Citation
Laser-induced breakdown spectroscopy

Material Information

Title:
Laser-induced breakdown spectroscopy fundamentals, instrumentation and applications
Creator:
Castle, Bryan C., 1971-
Publication Date:
Language:
English
Physical Description:
x, 244 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Aluminum ( jstor )
Calibration ( jstor )
Delay lines ( jstor )
Laser beams ( jstor )
Lasers ( jstor )
Lead ( jstor )
Plasmas ( jstor )
Signals ( jstor )
Wavelengths ( jstor )
Zinc ( jstor )
Chemistry thesis, Ph. D ( lcsh )
Dissertations, Academic -- Chemistry -- UF ( lcsh )
Emission spectroscopy ( lcsh )
Mineral industries ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1998.
Bibliography:
Includes bibliographical references (leaves 234-243).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Bryan C. Castle.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
029223962 ( ALEPH )
41372612 ( OCLC )

Downloads

This item has the following downloads:


Full Text







LASER-INDUCED BREAKDOWN SPECTROSCOPY: FUNDAMENTALS,
INSTRUMENTATION, AND APPLICATIONS
















By

BRYAN C. CASTLE


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




























To my wife, Cherie, and my parents, Leighton and Suzanne.













ACKNOWLEDGMENTS

I would like to express my sincere appreciation to Jim Winefordner for his guidance

and wonderful support extended to me over the last four years. Jim has created a sound

working environment that encourages experimental freedom and, therefore, increases both

productivity and creativity. The Winefordner laboratories, with the wide variety of

instrumentation and materials, are second to none! My gratitude also is extended to Benjamin

Smith for his friendship, assistance, and always calming presence. When things began to look

overwhelming, Benny was always there to straighten the curves in the road. I have been very

fortunate to have two superb research advisors and friends.

I am grateful to all the members of the Winefordner research group for their support

and friendship. I would especially like to acknowledge those who contributed directly to this

work. These people include Andrea Croslyn, Leslie King, Gretchen Potts, David Rusak,

Ricardo Aucelio, Andrew Knight, Hossein Nasajpour, and Omar Hamori. Additionally, I

would like to acknowledge the following visiting scientists for their input to the presented

work: Kobus Visser, Piet Walters, Nico Omenetto, and Karine Talabardon. In particular,

Kobus, is the only person, other than Jim and Ben, who has molded me into the scientist that

I am today. In the eighteen months we spent performing research together he has taught me,

through example, the ways of science.

My family has always provided me with love and support. I am grateful to my parents

for encouraging me to continue my education, and I am very fortunate to have their guidance








and wisdom. I can not express enough how much the love, support, friendship, and the

sacrifice of my wife, Cherie, means to me each and every day. I hope and pray we live a long

and happy life together so I can repay her for these selfless gifts. I would also like to

acknowledge my God who has always been the rock that makes up my foundation, and the

calm in the storm.

Finally, I would like to acknowledge the support of the chemistry department staff

including Jeanne Karably, Steve Miles, Larry Hamly, Joe Shalosky, Gary Harding, and Dailey

Burch. Each of them have made a contribution to the construction of the experimental

apparatus. For financial support I would like to acknowledge the National Science

Foundation Engineering Research Center for Particle Science and Technology and their

industrial partners, the United States Army, and the Florida Institute of Phosphate Research.














TABLE OF CONTENTS


ACKNOWLEDGMENTS...............................................................................iii

ABSTRACT.................. ...........................................................................ix

CHAPTERS

1 INTENT AND SCOPE OF DISSERTATION...................................................1

2 BACKGROUND OF LASER-INDUCED BREAKDOWN SPECTROSCOPY....3

Introduction to Emission Spectroscopy.............................. ...............3
Introduction to Laser-Induced Breakdown Spectroscopy......................5
Review of Laser-Induced Breakdown Spectroscopy Literature................12
Solids.................................... .... .......................... 15
Fundamental studies........................................ .............15
Analytical results......................... ..........................22
L iquids................................................... ...............................3 1
Fundamental studies..................... ..........................31
Analytical results......................... ..............................39
Gases ............... ...................................................46
Fundamental studies....................................... .............46
Analytical results........................ ................. ...........49

3 FUNDAMENTAL INVESTIGATIONS OF LASER
PRODUCED PLASMAS..................................................58

Part I: Spatial and Temporal Dependence of Lead Emission in Laser-
Induced Breakdown Spectroscopy......................... .................................58
Introduction...................... ............ ...........................58
Experimental..................... ............ .........................59
Results and Discussion............................. ......................... 63
Spectra....................................................................................63
Plasma Imaging......................... ... .................................63
Angular Dependence........................ ........... ..............75
Conclusions........................ .......... ...........................77








Part II: Level Populations in a Laser-Induced Plasma on a Lead Target............81
Introduction..................... .......... .......................81
Theory.............................................................................................. 83
Temperature Determinations Using Spectral Line Ratios..............83
Temperature Determination Using Boltzmann Plots...................84
Experimental................................................ ................................. 85
Lead Spectrum, Energy Level Diagram and Line Selection...........85
Data Capture and Graphical Presentation of the Raw Data...........86
Temporal Development of the 360 nm Series of Pb I Line
Intensities........................ .................. ......88
Background Correction........................................... ...............96
Stripping the 367 and 368 nm Line Spectral Overlap....................97
Height Dependence of the 360 Series ofPb I Line Intensities........99
Boltzmann Plot........................... .......................99
Height and Temporal Development of the Boltzmann
Temperature.................................................108
Conclusions........................ ......... ......................... 108

4 LIBS BENCHTOP INSTRUMENT: DEVELOPMENT AND
EVALUATION ................................... ................................. ................110

Introduction................................................110
Developm ent...................................................... 111
Instrument Overview.......................... .......................111
Pierced Mirror.................... ................ 15
Fundamental Evaluations............................ ....... 18
M ass Removal..................................... 118
Plasma Temperature............................................ 120
Optimized Gate Settings................. ......... ....121
Analytical Evaluations.......................... ....... ...126
Sampling Procedure.............................. ..... ...... ....126
A lloys................................. 126
Soils.... ...................... ......... 128
Paints.................................................... 133
Organics...................................................133
Detection Limits...........................................133
Day-to-Day Reproducibility....................... ............. ... 137
Depth Profiling.............................................139
Conclusions.......................... ................. ...............145










5 LIBS PORTABLE INSTRUMENT: DEVELOPMENT AND
EVALUATION ........................................................... .......................146

Introduction............................................ ......... 146
Experim ental........................................... ......... 147
Instrum ent Design................................................................. 147
Spectral Imaging Apparatus........................................ ...........149
Sample Preparation...................... ................................150
Results and Discussion............................ .. ........... ............ 150
Plasm a Imaging.......................... ......................................... 150
Lens-to-Sample Distance.......................................................151
Evaluation of Probe Design..................................................... 154
Lens-to-sample distance........................................154
Spatial filtering......................... ............................... 156
Analytical Applications...................... ................... .............. 156
Analysis of paint samples.......................................156
Analysis of NIST steel samples...................................160
Analysis of NIST iron ore samples................................164
Analysis of NIST organic samples.............................166
Conclusions........................ .... ..................... ........................166

6 VARIABLES INFLUENCING THE PRECISION OF LASER-INDUCED
BREAKDOWN SPECTROSCOPY MEASUREMENTS............................... 169

Introduction..................... ..... ............................... 169
Experimental...................... ........... .........................172
Results and Discussion. ...................................................................173
Choice of Analytical Line......................................................173
Signal Intra-Measurement Development.................................. 173
Dependence of Precision on Sample Movement.......................178
Dependence of Precision on Number of Laser Shots
A ccum ulated....................................................................... 184
Pulse to Pulse Stability of the Laser, and its Effect on Emission
as a Function of Pulse Energy.........................................185
Dependence of Precision on Gate Delay..................................... 188
Dependence of Precision on Surface Roughness.......................189
Dependence of Precision on Background Correction................191
Conclusions........................ .......... .........................193

7 MATRIX EFFECTS IN LASER-INDUCED BREAKDOWN
SPECTROSCOPY................................. ..........................194

Introduction...................................................... 194
Background.................................. ..........................194








Experim ental........................................................ ............................ 197
Results and Discussion...................... .... ........................199
Bulk M atrix Effects............................... .......................199
Internal Standards....................... .............................205
Speciation Effects............................. .. .....................214
Zinc Matrix Effect.............................. .....................215
C onclusions................................................................... ..........219

8 ANALYSIS OF ORES: PROGRESS TOWARDS DEVELOPMENT OF A
PROCESS M ONITOR.............................................................. .......220

Introduction................................................................ ........................220
Experim ental...................................................................... ....... 222
Results and Discussion............................... ................... 222
Phosphate Rock Analysis.........................................................222
Iron Ore Analysis............................ ........................224
Future Directions.......................... ................... ...................... 228

9 CONCLUSIONS AND FUTURE WORK........................................................231

A PPEN D IX ........................................................................ ....................................233

REFER EN C E S............................................. ........................... ............ ..............234

BIOGRAPHICAL SKETCH...................................................244
























viii













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


LASER-INDUCED BREAKDOWN SPECTROSCOPY: FUNDAMENTALS,
INSTRUMENTATION, AND APPLICATIONS

By

Bryan C. Castle

May, 1998


Chairman: Prof. James D. Winefordner
Major Department: Chemistry


In laser-induced breakdown spectroscopy (LIBS), a laser pulse of sufficiently high

power is focused to produce an irradiance that exceeds the material's breakdown threshold

of 10-100 MW cm2. If the electric field in the focused laser beam is greater than the

dielectric field strength of the material it encounters somewhere in the focal volume,

breakdown occurs at that point and eventually results in the formation of a transient, highly

energetic plasma. For analytical purposes, the emitted electromagnetic radiation is spectrally

resolved, and the emitting species in the laser-induced plasma are identified and quantified by

their unique spectral wavelengths and their line intensities.

Fundamental investigations of laser-induced plasmas have resulted in a better

understanding of their spatial and temporal development. When exploring the lead level








populations in a LIBS plasma, it was determined that over the first 15 ps of the plasma's life

time the energy distribution of the excited atoms does not follow a Boltzmann distribution.

Two different types of LIBS instrumentation have been constructed and evaluated.

A benchtop instrument for laboratory use, and a battery-powered, portable instrument for

field use. Both were evaluated for elemental determinations in alloy, soil, paint, ore and

organic matrices.

Investigations were conducted concerning the variables that influence the precision

of LIBS measurements. These included the choice of analytical line, sample movement,

number of measurements, laser stability, detector settings, surface morphology, and data

treatment.

The effect of a change in matrix on LIBS measurements was explored. Five different

sensitivities were observed for the analysis of zinc and chromium in aluminum, copper,

graphite, sand, and potassium bromide matrices. One possible method for compensation was

through the use of an internal standard. The internal standard can be selected based on the

standard criteria, with the addition of the evaluation of the temporal development of the

analysis and internal standard elements signals.

We also report on the investigation of the usefulness of LIBS as a process monitor in

a mining facility. Satisfactory results have been obtained for both phosphate rock and iron

ore minerals in the laboratory. Current research efforts have focused on the transferring of

the LIBS technology from the laboratory to the processing site.













CHAPTER 1
INTENT AND SCOPE OF DISSERTATION

It is my impression that the forces directing university research have undergone some

drastic changes in recent years. At one time the majority of experiments conducted in the

university setting were driven by the needs of fundamental science. More recently, the focus

has shifted from being fundamentally driven to being industrial driven, that is, research is

accomplished with a set goal in mind. It is not my intension to criticize or praise this change,

it is only mentioned to help place the research presented in this dissertation into the right

context.

From the beginning, my research has been industrially directed. Through our

involvement with the National Science Foundation's Engineering Research Center for Particle

Science and Technology we have had the opportunity to interact with several mining

industries. These include the phosphate mining companies located in Florida, PCS Phosphate,

Inc. (White Springs), Cargill Fertilizers (Fort Meade), and CF Industries (Wauchula), and also

one iron ore company located in Michigan, Cleveland Cliffs Mining (Marquette). Through

these interactions we learned of the industry wide need for an analytical method capable of

providing quantitative, elemental information, on-line, in real time. We believed that the

technique of laser-induced breakdown spectroscopy, as known as LIBS, could be a possible

solution to their problems. It was through this interaction that the research contained in this

dissertation was founded.






2

Chapter 2 provides a brief overview of both emission spectroscopy and laser-induced

breakdown spectroscopy, and concludes with a through review of the recent literature

concerning the fundamentals and applications of LIBS. Chapter 3 presents research on the

fundamental investigations of laser produced plasmas. This was a collaborative project with

Kobus Visser, a visiting professor from University of Stellenbosch in the Republic of South

Africa. This chapter includes all of our initial investigations into LIBS completed over a one

year period, including our research into the spatial and temporal development of lead emission

in LIBS plasma, as well as investigations into the lead level populations in a LIBS plasma.

