate was used as the plasma source for all

experiments. The expanded laser beam (12-mm in diameter) was focused using a 75-mm

focal length UV grade plano-convex lens (CVI, PLCX-50.8-38.6-UV-1064) to create an

approximately 20-[tm diameter focal spot for plasma formation. The plasma emission

was collected along the incident beam in a backward direction and separated using a

50-mm diameter elliptical pierced mirror. The collected light was launched into an









optical fiber bundle coupled to a spectrometer (2400-grove/mm grating, 0.12-nm optical

resolution), and recorded with an intensified charge-coupled device (iCCD) array.

Transmission Measurements

For the transmission and absorptivity measurements the plasma was formed in the

laboratory air. A probe Nd:YAG laser beam was synchronized to the plasma-creating

laser by externally triggering both the flashlamp and Q-switch. The temporal jitter

between the plasma-initiating laser pulse and the probe laser pulse was about 1 ns. The

probe laser was operated using either the frequency-doubled 532-nm line (-6 ns fwhm

pulse width), or the fundamental 1064-nm line. The pulse energy of the probe laser was

maintained at 15 mJ for both the 532 and 1064-nm probe beams. However, the flashlamp

pump energy was significantly reduced for the 1064-nm probe beam, which resulted in a

broader temporal pulse-width, namely -30 ns fwhm.

The probe laser beam was directed orthogonal to the plasma-creating laser beam,

and focused at the center of the resulting plasma using a 250-mm focal length lens. The

probe laser beam alone had insufficient energy (15 mJ) to create a plasma, which was

verified by the absence of breakdown with the probe laser alone, as well as the absence of

breakdown at long delay times following the 1064-nm laser initiated plasma. The probe

laser was translated both horizontally and vertically until the minimum transmission was

recorded at a delay time of 20 ns with respect to the plasma-initiating 1064-nm laser

pulse. The spatial positioning ensured that the probe laser beam was passed through the

most optically dense region of the plasma. It is noted that in a previous study using a

similar probe beam arrangement, the plasma was found to expand to its characteristic

volume of 1.4 mm3 by 20 ns following plasma initiation (Carranza and Hahn 2002c).

The transmitted probe laser pulse energy was recorded with a volume-integrating









calorimeter (25-mm detector area), using a 1-minute average at a 5-Hz repetition rate.

The detector was placed about 0.5 m from the plasma to eliminate any detection of direct

plasma emission, which was verified by the lack of signal in the absence of the probe

laser. The large detector area functioned to mitigate any effects of beam steering on the

recorded transmission, although no probe beam displacement was observed during the

experiments.

The probe laser beam transmission was determined as the ratio of the average laser

pulse energy transmitted through the plasma divided by the average laser pulse energy

recorded in the absence of the laser-induced plasma (i.e. probe laser only). The latter

value functioned as the reference pulse energy, and was taken as the average of the values

recorded before and after each plasma measurement. The average relative standard

deviation of the reference pulse energy was 1.6% for all experiments; hence probe laser

energy drift was not significant. All transmission measurements were recorded a

minimum of four times for each reported delay time.

Plasma continuum emission was measured by sampling three locations, each

approximately 20 nm wide, along the baseline of an ensemble average of 100 spectra

taken from the transmission data.

Electron Density and Plasma Continuum Measurements

For electron density measurements, the plasma was formed in an atmospheric

pressure gas cell (Huntington stainless steel 6-way cross) through which a gaseous stream

of 42 1pm of purified (HEPA filtered), dry air was passed. An additional flow rate of

high purity methane was added to the airflow to enhance the hydrogen emission line

signal. The overall methane/air flow rate was maintained at 3% methane, below the

lower flammability limit (5%) of methane in air, to prevent ignition by the laser-induced









plasma. To obtain the electron density, Stark broadening of the Ha line was measured.

For these measurements, a calibrated blackbody source was used to convert the recorded

plasma emission to a relative spectral irradiance scale across the measured spectral

bandwidth. All electron density measurements were based on the spectral analysis of an

ensemble average of 100 spectra.

Laser Shot Stability Study

The experimental system for the laser-induced plasma stability study is shown

schematically in Figure 2.5.


To Filtered Air
Pierced Mirror
Beam Splitter


Nd:YAG

Plasma Volume



Fiber
Optic
Oscilloscope CCD


st D Spectrometer
Fast Detectors



Figure 2-5. LIBS system incorporating fast detectors and a digital oscilloscope for
additional beam measurements.

For all experiments, the third harmonic (355 nm) of an injection-seeded Nd:YAG

(10 Hz repetition rate and 280 mJ/pulse) was used to create the plasma using a 50-mm

diameter, 75-mm focal length lens. The laser could be operated with or without the









seeder, while maintaining nominally constant pulse energy and timing. The plasma

emission was collected on axis with the incident laser beam using a pierced mirror (Rolyn

Optics, #60.2475) and 75-mm focal length condensing lens. The plasma emission was

subsequently fiber-coupled to a 0.3-m spectrometer (2400 groove/mm grating, 0.022

nm/pixel dispersion, -0.15-nm resolution), where spectral data were recorded using an

intensified CCD detector array. For all experiments, LIBS spectra were recorded using a

signal integration of 10 jts width following a delay of 10 jts with respect to the incident

laser pulse.

The laser pulse waveform was sampled prior to the plasma focal volume and

following transmission through the plasma using a series of optical wedges as beam

samplers, as shown in Figure 2-5. The sampled beams were recorded using matched, fast

photodiode detectors (-1 ns rise time) coupled to a digital oscilloscope (500 MHz

bandwidth and 4Gs/s sampling rate). A 355-nm laser line filter was used with both

detectors to eliminate possible signals originating from the plasma emission.

LIBS measurements were recorded either directly in ambient air or within a

uniform jet of filtered air. For the filtered air measurements, compressed air was dried

using an activated alumina desiccant, and passed through a high-efficiency particle

arrestor (HEPA) filter to effectively remove all particulates. It is noted that the amount of

water vapor and carbon dioxide in the compressed air were markedly reduced by the

presence of the activated alumina desiccant. The conditioned air stream was expanded as

a free jet with an exit diameter of 1 cm directly into ambient air approximately 1 cm

below the laser-induced plasma kernel (L/D = 1 at plasma center, laminar flow expected

within plasma volume).








These two configurations allowed LIBS data to be recorded under the aerosol

loadings of ambient air, and under identical experimental conditions but in the absence of

particulate loadings by using the HEPA-filtered air jet. Independent aerosol size and

concentration data were recorded using a commercial light scattering instrument with

sensitivity over a range from 100 nm to 2.5 |tm particle diameter. This instrument was

used to measure the ambient air aerosol loadings in the laboratory air, as well as to verify

the absence of any aerosol particles in the HEPA filtered air stream.

Carbon Calibration Study

The experimental system used for advanced LIBS calibration study is shown

below:


Exhaust


t


LIBS sample chamber: *
Six-way cross I


Mixing-drying
chamber
Flow controller
(N2)
m 4


Condensing Nd-YAG laser
Lens





Flow controller
S(CO, C02, CH4)
rn


Figure 2-6. Nebulization system coupled with a LIBS system (only Nd:YAG shown) for
aerosol analysis.









For all experiments, a 1064-nm Q-switched Nd:YAG laser running at a 5-Hz

repetition rate, 10-ns pulse width, and 290-mJ/pulse was focused using a 50-mm

diameter, 75-mm focal length lens to create the plasma. The plasma emission was

collected on-axis with a pierced mirror and fiber-coupled to a 0.275-m spectrometer with

a 2400-groove/mm grating (0.12-nm optical resolution). Spectral data were recorded

using an iCCD detector array. All carbon emission spectra were recorded using a delay

of 10 us with respect to the incident laser pulse and a signal integration time of 12 us.

This detector gating optimized the analyte signal of interest, namely the carbon (I) line at

247.86 nm. Additional Ha (656 nm) emission spectra were recorded using a temporal

delay of 4 |ts and a gate width of 4 hts, and CN emission spectra were recorded using a

temporal delay of 40 |ts and a gate width of 60 [ts.

Five different carbon-containing species were introduced to the LIBS sample

chamber, as shown in Figure 2-6, with the goal of providing different physical and

chemical states of the carbon carrier species. Calibration stream flow rates and analyte

concentrations were adjusted to provide a comparable range of atomic carbon mass

concentrations for all experiments.

Solid Phase Carbon Experiments

Two different types of solutions were nebulized to provide a carbon-rich aerosol

stream (i.e. solid-phase analyte) to the LIBS sample chamber. For the first set of

experiments, carbon solutions were prepared in the range from 500 to 5000 [tg/ml of

atomic carbon by dilution of Atomic Absorption (AA) and Inductively Coupled Plasma

Atomic Emission Spectroscopy (ICP-AES) grade carbon standard solutions (SPEX

CertiPrep, 10,000 [tg/ml of carbon as oxalic acid, (COOH)2). For the second set of









experiments, carbon solutions were prepared in a similar mass concentration range using

a monodisperse particle suspension of 30-nm polystyrene (C8Hs) spheres (Duke

Scientific). To avoid foaming within the nebulizer, the polystyrene suspensions were

first washed with a resin (BIO-RAD, AG 501-X8, 20-50 Mesh, Molecular Biology

Grade) to remove the surfactant. Specifically, 50 g of twice-washed resin was added to

300 ml of the polystyrene suspension (1% solids) and stirred for 18 hr, after which the

polystyrene suspension was poured off. For all experiments, the carbon mass

concentration for the polystyrene suspensions was adjusted by dilution with ultra-purified

deionized (DI) water, using the manufacturer's initial specification (10% solids). The

exact mass loadings were then verified by evaporating the water and weighing the

remaining dried polystyrene mass corresponding to a known initial suspension volume.

The preliminary estimates of carbon concentration based on the manufacturer's mass

loadings and the measured mass loadings agreed to within 5%, with the small difference

attributed to the loss of some particles to the resin beads during surfactant removal.

Aerosol flows were generated using a pneumatic nebulizer supplied with 5-

liters/min (1pm) of dry nitrogen. A 46-lpm dry nitrogen co-flow was introduced to the

aerosol stream to ensure uniform bulk flow and to facilitate complete desolvation of the

nebulizer droplets, resulting in primarily submicron-sized aerosol particles at the LIBS

sample point. For the given flow scheme approximately 103 particles exist in any given

plasma volume. The nitrogen supply was HEPA filtered prior to use, and all gases were

metered with precision mass flow controllers. Prior to nebulization, the polystyrene

suspensions were well mixed and then sonified for 5 minutes to minimize any particle

agglomeration. The overall aerosol generation system was described in full detail in a









previous publication (Hahn et al. 2001), where nebulization of similar AA-ICP standard

solutions provided aerosols with a modal diameter less than 100 nm based on particle

sampling and transmission electron microscopy (TEM) analysis.

The total mass loading of carbon in the LIBS sample stream was defined by the

liquid nebulization rate (nominally 0.12 ml/min), the carbon mass concentration in the

liquid solution/suspension, and the total gas flow rate (nebulizer flow plus co-flow). The

solution/suspension concentrations were adjusted to provide a nominal range from 1,000

to 10,000 atg carbon/m3 through the LIBS sample chamber, which corresponds to about

900 to 9,000 parts per billion (ppb) of carbon on a mass basis.

Gas Phase Carbon Experiments

Three different carbon-containing gaseous species were used to provide a carbon-

rich sample stream (i.e. gas-phase analyte) to the LIBS sample chamber. The three gas

phase species, carbon dioxide (C02), carbon monoxide (CO), and methane (CH4), were

purchased as nominally 50-ppm mixtures in a dry nitrogen balance (Spectra Gases). The

actual concentrations ranged from 49.3 to 50.8 ppm, with the true concentrations used for

all calculations. The gases were metered with precision mass flow controllers and mixed

with the nitrogen co-flow gas stream to provide a range of carbon mass loadings

consistent with the values used for the solid-phase experiments. To provide consistency

with the solid-phase experiments, the nebulizer was operated with ultra-purified

deionized water, and the total gas flow rate was maintained at the same value for all

experiments. In addition, valves on the experimental apparatus allowed the analyte gas

streams (CO, C02, CH4) to be diverted through the nebulizer for the case of 5 1pm of

analyte gas flow rate.









The total mass loading of carbon in the LIBS sample stream was controlled by

adjusting the relative flow rates of the analyte gas stream and the nitrogen co-flow

stream. The experimental conditions were adjusted to provide a nominal range from

1,000 to 7,800 [tg carbon/m3 through the LIBS sample chamber.

LIBS Data Collection

The analyte peak of interest was the carbon (I) atomic emission line at 247.86 nm

(21,648 61,982 cm-1). The recorded spectral window was approximately 30-nm wide

(230 to 260 nm), and was nominally centered on the 247-nm carbon line. For each

carbon species and carbon mass concentration, a series of 6000 spectra (6 x 1000 shots)

were recorded and ensemble-averaged. The experiments were repeated between four and

eight times for each carbon species, with each experiment recorded on different days in

an attempt to average any fluctuations in the overall system. Spectra were also recorded

for the nebulization of DI water only for all experiments to serve as spectral blanks (i.e.

no carbon emission line).

In addition to data recorded near the 247-nm carbon line, emission spectra were

also collected for each experimental condition, including DI water only, using 30-nm

spectral windows centered at 656-nm and 385-nm to capture emission features

corresponding to the Ha atomic emission line and the cyanide (CN) molecular emission

band, respectively. Free electron densities were measured from the Stark broadening of

the 656-nm Ha line using the same procedures reported recently (Carranza and Hahn

2002). Briefly, following the work of Griem, the Ha line intensities at full-width half-

maximum (fwhm) were related to the electron density using the fractional half-widths of

the reduced Stark profile (Griem 1974).










Plasma-Aerosol Interaction Study

Schematic representation of the experimental system for the study of plasma-

aerosol interaction is given below:



Fiber
Optic
Spectro Aerosol
meter Stream
iCCD Narrow-band Line Filter
TT (396.2nm/3nm FWHM)




iCCD
Nd:YAG
(10Ons/300mJ/pulse)
Pierced
LIBS Sample Turning mirror:
Chamber: protects CCD from
6-way cross 1064nm laser
emission not
absorbed by plasma

Figure 2-7. Intensified CCD for imaging combined with a LIBS system for sampling
aerosols within the laser-induced plasma.



For all experiments, a 1064nm Q-switched Nd:YAG laser with a 10ns pulse width

and 300mJ/pulse was focused using a 50mm diameter, 75mm focal length lens to create

the plasma inside a sealed, stainless steel six-way cross vacuum section (Huntington)

forming a LIBS sample chamber. A nebulized solution ofborosilicate glass

microspheres (Duke Scientific, 9002) in ultra-pure deionized water was fed into the

chamber with a dry HEPA-filtered air co-flow stream. The nebulized solution and co-

flow stream, as in previous work, mixed inside a 0.8m drying tube ensuring complete

dryout of aerosols upon entry to the LIBS sample chamber (Hahn et al. 2001). The

diameter of the microspheres was given by the manufacturer as 2.0+0.7[tm and was









confirmed to be accurate using optical microscopy, although no quantitative analysis was

performed. The density of the glass was 2.50g/cc and spheres dry number density was

9.5x1010/g. Calcium, a minor constituent ofborosilicate glass present as residual lime

(CaO), was the analyte of interest. Typical lime concentrations for borosilicate glass (in

which most of the lime present in soda-lime glass has been replaced with boric oxide,

B203, for modification of material properties) range from -0.05 to -1% by mass.






















Figure 2-8. Plasma image as visualized from a co-axial orientation to the laser beam.

The Imaging System

An imaging system was used to record the position of an entrained particle relative

to the plasma kernel, a wide array of experiments can be performed to measure the effect

of variations in particle position on spectral measurements. To that end, an iCCD

synchronized to the Q-switch of the plasma-inducing laser was set up according to the

schmatic in Figure 2-7. A 79.6 mm focal length, UV-grade achromatic lens (OptoSigma,

#027-3015) provided a magnification of 3.6 to the plane of the iCCD with respect to the









plane containing the plasma. In this configuration the iCCD resolution was found to be

5.3 utm/pixel (image targets and derivation of optical resolution is given in Appendix B).

Additionally, a high-efficiency 1064-nm dichroic turning mirror (Newport,

#20QM20HM. 15, UV-grade fused silica substrate) was placed in the image path

protecting the CCD and intensifier from laser energy at 1064 nm which propagates

through the plasma.

Plasma emission was forward-collected on the axis of the laser using 1024x1024

pixel iCCD camera (Andor iStar) with the intensifier gate triggered by the Nd:YAG Q-

switch. The gate delay of the camera (from the tail end of the laser pulse) and its width

were adjusted in time increments ranging from 2.0-30.0 [is and from 0.2-3.0 [is

respectively-the width maintained at 10% of the delay to allow adequate collection of

light by the iCCD as the plasma weakened in time. A 1064 nm mirror was placed in

front of the camera to protect the CCD and intensifier from laser radiation transmission

through the plasma which the authors measured to be 10% of total laser power.

Additionally an atomic line filter (CVI, #F03-396.1-4-1.00)-3nm full-width at half-

maximum centered at 396.2nm-corresponding to a strong Ca II line at 396.85nm, was

placed in the beam path so as to collect only calcium emission and background signal

over that same narrow spectral range. An achromatic lens with 1064nm anti-reflective

coating was used to focus the plasma image onto the CCD. Using a finely spaced and

static grid, the magnification and resolution of the imaging system were found to be 3.6

and 5.2am/pixel respectively. The images were sufficiently large that the laser repetition

rate which allowed the capture of every laser pulse was limited to 1Hz. For the

measurements, one gram of microspheres (9.5x1010/g) was diluted with 400cc ultra-pure









DI water to make a stock solution which was fed into the system by a nebulizer which,

when pumped with 5L/min dry HEPA-filtered air, delivered 0.14g/min of solution. The

resulting concentration of solids, as verified by total dryout weight, was 250[ag/mL

corresponding to a number density of 2.4x107/cc, assuming a mean particle diameter of

2am and a density, given by the manufacturer, of 2.5g/cc. The aerosol system was run

with a co-flow of 20L/min dry HEPA filtered air. At each gate delay, a series of image

frames were collected while running the aerosol system at steady state. Additionally at

each delay, the system was run with pure DI water only in order to collect a background

signal. A measure of the dark current signal was also measured by running the camera

with the lens cap on.

The Spectroscopy System

Plasma emission was back-collected on-axis with the laser using a pierced mirror

and fiber coupled to a 0.275m spectrometer with a 2400 groove/mm grating. Spectral

data were recorded over a 30nm window centered at 396nm with an iCCD camera-

again, the intensifier gate triggered by the laser Q-switch. As the spectra could be

recorded more rapidly than the images, the laser was run at 5-Hz. A dilute aerosol stream

(one aerosol per plasma shot was statistically desired for spectral calibration) was

generated by diluting the stock solution, noted above, 10:1 with ultra-purified DI water

and the dry air co-flow was increased to 40L/min. The nebulizer was again supplied with

5L/min dry HEPA-filtered air. With the aerosol system running as described, five

separate 1000-shot spectral sequences were taken over the same range of gate delays and

corresponding widths given above. In order to calibrate the spectral response of the

aerosol stream, the aerosol system was seeded with an inductively-coupled plasma (ICP)

standard solution of Calcium diluted to known concentrations of 0.0 (DI water blank),









2.5, 5.0, and 10.0tg-Calcium/mL-solution. Again, five 1000-spectrum sequences were

taken over the identical range of intensifier gate delays and corresponding widths as

noted above. Previous work demonstrated that mean aerosol diameter from an identical

system was approximately 100 nm for a range of ICP liquid analytes (Hahn et al. 2001).














CHAPTER 3
PLASMA TRANSMISSION PROPERTIES

Understanding the parametric changes within a laser-induced plasma with respect

to time is important in explaining connections between plasma energy and analyte

dissociation. Specifically, the ability of a plasma to absorb energy relates directly to its

ability to re-radiate that energy (Kirchoff's Law) both internally and externally. The

current supposition is that the plasma is most energetic and therefore most capable of

dissociating entrained particles during the times that it is most absorbing and/or energetic.

Time-Resolved Plasma Transmission

For the following portion of the study, a two-laser system was used to measure

temporal changes in transmission (and, hence, absorption) of a plasma produced in air.

The time-resolved plasma transmission measurements are presented in Figure 3-1 as a

function of delay time between the plasma initiating laser and the probe laser.