Chapter 4 presents the research on the construction and evaluation of a LIBS benchtop

instrument. Chapter 5 presents the research on the construction and evaluation of a battery

-powered, portable LIBS instrument. The following three chapters report on further

evaluations conducted using the LIBS benchtop instrument. Chapter 6 reports on the

investigations of the variables influencing the precision of LIBS measurements. Chapter 7

presents studies on matrix effects in LIBS. Finally, chapter 8 reports on the preliminary

results obtained from the analysis of mining minerals, including phosphate and iron ores.

The presentations, publications, and manuscripts that have resulted from this research

are listed in the Appendix.













CHAPTER 2
BACKGROUND OF LASER-INDUCED BREAKDOWN SPECTROSCOPY

Introduction to Emission Spectroscopy

Atomic spectroscopy is based on the concept that atoms, when properly energized,

may re-emit the absorbed energy as electromagnetic radiation having frequencies

characteristic of the radiating species, and with an intensity proportional to the number of

atoms radiating. Therefore, the frequencies at which emission is observed serve to identify

the radiating species, and the intensity of the emitted radiation is a measure of the number of

atoms responsible for the emission. These two properties are the basis of qualitative and

quantitative atomic spectroscopy.

In the late 1600s, Sir Isac Newton was credited with the first investigations of visible

radiation, with his discovery of the seven colors of white light [1]. It wasn't until the mid-

1800s through the work of Kirchhoff and Bunsen, that the qualitative usefulness of analytical

spectroscopy was discovered [2]. Through their work with flame emission, they

demonstrated the identification power of spectroscopy by identifying several elements based

on their flame emissions. The quantitative usefulness of spectroscopy was not totally realized

until the 1920s with the work of Gerlach and Schweitzer [3]. Since this time, the field of

spectroscopy has really blossomed, and currently analytical chemists are realizing the fruits

of these, and many more earlier discoveries.

The three basic components of an emission spectroscopy apparatus include the

excitation source, wavelength selector, and the radiant power detector. The excitation source








is required to produce free atoms from any sample and to transform these atoms into excited

species. The wavelength selector is needed to provide a specific frequency(s) to the

detector. The radiant power detector is capable of quantifying the intensity of

electromagnetic radiation emitted from the excited atoms at a given frequency determined by

the wavelength selector.

Early excitation sources were primarily limited to flames [4]. Since they proved too

inefficient for the excitation of certain materials, in particular refractory materials, arcs and

sparks were developed [4]. Arcs and sparks grew in popularity primarily because of their

ability to analyze most conductive samples with minimal sample preparation. The

disadvantage of these sources is their poor stability. In the 1960's, the development of stable

high-frequency and de plasma devices offered an even more energetic excitation source [5].

These sources offered improved sensitivity which translated into lower detection limits. One

other emission source which has grown in popularity, is the low pressure discharge, one of

which is the glow discharge [6].

There have been several different types of wavelength selectors used in emission

spectroscopy. These devices are primarily based on the principles of absorption, interference,

spatial dispersion, or interferometry. The most commonly used are the spatial dispersion

devices, typically consisting of a ruled grating [7]. The grating is placed in a sealed box

(spectrometer) that allows light to enter and exit through the entrance and exit slits. These

are adjustable to vary the resolution and light collection efficiency of the device. Collimating

optics are used to direct a parallel beam of radiation onto the grating. The grating is a plane

mirror that is ruled with closely spaced groves. The grating operates under the principle that

when different wavelengths of light striking the grating are reflected, they constructively





5

interfere at different angles relative to the normal. Therefore, as the grating is rotated,

different wavelengths of radiation will be directed onto the exit slit of the spectrometer.

The radiation at the exit slit of the spectrometer is monitored by the radiant power

detector. Examples of these are photomultiplier tubes (PMT), photodiodes (PD), photodiode

arrays (PDA), charge-coupled devices (CCD), and charge injection devices (CID) [7]. These

detectors convert the incident photon flux into a measurable electrical signal The PMTs and

PDs are single-channel detectors, while the PDAs and CCDs are multi-channel detectors.

The multi-channel detectors are more complex and expensive, but have the advantage of

simultaneously monitoring several wavelengths and the background between them.

Introduction to Laser-Induced Breakdown Spectroscopy

Soon after the development of the ruby laser, it was realized that when the laser

radiation was focused, the intense light beam was capable of vaporizing and exciting solid

material into a plasma. The possibility of using lasers as excitation sources in atomic emission

spectroscopy was first demonstrated by Breech and Cross in 1962 [8]. This technique,

known as laser-induced breakdown spectroscopy (LIBS), has been reviewed by several

researchers [9-12]. LIBS is based on the concept that when a nanosecond laser pulse of high

energy density strikes the surface of any material, the surface temperature is instantly

increased beyond the vaporization temperature. The coupling of the pulse energy into the

sample occurs through several (often unknown) mechanisms including single and multi-

photon absorption and dielectric breakdown [13-14]. The dissipation of this energy through

vaporization is slow relative to the rate at which energy is deposited. Therefore, before the

surface layer can vaporize, the underlying material reaches critical temperatures and pressures,

causing the surface to explode. The ablated material, in the form of particles, free electrons,





6

atoms, and ionized atoms, expands at a supersonic speed and forms a shock wave in the

surrounding atmosphere. After several microseconds, the plasma plume slows down via

collisions with ambient gas species, and at this point the shock wave separates from the

plasma front and continues propagation at a speed approaching (or exceeding) the speed of

sound. Electron number densities on the order of 10" to 10"9 cm3 and plasma temperatures

in the range of 104 to 105 K have been reported [15-16]. At this stage, the plasma begins to

decay through radiative, quenching, and electron-ion recombination processes that lead to a

formation of a high dense neutral species zone in the post-plasma plume. The decay typically

ends with the formation of clusters, which usually occurs within hundreds of microseconds

after the plasma has been ignited. For analytical purposes, the emitted radiation (integrated

over the first tens of microseconds) is spectrally resolved, and the emitting species in the

laser-induced plasma are identified and quantified by their unique spectral wavelengths and

line intensities.

A simplified experimental apparatus required for LIBS analysis is shown in figure 2-1.

A pulsed laser source is used to vaporize and excite the material present at the focal point of

the focused laser beam. The emitted radiation from the laser-produced plasma is collected

and dispersed by the spectrometer and then quantified by the detector. A digitally enhanced

photograph of a laser-produced plasma is shown in figure 2-2. The plasma consist of three

primary regions, the high temperature core, the lower temperature middle, and the expanding

outer edge known as the shock wave. The total volume occupied by a laser plasma formed

under atmospheric conditions is approximately 3 mm3. The plasma is a transient source with

a typical lifetime ofapproximately 50 ps. Figure 2-3 shows the decay of the emission spectra

observed from a pure lead sample. The front most spectrum was captured at a delay of 0.8











Detector


Spectrometer L




.N d-A Nd-YAG Laser


Figure 2-1. Schematic of basic LIBS experimental apparatus.









HOT, VAPOR PLASMA






SHOCK WAVE




RADIATION


Figure 2-2. Digitally enhanced photograph of a laser-induced breakdown plasma.








t
Pbl
368.35 nm


Pb I
363.96 nm
Pb I
357.27 nm Pb I
373.99 nm
Pb I
367.15 nm












355 360 365 370 375
Wavelength (nm)

Figure 2-3. Temporal development of a series of lead emission lines observed in a LIBS plasma.





10

ps with respect to the laser pulse, and successive spectra at 0.5 js additional increments. It

can be noted from this figure that over a 14 Ps interval (as shown in this case) the emission

signal reaches a maximum and then begins to decay. This indicates that by using gated

detection, the signal-to-noise ratio can be optimized.

There are several attractive features to LIBS. One of the most attractive is that only

optical access to the sample is required. This has allowed LIBS to be used for remote sensing

and process monitoring. Also, this is what distinguishes the LIBS technique from

conventional plasma emission spectroscopy. The sample doesn't need to be transported to

the plasma source, because the plasma is formed within the sample or at the surface of the

sample. Another feature of the technique, from a spectrochemical analysis point of view, is

that it provides simultaneous multi-element capability with minimal, if any, sample

preparation. However, there are some severe problems to overcome with this method

including the variable mass ablation from heterogeneous samples which causes elemental

quantitation to be problematic. This could account for the fact that the LIBS technique has

not yet reached its full potential. An overview of the advantages and disadvantages of LIBS

is listed in table 2-1.

LIBS appears at first glance to be quite simple because the vaporization, atomization,

and excitation processes are carried out in one step by the laser pulse. However, the

mechanism under which these occur is not clearly defined. The reason for the complexity is

based on the large number of variables that influence the ablation process. Several of these

variables and their effects have been studied including: laser pulse temporal duration and

shape,[17] laser wavelength,[18-20] laser energy,[21-23] physical and chemical characteristics

of the target material,[24-25] composition and pressure of the surrounding environment,[26-

28] and effect of magnetic field [29-30].


















Table 2-1. Advantages and disadvantages of laser-induced breakdown spectroscopy.

Advantages Disadvantages

1. Minimal (no) sample preparation 1. Variation in the mass ablated caused by
changes in the bulk matrix.

2. All states of mater can be analyzed, as 2. Difficulty in obtaining matrix matched
well as both conductive and standards.
nonconductive samples

3. Very small amount of material is 3. Detection limits higher (poorer) than
vaporized (10s ofng) standard solution techniques (i.e. ICP-
OES)

4. Easy analysis of refractory materials 4. Poor precision, typically 5 10%
such as ceramics

5. Micro analysis is possible with spatial 5. Standard emission disadvantages, such
resolution of 1-10 pm as spectral interference and self absorption

6. Capability of remote analysis in harsh 6. Possibility of optical component
environments damage from high energy density lasers

7. Atomization and excitation are in one 7. Complexity
step

8. Capable of simultaneous multi-element
analysis








Review of Laser-Induced Breakdown Spectroscopy Literature

There have been several publications in the last decade which have reviewed laser-

induced breakdown spectroscopy either as a unique method of elemental analysis, or as a

member of the family of atomic emission techniques. LIBS has also been addressed in

reviews on laser applications, process monitoring, and materials processing. Many

spectroscopic studies of laser plasmas have also appeared in the physics journals. In order

to construct a meaningful review, the literature must first be reduced to a subset. In this

instance, LIBS in analytical chemistry is chosen.

This review will cover fundamental studies and analytical results and applications of

LIBS related to the field of analytical chemistry. The review is divided according to target

phase. Solids, liquids and gases are treated in sections devoted to each. Articles from the

physics literature are included when they are of interest to the analytical chemist.

Spectroscopic studies of laser-induced plasmas created in pulsed-laser deposition (PLD)

experiments have, for the most part, been omitted under the assumption that these papers are

more pertinent to materials science than to analytical chemistry. Fundamentals and

applications of laser ablation as a sampling technique are also ignored on the grounds that,

in these instances, the laser plasma does not serve as the excitation source.

The time frame to be covered has not been specifically defined because the amount

of literature dealing with each of the three phases of targets differs greatly. The literature on

solids is easily the largest, and so the review of solids is taken almost entirely from

publications within the last 5 years.

Literature on laser-induced breakdown in gases is the oldest. Many of these

publications are in the physics literature and deal with mechanisms of breakdown and plasma





13

diagnostics. There are a smaller number of papers that report upon the determination of trace

metals in gas phase matrices. Most of these are intended to prove the usefulness of LIBS in

hazard monitoring. In the section on gases, the analytical chemistry literature over the past

10 years is reviewed and appropriate additions are made from the physics literature.

The section dealing with liquids includes bulk liquids, isolated droplets, and aerosols

generated from liquids. Among the chemistry journals, this literature is the smallest. The

physics is of interest primarily because of medical applications of laser-induced breakdown,

and publications which are judged to be of interest to the analytical chemist have been

included. All publications related to the elemental analysis of liquids by LIBS during the last

14 years are included.

Previous reviews on this topic begin with a 1984 review by Adrain and Watson titled

"Laser Microspectral Analysis: A Review of Principles and Applications" [31]. This was

followed by Cremers and Radziemski who published "Laser-induced Breakdown

Spectroscopy: Principles, Applications, and Instrumentation" in 1990 [32]. This paper

included a brief review of theory and instrumentation for LIBS followed by a few industrial

applications. In 1992, Thiem et al. reviewed LIBS theory as part of their paper titled "Lasers

in Atomic Spectroscopy: Selected Applications" [33]. The section on laser-induced plasma

covered laser-material interaction, plasma production factors (wavelength, energy, and

properties of target), and emission factors (temperature and electron density).

Majidi and Joseph published "Spectroscopic Applications of Laser-Induced Plasmas"

in 1992 [34]. This publication has perhaps had the most influence in terms of style and

content on the review that follows. Majidi and Joseph reviewed analytical results on solids,

liquids, gases, and mixed phase systems for the years 1987-1992. Emphasis was on





14

applications of LIBS such as determination of hazardous elements in air, toxic elements in

wastewater, and elements of interest in coals and iron ores. The authors stressed the fact that

LIBS required only optical access to samples.

In 1993, Ibrahim and Goddard published "An overview of Laser-Induced Breakdown

Spectroscopy" [35]. They concentrated on the topics of laser-material interaction, local

thermodynamic equilibrium, and plasma diagnostics. The instrumentation was described, the

use of gated detection was explained, and a few applications were presented. It was not a

literature review.

In a 1993 review by Darke and Tyson entitled "Interaction of Laser Radiation with

Solid Materials and its Significance to Analytical Spectrometry," applications of LIBS were

reviewed in the section on laser ablation [11]. In this paper, LIBS was referred to as laser

microprobe optical emission.

The following year application of LIBS to process control was briefly reviewed by

Noll et al [36]. The authors discussed fundamentals of plasma formation and laser-material

interaction with respect to optical and heat penetration, laser energy, density, and absorbance.

The use of fiber optics and on-line sampling was discussed, and a periodic table was presented

showing elements which had been determined and their detection limits in iron ore. This idea

was incorporated into the review that follows. Each section was concluded with a periodic

table or tables with elements that have been determined and shaded according to detection

limit.

Most recently, LIBS was briefly reviewed as a section in the comprehensive review

of atomic emission by Sharp et al. published in 1995 [37].








Solids

Fundamental studies

Many papers in this area stem from the need to more fully understand the LIBS

plasma in order to obtain useful quantitative results. One such paper is the 1988 publication

by Chen and Yeung who used the acoustic signal generated by a laser-induced plasma as an

internal standard [38]. They reported that the magnitude of the acoustic wave was

proportional to the emission signal for major and minor elements within the solid target.

Furthermore, they found that this proportionality was independent of laser power and focus

spot size.

In the same year, Wood et al. studied the effect of laser pulse duration on soft x-ray

emission from a tantalum target [39]. Using a colliding-pulse mode-locked dye laser at 620

nm with pulse durations of 100 and 600 fs and a Nd:YAG laser at 1064 nm with a pulse

duration of 70 ps, the plasma emission in the range from 10-71 nm was observed. It was

found that longer pulses gave relatively shorter wavelength emission and longer emission

lifetimes than short pulses.

In 1989, Coche et al. used laser-enhanced ionization detection in a laser plasma to

study the processes of ionization and recombination [40]. They used a N, laser (337.1 nm,

5 mJ, 10 ns) to ablate a solid target. A dye laser was used to selectively ionize species in the

plasma at different delays relative to the ablation pulse. Optogalvanic detection was used to

give an indication of the number of atoms in the probe volume at the chosen time delay. In

this way, ionization and recombination rates could be inferred.

Also in 1989, lida studied the atomic emission characteristics of a laser plasma in

reduced pressure argon [41]. Using a ruby laser (1.5 J, 20 ns), a plasma about 10 mm in





16

height, and more than 100 (is in duration was formed at pressures between 0 and 50 Torr.

The plasma at 50 Torr had greatly increased line emission and background due to confinement

by the Ar atmosphere. The plasma at lower pressure showed less background and less line

emission.

Another investigation of plasma expansion was done by Balazs et al. in 1991 [20].

These authors investigated ruby laser pulses on a copper target, and developed a two-part

model for the interaction. The first part of the model dealt with the heating and melting of

the solid and included parameters such as thermal diffusivity, pulse duration, and material

density. The second part of the model described the plasma expansion into a vacuum.

Mason and Goldberg have characterized a laser plasma in a pulsed magnetic field [30].

The first part of the paper included spatially resolved emission studies. The authors found

that when a pulsed magnetic field was oriented normal to the laser beam, it caused radial

compression and axial expansion in the plasma. The emission intensities of both atoms and

ions were also increased in this magnetically confined plasma.

In the second part of the paper, the authors discussed time resolved emission and

absorption studies. They concluded that the increase in emission seen in the plasma was due

to Joule heating caused by the induced secondary current in the plasma. This was evident by

the fact that the increase in emission was seen later in time than the maximum of the applied

magnetic field. Also, the intensity increase was attributed to increased atomization efficiency

and longer residence times in the plasma.

The effects of buffer gas type on the plasma produced by a Nd:YAG laser (1064 nm,

100 mJ, 7 ns) on a metal target were studied by Owens and Majidi in 1991 [27]. They

observed an increase in the ratio of Al II / Al I intensity in helium gas relative to argon gas






17

and air. This increase was attributed to the ability of excited helium atoms to transfer energy

to a similar energy level in the aluminum ion.

More metal target studies were evaluated by Lee et al. in 1992 [25]. These authors

used an ArF laser (193 nm, 100 mJ, 10 ns) to produce plasmas on copper and lead. The lead

plasma was much larger than the copper plasma (5 mm vs 2 mm) and had a slightly lower

excitation temperature as determined by Boltzmann plots. The temperature of the copper

plasma was 13200-17200 K, and the lead plasma was 11700-15300 K.

In 1992, Marine et al. studied plasma expansion by optical time-of-flight

measurements [42]. They determined that the velocity distribution of ions produced by a UV

pulse of several nanoseconds duration was broad and not well defined. In contrast, the ions

produced by a picosecond IR pulse traveled with a velocity inversely proportional to the

square root of their mass. Neutral atoms still had a poorly defined velocity distribution.

These authors also noted the appearance of a bi-modal temporal profile for YO' emission and

attributed this to two possible mechanisms for formation of YO'.

Kuzuya et al. studied the effect of laser energy and atmosphere on the emission

characteristics of laser-induced plasmas [22]. They used a Nd:YAG laser at 1064 nm and

pulse energies from 20-95 mJ in atmospheres of He, Ar, and air from 1 Torr to 1 atm. The

authors reported that the maximum emission intensity was observed at 95 mJ in 200 Torr of

Ar. However, the maximum signal to background was obtained in helium at 40 Torr and 20

mJ of power. Images showed the different sizes and shape of these plasmas.

Okana et al. studied mass removal in non-metallic inorganic solids and determined

relationships between laser power and atom yields [43]. Their paper described vacancy

initiated laser ablation as a process by which weakly bound atoms were released from around






18

vacancies and vacancy clusters. The atom yield was determined to be an exponential function

of laser fluence.

Time resolved emission studies from a laser plasma on sodium chloride were reported

by Yago et al. in 1993 [44]. They used a Nd:YAG laser of 150 ns duration focused onto a

NaCI pellet. The emission spectra showed self-reversal in air but not in a vacuum. The

plasma was divided into two zones, a hot core behaving as an emission zone, and a low

temperature periphery behaving as a reabsorption zone. The plasma expansion rate was

shown to be determined by ambient gas pressure.

In 1994, Kagawa et al. used a XeCI laser (308 nm, 15-70 mJ, 20 ns) to produce

plasma on a Zn target in vacuum [45]. Time resolved studies showed a number density jump

which represented the blast wave expansion into the observed volume. This outermost

portion of the plasma was shown to be ideal for analytical measurements because the

background in this area was greatly reduced.

In 1994, Tambay and Thareja studied emission in a laser plasma of Cd metal vapor

formed in a heatpipe [46]. They showed that emission from the vapor was stronger when the

plasma was formed on a tungsten target than it was when the plasma was formed in the gas

itself The authors claimed that this was due to pumping of the vapor by soft x-rays formed

on the tungsten target.

Transition probabilities of 28 Si ion lines were determined using a laser-induced

plasma as a source by Blanco et al. in 1995 [47]. Using a Nd:YAG laser (1064 nm, 280 mJ,

10 ns) to produce a plasma on pure silicon, these authors observed emission in Ar and Kr

atmospheres. The plasma produced was found to have an excitation temperature of 20000

K and an electron density of 10"7 cm'. Absolute transition probabilities for the Si ion lines

were calculated.






19

Jensen et al. published mechanistic studies of laser-induced breakdown on model

environmental samples in 1995 [24]. They used a KrF laser (248 nm, 30 ns) to produce

plasmas on SiO2 containing Eu and Cr, which were added as the solids Eu2O3 and K2Cr2O,

to the sand. Detection limits of 100 ppb for Eu and 2 ppb for Cr were reported. A sample

was also prepared in which the source of Cr was a solution which was added to the sand and

then evaporated. This method of sample preparation gave an order of magnitude less signal

and a different temporal profile for the Cr emission.

In 1995, Tasaka et al. studied the emission of a laser-produced plasma on graphite

[48]. They used a Nd:YAG laser in He and in air to form a plasma in which they observed

a "triple plume" composed of three distinct regions. The authors claimed that these regions

appeared because of three different speeds in the expanding plasma. The fastest region was

composed of carbon ions from the target and N and O ions when the experiment was done

in air. The second plume was composed of the compressed neutrals in the vicinity of the

shockwave. The slowest plume was then the target vapor composed of larger molecules.

Thareja et al. also studied graphite plasmas at low pressure and found similar temporal

profiles [491.

Intense emission from the CN radical has been observed in plasmas produced on

graphite. With the use of a high resolution spectrometer, several authors have resolved the

vibrational and rotational structure in the CN emission bands. The emission from the violet

band of CN around 388 nm has been used to calculate vibrational and rotational temperatures

in the laser plasma [50-531.

In 1996, Al-Wazzan et al. studied three-dimensional number densities of species in

laser produced plumes [541. They used absorption of an expanded dye laser beam to form

shadow images of Ba ions in a plume produced by excimer laser ablation at 248 nm. They






20

also used fluorescence from planar slices of the plasma to obtain sequential cross section

images which could be built into a three-dimensional image.

Al-Wazzan et a. also carried out an experiment in which they observed plasma in

vacuum and in ambient oxygen [551. In oxygen, the expanding plume showed increased

temperature and electron number density at the shock front due to increased collisional

excitation rates. In a vacuum, enhancement at the shock front was not observed.

Bulatov and Liang obtained full spectra at each pixel in the image of a laser-induced

plasma as depicted in figure 2-4 [561. They used this technique to create classification maps

which gave location of any species of interest within the plasma. They studied the effects of

different focal length lenses and different sampling geometries on plasma formation and

location of species within the plasma.

In 1996, Mulatari et al. obtained time resolved images from a laser plasma formed on

a sample at non-normal incidence [571. They varied the angle of the incident laser beam with

respect to the target from perpendicular to nearly parallel. It was found that the plume was

generated in the perpendicular direction regardless of the laser angle. However, the maximum

emission signal was obtained with normal incidence.

Mulatari and Cremers also published a second study that year reporting on the use of

an acousto-optic tunable filter to capture spectrally resolved images [581. They used a series

of different lenses to examine the different distributions of elements within each plasma. By

collecting light from the outermost edge of the plasma, they were able to use ungated

detection to obtain analytically useful spectra with low background.

Nemet and Kozma studied the shape of emission lines produced on gold targets at

different delays relative to the ablation laser [591. The 406.51 and 389.79 nm lines were









Jl1/C


Figure 2-4. Spectroscopic imaging of a laser-induced plasma.


~ -~J~h






22

observed. These lines were asymmetric and shifted relative to their natural wavelength at

times up to 1000 ns after the laser pulse. These lines could be described by asymmetric

Lorentz-type profiles. After 1000 ns, the lines appeared to be very close to their natural

wavelength and were Lorentzian and symmetric.

In 1997, Kuriawan and Kagawa used a long pulse Nd:YAG laser to produce a

plasma on a brass target in vacuum [601. The authors were especially interested in the

secondary plasma formed by compression in the vicinity of the shock wave. The emission in

this secondary plasma was captured, and it was observed that when a wedge of aluminum was

placed very close to the target, the emission signal increased in the vicinity of the wedge. The

authors attributed this effect to increased compression provided by the shock wave interaction

with the immovable aluminum wedge.

In 1997, Martin et al. used a laser-induced plasma to establish spectral calibration of

their detector [611. A Nd:YAG laser (1064 nm, 350 mJ) was focused onto targets of

ceramic, polyethylene, ZnS, and aluminum, all in vacuum. Line pairs in the deep UV were

used with known transition probabilities to determine the relative efficiency of the detection

at each wavelength.

Most recently, Granse et al. modeled a laser-induced plasma and compared their

model to experiments with different lasers and different materials [621. The model they

derived accounted for fluid dynamics of the plasma, absorption of laser energy via inverse

bremsstrahlung, and the dynamics of ionization and recombination.

Analytical results

The number of papers describing analytical results of LIBS studies on solids is easily

larger than the number of papers dealing with either liquids or gases. For this reason, only





23

publications after 1992 are considered in this section. Figures of merit given for specific

determinations differ greatly from one author to the next. In the periodic tables which follow

this section, the lowest published detection limits for the elements in the matrix of interest are

used to determine the shading.

Carbon content in steel was determined by Aguilera et aL in 1992 [631. The authors

observed the emission from the 193.1 nm carbon line because the 247.9 line had a spectral

interference. The experiment was carried out in a CO2 free environment, and a neighboring

iron line was used as an internal standard. A Nd:YAG laser at 1064 nm was used, and it was

found that the slope of the calibration curves decreased with increasing laser power. Also,

there was a slight difference in slope for stainless and non-stainless steels. At a laser power

of 100 mJ, the detection limit for carbon was 65 ppm and the RSD was 1.6 %.