Data are presented for the 532 and 1064 nm probe beams, along with the temporal

profile of the plasma-initiating laser pulse as a reference. Zero delay time corresponds to

the peak-to-peak temporal alignment of the two lasers. The transmission profiles reveal

the rapid increase in plasma opacity, with the minimum transmission values of 9 and 16%

recorded 30 and 40 ns following the plasma initiating pulse for the 532 and 1064 nm

probe beams, respectively. It is noted that the 1064 nm probe laser had a considerably

wider pulse width, as discussed above; hence the temporal resolution of the 1064 nm data

is less than the 532 nm data. These observed time scales for the rapid increase in plasma

absorptivity correspond well with the full-width of the plasma initiating laser pulse,














100 - -



80



o 60
E


E 40



20 -- 532 nm probe
---- 1064 nm probe


0
-50 0 50 100 150 200 250 300 350 400 450 500
Delay Time (ns)


Figure 3-1. Plasma transmission as a function of time following plasma initiation using
532 and 1064 nm probe laser beams. The inset laser pulse profile corresponds
to the plasma initiating laser pulse at zero delay time.

which decays by 99% of its maximum value in 30 ns from the peak intensity. The Figure

3-1 data are further analyzed below in the context of electron number density; however,

the transmission data alone are evidence that initial plasma processes are closely coupled

temporally to the plasma initiating laser pulse. The abrupt change in slope of the plasma

transmission profile at the tail end of the initiating laser pulse may be interpreted as

denoting the change from net plasma energy input to net energy dissipation, the latter for

example via radiative transfer and electron recombination (free-bound).

An additional interesting feature of the Figure 3-1 data is the return of the plasma

transmission to nearly 100% (e.g., 98% and 94% at 532 and 1064 nm, respectively) by









500 ns following plasma initiation. This relatively rapid return to an essentially

transparent plasma is significant in the context of radiative transfer and in consideration

of new LIBS configurations such as dual pulse LIBS. If a second laser beam is to be

coupled to an existing laser-induced plasma, temporal delays beyond the first 0.5 |ts will

most likely result in significantly different laser-plasma interactions than those following

the first laser pulse within a few hundreds of nanoseconds. The transmission

measurements may also be considered in terms of the plasma expansion. As reported in

an eariler paper (Carranza and Hahn 2002a), the plasma expands to its characteristic

volume of 1 mm3 within approximatley 20 ns, corresponding to an initial plasma

propagation velocity on the order of 10 km/s.

Time-Resolved Continuum Measurements

Further insight into the plasma processes are revealed by the time-resolved

continuum emission measurements presented in Figure 3-2. The continuum emission

data correspond to the integrated sum of approximately 20 nm bandwidths centered at

either 270, 400 or 530 nm. The data are all normalized to their respective maximum

temporal emission value. The continuum emission data represent radiative transfer from

the plasma, primarily by recombination radiation (free-bound) and Bremsstrahlung

emission (free-free). The temporal scale of the continuum emission corresponds

markedly with the transmission data discussed above. The continuum emission of all

three bands peak between 20 and 40 ns following the peak of the plasma initiating laser

pulse, with the 270 nm continuum band peaking -10 ns before the peak of the 400 and

530 nm bands. A fixed detector gate of 20 ns was used for all measurements; hence

temporal resolution is limited to such a value. For all three spectral bands, the continuum










emission was observed to decay by one order of magnitude during the first 200 ns, which

represents a decrease in effective blackbody temperature of about 70% based on emissive

power per the Planck distribution. Although thermodynamic equilibrium is necessary for

the plasma radiative transfer to approach the blackbody function, such a 70% decrease in

temperature is quite consistent with reported temperature measurements (see above

references) over similar time scales.


270nm
-.*- 400nm
-A- 530nm


0 100 200 300 400 500
Time (ns)

Figure 3-2. Plasma continuum emission (20 nm bandwidth) for three spectral regions
centered at 270, 400 and 530 nm. All data have been normalized to maximum
emission intensity for each respective spectral region.









Electron Density Measurements

To better characterize the laser-induced plasma over the time scales discussed

above, electron number density measurements were also performed. Electron densities

were calculated based on the Stark broadening of the Ha line (656.3 nm), which is well

suited to electron density measurements between 1017 and 1019 cm-3 (Griem 1997).

Following the work of Griem, the Stark line intensities at full-width half-maximum

(fwhm) are related to the electron density through (Griem 1974)

AAfwhm = 2a1/2 (27r)(415)2/3 e (n)273 3-1

where aG 2 is the fractional half-width of the reduced Stark profile (see Appendix C for

al 2 values), e is the electron charge (4.8032x10-10 esu), and ne is the free electron number

density. The present data were reduced using the values of al/2 as reported by Griem

(1974), assuming a plasma temperature of 30,000 K. The inversion of electron densities

from Stark widths is rather insensitive to temperature, as related to the corresponding

reduced line widths. To assess sources of error due to temperature uncertainty in the

present data, the measured Stark line widths were also inverted using temperatures of

20,000 and 40,000 K and the corresponding fractional half-widths. Accordingly, the

error bars reported with the measured free electron densities represent the maximum error

associated with the uncertainty of + 10,000 K. The data were also analyzed using the

more recent Stark broadening constants reported by Gigosos and Cardenoso (1996).

Using the same temperature of 30,000 K, the calculated electron densities were increased
18 3
by 27% at the maximum value (2.6x101 cm-3) and by 12% at the minimum value

(8.9xl017 cm-3), with an average increase of 20% over all measurements. Such

differences are assumed within the expected absolute accuracy of the reported electron










density measurements based on the data analysis as performed in the current study.

Finally, as noted above, recent theories have addressed the indirect and direct coupling

between electron and ion broadenings and the impact on Stark broadening (Oks 2000).

Over the range of the current Stark line widths, application of this more generalized

theory is expected to change the electron densities by 10 to 20 percent based on recent

published comparisons (Parriger et al. 2003).

The measured free electron number densities are presented in Figure 3-3 as a

function of time following the plasma initiating laser pulse.



3.0 1018



2.5 1018


E
0 18
2.0 10


o I
2 1.5 1018

LU

1.0 1018



5.0 1017 I I
0 100 200 300 400 500 600
Delay Time (ns)


Figure 3-3. Free electron number density as a function of time following plasma
initiation, as measured using the Stark broadened Ha line width.









The hydrogen line widths were not sufficiently resolved from the significant

continuum emission for times less than 100 ns. From 100 to 500 ns following plasma

initiation, the free electron number density decreases from 2.6x1018 cm3 to less than 1018

cm-3. As observed by the error bars in Figure 3, the uncertainty associated with the

assumed plasma temperature is greater in the region of more significant Stark

broadening, however, even at the greatest electron number density measured the

corresponding relative error is about 5%. The measured electron density of 1.2x1018 cm-3

at a delay of 350 ns is in exact agreement with the value reported by Yalcin et al. (1999),

as discussed above. It is noted that Yalcin et al. used a similar laser energy and optical

set-up, and reported a similar coupling of -2/3 of the incident laser energy into the

plasma, which also agrees well with the value measured with the current LIBS setup

(Carranza and Hahn 2002b). Overall, the electron density profile reveals a near linear

decay in electron number density between 100 and 400 ns, following by a more gradual

decay over the last three data points. This is somewhat in contrast to the non-linear

changes in plasma transmission and plasma continuum emission during a similar

temporal region, namely between 100 and 400 ns. However, it is noted that the

continuum emission is a highly non-linear process (i.e. Planck distribution), while the

transmission data must be further reduced to make quantitative comparisons.

Plasma Absorption Cross-Section

Modeling Plasma Transmission

The plasma transmission c may be modeled using the Beer-Lambert relationship,

namely:


r = exp{(oNL) + (oNL)2 ...+( NL) }









where (oNL)i represents the optical depth or turbidity of the ith species, with C the

absorption cross-section, N the absorbing species number density, and L the absorption

path length. For the present analysis, it is assumed that free electrons are solely

responsible for attenuation of the probe laser beam, which results in the following

relation between the measured plasma transmission and the absorption coefficient

(defined as Kabs = o ne), namely:

KaL=ln(r). 3-3

While the absorption cross-section is correctly the sum of the absorption and

scattering cross-sections, scattering of the probe laser beam by plasma free electrons,

Thomson scattering, is neglected due to the state of the plasma frequency in relation to

the incident frequencies of the probe laser (Warner and Heiftje 2002). Specifically, for

the first few hundred nanoseconds, the measured electron density is nominally

2x1018 cm-3 which yields a plasma frequency of 1013 Hz using the relation

v = 9x103 n /2 (Thorne et al. 1999). The plasma frequency is the reciprocal of the

Debye length and represents the resonance frequency of the free electron oscillations

about their equilibrium positions. For the 532 and 1064 nm probe lasers, the

corresponding frequencies of 5x1014 and 2x1014 Hz, respectively, are sufficiently close to

the plasma frequency such that significant absorption is expected, thereby dominating

any scattering effects. Equation 3-3 enables conversion of the Figure 3-1 transmission

data to a more quantitative measure of the plasma properties, namely the plasma optical

depth absorbancee) or turbidity.







42


Plasma Absorption

The reduced plasma turbidity data and the measured electron density data both

represent the temporal evolution of the plasma, hence it is interesting to examine the

correlation between these two parameters. To this end, the measured plasma turbidity

and electron density data were paired at each similar temporal delay over the range from

100 to 300 ns following plasma initiation. The result is presented in Figure 3-4, which

reveals a linear relationship between the 532 nm probe laser turbidity and the free

electron density (R=0.998).


1.6


1.4


1.2


y = -1.3542 + 1.0402e-18x R= 0.99795

0 .0 I I I I I I I I I I I I I I I I I I I I I-
1.41018 1.61018 1.8 1018 2.01018 2.21018 2.4 1018 2.6 1018 2.81018

Electron Density (cm3)


Figure 3-4. Measured absorption coefficient for the 532 nm probe laser beam as a
function of the corresponding measured free electron density.









An interesting result of the Figure 3-4 data is that based on the assumption of free

electron absorption and the additional approximation of constant optical path over this

200 ns time period, the slope of the linear fit yields the free-free (inverse Bremsstrahlung)

absorption cross-section. In a complementary recent study, the two-laser approach was

used to profile the overall plasma shape under similar experimental conditions (Carranza

and Hahn 2002a). It was reported that by about 20 ns following plasma initiation, the

plasma had expanded to its characteristic size, with a linear dimension of 1.1 mm along

the path of the orthogonal probe laser beam. Using this path length, the calculated free-

free absorption cross-section is 9.5x1018 cm2 for 532 nm radiation. This value compares

very well with a calculated effective absorption cross section of about 3x10-7 cm2 at 500

nm based on the sum of free-bound and free-free cross sections per H- ion (Griem 1997).

The absorption coefficients Kabs corresponding to the Figure 3-4 data range from 2 to 12

cm-1. These values agree very well with theoretical calculations reported for 532 nm

radiation in a fully ionized hydrogen plasma (Hora and Wilhelm 1970).