Also in 1992, Hader used LIBS for on-line quality control of rubber mixing [641. The

author called the technique "remote laser microanalysis" and used the acronym RELMA. The

existing technique for quality control of rubber mixing involved a discrete sampling step in

which a small portion of rubber was taken from a batch and analyzed. This led to problems

when it became necessary to distinguish between bulk composition fluctuations and

inhomogenieties; with on-line sampling by LIBS in a number of randomly chosen positions

in the rubber mix, the two problems could be distinguished. Elements were found in the

cross-linkers and plasticizers added to the raw rubber.

Lorenzen evaluated other on-line applications of LIBS in the same year [651. This

time the technique was referred to as "laser-induced emission spectral analysis" and given the

acronym LIESA. The authors described applications such as the determination of minor

elements in liquid steel, and depth profiling of layers on metallic substrates.






24

In 1993, Sabsabi et at. used a KrF laser (248 nm, 100 mJ) to analyze aluminum alloys

[661. They carefully optimized the delay time by using a PMT to look at the temporal

evolution of several different emission lines. An aluminum line was used as an internal

standard and magnesium was determined at a few ppm.

The following year Thiem et al. investigated LIBS of alloy targets [67]. They

determined Al, Cu, Fe, Ni, and Zn using a Nd:YAG laser at 532 nm in a vacuum chamber.

Using non-resonant lines for all the elements, the authors were able to generate linear

calibration curves from a few hundred ppm to a few percent for each element. For the

determination of these elements at the percent level, the authors found many potential

emission lines in the spectral window from 300 to 400 nm.

Thiem and Wolf used LIBS to analyze mining ores and compared the results to those

obtained by a digestion ICP-AES method [681. AI, Ca, Cu, Fe, K, Mg, Mn, Si, and Ti were

determined in aluminum and manganese ore. The comparison showed that both methods

gave roughly the same accuracy, but the ICP technique had better precision. The authors also

pointed out the similarity in cost for the two techniques, about $100,000 for either.

Investigations of the use of resonant wavelengths in laser ablation had been made as

early as 1992 by Borthwick et at [69]. Although most of the work was carried out with time-

of-flight mass spectrometry, the paper is included here because of its relevance to LIBS. The

authors noted that when an ablation laser was scanned through specific wavelengths, an

enhanced ion yield was detected. More specifically, an enhancement in Ga ions in the ablation

of GaAs, and an enhancement in Al ion yields in ablation of steels were noted. The ion yields

were even more pronounced when grazing angles of incidence of the ablation laser were used.





25

In 1995, Allen etal. employed resonant wavelengths in LIBS of thin films [70]. The

authors used resonant laser ablation time-of-flight mass spectrometry to investigate copper

thin films on a silicon substrate. Using a XeCl excimer laser pumped dye laser (5 mJ pulse

energy at 463.51 nm) multi-photon ionization of copper was observed. By using an

unfocussed laser continually and watching the decay of the copper signal, the authors were

able to determine the thickness of the film between 20 and 100 A. It was calculated that

between 10" and 102 A per shot was removed by the laser, although this was at a power

below the breakdown threshold, so emission could not be collected.

Anderson et al used LIBS for depth profiling without the resonance feature of Allen's

experiment [711. A Zn/Ni coating between 2.7 and 7.2 Im thick on a steel substrate was

analyzed and a calibration curve of signal duration with thickness was found to be linear. The

depth resolution was far poorer than in the mass spectrometric experiment of Allen because

the energy density at the target was much greater. Nonetheless, Sn coatings of less than 1 uim,

and Cr coatings on the order of a few nm on steel could be determined by this technique.

Arnold and Cremers used LIBS to determine metal particles on air sampling filters

[72]. TI was collected on filter paper by passing contaminated air through a filter or by

wiping a filter on a TI surface. The laser beam from a Nd:YAG laser at 1064 nm was formed

into a line at the focus by a pair of cylindrical lenses. The detection limit for TI was 40 ng/cm2

of filter paper, and the calibration curve was linear up to 40 pg/cm2. The 535.05 nm TI line

was used for detection.

Sattmann and Sturmalso investigated the use of a multiple Q-switch Nd:YAG laser

for analysis of steel samples [73]. Single, double, and multiple pulses were used to produce

plasmas on a low-alloy steel. Material ablation, emission intensity, electronic temperature,





26

and electron number density were determined for each plasma type. All of these parameters

were greater for the double and multiple pulses than for the single pulses presumably because

of a smaller shielding effect in these cases. Calibration curves were correspondingly steeper.

Davies and Telle used a 100 m fiber optic pair to perform remote LIBS on ferrous

targets [74]. A pair of 550 im (OD) fiber optics were used to transmit laser light to the

target and to return plasma emission to the spectrometer. Detection limits of 200 ppm or less

were found for Cr, Cu, Mn, Mo, Ni, Si, and V. A number of potential emission lines to use

for calibration were given for these elements.

Cremers et al. used a fiber optic probe to determine Ba and Cr in soil [75]. A single

1.5 mm fiber was used both to deliver laser light and collect emission. A glass plate was used

to reflect the emission to the spectrometer. The Ba ion line at 493.41 nm and the Cr atom

line at 425.44 were used for calibration which was linear over 4 orders of magnitude for both

elements. The detection limit for Ba was 26 ppm, and the detection limit for Cr was 50 ppm.

Cremers also determined at what distance LIBS could be performed using

conventional optics rather than fiber optics [76]. By using a beam expander to increase the

diameter of the laser beam and a pair of lenses with adjustable distance to focus this expanded

beam onto a target, he was able to produce a plasma and collect light at a distance of 24 m.

Detection limits in a simulated moon rock were at the level of a few percent due mostly to the

small solid angle of collection at this distance.

Bescos et at. analyzed aluminum samples in 1995 [77]. The authors simultaneously

determined Mg, Mn, Fe, and Pb in aluminum using a spectral window from 380 to 410 nm.

Detection limits were around 100 ppm and calibration curves were linear up to about 1%.

Gonzales et a. also analyzed steel samples [78]. Their 1995 publication dealt with

sulfur determination. They reported a detection limit of 700 ppm using the 180.73 nm atom






27

line. The analysis was done in a N2 atmosphere, and the Fe ion line at 186.47 was used as an

internal standard. Precision was 7%.

Hakkanen and Korppi-Tommola used LIBS to study elemental distributions of paper

coatings [791. A XeCI laser at 308 nm was used to generate a plasma on paper, and Ca or

Si was used as an internal standard. A spatial resolution of about 250 pm could be achieved

with a laser pulse power of 200 mJ. It was estimated that 2 ng of paper coating were

vaporized per shot. Al, Si Mg, Ca, and C were monitored quantitatively in a spectral

window from 220-290 nm.

Sabsabi and Cielo analyzed aluminum alloys by LIBS in 1995 [801. The authors

characterized the laser plasma on aluminum targets using Stark broadening of Al ion lines to

determine electron number densities and using Boltzmann plots a series of Fe atom lines to

determine electronic temperature. Calibration curves for Mg, Mn, Cu, and Si were

constructed and detection limits were as low as 10 ppm, for Mg.

Soil was analyzed by Ciucci et at. in 1996 [81). They used both a Nd:YAG laser at

1064 nm and a XeCI laser at 308 nm. The authors found that the background decay in the

laser plasma was faster for the 308 nm pulse than for the 1064 nm pulse, and used a

correspondingly shorter delay time with the XeCl laser. Cu, Pb, and Cr in soil were

determined, and an entire spectra from 350-700 nm was shown with lines identified for the

geological survey soil sample GXR-2.

Ernst et al. used LIBS to determine Cu in A533b steel [821. The first attempt was

carried out with fiber optic delivery of the laser. This method was significant because in the

hazardous environment of a nuclear reactor pressure vessel, Cu concentration in the steel is

an indicator of radiation embrittlement and of expected material lifetime. Because the fiber






28

optic could not deliver enough power to produce sensitive detection, beam delivery was done

with conventional optics. The Cu line at 324.75 was used for calibration, and by using a

second order calibration curve, Cu could be determined in the range between 100 ppm and

5%.

Geertsen et al. revisited LIBS for aluminum samples in 1996 [831. Using a Nd:YAG

at 1064 nm and a pulse power of 230 mJ the authors formed craters ~5 im deep in aluminum

with a single shot. The detection limits for Mg and Cu were 4 and 40 ppm, respectively, with

a precision of about 8 %. Spatial resolution was assessed by rastering the beam over a sharp

Al-Cu junction. The best spatial resolution obtained was 6 pm.

Marquardt et at. determined Pb in paint using a Nd:YAG laser at 532 nm coupled into

a fiber optic probe [84]. The probe consisted of two fibers, excitation and collection,

terminated in a probe head with an aspheric lens to focus the laser light to a spot and focus

plasma emission into the collection fiber. The common end of this fiber is shown in figure 2-

5. Different combinations of excitation/collection fiber diameters were tried. The detection

limit for Pb was 140 ppm, and precision was 5-10 % even when analysis was carried out

through layers of non-lead containing paint.

Miziolek also reported on a LIBS probe in 1996 [851. This application was for

determination of heavy metals in soils, and the probes were to be used in a cone penetrometer

truck. In one probe, a compact laser was mounted in the probe head. In the other probe,

laser light was carried through a fiber optic to the probe head. The compact laser head

provided the more sensitive probe with detection limits for Pb, Hg, Cr, Cd, and Zn of 1-10

ppm in sand and silt. The fiber optic delivery provided less power but a higher laser repetition

rate; 15 mJ vs 28 mJ, 30 Hz vs 1/3 Hz.














f/2 Lens

Sample
Collection Fiber <.../ ...



Excitation Fiber I- -


Aspheric Lens


Figure 2-5. Common end of fiber optic LIBS probe used by Marquardt et al. for determination of lead in paint.





30

Also in 1996, Cremers et al. introduced a portable LIBS instrument and used it to

determine hazardous elements in soil, in paint, and in samples on filter paper [861. A compact

laser head was used as the probe. Ba, Be, Pb, and Sr were determined in soil with detection

limits of 265, 9.3, 298, and 42 ppm respectively. In paint, the detection limit for lead was

much greater because the 405.8 nm line could not be used due to spectral interference. On

filter paper, Be and Pb were determined with detection limits of 21 ng/cm2 and 5.6 pg/cm2

respectively. In the filter paper studies, some particle size effects were evident.

Palleschi et al. used a 400 mJ Nd:YAG laser at 1064 nm for a variety of

determinations including Hg in air (5 ppm detection limit), and pollutants in power plant

smoke and soil [87]. These authors analyzed the geographical survey GXR-2 sample and

determined concentrations for 18 elements.

Vadillo and Lasema analyzed geological samples of vanadinite, pyrite, garnet, and

quartz [881. These determinations were done in vacuum to increase the lifetime of the ionic

species and of the ion lines observed. Fe, Mn, Mg, and Si were determined in each of the

rocks, and At was also determined in garnet. Differences in composition were as expected.

Lasema et al. used LIBS to analyze the surface of solar cells [891. A N2 laser at 337.1

nm was focused onto the cell and the cell was rastered under the focus providing spatial

resolution of about 30 pm. The C ion line at 588.9, Ag atom line at 546.5, Si atom line at

634.7, and Ti atom line at 625.9 were used in determinations of these four elements. The

concentration of each element was mapped across the surface of the solar cell.

In another application of LIBS to alloys, Kim et al. determined aluminum in a zinc

alloy [901. The authors used a Nd:YAG laser (1064 nm, 105 mJ, 3 ns) in vacuum, air and

argon. The presence of Al in the Zn alloy was important because of its detrimental effect on





31

its welding properties. The delay and the distance from the surface of the target were

optimized in each atmosphere. Because the Al atom line at 308.22 nm was used for

calibration, large delay times were used to insure no contribution from a shorter-lived Zn

interference. In air, a 30 us delay was used and in Ar a 50 us delay was used. The Zn line

at 307.59 was used as an internal standard.

In 1997, Maravelaki et al. used LIBS to monitor the laser cleaning of marble artifacts

[91]. Although lasers had been used previously for the cleaning of such treasures, these

authors were the first to examine the emission from the plasma. Crusts of 20-600 mr

consisting of gypsum, iron oxides, soot, and calcite were ablated from the marble by a

Nd:YAG laser. The plasma formed in this ablation was analyzed spectroscopically to

determine the endpoint of the cleaning.

Figures 2-6 and 2-7 show graphically the detection limits for elements that have been

determined using LIBS in sand, ore, or soil and in steel or alloy.

LiQuids

Fundamental studies

Besides papers in the field of analytical chemistry, there have also been a number of

papers which have dealt with the physics of laser-induced breakdown in aqueous solution.

Most of these deal with the processes of cavitation and the factors which can affect the

breakdown threshold.

In 1984, Armstrong identified three time domains in the heating of an aerosol: the

acoustic regime, the internal conductive regime, and the external conductive regime [921. He

explained the phenomenon of aerosol-enhanced air breakdown as being caused by increased

electron collision frequency in the hot, dense, vapor surrounding the heated aerosol particles

in the beam.










BC N 0 F Ne

Si P S CI Ar


H Cr Fe

Zr MNb Mo Tc Ru


Co G Ge

Rh Pd Ag Cd In Sn


La Hf Ta W Re Os Ir Pt Au Hg TI

Ac


Key: i < 10 ppm
S10 -100 ppm
> 100 ppm


Figure 2-6. Elements determined in ores, rocks, and soils shaded according to detection limit.


Sc

Y


K

Rb

Cs

Fr Ro


As Se Br Kr

Te I Xe

Bi Po At Rn














Ti V Cr Mn

Zr Tc

Hf Ta Re


Os Ir Pt Au Hg


< 50 ppm
50 200 ppm
> 200 ppm


Figure 2-7. Elements determined in metals shaded according to detection limit.


K

Rb

Cs

Fr


Rh Pd


Zn77


Key: D





34

Three years later Armstrong used a CO, laser at 10.6 pm to explode aerosols

generated by a vibrating orifice aerosol generator [931. A UV laser was used to produce

shadow images of the exploding droplet, and a phase Doppler particle analyzer system

determined size and velocity of the expelled particles at a distance of 2 mm from the original

drop. It was found that the CO2 laser power affected the size of particles produced in the

breakdown for original drop sizes between 30 and 50 tm. The speed of the particles

produced was independent of laser power.

Chylek et al. studied the effect of size and material of liquid aerosols on breakdown

thresholds in 1986 [941. The setup that they used is shown in figure 2-8. A Nd:YAG laser

(532 nm, 10 ns) was focused onto droplets generated by a Berglund-Liu vibrating orifice

aerosol generator. Breakdown thresholds were determined on aerosols of different size

generated from liquids of various refractive index, density, surface tension, and chemical

structure. Thresholds were seen to decrease with droplet size. Refractive index had no

apparent effect. Density, surface tension, and chemical structure had effects which could not

be independently examined.

In 1987, Hsieh et al. used a Nd:YAG laser at 532 nm to examine the exact location

of breakdown initiation on water droplets generated by a Berglund-Liu vibrating orifice [951.

They found that as the energy of the laser increased, the breakdown moved from outside the

droplet to the inside. Theoretical calculations showed that the curved liquid gas interface of

a water droplet in air focused laser energy at one point just inside the illuminated face and at

another point just beyond the shadowed face.









Pyroelectric Detector


*
Acoustic Detector Microscope Viewing
A,6 from the top


Beam


Orifice


Frequency Generator

Figure 2-8. Schematic of experimental setup used by Chylek et al. to investigate LIBS of aerosols.





36

The effect of laser wavelength and irradiance on spectra from laser-induced

breakdown of single levitated aerosol droplets was examined by Biswas et al. in 1988 [96].

Droplets of glycerine saturated brine solution approximately 18 pm in diameter were optically

levitated and probed with a Nd:YAG laser at 1064, 532, and 355 nn. Emission lines of Na,

C, and N were observed. It was shown that the energy required for breakdown, and the time

of plasma emission increased with increasing wavelength. The optimum time delay for

determination of these elements increased with laser wavelength and showed more variance

from element to element at longer laser wavelength.

Also in 1988, Zheng et at acquired temporally and spatially resolved spectra of laser-

induced breakdown of a 40 pn 4 M NaCI droplet using a Nd:YAG laser at 532 nm [97].

Using fiber optic ribbons and a streak camera, these authors observed that emission began

first at the shadowed face of the droplet and propagated toward the illuminated face.

Hammer et al. also examined thresholds with ultrashort laser pulses [981. Using

pulses of 2.4 ps, 400 fs, and 100 fs from a pulsed dye amplifier, they determined breakdown

thresholds in saline solution, high purity water, and tap water. The thresholds in these media

were not significantly different. With the 2.4 ps pulse the breakdown threshold was found to

be 5 x 10" W crr, with a400 fs pulse it was 1.3 x 10" W cm2 and with a 100 fs pulse it was

5.65 x 10" W cmO.

Kitamori et at. used the acoustic wave generated during particle-induced breakdown

to detect polystyrene particles in aqueous solution [991. A Nd:YAG laser (532 nm, ~1 mJ,

6 ns, 10 Hz) was focused into a solution. In pure water, the energy density produced by this

beam was not sufficient to cause a breakdown. When a particle entered the probe volume,

however, a breakdown occurred and subsequently an acoustic wave was generated. By





37

detecting these acoustic waves, the probability of breakdown was determined. By analyzing

solutions of known particle concentration, a calibration curve of breakdown probability vs

concentration was generated. This curve was used in the analysis of solutions of unknown

particle concentration.

Vogel et al. used a Nd:YAG laser to produce and image a breakdown in water [1001.

The fundamental wavelength of 1064 nm was focused into the water to produce a breakdown.

A portion of the beam was frequency doubled to produce 532 nm light which passed through

a beam expander and through the underwater breakdown, providing a shadow of the plasma

and the cavitation bubble. The size of the plasma and cavitation bubble were examined for

1064 nm laser pulse durations of 30 ps and 6 ns.

Sacchi discussed the mechanisms of laser breakdown in water and described the

differences in threshold for short pulse and long pulse lasers in 1991 [1011. He claimed that

the avalanche breakdown process or inverse bremstrahlung was responsible for breakdown

when nanosecond laser pulses were used. In this case, the breakdown threshold behaved

probabilistically. The probability of breakdown scaled linearly with the log of laser power for

low powers, and scaled parabolically with the log of laser power at higher power. In contrast,

with laser pulses of less than nanosecond duration, multi-photon ionization appeared to be

the mechanism ofplasma formation, and a definite threshold above which breakdown would

occur could be determined.

Sacchi returned to this topic in 1996 and produced images of laser-induced

breakdown on solids under water. By using a dye laser oriented parallel to the underwater

target and delayed relative to the ablation laser, the authors obtained shadow graphs of the

evolving plume and shock wave. A series of images followed the growth and decay over a

total time of about 7004is. A schematic of the setup used is shown in figure 2-9.































Beam Expander Target


Figure 2-9. Schematic of experimental setup used to obtain shadow graphs of a laser-induced plasma.





39

In 1992, Pinninck et al. studied the effect of resident particles on laser-induced

breakdown thresholds [1021. It was found that even at wavelengths where the particles did

not absorb laser light, there was a decrease in breakdown threshold in particle laden air versus

clean air. This was taken to mean that particles were focusing laser light. Location of plasma

formation (inside, shadowed side, or illuminated side) was found to depend on ionization

potential, gas pressure and laser wavelength. Refractive index had no effect on breakdown

threshold, in agreement with the observations reported by Chylek in 1986 [941.

In 1993, Nyga and Neu created a plasma on calcite submerged in water using two

fiber optics to deliver two pulses from two XeCI lasers [1031. Both pulses were 30 ns pulses

at 308 nm through 600 um fibers which were brought in close proximity to the target with

no focusing lenses. The pulses were separated in time by -300 us, and the second fiber was

equipped with a 308 nm dielectric beam splitter and used to collect emission from the calcite.

No quantitative results were reported.

Feng et al. also examined the effect of laser pulse duration on breakdown thresholds

in water [1041. In their 1997 paper, they determined the mechanism of breakdown with

nanosecond pulses to be due to cascade ionization, and the mechanism of breakdown with

ultra-short pulses to be multi-photon ionization, similar to Sacchi's findings [101]. These

authors, however, went on to model the breakdown phenomenon in water relating the

breakdown threshold to laser power, spot size, and pulse duration. They derived an extended

non-linear Schrodinger equation to describe the relationship.

Analytical results

Quantitative results for LIBS with liquid samples will be presented in two sections

based on the nature of the analysis. The first section will cover determinations done in bulk





40

liquid. This implies that a laser spark is generated either on the surface or beneath the surface

of a relatively large liquid sample. The second section will cover determinations done in

aerosols or droplets generated from liquid samples. Analysis of dry aerosols generated from

solids are included in the last section on gas phase determinations.

Bulk liquid. In 1984, Cremers et at. reported the use of a repetitive single spark

(RSS) and repetitive spark pair (RSP) for the determination of Li, Na, K, Rb, Cs, Be, Mg, Ca,

B, and Al in aqueous and organic solutions [105]. With the single spark, detection limits for

all elements except Li were >1 ppm. The detection limit for B was 1200 ppm. When a

second spark which was generated from a second laser and delayed by 18 ps relative to the

first was used to form the plasma, the detection limit for B was reduced to 80 ppm. This

improvement was attributed to formation of the plasma within a cavitation bubble when the

RSP method was used. The experimental setup is shown in figure 2-10.

In 1987, Wachter and Cremers reported a detection limit of 100 ppm for uranium in

solution by LIBS [106]. The plasma used was formed by a Nd:YAG laser (1064 nm, 260 mi)

on the surface of a 4 M nitric acid solution contained in a small glass vial. Each analysis was

done by averaging 1600 laser shots. This averaging was necessary to overcome poor shot-to-

shot precision due mainly to small variations in lens to sample distance.

In 1987, Cremers conducted an experiment to determine the maximum lens to sample

distance (LTSD) that could be used in a LIBS experiment [761. With a 250 mJ pulse from

a Nd:YAG laser at 1064 nm, it was found that a 2 m focal length lens could produce

breakdown on molten metal The larger problem was collection of the emitted light. A fiber

optic could collect enough light at 0.5 m from the plasma for modest figures of merit to be

obtained. At distances farther than this, the sensitivity of the experiment decreased rapidly.


















5 cm Focal Length
Lenses


1064 nm Pulsed
Laser Beam





Teflon Cell
Spark
Fill Hole


Figure 2-10. Experimental setup used by Cremers et al. to analyze aqueous solutions with LIBS.






42

In 1993, Aragon et al used LIBS to determine carbon content in molten steel [107].

They used a Nd:YAG laser (1064,200 mJ, 8ns) focused onto molten steel in a crucible under

an argon atmosphere. In order to provide a homogeneous sample, a jet of argon gas was

directed downward into the crucible. This served to remove the topmost layer of molten

liquid which would be enhanced in the lighter elements. Carbon was determined by ratioing

of the C 193.09 nm to Fe 201.07 nm intensity and a detection limit of 250 ppm was found

with an RSD of 6%.

In 1995, Stolarski et al. produced sub-surface plasmas in saline solution, triple-

distilled water, and tap water with Nd:YAG laser pulses of <5 mJ and pulse durations of 5

ns and 80 ps [108]. The focussing lens used was 2.54 cm. in diameter and had a focal length

of 17 mm. This lens was chosen to approximate the lens in the human eye. The authors

found that an energy of 1.5 mJ was enough to produce a breakdown at either pulse duration,

and found that the plasma temperatures and electron number densities depended little on pulse

duration for the two studied. These authors observed sodium emission around 590 nm in a

0.9% NaCI solution, but no quantitative analysis was performed.

Also in 1995, Ito et al. used LIBS to determine colloidal iron in water [109]. They

focused aNd:YAG laser (1064 nm, 100 mJ, 10 ns) into a flowing stream of water containing

FeO(OH) as submicron particles. The detection limit for iron was in the ppm range. This

technique differed slightly from previous techniques in that it relied on particle-induced

breakdown for the formation of the plasma. A plasma was formed only when a FeO(OH)

particle was present in the probe volume. Because of the extremely small size of the particles,

however, it was essentially a determination of iron in water. The following year, the same

authors used a second laser delayed by 1 uts relative to the first in a technique similar to






43

Cremer's repetitive spark pair. With this technique, they were able to improve the detection

limit for iron to 16 ppb [110].

In 1996, Knopp and Scherbaum used a dye laser at 500 nm with a pulse energy of 22

mJ to produce a sub-surface breakdown in aqueous solution [111]. They reported detection

limits for Cd, Pb, Ba, Ca, Li, and Na of 500 ppm, 12.5 ppm, 6.8 ppm, 130 ppb, 13 ppb and

7.5 ppb respectively. No signal was obtained for either Hg or Er at 0.1% in solution. When

micron-sized particles of ErBa2Cu30,, rather than a soluble salt, were used as the source of

Er, the detection limits decreased by a factor of 103. The improvement illustrated the

increased sensitivity of particle-induced breakdown as compared to breakdown in particle-free

solution. This experiment also suggested that 500 nm laser light had analytical advantages

over 1064 nm light.

In 1997, Ho and Ng used both a Nd:YAG laser (532 nm, 12 ns) and an ArF laser (193

nm, 15 ns) to determine Na in aqueous solution [112]. In order to increase absorbance of the

laser light, methyl violet was added to the solution to be analyzed. It was found that the 532

nm laser light produced a more visibly intense plasma; because the continuum emission was

higher in the 532 nm produced plasma, the 193 nm laser light actually gave better detection

limits for Na. A detection limit of 230 ppb was reported for Na in the methyl violet solution

with the plasma formed by the ArF laser.

In 1996, a similar experiment was performed by Paksy et al. who used a Nd:YAG

(1064 nm, 15 mi, 4 ns) focused onto the surface of a molten alloy in an argon atmosphere

[113]. Because the laser power used was significantly less than that used by most authors,

the delay of the detection relative to the laser was also less. A 100 ns delay and 1000 ns gate

width were used. Detection limits were 0.001% for Si in Fe, 0.006% for Cr in Fe, 0.06% for

Si in A, and 0.007% for Mn in AL






44

Area et al. determined Cr, Pb, and Cu in aqueous solution with a single laser shot on

the surface of the water [114]. Detection limits were 100, 100, and 50 ppm for the three

elements, respectively. Calibration curves were linear over 1 order of magnitude. The

authors also analyzed aqueous solutions by placing a drop of water to be analyzed on a KBr

pellet. When this technique was used, detection limits were improved by almost an order of

magnitude.