Extrapolation for Small Time Scales

Another interesting result of the measured absorption cross-section is that it enables

calculation of electron densities from transmission data at time scales below 100 ns, the

lower limit realized in the present study for direct Stark broadening measurements. The

extrapolated electron density measurements are presented in Figure 3-5 down to 15 ns

following plasma initiation. The free electron density profile peaks at 30 ns

(corresponding to the minimum in 532-nm transmission) at a value of 3.6x1018 cm3 and

18 3
reveals a value of about 2x1018 cm-3 at 15 ns delay. It is noted that the delay of 30 ns

corresponds to the 99% decay point of the incident laser pulse, hence essentially the full-

width. Accordingly, based on the extrapolation of the measured transmission, the










electron density profile peaks at the tail end of the incident laser pulse. Overall, the

change in electron number density during the first few hundred nanoseconds of plasma

evolution is not that great, hence an aerosol particle engulfed by the plasma is subjected

to a relatively consistent frequency of electron-particle collisions. This suggests that the

plasma temperature (i.e. electron energy temperature) could be more important than

absolute free electron density for plasma-particle interactions, namely particle

dissociation and vaporization.


4.0 1018




3.5 1018


3.0 1018




2.5 1018


2.0 1018




1.5 1018


50 100 150 200 250 300

Time (ns)


Figure 3-5. Predicted electron density values as a function of time following plasma
initiation. The data are based on the calculated free-free absorption cross-
section of 9.5x1018 cm2 and the measured 532-nm transmission data.









While temporal regions limited to the first few hundred nanoseconds of plasma

evolution are characterized by significant electron densities, plasma frequencies, and

plasma absorptivity, as described above, the longer delay times (500 to 1000 ns) reveal a

plasma frequency vp -1012 Hz. Such a relatively low plasma frequency is consistent with

the corresponding near transparency of the plasma to the probe laser wavelengths (i.e.

Vincident), as expected for the condition of Vp << Vincident.

The laser-induced plasma investigated-typical of conditions used for LIBS-based

gas and aerosol analysis-is characterized by significant transient changes in electron

density and plasma absorptivity during the first few hundred nanoseconds. The

fundamental change from an absorbing (i.e. optically thick) plasma to a non-absorbing

plasma during this period is important with respect to the radiative transfer within and

from the plasma, and is therefore expected to play a role in the interactions between the

laser-induced plasma and incorporated particles. Hence a better understanding of such

processes provides a starting point for more complex plasma-particle interactions, as well

as furthering the development of plasma modeling capabilities, and advancing LIBS as an

analytical technique.














CHAPTER 4
LIBS PLASMA STABILITY

The shot-to-shot precision of LIBS-based spectral analysis is considered in the

context of laser-induced plasma temporal and spectral characteristics, including the

effects of laser cavity seeding and the presence of ambient aerosol particles within the

laser-induced plasma volume.

Effects of Laser Cavity Seeding

All experiments were recorded using the same spectral window centered at 490nm,

which enabled the simultaneous measurement of atomic emission from four pronounced

peaks corresponding to three gas phase species, namely the hydrogen beta line at

486.13nm (82,259 102,824cm-1), two neutral nitrogen lines at 491.49nm (86,137 -

106,478cm-1), and 493.50 nm (86,221 106,478cm-1), and the 247.86nm neutral carbon

line (21,648 61,982cm-1), which is observed in second order at 495.71nm. A

representative laser-induced plasma spectrum recorded in ambient air for this spectral

window is shown in Figure 4-1. Because all three species correspond to gas-phase

ambient air species, with the carbon attributed to the approximately 300 ppm of ambient

air carbon dioxide and the hydrogen attributed to water vapor, spectral signal variations

due to sample non-homogeneity (e.g. as with aerosol-derived analyte species) is

eliminated. It is noted that the intensity of the carbon and hydrogen atomic emission

lines were significantly reduced in the HEPA filtered air stream due to the reduction in

water vapor and carbon dioxide, as discussed above.












25000


v

>.,
15000
c
o
C=


480 485 490 495 500
Wavelength (nm)

Figure 4-1. Sample spectrum showing atomic emission lines from the plasma spark
recorded in ambient air. The 495.7 nm carbon peak is the 247.86 nm carbon I
line observed in second order.

A series of 400 sequential LIBS spectra were recorded in HEPA filtered air with

the laser operating with the injection seeder. Immediately following this set of data, the

seeder was turned off, and a second series of 400 spectra were recorded with unseeded

laser operation. The full-width, baseline-subtracted peak areas were calculated for each

of the four atomic emission lines on a shot-to-shot basis. The various peak-to-base (P/B)

and peak-to-peak ratios were calculated for each laser shot, and compared for the seeded

and unseeded experiments to assess the effects of laser cavity seeding on LIBS shot-to-









shot precision. For each emission peak ratio, the average value and standard deviation

(N=400 shots) were calculated for both the seeded and unseeded data sets.

Effects on Shot-to-Shot Precision

The results are summarized in Table 4-1 for all four peak-to-base ratios and all six

peak-to-peak combinations, with the latter including both intraspecies and interspecies

comparisons. As an example, the N491.5/N493.5 and the N491.5/C247.9 atomic emission ratios

are presented in Figure 4-2 for the entire 800 shot sequence. For both cases, Figure 4-2

reveals no apparent differences between the shot-to-shot precision due to seeded or

unseeded laser operation. For example, the nitrogen-to-nitrogen emission ratio

(N491.5/N493.5) yields an average value of 0.458 with a relative standard deviation (RSD)

of 3.39% with laser cavity seeding, while the average value with unseeded operation was

0.457 with a RSD of 3.45%.



Table 4-1. Ratios of the area-integrated spectral peaks (peak-to-base and peak-to-peak)
for seeded and unseeded laser operation. Each data entry represents the
average and standard deviation based on 400 single-shot spectra recorded in
HEPA filtered air. The 247.9 nm carbon peak was observed in second order.

Seeded Unseeded
Average Std Dev RSD(%) Average Std Dev RSD(%)
H(486)/Base 24.5 6.37 26.0 24.5 6.74 27.5
N(491)/Base 13.6 0.535 3.93 13.7 0.537 3.93
N(493)/Base 27.7 0.745 2.69 27.8 0.700 2.52
C(248)/Base 2.63 0.416 15.8 2.58 0.419 16.3
H(486)/N(491) 1.95 0.513 26.3 1.95 0.548 28.1
H(486)/N(493) 0.892 0.233 26.1 0.890 0.248 27.9
H(486)/C(248) 8.94 1.98 22.2 9.13 2.13 23.3
N(491)/N(493) 0.458 0.015 3.39 0.457 0.016 3.45
N(491)/C(248) 4.69 0.785 16.8 4.81 0.796 16.6
N(493)/C(248) 10.2 1.67 16.3 10.5 1.70 16.2













10
Seeded Laser Unseeded Laser
0


o o o



-W o o 0 @0 a" U
0 0

0 00
o
S( N 491.5 / C 247.9)



1o





(N 491.5 / N 493.5)






0 100 200 300 400 500 600 700 800
Shot Count

Figure 4-2. N491.5/N493.5 and N491.5/C247.9 peak-to-peak atomic emission ratios for an 800
shot sequence. The data were recorded in HEPA filtered air for both seeded
operation (shots 1-400) and unseeded operation (shots 401-800).

It is noted that the ratio of the statistical weight and the transition probability (gA)

for these two lines (1.616x106 s-1/3.520x106 S-1) predicts an intensity ratio N491.5/N493.5 Of

0.459, which is in excellent agreement with the experimental value of 0.458. However,

because the two emission lines share a common upper energy state (106,478 cm-1), the

intensity ratio is expected to be rather insensitive to fluctuations in plasma properties

such as temperature. The relative precision of these two lines is further discussed below

in the context of the respective peak-to-base ratios. Comparison of the nitrogen-to-

carbon emission ratios between the unseeded and seeded laser operation revealed similar









results. These two atomic transitions are characterized by significantly different upper

energy levels, namely 106,478 cm-1 and 61,982 cm- for nitrogen and carbon,

respectively, hence the ratio is expected to better reflect plasma fluctuations. As

observed in Figure 4-2, the shot-to-shot fluctuations were observed to vary more than the

nitrogen-to-nitrogen peaks, but were not observed to change when comparing seeded and

unseeded operation. Specifically, the N491.5/C247.9 emission ratio yielded an average value

of 4.69 with a RSD of 16.8% with laser cavity seeding, while the average value with

unseeded operation was 4.81 with a RSD of 16.6%, the difference being statistically

insignificant.

An examination of all peak-to-peak combinations presented in Table 4-1 reveals no

significant variation in average values and RSD between the seeded and unseeded laser

operation. Of the six peak-to-peak ratios, four showed slight increases in RSD with

unseeded operation, while two revealed slight decreases in RSD with unseeded operation.

Overall, the average change in RSD was equal to a 3% increase when changing from

seeded to unseeded laser operation. The data support the present conclusion that the

effects of laser seeding are negligible with respect to laser shot-to-shot precision based on

both interspecies and intraspecies atomic emission intensity ratios.

The data also support the same conclusion if the single atomic emission line

intensities are explored, specifically using the integrated emission peak to continuum

intensity ratios, referred to as the peak-to-base (P/B) ratio. This metric is widely used for

LIBS-based spectral analysis as it enhances data precision by normalizing for changes in

absolute plasma emission signal levels (Radziemski et al. 1983, Carranza and Hahn

2002a), which can be considerable. Using the P/B ratio as a comparison, the 491.5 and









493.5 nm nitrogen P/B ratios yielded an average value of 13.6 with 3.9% RSD and 27.7

with 2.7% RSD, respectively, for the seeded laser operation. Unseeded operation yielded

an average of 13.7 with a 3.9% RSD and 27.8 with 2.5% RSD, for the 491.5 and 493.5

nm lines, respectively. As with the peak-to-peak ratios, the differences in the peak-to-

base ratios between unseeded and seeded laser operation are statistically insignificant. It

is also noted that the average RSD of the two nitrogen peak-to-base ratios is 3.3%, which

is in agreement with the RSD recorded for the peak-to-peak ratio of these same two lines.

The overall agreement in analyte precision between the peak-to-base and peak-to-

peak ratios, given the common upper state of these two nitrogen lines, suggests a limiting

precision dictated by spectral shot noise, which is significant with single-shot intensified

CCD spectra. Overall, when comparing seeded and unseeded laser operation, similar

results were also obtained with the peak-to-base ratios of the hydrogen and carbon

emission peaks, as shown in Table 4-1, although the relative standard deviations of these

two species are greater than the values obtained with nitrogen. As an example, the peak-

to-base ratios of the 491.5-nm nitrogen line and the carbon line are presented in Figure 4-

3 for the entire 800 shot sequence. It is noted, however, that the increased RSD with

carbon and hydrogen reflects the relatively low concentrations of these two species in the

HEPA filtered stream, hence the spectra are characterized by poor analyte signal-to-noise

ratios overall. In ambient air, for example, the RSD of the hydrogen peak-to-base ratio is

reduced to 3.4%, which is comparable to the nitrogen values. In aggregate, the above

results are evidence of the limited effect of laser cavity seeding on the analyte signal

precision, both with peak-to-peak and peak-to-base ratios, and also demonstrate with










nitrogen the high precision obtainable with the LIBS technique under homogenous

sample conditions and high analyte signal-to-noise.


18


16


14


12

0
y 10

Co
S8
o2

(D 6
0


0 100 200 300 400 500 600 700 800
Shot Count

Figure 4-3. N491.5 and C247.9 peak-to-base atomic emission ratios for an 800 shot
sequence. The data were recorded in HEPA filtered air for both seeded
operation (shots 1-400) and unseeded operation (shots 401-800).




Temporal Effects on the Plasma Formation Process

While the effects of laser cavity seeding were not apparent in the resulting LIBS

spectra of homogenous gas phase analytes, it is still worthwhile to explore the potential

effects of seeding on the overall plasma formation process. For example, the analysis of

aerosol species involves the coupling of the discrete particles with the breakdown and










plasma formation processes; hence temporal fluctuations following breakdown initiation

may be important. The experimental system was designed to capture individual laser

pulse waveforms prior to breakdown (i.e. incident waveform) and following each plasma

event (i.e. transmitted waveform). In HEPA filtered air, reference incident waveforms

and transmitted waveforms were recorded for each laser pulse for a sequence of 100 shots

under seeded operation. The seeder was turned off and a second set of reference and

incident waveforms were recorded for unseeded laser operation. A representative pair of

incident and transmitted waveforms is shown in Figure 4-4, as averaged over a series of

100 shots.


0.090



0.075



0.060



0.045



0.030



0.015



0.000


0 8 16 24 32 40 48 56 64
Time (ns)

Figure 4-4. Representative waveforms for the incident laser pulse and the transmitted
laser pulse following laser-induced breakdown. Each waveform represents an
average of 100 shots.









The plot illustrates the observed time-dependent plasma absorption of a significant

fraction of the incident laser irradiance. The appearance of a second local maximum is

noted in the transmitted waveform about two-thirds of the way into the laser pulse.

Similar behavior was recorded in a series of recent studies (Bindhu et al. 2003, 2004),

and is attributed to a brief saturation or self-regulating condition during the plasma

formation process.

The temporal onset of significant plasma absorption of the incident waveform may

be defined by the location of the first local maximum in the transmitted waveform with

respect to the local maximum of the incident waveform. This temporal truncation is

defined in Figure 4-4. Examination of the truncation time on a shot-to-shot basis

revealed a degree of variation of the truncation time about the mean truncation time. This

variation may be interpreted in terms of the nature of the breakdown process, where the

exact spatial and temporal point of breakdown initiation is expected to fluctuate due to

variations in the laser pulse field coupled with the statistical nature of multi-photon

ionization processes. It is expected that the degree of variation may be influenced by the

presence or absence of laser cavity seeding, in that with cavity seeding the temporal laser

mode is expected to be more uniform. The temporal variation of breakdown initiation

was quantified for 200 laser pulses (100 seeded and 100 unseeded) recorded in the HEPA

filtered air stream, with the exact truncation time calculated on a shot-to-shot basis. The

mean truncation time was observed to change by slightly less than 2 ns between seeded

and unseeded operation. However, it was observed that the laser cavity seeding produced

a slight change in the shape of the overall pulse waveform. Specifically, the average

point of breakdown initiation remained relatively constant with respect to the leading









edge, while the position of the maximum intensity in the incident waveform shifted

further into the pulse with seeded operation. Therefore, to provide the more

representative measure of variation in the truncation time on a shot-to-shot basis, the data

were further processed by calculating the variation in truncation time about the mean

truncation time for both seeded and unseeded operation.

Cumulative probability plots of these two data sets were prepared and are shown in

Figure 4-5. Both unseeded and seeded operation produced a set of truncation times that

cluster primarily within + 1 ns of the mean, however, careful analysis revealed a broader

variation of the breakdown initiation point with unseeded operation. Specifically, the

absolute range in truncation time about the mean was found to be 3.2 ns for the seeded

data and 3.6 ns for the unseeded data. The standard deviation about the mean was 0.26 ns

for seeded operation, and 0.29 ns for unseeded operation. Furthermore, four data points

corresponded to truncation times greater than 1.5 ns behind the mean truncation time for

unseeded operation, compared to only 1 such point for seeded operation, while five data

points corresponded to truncation times more than 1 ns ahead of the mean time for

unseeded operation as compared to only two for seeded operation. Hence the extreme

variations in the distribution of truncation times are dominated by unseeded cavity

operation data by a factor of three as compared to seeded operation. The conclusion is

that the operation of the cavity seeder does produce a more repeatable breakdown process

with respect to the temporal onset of laser-induced breakdown. As discussed above, such

effects may play a role in further understanding of the overall breakdown process,

notably in the presence of non-homogeneous samples such as aerosol sampling. Overall,

cavity seeding appears to play a role in enhancing the reliability of the breakdown











process, although as seen above, such an effect does not translate into any significant


increase in precision with LIBS-based spectral analysis. These data suggest that the early


portion of the incident laser pulse, which is coupled strongly to the breakdown process,


may be an important factor in the resulting precision of the analyte signal. Therefore,


overall laser pulse temporal stability, as observed with cavity seeding, may not noticeably


affect signal precision because of the relative unimportance of pulse stability following


the point of breakdown initiation.




2 .0
---- Seeded d
1.5 ---Unseeded


--------------------------- - - -
1.5
0 .0


(.3

0.0

0 -0.5

.o
zr -1 .0


-1 .5


-2 .0
P 0 0 0 0 0 0 U)

P e rce nt
Figure 4-5. Cumulative probability distribution plot for both seeded and unseeded laser
operation.

Effects of Aerosol Loading

To observe the potential effects of aerosol loadings on LIBS precision, it is


necessary to compare a constant analyte concentration under changing loadings of


concomitant particles. Recall that the HEPA filtered air stream contained significantly









less carbon and hydrogen, hence only the nitrogen emission lines were compared to

assess the effects of aerosol loading. The removal of carbon and hydrogen only slightly

altered the nitrogen signals, with the 400-shot N491.5/N493.5 average equal to 0.44 in

ambient air (cavity seeding) and 0.46 in HEPA filtered air (cavity seeding). As before,

a series of 400 sequential LIBS spectra were recorded with the laser operating directly in

ambient air with injection seeding. Immediately following this set of data, the HEPA

filtered jet was turned on and a second series of 400 spectra were recorded with seeded

laser operation. The full-width, baseline-subtracted peak areas were calculated for each

spectrum, and were used to calculate the N491.5/N493.5 intensity ratio on a shot-to-shot

basis.

Effects on Shot-to-Shot Precision

The intensity ratios are shown in Figure 4-6 for the entire 800 shot sequence. The

nitrogen-to-nitrogen emission ratio (N491.5/N493.5) yielded an average value of 0.440 with

a RSD of 5.93% in ambient air with laser cavity seeding, while the average value in the

HEPA filtered air stream was 0.458 with a RSD of 3.39% with cavity seeding. Near

identical results were obtained with unseeded laser operation, specifically, the average

(N491.5/N493.5) intensity ratio was 0.440 with a RSD of 5.81% in ambient air, while the

average value in the HEPA filtered air stream was 0.457 with a RSD of 3.45%. The

LIBS shot-to-shot precision, as characterized by the relative standard deviation, was

reduced from 5.93% to 3.39% with seeded operation, and from 5.81% to 3.45% with

unseeded operation, by the exclusion of the ambient air particles as realized with the

HEPA filtered air stream. Identical results were obtained when comparing the influence

of ambient air on the individual nitrogen emission lines, namely the peak-to-base ratios.

The 491.5-nm emission line produced an average P/B of 11.0 with a RSD of 6.2% in










ambient air with cavity seeding, while the average P/B in the HEPA filtered air stream

was 13.6 with a RSD of 3.9% also with cavity seeding. With unseeded laser operations,

the RSD of this emission line was found to similarly decrease from 6.4% to 3.9%.

Overall, both seeded and unseeded laser operation were consistent with realizing a nearly

60% improvement in shot-to-shot precision in the absence of concomitant aerosol

particles from the sample stream. Clearly the effects of aerosol loading dominate the

effects of laser cavity seeding as manifest in the shot-to-shot precision of both the analyte

peak-to-peak and peak-to-base emission ratios.


0.70
Ambient Air HEPA Filtered Air



0.60


0

OO
0.30 -
S 0.0 .o I I I I

0.4 0 0 0 0o6 O

C") 0
z

0.30





0 100 200 300 400 500 600 700 800
Shot Count

Figure 4-6. N491.5/N493.5 peak-to-peak atomic emission ratios for an 800 shot sequence.
The data were recorded in ambient air (shots 1-400) and in HEPA filtered air
(shots 401-800). For all shots, the laser was operated with the cavity seeder.









A commercial light scattering based instrument was used to assess the average

ambient air particle loadings in the laboratory air. Typical loadings were approximately

900 particles per cubic centimeter in the size range from 100 nm to 2.5 |tm, with about

80% of the particles between 100 and 200 nm, 18% between 200 and 300 nm, and the

remainder between 300 nm and 2.5 |tm, with the largest fraction of the remainder

weighted toward the smallest size range. In an earlier study, estimates were made for the

effective laser-focal volume and the laser-induced plasma volume for aerosol interactions

(Carranza and Hahn 2002b), namely -0.01 mm3 and 1.4 mm3, respectively. The

probability of a direct laser beam/aerosol particle interaction is about 1%, while the

probability of laser-induced plasma/aerosol particle interaction is about 70%, using

Poisson sampling statistics based on these effective sample volumes and aerosol loadings,

as described previously (Carranza and Hahn 2002b). Clearly the ambient air particle

loadings are such that the probability of interaction with the resulting plasma process is

significant.

Temporal Effects on the Plasma Formation Process

Finally, the effects of aerosol presence on the temporal onset of laser-induced

breakdown were assessed in a similar manner as described above. Incident and

transmitted waveforms were collected in both ambient air (100 shots) and in HEPA

filtered air (100 shots) with the laser operated with cavity seeding for all measurements.

For both cases, the range about the mean truncation time was identical, namely 3.2 ns.