Aerosols and droplets. In 1983, Radziemski et al. used a nebulizer/heat chamber

system to produce an aerosol from a liquid solution and analyzed the aerosol with LIBS

[115]. Detection limits for Na, P, As, and Hg were reported to be 6 ppb, 1.2 ppm, 0.5 ppm,

and 0.5 ppm, respectively. The authors determined that local thermodynamic equilibrium

existed in the aerosol spark for time delays greater than 1 is relative to the laser.

In 1987, Eickmans et al. spatially resolved the emission from laser-induced breakdown

of an aerosol [116]. Approximately 45 gm diameter droplets were formed by a vibrating

orifice aerosol generator which was synchronized with the Nd:YAG laser beam at 532 nm.

Emission of Li, Na and H was observed in 5 M salt solutions of either NaCI or LiC1. Spatial

resolution of the emission revealed that in the plasma plume in the vicinity of the illuminated

face of the droplet, emission lines showed Stark broadening and self-reversal. Near the

shadowed face of the droplet, emission lines were narrower and less intense.

In 1988, Essien et al. used LIBS to determine Cd, Pb, and Zn in an aerosol [117]. A

Nd:YAG laser (1064 nm, 100 mJ, 5 ns) was focused onto an aerosol generated by a

nebulizer/heat chamber. Detection limits were 19 ppb, 210 ppb, and 240 ppb for Cd, Pb, and

Zn, respectively.






45

Also in 1988, Archontaki and Crouch used an isolated droplet generator to produce

equally spaced, uniformly sized droplets [118]. The experimental setup was such that a single

drop of known size was always in the probe volume of the laser for a given repetition rate.

A Nd:YAG laser (1064 nm, 100 mJ) was used to determine detection limits for Li, Na, Mg,

Ca, Mn and Al of 0.3, 2.2, 1.9, 0.4, 7.2, and 5.2 ppm, respectively. Droplet diameter had

little effect on the detection limit for sizes between 58 and 75 nm.

In 1992, Ng etal. used an ArF laser (193 nm, 150 mJ) to analyze an aerosol generated

by a concentric glass nebulizer/spray chamber traditionally used for ICP-AES [119]. The

authors determined the optimum delay for detection to be 6 ps relative to the laser pulse.

Detection limits for Na, Li, In, Al, Ga, Ca, Mg, K, and Sr were determined to be 0.9, 0.3, 10,

3, 3, 8, 3, 2, and 20 ppm, respectively.

The following year Parigger and Lewis used a picosecond XeCI laser (308 nm, 4 mJ)

to produce LIB in 66 pm water droplets [120]. Plasma diagnostics showed a temperature of

-10 000 K and an electron number density of 10"i cm3. The H, line at 656 nm was used as

an internal standard and the detection limit for sodium was ~1 ppm. Poulain and Alexander

performed a very similar experiment in 1995 using a KrF laser (248 nm, 200 mJ). The

detection limit for Na was found to be 165 ppm [121].

Haisch and Paine reported on the characterization of colloidal particles by LIBS in

1996 [122]. Field flow fractionation was used to sort particles by size and deposit them onto

filter paper using partitioned pumping to retain time resolution. A cylindrical lens was used

to focus the laser onto the paper to avoid ablation of the paper which could occur if the

energy density became too high. Detection limits for Si were in the ppb range.






46

Figures 2-11 and 2-12 show graphically the detection limits for elements that have

been determined using LIBS in bulk liquids and in aerosols generated from liquids.



Fundamental studies

In 1988, Kumar and Thareja studied laser-induced gas breakdown in the presence of

an electric field [123]. They used a XeCI laser (308 nm, 60 mJ, 8ns) focused between two

electrodes separated by 6.5 mm. A transverse static electric field was produced by applying

different potentials to the electrodes. In order to study the effect of the field on breakdown

thresholds, the current on the electrodes was adjusted so that no breakdown occurred using

only the electric field or only the laser. The authors were then able to determine that pre-

breakdown electron densities of 10' cm3 created by the field were sufficient to lower the

laser energy density needed to produce breakdown.

Parigger and Lewis used a Nd:YAG laser (1064 nm, 300 mJ, 7.5 ns) to produce a

plasma in CO [124]. They observed the emission of the C2 Swan band around 565 nm to

determine vibrational temperatures in the decaying plasma. The observed spectra was fit with

a model in which temperature was a parameter. At a time delay of 30 ps relative to the laser

pulse, a temperature of 6745 K was obtained. This temperature was confirmed by use of the

CN violet band around 388 nm to calculate the vibrational temperature in a similar way.

Yagi and Huo investigated breakdown in H2 gas at different pressures initiated by a

KrF laser (248 nm, 250 mJ, 20 ns) [125]. This work was related to Raman spectroscopy in

that the breakdown threshold determined the maximum energy with which the Raman

transitions could be pumped. The authors found that below 600 Torr, breakdown thresholds

were pressure independent. Thresholds decreased between 600 and 3000 Torr, and increased

slightly above 3000 Torr.














Ti V Cr Mn Fe Co Ni


Y Zr Nb Mo Tc Ru Rh Pd


Re Os Ir Pt


H

Li

No


Ag In Sn Sb

Au Hg TI Bi


D < 1 ppm
S1 -10 ppm
S> 10 ppm


Figure 2-11. Elements determined in bulk water shaded according to detection limit.


C NO

Si S

Cu ZnGa Ge As Se


Ca I Sc


Rb Sr


Ba La Hf Ta W


Fr Ra Ac


S--








H He

Li B C N Ne

Na Mg Si Ar
K Sc Ti V Cr Mn Fe Co Ni Cu ZnGa Ge Se Br Kr

Rb Y Zr Nb Mo Tc Ru Rh Pd Ag Cd Sn Sb I Xe

Cs Ba La Hf Ta W Re Os Ir Pt Au TI Pb Bi Po At Rn

Fr Ra Ac


Key: l < 10 ppm
S10 -100 ppm
> 100 ppm


Figure 2-12. Elements determined in liquid aerosols shaded according to detection limit.






49

Yalcin et al. investigated the sensitivity of a laser spark produced by a Nd:YAG at 532

nm to ambient conditions such as variation in background gas, presence of particles, and

humidity [126]. This was done in order to assess the potential usefulness of LIBS in toxic

metal monitoring. The authors found that excitation temperature and electron number density

in the plasma remained essentially constant with variations in the ambient conditions.

Analytical results

Cremers and Radziemski again carried out, pioneering work on gasses. In their 1983

publication, they detected chlorine and fluorine in air [127]. Detection limits for these two

gases were 8 and 38 ppm by weight, respectively. These values corresponded to absolute

detection limits of 80 and 2000 ng. In a helium atmosphere, the absolute detection limits

improved to 3 ng for both. Experiments were carried out in which molecular gases containing

both chlorine and fluorine were introduced to the sampling chamber. The relative intensities

of the chlorine and fluorine signal were found to be an indication of the number of each of

these atoms in the molecular gas. All experiments were carried out using a Nd:YAG laser

(1064 nm, 100 mJ, 15 ns) and detecting Cl at 837.6 nm and F at 685.6 nm.

These authors also detected Be in air [128]. The limit of detection in this experiment

was 0.5 ppb by weight. The Be was introduced into a chamber by laser ablation and diluted

by air. A second laser produced a breakdown in the chamber from which the emission was

measured. It was found that the emission from the Be II doublet at 313.1 nm was stronger

than any atomic emission even at delay times of 20 Its. This doublet was used for calibration

which was linear over 4 orders of magnitude.

Cremers et al. used LIBS to determine sodium and potassium in a coal gassifier

stream as early as 1983 [127]. These elements were present as particulates in the stream and






50

so again the question of whether or not the target was a gas or solid arises. Regardless, the

authors were able to determine Na with a detection limit of 4 ppb by weight using the doublet

at 589 nm. Figures of merit for potassium were not given. In conclusion, the authors noted

that sulfur could also be monitored.

Sneddon extended this technique to the determination of P in air and lowered the

absolute detection limit of Cl in air. In a 1988 paper, Sneddon detected phosphorous at 15

ppm and Cl at 60 ppm by weight in air [129]. A calculated plasma volume of 0.010 cm3 was

used to determine an absolute detection limit of 60 pg for Cl.

Ottenson et al. reported on the application of LIBS to analyze particulates in a coal

combustion vent in a 1989 publication [130]. A Nd:YAG laser was triggered on the

scattering from a HeNe laser to explode particles detected in the probe volume. The

experiment used on-line temperature correction to account for different excitation conditions

in plasmas formed on different size particles. Moreover, the size of particles to be sampled

could be adjusted by varying the laser power. With lower laser power, only larger particles

were sampled; the smaller particles were not sufficient to provide plasma formation. The on-

line temperature correction automatically accounted for differences in excitation conditions

with varying laser power.

In 1990, Morris et al. used laser-induced breakdown as a detector for gas

chromatography [131]. The UV laser light was focused into the effluent from a gas

chromatograph at a power below the breakdown threshold of the carrier gas. When eluting

carbon containing molecules were present in the effluent, the threshold for breakdown was

lowered and a plasma formed. The organic analytes were indirectly detected by optically

detecting plasma formation. A schematic is shown in figure 2-13. No emission from the

plasma was measured, and so the potential of this technique was only as a universal detector.









Optogalvanic Probe


Microplasma


Capillary


Gas Sheath


Figure 2-13. Schematic of setup used by Morris et al. for use of laser-induced breakdown as a detector in gas chromatography.


Focused LE
193 nm




Screen






52

In 1991, Cheng et al. studied polyatomic molecular impurities in helium gas at ppm

levels [132]. Using ungated detection, they determined B2z-l, PH3, and AsH, with detection

limits of 1, 3, and 1 ppm, respectively. The ionic phosphorous lines at 602.4, 603.4, 604.3,

and 605.5 nm were used. As lines were observed at 228.8, 235.0, 278.0, and 286.0 nm. B

lines were observed at 336.0 and 434.5 nm. The plasmas were produced by a Nd:YAG laser

at 532 nm.

Also in 1991, He and co-workers focused a pulsed laser onto a metal rod in an

atmosphere of helium or argon seeded with a reaction gas to produce ion-molecule complexes

and observe their emission [133]. Ion lenses were used to preferentially draw ions into the

He or Ar gas for reaction with the seeded gas. Emission from species such as A10, AIO-CO,

Al-Ar and AIH, was observed.

Joseph and Majidi used an electrothermal atomizer to create a gas phase sample from

a liquid for analysis by LIBS [134]. A few 4L sample was deposited in the furnace and was

dried, ashed, and atomized. The setup is shown in figure 2-14. At the onset ofatomization,

a Nd:YAG laser (1064 nm, 100 mJ) was focused into the furnace and fired at a repetition rate

of 10 Hz. Emission from the plasma was collected through the dosing hole of the tube.

Experiments with cobalt and cadmium gave absolute detection limits of 5 and 50 pg.

Casini et al. also determined several elements of interest in air using LIBS [135].

Their experiment used a Nd:YAG laser (1064 nm, 40 mJ, 7 ns) to produce a plasma in an

atmospheric pressure chamber as shown in figure 2-15. Using the saturated concentrations

of air over chosen liquids at the given temperature, and then successively diluting these

concentrations in the chamber, yielded calibration curves for a number of elements. Cl was

detected at 449.0 nm with a detection limit of 60 ppm, S at 415.3 nm with a DL of 200 ppm,






Computer
mI


Tube


\
High Energy
Mirror


Photodiode Array


Figure 2-14. Schematic of setup used by Joseph and Majidi for LIBS in a graphite furnace.









Parabolic Mirror


Laser Controller Computer

Figure 2-15. Schematic of setup used by Cassini et al. for air analysis using LIBS.


Lens


Monochromator






55

P at 442.1 nmwith a DL of200 ppm,Naat 371.1 nm with a DL of 110 ppm, Hg at 404.7 nm

with a DL of50 ppm, Be at 381.4 nm with a DL of 130 ppm, and As at 454.4 nm with a DL

of 130 ppm.

In 1994, Flower et al. used LIBS to monitor metal aerosol emission generated by a

pneumatic nebulizer [136]. A chromium salt solution was passed through the nebulizer and

a heating pipe to produce a gas with known concentration of Cr. The detection limit was 200

ng Cr / standard cubic meter of air (scm). The laser used was a Nd:YAG (1064 nm, 180 mJ)

and detection was at 312 nm.

In 1994, Lazzari et al. detected mercury in air using LIBS with a Nd:glass laser (400

mJ, 8 ns) to produce plasma in a chamber filled with saturated Hg vapor [137]. With

successive dilution of the mercury vapor, they obtained a detection limit of 10 ppb by

observation of the 253.7 line.