The relative standard deviation about the mean time was 0.261 ns in the HEPA filtered

air and 0.254 ns in the ambient air. Therefore, in contrast to the effect on plasma spectral

precision, it appears that aerosol inclusion does not contribute significantly to temporal






60


variations in the onset of plasma spark inception. This is in agreement with the

probability of direct laser beam/particle interaction (-1%) as discussed above. While the

1% value represents the average probability of the laser beam encountering a single

aerosol, the probability of such an encounter directly within the temporal/spatial regime

of breakdown initiation must be much lower, thereby making the direct impact of such

interactions of second-order importance in the actual breakdown initiation, as reflected in

the temporal variation of the breakdown onset.














CHAPTER 5
LIBS CALIBRATION

Atomic line spectra, measured using the LIBS technique, are expected to increase

linearly with an increase in analyte concentration within the plasma volume. Most

studies using LIBS for quantitative measurement construct such a calibration curve. A

widely used assumption in LIBS research is that the slope of the curve remains

unaffected by size or phase differences in analyte species. The following experimental

results were compiled for the purpose of analytical examination of this assumption.

Experiments required the use of several different analyte sources. For discussion

purposes, the nebulized Atomic Absorption (AA) and Inductively Coupled Plasma

Atomic Emission Spectroscopy (ICP-AES) grade carbon standard solutions will be

referred to as the AA-ICP case, the nebulized 30-nm polystyrene suspensions will be

referred to as the 30-nm PS case, and the three gas-phase experiments will be referred to

as the CO2, CO, and CH4 cases.

Raw and Baseline-Subtracted Spectra

Representative spectra in the range of the 247.86 nm carbon line are shown in

Figure 5-1 for both gas-phase and solid-phase carbon sources. The upper set of four

spectra represents the unprocessed plasma emission as recorded, demonstrating the

presence of the carbon atomic emission line in combination with significant continuum

emission. The lower set of three spectra represents background-subtracted spectra, which

are characterized by the presence of the distinct carbon emission line and an essentially

zero baseline. The baseline-corrected spectra were obtained by scaling the average DI







62


water only spectrum (also shown in Figure 5-1) to the carbon-containing spectrum of

interest for each data set, and then subtracting the scaled background. This process was

very successful in removing the continuum emission, as evidenced in Figure 5-1 by the

flat spectral response surrounding the distinct, isolated carbon emission peak.


7000



6000



5000



4000



3000


2000



1000


0 ... I I I I I I I I I.-

238 240 242 244 246 248 250 252 254 256 258
Wavelength (nm)




Figure 5-1. LIBS spectra in the vicinity of the 247.86 nm carbon atomic emission line for
different sources of carbon. The upper four spectra represent raw data, the
lower three curves are background-subtracted spectra using the deionized
water spectrum (DI) as a blank. All spectra have the same scale and have
been shifted vertically for clarity.









Calibration Using Atomic Spectra

The carbon signal used for this study was the calculated peak-to-base (P/B) ratio.

This metric provides a precise measure of laser-induced plasma atomic emission by

normalizing out fluctuations in the absolute plasma emission signal (Radziemski et al.

1983). The reported P/B was calculated using the integrated full peak area of the 247.86

nm carbon emission line divided by the integrated signal of the scaled background

spectrum over the same spectral width.

Using this procedure, calibration curves were calculated over the range of carbon

mass concentrations for each of the carbon analyte sources. The resulting calibration

curves are presented in Figure 5-2 for all five analyte sources. All five curves display

high linearity, with the slopes and corresponding regression coefficients summarized in

Table 5-1.

Despite the constant experimental conditions and similar range of atomic carbon

mass loadings for all five species, the resulting slopes of the calibration curves reveal

marked differences. Most significantly, the calibration slope for the AA-ICP standard

was roughly double that of the 30-nm PS curve, and eight times that of the three gas-

phase species curves. Consistent with previous findings, the three gas-phase calibration

curves group very closely to one another, with the differences in slope being statistically

insignificant. In addition, experiments for the three gas-phase species were repeated at

discrete data points with the carbon-containing gas stream fed through the nebulizer.

This exercise did not reveal any significant changes in the resulting P/B ratios with

respect to the Figure 5-2 data, confirming that nebulization of the analyte species is not

responsible for the marked difference in analyte response between the gas-phase and

solid-phase carbon species. Clearly, the current results contradict the assumption of






64


analyte signal independence on the analyte source when one considers both solid-phase

and gaseous-phase analyte sources, or even between varying sized solid-phase analyte

sources. Prior to further analysis and discussion of the observed analyte source effect, it

is useful to consider possible changes in the plasma itself (i.e. true matrix effects).


2,000


4,000 6,000
Carbon Concentration (|jg/m3)


8,000


10,000


Figure 5-2. Calibration curves based on the 247.86 nm (C I) atomic emission line for the
five carbon analyte sources investigated. Error bars represent one standard
deviation.









Table 5-1. Summary of the carbon calibration response and plasma conditions.

Free Elec. Plasma CN
Analyte Calibration Curve Free Elec. Plasma CN
R-value Density Temp. Signal
Source Slope 1017 cm-3) (K) (P/B)


DI Blank N/A N/A 1.06 4.4% 5189 + 80 N/A


AA-ICP 8.24E-04 + 5.2 % 0.999 1.03 + 4.6% 4973 + 10 22.27


30-nm PS 4.25E-04 5.8 % 0.998 1.06 4.4% 4799+ 100 15.03


CO 1.05E-04 9.6 % 0.997 1.04 4.5% 4952 + 40 6.98


CO2 1.16E-04 3.2 % 0.997 1.00 4.7% 5114 +20 6.07


CH4 9.38E-05 4.2 % 0.998 1.06 4.4% 4994 + 70 10.18


Calibration Using Molecular Bands

In addition to the carbon response as measured with the 247.86 nm atomic emission

line, it is also useful to consider additional carbon-related emission bands. Specifically, if

the differences in analyte response observed with the 247.86 nm line are the result of

differences in the amount of atomic carbon generated within the plasma volume, then one

might expect a similar trend with other carbon emission features such as those from CN.

Measurements of the CN violet system (B2E X2E) have been widely observed in laser-

induced plasmas and have been used for quantitative analysis (Boyain-Goitia et al. 2003,

Ferioli et al. 2003). In the current study, measurements of the CN violet band were

performed and analyzed using the 388.3 nm (0,0) and 387.1 nm (1,1) emission lines.










Representative background-subtracted (using DI water only) CN emission spectra are

shown in Figure 5-3.


2000





1500





1000
U,




500


370 375 380 385 390 395 400
Wavelength (nm)

Figure 5-3. LIBS spectra in the vicinity of the CN violet system for the AA-ICP and CO2
analyte sources. The curves have been background-subtracted using the
deionized water spectrum as a blank. All spectra have the same scale.

The P/B values were calculated for the two lines as described above and then

averaged to quantify the CN emission, with the results summarized in Table 5-1 for all

five carbon analyte sources. Analysis of the CN emission data reveals a trend that is in

agreement with the analyte response observed with the carbon atomic line, namely a

significant reduction in analyte signal when comparing the gas-phase species to the solid-

phase species. The primary differences between the atomic carbon and CN emission









results are (i) that the ratios of the various solid-phase to gas-phase P/B values were

reduced by about of factor of two, (ii) and that while the CN emission response of the

CO2 and CO experiments was consistent, the CN response of the CH4 experiment was

about 50% greater than the CO2 and CO result.

In summary, the above results clearly support a significant effect of the analyte

state (i.e. solid vs. gas phase) on the resulting analyte response, as measured by the

corresponding analyte emission peaks. The similar response of the 247.86 nm carbon

atomic emission line for all three gas species (which have markedly different C-H-O

ratios), coupled with the identical free electron densities for all plasma conditions, lead to

the conclusion that the observed differences in analyte response are not the result of

differences within the plasma following dissociation (i.e. not emission quenching or

analyte recombination effects). In contrast, the differences in analyte response are

concluded to arise from differences in the amount of analyte originally dissociated to

atomic carbon within the laser-induced plasma volume. To discuss the proposed reasons

for such a disparity in analyte dissociation, it is first necessary to consider in more detail

the nature of the solid-phase carbon and the overall role of particulates in laser-induced

plasmas.

Analysis of Matrix Effects

Although care was taken to provide a consistent overall matrix for all experiments,

including the nebulization of DI water for all cases, it is necessary to examine whether

the different analyte species effected the bulk plasma. For example, large variations in

plasma properties, such as electron density, resulting from the different analyte species

could explain the different analyte response curves. One measure of consistent plasma

formation is the continuum plasma emission. Careful examination of the Figure 5-1 data









reveals identical background-subtracted spectra about the carbon emission peak. This

was also the case for the remaining analyte sources, namely CO and CH4. Based on the

consistent shape of the plasma continuum emission between the DI water case and all

five analyte cases, one may conclude similar plasma properties. However, to provide a

more quantitative measure of plasma consistency, the free electron densities, from the

measurements of H, line Stark broadening, were also evaluated. The fractional half-

widths used to reduce the Stark broadening data were those reported by Griem (1974),

assuming a plasma temperature of 10,000 K, which is appropriate for the delay times and

laser pulse energy used. As discussed previously, the inversion of electron densities from

Stark widths is rather insensitive to temperature, as related to the corresponding reduced

line widths. Nonetheless, to assess sources of error due to temperature uncertainty in the

present data, the measured Stark line widths were also inverted using a temperature of

20,000 K and the corresponding fractional half-widths. Accordingly, the uncertainties

reported with the measured free electron densities in Table 5-1 represent the maximum

error associated with the uncertainty of 10,000 K and the uncertainty associated with the

measured line width (-0.02 nm). This analysis revealed that the free electron densities

were statistically identical (lxl017 cm3) for all five analyte sources as well as for the

baseline DI water case (see Table 5-1), indicating no fundamental changes in the plasma

state with the various carbon species. To confirm this, independent measurements of

plasma temperature were made using a ratio of N2 lines at 375.95 nm and 391.50 nm

using the method ofLaux et al (2001). The results are summarized in Table 5-1. The

temperatures are considerably lower than 10,000 K as the spectra were recorded at longer









delay (40 [ts) to capture the molecular emission, however the relative uniformity over all

species suggest a rather invariant plasma over the range of experimental conditions.

Aerosol Loading and Size Considerations

The nebulizer system used in the current experiments was investigated in detail in a

previous study (Hahn et al. 2001). In that earlier work, AA ICP-AES grade solutions of

iron and titanium were nebulized under essentially identical conditions as the present

study. The resulting solid particulates were collected and analyzed using TEM. The

particles produced following desolvation (i.e. particles at the LIBS sample point) were

found to group about a mean particle size less than 100 nm, and as expected, were found

to scale with the mass concentration of analyte in the nebulized solution. Based on this

earlier study, the current nebulization of solutions of carbon as oxalic acid over the range

of 500 to 5000 [tg-C/ml is expected to produce carbon-rich aerosols primarily in the size

range from about 50 to 100 nm. Qualitative light scattering measurements through the

LIBS sample chamber revealed the presence of a uniform aerosol cloud of similar

scattering intensity for both the AA-ICP and 30-nm PS experiments. This uniform

scattering was absent with the three gas-phase carbon species. The chemical nature of

the submicron-sized aerosols created from nebulization of the AA-ICP solutions was not

determined, but it is expected that the particles would be rich in oxalic acid, which takes

the form of a transparent, colorless crystal in the solid phase. TEM analysis of the 30 nm

PS suspension particles confirmed a mean size of 28 nm with a relative standard

deviation of approximately 16%.

Assuming that gas molecules have an effective size on the order of one Angstrom,

there was a pronounced range of sizes realized with the three groups of analyte species:









namely -0.1 nm for the gas phase species, -30 nm for the polystyrene particles, and -100

nm for the nebulized AA-ICP standards. It is noted that this entire size range falls well

below a previously established 2.1 |tm upper size limit for complete dissociation of

silicon oxide particulates in a comparable laser-induced plasma (Carranza and Hahn

2002c); hence the current results support the earlier conclusion that the varied analyte

response is not the result of incomplete dissociation within the plasma stemming from

size effects.

Proposed Mechanism for Particle-Plasma Interaction

As an alternative to the incomplete dissociation argument, it is proposed that a

mechanism is in effect that selectively (i.e. dependent on analyte state) perturbs the

analyte species during plasma formation and growth, thereby affecting the resulting

amount of analyte present within the subsequent laser-induced plasma.

Such a mechanism must selectively remove molecular species from within the

resulting plasma volume to account for the analyte response data observed in this study.

In addition, one would expect the depletion of molecular species to be combined with a

secondary size effect, such that solid-phase species preferentially remain within the

plasma volume in a manner that scales with particle size or mass. Previous

measurements of the plasma volume using a temporal probe laser revealed that the laser-

induced plasma grew from its initial breakdown volume (i.e. laser focal volume) to an

effective plasma volume of about 1 mm3 over a time-scale of about 20 ns (Carranza and

Hahn 2002b). One may consider that the rapidly expanding plasma can be thought of as

a shockwave-like phenomenon, in which the highly energetic pressure and electron wave

rapidly expands from the plasma kernel. As the plasma wave expands, molecular and











particulate species are pushed toward the edge of the plasma volume. However, due to

the many orders of magnitude difference in the mass of gas-phase species (molecules)

and solid-phase species (particulates), it is expected that the efficiency at which a given

species is carried by the plasma wave will scale inversely with particle mass. In other

words, an effective analyte slip factor will exist as the plasma wave expands. A

schematic of this proposed effect is presented in Figure 5-4.


Figure 5-4. Schematic detailing the proposed interactions between the rapidly expanding
plasma wave and the analyte species. Large particulates or aerosols (upper
image pair) are shown to resist outward radial transport by the plasma wave,
whereas nano-scale analyte species (lower image pair) will tend to be depleted
from the plasma core.

Within such a model, analyte species at the molecular level would tend to be


selectively swept out of the plasma core toward the plasma edges, where temperatures

and electron densities are significantly lower. Analyte species at the particulate scale


S 0
0
0
0
*
S 0
0 0
0. *
0 *
S 0
S
0 0 0









would be less affected by these drag forces due to inertial effects, and would therefore

selectively remain in the plasma core volume, leading to more complete dissociation and

therefore a greater effective analyte concentration and emission response. Some

preliminary LIF imaging of LIBS plasmas using gold aerosols and particulates support

this model (Nakata and Okada 1999).

Based on such a working model, it is expected that a finite particle size (i.e. mass)

exists where the efficiency of remaining within the plasma core reaches a plateau,

essentially the point in which all such particulates effectively remain within the core

plasma volume. While determination of such a particle size requires further validation of

the hypothesis itself, as well as additional experimental work and modeling, it is useful to

consider some limiting values based on the available data. A linear analyte response was

observed for silicon oxide particles over the size range 1-2.1 |tm, as reported previously

(Carranza and Hahn 2002c). Hence one micron may be considered an upper limit for the

estimated particle size necessary to minimize analyte depletion within the plasma

volume. Based on the current study, particle size effects are concluded to exist over the

30 to 100 nm size range. Accordingly, one may reasonably conclude that the critical size

for an analyte species such that preferential removal from the plasma core is minimized is

in the range of 0.1-1 |tm. Particles above this range are expected to display a linear

analyte response (i.e. no particle size effect) up to the limit for complete vaporization.

Particles below this range, including down to the molecular regime, are expected to

display significant size effects with respect to the analyte response.

Overall, the range of data reported in the current study supports the conclusion that

there is preferential accumulation and dissociation of solid-phase analyte species within









the plasma center resulting in an enhanced analyte response, as measured by atomic and

molecular emission, as compared to gas-phase analytes. It is proposed that such an effect

is realized by the interaction of the analyte species (solid or gas phase) with the

expanding plasma, with relative inertial effects playing a key role. To date, published

studies have explored the interaction of aerosols in laser-induced plasmas (Lushnikov and

Negin 1993, Schoolcraft et al. 2000, Gomushkin et al. 2004); however, the coupling of

such interactions, including particle dynamics and plasma emission processes, remains an

ongoing issue. An important outcome of this study is demonstration of the need to

produce calibration schemes for use with LIBS that reflect as much as possible the

physical state of the analyte species of interest.














CHAPTER 6
SPATIAL CONSIDERATIONS IN PLASMA EMISSION

The physical process and time-scale of particle dissociation within a laser-induced

plasma and the subsequent spatial diffusion of analyte atomic species released from the

particle throughout the plasma has not been well understood to date. Even recent

theoretical models of laser-induced plasma behavior treat species dissociation and

spectral analyte emission within a laser-induced plasma as temporally instantaneous and

spatially homogeneous phenomena. In an effort to further explore these issues of basic

plasma science which underlie LIBS methodology-knowledge of particle vaporization

and dissociation and the presence and significance of spatial inhomogeneities in analyte

emission; namely, experimental data and analysis that considers and accounts for spatial

and temporal variability in atomic emission from aerosol sources within a laser-induced

plasma is presented. It should be noted that the term dissociation is used herein as a

general description to denote a variety of complex processes including particle melting

and vaporization, fragmentation, sublimation, molecular ionization, atomic ionization, as

well as concomitant heat and mass transfer.

Background Emission

A characteristic image of the laser induced plasma for the experimental blank is

shown in Figure 6-1, below. The figure represents the on-axis projected image of the

three-dimensional plasma, and gives a measure of both the optical size and intensity of

the plasma. The plasma images were recorded for a series of delays and detector gate

widths, as presented in Chapter 2. To first gain an understanding of the plasma









characteristics in the absence of aerosols, images were recorded in the presence of only

HEPA filtered air and nebulized DI water, as described previously.


Figure 6-1. Two-dimensional on-axis projection of the laser-induced plasma in an
aerosol-free environment.

A series of 20 individual images were averaged together for each delay time. When

compared over time, the averaged images revealed that the plasma size grew essentially

linearly with time, while the plasma emission intensity, as measured by ICCD counts,

decayed exponentially in time. It should also be noted that beyond the longest reported

delay time in the study, 30s, the plasma shape was observed to deviate from a spherical

structure toward a more annular intensity profile. This atypical structure was observed

between delay times from about 30 to 50[ts, after which the measured intensity profile

once again took on a spherical structure. The overall intensity of the image in the range

of 30 to 100[ts was characterized by a marked decrease, making qualitative analysis









difficult. The primary region of analysis was therefore confined to delay times up to

30[ts, leaving the latter delay times and related plasma behavior as a subject for future

work.

By directly measuring the cross-sectional area of the plasma image, the growth of

the plasma was quantified over the range of delays, with the results shown in Figure 6-2

below. As observed in the figure, the plasma growth in size is basically linear with time,

other than a brief deviation at a delay of 8[ts.


0 5 10 15 20 25 30 35
Intensifier Gate Delay [jis]

Figure 6-2. Change in observed diameter of the plasma image as a function of time delay
from the end of the laser pulse. Error bars represent an estimated 0.2mm
uncertainty in locating the plasma edge above intensity fluctuations.

The overall plasma size in the temporal region of interest, -3.5mm diameter, is

somewhat larger than the emission-based plasma diameter of 1.7mm reported previously






77


by Carranza and Hahn (2002b). However, in the earlier study, the reported volume was

an effective plasma volume based on a spectral-equivalent concentration and an apriori

determined analyte mass released from the aerosol, hence the absolute agreement

between the two different measurements is not necessarily expected.

In addition to measuring the plasma size, the images were also used to calculate the

emission intensity profile. A 400-pixel intensity average about the center of each plasma

image was taken and the results are shown in Figure 6-3.


1.20



1.00


0.80



0.60



0.40



0.20


0.00


5 10 15 20 25 30
Delay (us)


Figure 6-3. Center-averaged plasma intensity, normalized to the highest measured value
at 21as, as a function of time delay from the end of the laser pulse. Error bars
represent +1 standard deviation.









The plasma emission intensity, which corresponds to continuum emission, is

observed to decay exponentially with time, which is consistent with the intensity profiles

reported in Chapter 3. Clearly the plasma images are consistent with the spectral data

produced by collecting and dispersing the spatially integrated plasma emission, although

more detailed comparisons are made below.

Spatial Distribution of Atomic Emission

For investigation of the plasma-particle interactions using the plasma imaging

system, deionized water was replaced in the nebulizer with the stock suspension of

nominally 2-[tm borosilicate microspheres described in Chapter 2. The particle

suspension was adjusted to promote the creation of an appropriate particle number

density at the sample point such that the LIBS plasma was expected to sample an aerosol

particle with a reasonable probability, namely with a -10-20% sample rate. Introduction

of the nebulized aerosol stream to the LIBS sample chamber led to the appearance of

localized, luminous bursts or "hot spots" within the plasma image-the position of which

were observed to be spatially random throughout the plasma. An example of such an

image is shown below in Figure 6-4. That this emission was due to the spatially non-

homogeneous presence of calcium atoms is strongly supported by the insertion of the

narrow line filter (396.8nm) in the image collection path that corresponds to the ionized

calcium atomic emission line at 396.85nm (0-25,192 cm-1). Hence the localized,

luminous feature is concluded to result from the atoms released by a single borosilicate

particle that is sampled within the laser-induced plasma. It is noted that such hot-spots

(i.e. particle hits) were only observed with the nebulization of the particle suspension;

only background structure (Figure 6-1) was visible in the absence of the aerosol stream.


