Zhang et al. performed LIBS in a particle loaded methane/air flame and in an oil-

fueled combustor vent [138]. In their 1995 publication, they described the addition of a

flange with four movable windows to a pre-existing combustor vent. LIBS was performed

through these windows which were kept clean by a flow ofN, gas and could be switched out

periodically for cleaning.

In 1995, Nordstrom studied the laser-induced plasma emission spectra of N,, 02, and

ambient air from 350 to 950 nm [139]. This range was covered with approximately 0.5 nm

resolution and atomic, ionic, and molecular emission from a plasma created by a CO2 laser

(10.6 prm, 320 mJ, 200 ns) was observed. Spectra were computed for N, 0, and a few

molecular species, and a comparison between experiment and calculation was done.






56

The same year, Parigger et al. published a fundamental paper on laser-induced plasma

in the gas phase [140]. The breakdown was produced by a Nd:YAG (1064 nm, 220 mJ, 6

ns) in hydrogen at 150 and 810 Torr. The H, line at 656 nm was used to calculate electron

number densities in the decaying plasma with temporal resolution as high as 6 ns early in the

plasma and 1 js at later times. Spatial resolution was also achieved by focusing different

portions of the 1:1 plasma image onto the spectrometer entrance slit. Spatial resolution was

approximately 50 mn. Surface plots for electron number densities resolved in time and space

were then created. Electron number densities varied between 10"6 and 10"9 cm'.

In 1996, Haisch et al. used LIBS to detect chlorine in the gas phase [141]. Two

experimental setups were described. The first was a benchtop experiment using a 320 mJ

Nd:YAG laser. The second setup used a miniature Nd:YAG laser built into a sensor head.

This miniature laser provided 18 mJ of pulse energy. Detection of Cl in chlorinated

hydrocarbons was accomplished by focusing the laser onto a copper target in the presence of

the gas to be analyzed. This provided for formation of CuCl in the plasma. The luminescence

of the D-system of CuCl was then detected around 440 nm. Detection limits were a few ppm.

Figure 2-16 shows the elements determined in a gas phase matrix and their relative

detection limits. The unit scm refers to a standard cubic meter of gas.








H

Li Be

Na Mg

K Ca Sc Ti V Cr Mn Fe Co


Rb Sr


Cs I Ba


Fr


Y Zr Nb MoI Tc IRu


La Hf Ta W I Re Os


Ra Ac


Rh Pd Ag In

Ir Pt Au Hg TI


Key: El < 10 lg/scm
S10 100 lg/scm
> 100 gg/scm


Figure 2-16. Elements determined in gas phase matrices shaded according to detection limit.
The unit scm refers to a standard cubic meter of gas.


lil













CHAPTER 3
FUNDAMENTAL INVESTIGATIONS OF LASER PRODUCED PLASMAS

Part I: Spatial and Temporal Dependence of Lead Emission in Laser-Induced Breakdown
Spectroscopy

Introduction

Studies investigating the spatial and temporal development of the LIBS plasma will

provided an insight into the complexity of these pulsed sources. Through the use of time-

resolved spectral imaging, a better fundamental understanding of the plasmas behavior is

attainable. The results of these studies have the potential to considerably improve the

analytical usefulness of LIBS results.

Kuzuya et al. [22] reported on the change in spatial distribution of the LIB plasma

spectral radiation at 400 nm caused by gaseous atmospheres of Ar, air, and He at different

pressures. Their results indicated that breakdown in the gaseous atmosphere above the target

occurred at excessively high laser power densities leading to decreased ablation and emission

from the target material because a major portion of the laser radiation was absorbed by the

opaque gaseous plasma in front of the target material. Sattmann et al. [73] studied the effect

of double laser pulses on the spatial structure of the plasma integrated over time and

wavelength using photographic detection. Temporal and spatial resolution of the LIB plasma

structure was reported by Palau et al.[142] for three different laser wavelengths, without

wavelength resolution. Multari and Cremers[58] reported on the spatial and temporal





59

distribution of the Cr 425 nm line emitted by the LIB plasma when a flat target was tilted at

various angles with respect to the incident laser beam. They concluded that after the plasma

formed on the solid target, it expanded upward along the path of the incident laser pulse, but

once the chromium line emission became dominant over the continuum background, the

emission appeared symmetric about the normal to the surface rather than along the axis of the

incident laser pulse. Vadillo et al. [143] showed non-dispersed, time resolved LIB images

captured by means of a CCD and monochromator in mirror mode (zero order). They

incorrectly inferred the width of the plasma from the digital images, failing to recognize that

the measured image actually represented an image of the entrance slit with non-uniform

irradiance caused by the plasma image projected upon the entrance slit. Bulatov et al. [56]

reported 2D spectral images from the LIB plasma created by means of Fourier transform

visible spectrometry. The wavelength resolution was in 4 nm spectral windows over the range

400-800 nm, which was insufficient for spectral classification of single emission lines. The

results also lacked time resolution.

As part of a research project to develop a portable LIBS instrument for the

determination of toxic metals in the environment and in particulate samples, we have

investigated the temporal and spatial development of lead ionic and atomic emission from a

LIB plasma.

Experimental

A schematic of the LIBS system which was used, is depicted in figure 3-1. It consists

of a laser, motorized X,Y,Z translational stage carrying the sample, spectrometer, ICCD

detector, detector gating and control electronics, and a computer for control and data

acquisition. The Nd:YAG was a Quantel YG580 which delivered 1 J at the fundamental
























X,Y,Z Stage Stage Control Computer


Figure 3-1. Experimental apparatus for capturing time resolved LIBS spectra.






61

wavelength of 1064 nm and produced a 9 ns duration pulse at a maximum of 30 Hz

repetition frequency. A 80/20 beam spliter was used to attenuate the laser output so that only

20% of the laser power was used in the generation of the plasma. The laser beam was

focused on the target using a 2.5 cm diameter, 10 cm focal length lens. The resulting plasma

created a 200 pm diameter crater with an average incident laser irradiance of 18 GW / cm2.

The initial design directed the laser beam perpendicular to the target material. In a later

modification, the laser beam was incident at 450 with respect to the target surface to allow for

the desired movement of the fiber optical system used for radiation collection. Pressed

sample pellets, of 1.4 cm diameter and varying thickness were made with a pellet press at

4500 psi pressure. In pressing the pellets pure lead flakes (Fisher Scientific, Catalog# L-24)

and NIST soil standards (SRM 2711) were used as sample material.

A positioning system consisting of a helium-neon laser, lenses and a position sensing

detector was used to monitor the position of the sample surface with respect to the laser beam

focus and the light collecting system's focal point; control was accomplished with a motorized

X, Y, Z stage (Oriel, Model# 18011). For all measurements, the target was moved

horizontally between shots to allow sampling from a fresh location and to improve the

reproducibility of mass ablation.

A 2.5 cm diameter fused silica lens with a focal length of 5 cm was used as the

collection optic to produce a one-to-one image of the plasma as viewed from the side on the

entrance slit of a 0.5 m focal length spectrometer (Acton Research, Model# SP-500) with a

grating of 1200 grooves / mm. The reciprocal linear dispersion was 1.6 nm / mm at 250 nm

and an entrance slit width of 25 pm was used. The intensified charge coupled device (ICCD),

( Princeton Instruments, Model# ICCD-576S) being used, had pixel dimensions of 22.5 x

22.5 umr and a viewing area of 12.9 mm by 8.6 mm. It was oriented in such a way that the





62

longer dimension corresponded with the wavelength dispersion axis and the shorter dimension

with the entrance slit height. The ICCD was operated by a controller (Princeton Instruments,

Model# ST-130). A programmable pulse delay generator (Princeton Instruments, Model#

PG-200) was used to gate the ICCD. The PG-200 was computer programmable, and was

capable of gating the ICCD with integration times as short as 20 ns. The Q-switch trigger

signal from the laser was connected to the "Ext trig in" input of the programmable pulser

(PG-200) and served as the master control of the delay gate. This trigger signal provided

-250 ns pretriggering with respect to the onset of the laser pulse and enabled the system to

open the ICCD gate before the plasma actually started emitting. The PG-200 high voltage

gate pulse was connected to the ICCD, while its "Aux dly'd trig out" signal was used to

trigger the ICCD Controller ST-130 "Ext sync" input. The PG-200 was programmed to

activate the ST-130 by means of this signal to perform the data read-out from the ICCD and

to transfer data to the control computer 1 us after the completion of the gate pulse. The ST-

130 "not scan" output was connected to the PG-200 "Inhibit in" connector to prevent the PG-

200 from being pretriggered before the most resent data capture, transfer and real time

processing was completed. The measuring system was thus operated in a synchronous mode,

and the detector delay and integration time settings and selection of active pixels were

completely controlled by the software. The entire experimental apparatus was under the

control of a Gateway 2000 Pentium 133 MHz computer running the Winspec V1.4.3.6

(Princeton Instruments) software.

A miniature spectrometer (Ocean Optics, Model# S1000) consisting of collection

fiber, optical components, modular grating disperser, and linear CCD detector with control

electronics was used in the angular dependence measurements.








Results and Discussion



Since existing elemental spectral intensity tables [144-147] are based mainly on

sources such as electric arcs, sparks, and gas discharges, one often finds that they are not

directly applicable to the observed laser-induced plasma intensities. We therefore captured

and analyzed the spectrum of our element of interest, lead. Spectra were accumulated from

10 single laser shots on a solid lead target. Spectra were captured after a delay of 2 ls, using

an integration time of 15 us. Each spectrum covered a wavelength window of~20 nm and

successive sets of measurements were taken with the center of the window shifted by 15 nm

increments to cover a total spectral range from 200-450 nm. The resulting spectra are not

shown due to space limitations. However a classification of the observed lead spectral lines

are given in table 3-1. The relative intensities presented in the table are not corrected for the

detector's varying spectral efficiency due to a lack of a reliable radiance standard for

wavelengths below 250 nm.


Plasma Imaging

Time resolved images of the laser produced plasma were captured by using the

modified experimental setup shown in figure 3-2. The collecting lens was placed at its focal

length (250 mm) from the laser-induced plasma, to provide a nearly parallel beam of radiation

incident on the interference filter. The focusing lens (f=250 mm) focused this transmitted

parallel beam of radiation in its focal plane where the ICCD was placed to capture the

spectrally filtered plasma image. The beam combiner (a neutral density filter of0.3 optical

density) allowed radiation from the plasma as viewed from the front via one of the plane











Table 3-1. Table of lead lines with their relative intensities as observed in the LIB plasma
and their associated states.


Upper Stat Lower State
Wavelength Intensity


(nm) (Relative) Term


205 328

217.000

220.3534*

223.7425

224.689

239.3792

2440 1940

241.1734

244.3829

244.6181

247.6378

257.7260

261.4175

262.8262

266.3157

269.7541

280.1995

282.3189

283.3053

287.3311

357.2729

363.9568

367 1491

368.3462

373.9935

401.9632

405.7807

406.2136


15236 8s'P,

83017 6d'D,

159839 7stS,,

16023 7d'D,

39766 7d'D,

52941 7dVF,

31515 7s'P,

14552 7d'F,

22780 8s'P,

35871 8s3P,

59174 7s'P,

63035 7s'P,

151636 6d'D,

13882 8sP,

115415 7s'P,

9837 14s'P,

161677 6d'F,

90279 6d'D,

140343 7s'P,

109403 6d'F,

62390 7s'P,

64527 7s'P,

13181 8s'P,

72813 7s'Po

55761 7s'P2

10236 6d'F,

116066 7s'P,

23673 6d'D,


Energy Term
(cm ')

48686 87 6p2'p,

46068.57 6p23'P

59448.00 6pP,,

52499.53 6p2'P,

52311.37 6p2'P,

52412.30 6p2'P,

49439.73 6p2'P,

52101.77 6p2'Pi

48726.16 6p2'P,

4868687 6p2'P,

48188.67 6p2'P,

49439.57 9p2'P,

46060.90 6p2'P,

48686.87 6p2'P,

48188.67 6p23'P

58517.67 6p21'D

46328.81 6p2'P2

46060.90 6p2'P,

35287.24 6p2'P,

4544326 6p2'P,

49439.57 6p2'D,

35287.24 6p2'P,

48686.87 6p2'D,

34959.90 6p2'P,

48188.67 6p1D,

46328.81 6p2'D2

35287.24 6p2'P,

46068.57 6p2'D2


Energy
(cm-')

0.00

0.00

14081.00

781937

7819.35

10650.47

7819.35

10650.47

7819.35

7819.35

7819.35

10650.47

7819.35

10650.47

10650.47

21457.90

10650.47

10650.47

0.00

10650.47

21457.90

7819.35

21457.90

7819.35

21457.90

21457.90

10650.47

2145790









Plane mirror for side-on view


Beam combiner


Plane mirror for frontal view

Collecting lens


- Spectral filter


r Focusing lens


ICCD
fs

Figure 3-2. Schematic of the modified optical system for capturing time resolved, 2D spectral images of
the LIB plasma simultaneously from two orthogonal directions.