Figure 6-4. A spatially localized emission burst from an aerosol in the upper left-hand
corner of the plasma image. The delay time for this image was 2ts-
sufficiently early that background emission is comparable in strength to
analyte emission. The image was recorded using a 3nm interference filter
centered at a wavelength of 396.2nm.

The diameter of each luminous burst was averaged for each time delay for a series

of 30 such bursts to provide a quantitative measure of the diffusion length-scale of the

calcium atoms. A plot of the measured emission burst diameter as a function of delay

time is presented in Figure 6-5. Atomic diffusion coefficients were calculated from the

slope of this curve, yielding values of 0.063m2/s at 2[ts and 0.017 m2/s at 30[ts, with an

average value of 0.04 m2/s over the temporal range of 2-30!ts. These values suggest a

diffusion time-scale, whereby individual aerosols are dissociated with time and

subsequently act as expanding (via diffusion) localized atomic sources in which the









corresponding atomic emission is readily distinguished from the background plasma

emission.

Heiftje et al. (1987) reported such an approach for droplet desolvation and solute-

particle vaporization processes within an inductively coupled plasma. The current

diffusion rate, 0.04 m2/s, is consistent with the rates used by Heiftje; assuming a plasma

temperature (T) of 15,000K and scaling by T3/2. It should be noted that microsecond-

scale diffusion rates stand in stark contrast long-standing theoretical models and

empirical interpretations of aerosol-derived atomic emission, in which the analyte species

are assumed to uniformly diffuse through the plasma volume on a rapid, nanosecond

time-scale.

Rather, the current observations suggest that the rates of analyte diffusion within a

laser-induced plasma have much in common with diffusion mechanisms at work in much

lower temperature plasmas and even flames, given appropriate scaling to account for

temperature differences; yet another implication of energy-driven dissociation within

laser-induced plasmas. The plasma image sequence shown in Figure 6-6, demonstrates

the diffusion of localized atomic emission with time. Steady spatial growth of the clearly

defined emission burst occurs over the 2-30[ts range of delay times. Note also that the

burst does not fill the entire plasma even at the longest time, 30[ts, which stands in

contrast to the widely used assumption in LIBS practice that analyte emission occurs over

the entire plasma volume for a significant portion of the temporal emission process.

Another important observation is the clear decrease in burst intensity and plasma-center

intensity for the series of images in Figure 6-6, which confirms the measurement of

Figure 6-3, and supports established theory regarding temporal decay of LIP emission.









1.50-


1.25-


8 1.00 -
E

Ct
0.75


0 0.50

E T
S0.25-



0.00 I
0 5 10 15 20 25 30 35
Delay [us]
Figure 6-5. Average diameter (N=30) of emission bursts as a function of plasma delay
time. Error bars represent +1 standard deviation.

Correlation of Image and Spectral Data

While the imaging data provide a qualitative measure of the physical emission

process, it is important, in terms of the overall role of LIBS as an analytical technique, to

substantiate the validity of the image data through correlation with traditional, established

spectral metrics. Here, the peak-to-base (P/B) ratio is taken as the signal of interest and

compared between the spectral image sets and traditional spectral data sets.

Representative spectra with salient atomic and molecular line structure present within the

range 380-410nm are shown below in Figure 6-7. The corresponding P/B trends over all

delay times are shown below in Figure 6-8.

























IlI


(e) (I)
Figure 6-6. Images (a)-(e) demonstrate the observed growth of emission bursts for time
delays of 2, 4, 8, 15, 22, and 30 ts respectively. Both spatial growth and
intensity decay are observed with increasing time delay. All images reflect
the same spatial scale, shown in (a).












4000 Ca II 393.37

Ca II 396.85
CN Violet &
-i 3000 +
S3000 N + Bands
2
S/ I 394.7
C

S2000 30ps




1000 8-s


4ps


380 385 390 395 400 405 410
Wavelength (nm)

Figure 6-7. Three reference spectra taken at 4, 8, and 30us respectively demonstrate the
evolution of the primary calcium line at 396.85nm in addition to structure
belonging to other ions and molecules present in the plasma.

It is important to realize that the image data represents a local P/B ratio, such that

normalization to the continuum emission is performed in the immediate vicinity of the

emission burst. This is in contrast to the traditional spectral data, which is spatially

averaged over the entire projected plasma area. Given equal apparatus response, one

might assume that the localized P/B would be substantially increased as compared with

the spatially averaged P/B. The good agreement between the two at small delay times-

and as much as 100% greater relative P/B response from the spectral data at the

intermediate delays-however, suggests that the spatially-resolved and spectrally-










resolved plasma signals are characterized by slightly different temporal profiles. The

more general and important observation, though, is that the two trends correlate well and

suggest the use of image data to selectively analyze spatial regions could be utilized to

improve detection limits using LIBS.

1.20


1.00


0.80



0.60



0.40


0.20


n nn


w. w
0 5 10 15 20 25 30 35
Delay (us)
Figure 6-8. Comparison of measured P/B values for image data and spectral data. Error
bars represent one standard deviation. The data are normalized to the
respective P/B values at a delay of 30 [.s.



Rate of Aerosol Dissociation

To analyze the rate at which the experimental aerosols were dissociated within the

laser-induced plasma, the Ca II line at 396.85 nm was selected as the analyte peak of

interest for LIBS analysis. The calcium signal used for spectral analysis was the


A Normalized Spectra
Normalized Image a


















, t I I I .I I .









calculated peak-to-base (P/B) ratio. This metric provides a precise measure of laser-

induced plasma atomic emission by normalizing out fluctuations in the absolute plasma

emission signal (Carranza and Hahn 2002b). The reported P/B was calculated using the

integrated full peak-area of the 396.8nm calcium emission line divided by the integrated

signal of a scaled background spectrum over the same spectral width.

Using the above described calibration procedure for ICP-standard solutions of

calcium at 0, 2.5, 5 and 10tg/mL, a calibration curve was constructed at each delay time.

The P/B values for the experimental borosilicate aerosol streams were then measured and

compared with the calibration curve at each intensifier gate delay (2-3 0ts)-providing a

measure of the equivalent concentration of calcium at each of the six delays.

In order to proceed with calculating an experimental value for absolute mass of

calcium detected for a given aerosol particle, a clear relationship between the spectrally-

derived equivalent calcium concentration, the effective plasma volume, and absolute

mass of calcium in a single aerosol is needed. Following the analysis of Hahn and

Lunden (2000), an equivalent mass-concentration per plasma event corresponding to a

single aerosol hit is related to the average mass of a single particle through the relation

Meqf Mavg ND 6-1

where Meq is the equivalent analyte mass concentration (mass/volume) based on the

resulting LIBS spectrum per a traditional spectral calibration curve response function,fis

the frequency of particle hits per laser shot, Mavg is the analyte mass of an average aerosol

particle, and ND (particles/volume) is the number density of aerosols in the sample

matrix. For dilute aerosol streams (-1 particle/plasma volume or fewer, such as those









used in the current work) the application of Poisson statistics to estimate frequency is

valid and leads to the relation

f 1 exp(-p) 6-2

where p is the average number of particles per plasma volume. It is readily shown that

S= VpND 6-3

where Vp is a measure of the characteristic plasma volume. Again assuming a low

particle hit-rate corresponding to dilute aerosol streams (i.e. values of p much less than

unity), the exp(-u) term of Equation 6-2 can be approximated by 1 p (to order p2).

Combination of this approximation with Equations 6-1, 6-2, and 6-3 leads to useful

relationship for calculating an experimentally determined analyte mass for a single

aerosol particle, namely

Single = Meq Vp 6-4

Equation 6-4 is significant, in that it allows calculation of the absolute analyte mass

by multiplying the equivalent analyte concentration by the effective plasma volume. The

equivalent analyte concentration is readily calculated by using the resulting spectrum (or

average particle hit spectrum) in combination with a traditional calibration curve. For the

current study, calibration curves were constructed by nebulizing known concentrations of

calcium standards (SPEX ICP-AES solutions) over a range of resulting calcium

concentrations. The equivalent calcium concentration values where then calculated using

the ensemble-averaged spectrum corresponding to a number (-200) of particle hits

recorded for the dilute aerosol stream of the borosilicate particles. The equivalent

concentrations were then multiplied by the effective plasma volume, which in the current

work was taken as the sphere of rotation formed by a plasma using the directly measured






87


optical diameters as given in Figure 6-2. The resulting product provides a direct measure

of the absolute calcium mass released as a function of time, which is presented in Figure

6-9.


600


500



400



300



200


100


0 5 10 15 20 25 30 35
Delay Time [Ls]

Figure 6-9. Mass of calcium analyte detected using LIBS on a dilute aerosol matrix of
borosilicate glass microspheres. The first four data points show a logarithmic
increase with time-the subsequent plateau corresponds to a marked decrease
in plasma energy. The error bars represent +1 standard deviation.

The asymptote in the curve of Figure 6-9, taken to represent the maximum number

of calcium atoms within the plasma volume, indicates that approximately 400fg of

calcium are present in each aerosol. Using an average aerosol diameter of 2pm, 0.7 tpm

standard deviation, and density of 2.5g/cc (as reported by the manufacturer), -400fg


I









translates to 2% calcium by mass. No independent measurements of calcium

concentration in the aerosols were performed, but the 2% value is in the range reported

for borosilicate glasses industry-wide.

The plateau in Figure 6-9 suggests that the plasma has fully dissociated the calcium

aerosol at approximately 15[ts from the end of the laser pulse, and only partial

dissociation has occurred before this time. Hence, the concept of a rate-limit with regards

to particle dissociation within a laser-induced plasma, as first put forth by Carranza and

Hahn (2002c), is strongly supported. In that work, the authors observe that, while several

orders of magnitude more energy exists in the focused laser pulse than exists in the bond

energy of a given solid particulate, the measurable spectral signal (directly proportional

to atomically dissociated analyte mass in the plasma volume) experiences a plateau as

particle size reaches a critical value of about 2[tm. It is proposed in the current study that

dissociation processes may plateau at some critical value of plasma energy or electron

density (see Chapter 3 for full treatment of this relationship), which corresponds directly

to a critical time since the plasma decays at a known, measurable rate. Note in Figure 6-3

that the plasma emission at 15[ts has decayed to less than 10% of its value at 2ats-the

change in concavity for both Figure 6-3 and Figure 6-9 occurs at nearly the same delay

time. This is consistent with the rapid decrease in electron density as reported in Chapter

3 (see Figure 3-3). Such behavior indicates that the plasma has lost significant energy via

radiative transfer, hence may no longer be energetic enough to dissociate additional mass

(i.e. atoms) at the appreciable rate necessary for complete dissociation at these time

scales. This observation could give rise to an interesting theoretical model whereby both

thermal transport (via convective/conductive processes) and free electron interaction