Plasma





66

mirrors and a side view via the other plane mirror to reach the detector simultaneously. By

slight rotation of the plane mirrors around vertical axes, the two imagesfand s (representing

the frontal and side-on images respectively) could be laterally offset from one another. The

enlargement of the image was given by the ratio of the focal length of the focusing lens to that

of the collecting lens.

This setup allowed simultaneous registering of two perpendicular views of the plasma

during a programmable time interval A solid lead pellet was used as the sample. Interference

filters were used sequentially to produce narrow wavelength band images of the plasma. With

a proper choice of peak wavelength and spectral bandwidth, the interference filter isolated and

transmitted a single transition, providing a means of measuring the temporal and spatial

dependence of the wavelength-associated population distribution. An interference filter with

a spectral FWHM of 8 mn was used to observe the ionic line emission from the transition Pb

II at 220.3534 nm. The filter was tilted at 170 with respect to the optical axis to have the

filter's transmission peak correspond with the ionic line wavelength. This filter could not

completely isolate the ionic line since its full-width at half-maximum permitted a minimal

detection of the Pb I 217.000 nm atomic line transition. An interference filter with a peak

transmission at 280 nm and a FWHM of 8 nm was used to detect the lead atomic line. This

filter permitted radiation from the lead atomic transitions at the wavelengths 280.1995,

282.3189, 283.3053 and 287.3311 nm to be observed simultaneously, with the major

contribution coming from the 280 nm wavelength. It is important to note that during the

initial 500 ns of plasma existence, the observed radiation was mainly due to broadband

continuum radiation at the filtered wavelengths and not due to the elemental line emission.

This was verified by analysis of a sample which did not contain lead (or any spectral emitters






67

at 220 or 280 nm). To obtain a wavelength integrated image, a neutral density filter (optical

density 1) was used to attenuate the radiative flux sufficiently in order to prevent saturation

of the ICCD.

The top two images in figure 3-3 show a mm scale as viewed by the ICCD from the

front (left hand picture) and from the side (right hand picture), when the ruler was placed at

the position where the laser-induced plasma would have formed, using identical imaging. The

physical orientation for the left and right sides of the frontal view and the front and rear of the

side-on view is also shown. The ruler images were obtained from a double exposure blocking

each of the orthogonal paths successively and rotating the ruler through 90* between the

exposures. The clear plastic ruler was back illuminated with a mercury pen lamp. The lower

five frames in figure 3-3 show the wavelength integrated temporal development of the LIB

plasma using a neutral density filter with optical density 1 for radiative flux attenuation. In

each frame, the plasma image on the left is the front view, and the side view is on the right.

The laser beam is entering from the right on the side view, striking the sample at about a 45

angle. During these delay times, only continuum radiation was emitted. These images were

generated by subtracting two successive image files captured with a pulse width of 200 ns,

separated by a 10 ns delay. Even though the PG-200 could gate the ICCD with pulses as

short as 20 ns, it was not practical for full area ICCD imaging, due to the "iris effect"[148]

which caused the central pixels in the ICCD to have almost zero sensitivity at such short pulse

durations. For figures 3-4 thru 3-6, a pulse width of 200 ns was used. Each image at a

specific delay time is from a different laser produced plasma. The data shown here are from

sequentially captured laser shots. No attempt was made to use the "best matching" images.

A normalized intensity scale was used for each figure, 3-4 thru 3-6. This allows comparison







Left RiehrontR


Figure 3-3. Wavelength integrated lead plasma images. Top two images show a back illuminated mm ruler for scale. The
lower five images show the temporal development of the wavelength integrated plasma with increasing delay time.


Left


Right


Front


Rear
























C,, C n


1 1'7M


Figure 3-4. Wavelength integrated lead plasma images on a longer time scale than shown in figure 3-3.























5.ima on a loner time scale than shown in figure 3-4.
Figure 3-5. Wavelength integrated lead plasma images on a

















170 ns 370 ns 570 ns








770 ns 970 ns I 170 ns









1370 ns 1570 ns 1770 ns

Figure 3-6. Images showing temporal and spatial development of the ionic emission from the LIB plasma
transmitted by a 220 nm filter.





72

of images within one figure, but not comparison between figures. Figure 3-3 shows that the

plasma starts from the spot where the laser strikes the target and then continues to grow

along a preferential axis coinciding with the incident laser beam. Figure 3-4 and figure 3-5

shows the further development of the wavelength integrated plasma on a longer time scale

and with a different neutral density filter to show the temporal development of the combined

line emission. Lines have been drawn on the pictures to indicate the surface of the lead target.

Since the top and side of the pellet are highly reflective, anything below this line should be

ignored, since it represents reflected radiation. The reflection from the side of the pellet is

due to back reflection from the neutral density filter. These images show that at ~ 1 gs after

the initiation of the plasma, it has reached a maximum size and brightness, and has begun to

decay. It also shows that the development of the LIB plasma on a homogeneous solid sample

is very reproducible. Finally non-normal incidence of the laser beam onto the target material

leads to non-symmetrical (cylindrical) plasma formation. The protrusion seen in the side-on

view and starting to develop at delay times of > 500 ns occurs long after the termination of

the laser pulse, but along its' incident path.

Figure 3-6 shows the 220 nm ionic line emission. The first two plasma images at

delay times of 170, and 370 ns show primarily the continuum background emission. Note that

the lead ions are confined to a smaller volume, and that they do not extend out into the

plasma tail seen in the images in figure 3-4, thus resulting in a "cut off' shape at the rear of

the side-on view. The ionic emission is also shorter lived. This becomes more obvious

when comparing the ionic emission in figure 3-6 to the atomic emission shown in figure 3-7.

In figure 3-7 and figure 3-8 the atomic emission from the 280 nm atomic line and also

the contribution of emission from a few other minor lines is shown. Initially, during the




















l7O ns


370 ns


1 I'7n ,


I 1 IM I 1 I' 111 ",

Figure 3-7. Images showing temporal and spatial development of the atomic emission from the LIB plasma
transmitted by a 280 nm filter.

































5~7 6T7I.!I / is
n a longer time scale than shown in figure 3-7.

Figure 3-8. Continuation of the 280 mn filtered lead plasma images on a longertim
Figure 3-8. Continuation of the iw0 nm






75

continuum emission, the plasma images, have a "cut off" appearance, similar to the

corresponding images for the lead ionic images. Then, as the plasma decays, it develops a

more cylindrically symmetric shape. These images around 280 nm do not show the protrusion

in the side-on images for the wavelength integrated images of figure 3-4. This would mean

that the emitters responsible for the protrusion in figure 3-4 must be lead atomic or ionic

excitation states other than the ones associated with the 280 or 220 nm lines. Because of a

lack of appropriate narrow band wavelength interference filters, we could not identify the

wavelength(s) responsible for the protrusion.

From these measurements, it is obvious that the elemental spectral line emission in the

LIB plasma is a complicated three dimensional spatial and temporal function, which is

different for ionic and atomic lines and most likely different even for different wavelength lines

from the same element and ionization stage. This indicated that the assumption of local

thermodynamic equilibrium might not be valid [149].

Angular Dependence

Since the plasma images observed in the previous section are not spherically

symmetrical, the spectral brightness will be an anisotropic function. The line of sight along

which the spectral brightness is highest will also be the one providing the optimum sensitivity

in a single parameter optimization. Experiments were carried out to determine if there was

a dependence of the signal intensities on the polar angle of radiation collection when the

whole volume of the LIB plasma was observed. Since it was necessary in these measurements

to align the horizontal axis of rotation of the radiation collecting system with the plasma, the

experimental apparatus shown in figure 3-9 was used. It was necessary to keep the lens-to-

plasma distance constant while the collection optic was rotated along a horizontal axis,










Optical fiber with focusing lens
on polar rotational stage


\ I- -,
&i rT '' __


Laser beam incident at 450

Target on X,Y,Z stage


Figure 3-9. Schematic diagram of the modified radiation collection system used for the angular dependence measurements.





77

otherwise the focus would be changing. For these measurements the Ocean Optics

spectrometer was used. Since the spectrometer we used didn't have the capability of gating,

no temporal resolution was available. Therefore, the time integrated emitted radiation from

the plasma (including background) was used for the measurement. Six successive laser

produced plasmas from a solid lead target were averaged to produce the spectra shown in

figure 3-10(a). The plot shows the 10 background-corrected spectra collected at 100 intervals

as the collection optic was rotated from 90 to 0 with respect to the horizon. Figure 3-10(c)

shows the wavelength integrated net signals over the observed line profile for each lead line

at the corresponding viewing angle. As can be seen in Figure 3-10(c), the signal intensities

are roughly constant from 900 to 40, with a slight increase at 30 before they slope off. The

decrease in signal was due to blocking of the solid angle of collection by the target. We also

observed a greater enhancement at about 400-300 when this experiment was repeated with a

soil sample. This more or less constant behavior with a peak at about 300 with respect to the

horizon is consistent with a cylindrical or mushroom shaped structure.

Conclusions

We found that the spatial and temporal behavior of the transient emission generated

by a laser-induced breakdown plasma is rather complicated. The different spatial and

temporal distributions for the ionic and atomic lines is to be expected during the existence of

the plasma considering the dynamic shift in ionization equilibrium as the plasma initially heats

up and then cools down. What is quite surprising though is that different atomic lead lines

also seem to have vastly different but uniquely reproducible time and space dependencies.

Measured time resolved spectra, to establish the most appropriate time interval producing an

optimum signal to background ratio (or what ever parameter is optimized), will be dependent
















(a) (b)
1400 15000
1200 4 -
1000 10000
goo
600 5000
400
200 0
0 '280 rnm
g00 283 nm

20. 287nm

276 2'8 Z-l !s82 284 2,t 2118 4 20 40 80
Viewing Angle
Wavelength (nm)



Figure 3-10. Angular dependence plots.
(a) Background corrected lead spectra plot; (b) Background plot.














(c) (d)
20000

15000

10000

5000

7 0
1280 nm

287 4m

Viewing Angle Viewing Angle



Figure 3-10 continued.
(c) The wavelength integrated net signal plot; (d) Signal to background ratio plot.






80

on the specific volume of plasma being observed and the line of sight of the detector. These

additional variables affecting the LIB emission should be considered together with the well

documented independent variables such as the laser pulse duration and shape in time, [17]

the laser wavelength [18] and energy, [21] the physical and chemical characteristics of the

target material, [24] and the composition and pressure of the surrounding environment [26].

Multi-variable optimization techniques for increased sensitivity and decreased limits

of detection will consequently also be complex procedures, especially considering that the

physical dimension of the plasma is of the order of millimeters in size and microseconds life

time with steep spatial and temporal gradients in spectral brightness.

With homogeneous material in the solid phase as the target and constant laser and

imaging parameters, the LIB plasma is a very reproducible emission source which is a

promising characteristic for analytical applications. We conclude that time resolved, spectral

imaging has the potential to considerably improve LIBS results and lead to a better

fundamental understanding of the LIB plasma.








Part II: Level Populations In A Laser-Induced Plasma On A Lead Target

Introduction

In laser-induced breakdown spectroscopy the mass ablation rate and characteristic

elemental emission intensity vary non-linearly on a pulse-to-pulse basis (apart from the latter's

temporal and spatial variation during a single shot) because of variations in the laser energy,

the temporal pulse shape and the spatial profile of the laser beam on the irradiated material.

The chemical composition and homogeneity in the targeted volume, the physical properties

(phase, pressure, temperature) and the mechanical properties (smoothness of the surface,

crystalline orientation) of the material and the atmospheric conditions (gas composition,

pressure) play a role, suggesting the likelihood for strong matrix effects if not properly

controlled. Objective interlaboratory comparisons are further obscured by the use of different

(and often unreported) laser focusing conditions (affecting the energy and power density), the

angle of laser beam incidence, the volume or subvolume of plasma actually being observed,

the angle of light collection, and the presence of previously ablated particulates in the

breakdown volume creating a memory effect for subsequent laser shots.

An understanding of the characteristics of the processes involved will enable better

control and lead to improved analytical performance. Temporal and spatial resolved

observations of laser-induced plasmas have been reported [45,150]. In an often cited

publication [115], Radziemski et al. concluded that local thermodynamic equilibrium (LTE)

was established in the LIB plasma at delay times greater than 1 us. Their conclusion was

based on the agreement between calculations from a theoretical one-dimensional

hydrodynamic model and the intensity ratios ofC I / C I; N II / N I; Be II / Be I applying the

Saha equation and a Boltzmann plot for Be I lines in an air plasma at 580 torr. They found






82

time-resolved temperatures ranging from 22000 to 8000 K during the plasma decay. Spatial

resolution using Abel integral inversion was applied only to the Be II / Be I line pair.

Spatially resolved temperatures changed less than 5% compared to the line-of-sight integrated

values, leading Radziemski et al. to conclude that the uncertainties introduced by unfolding

a small source of only approximate cylindrical symmetry would not improve the temperature

or electron density values. Lee et al. [25] measured spatially resolved intensities from lead

lines at 357.27, 363.96, 368.35, 373.99 and 405.78 nm and applied the Boltzmann plot

technique to determine temperatures for the plasma induced in air at atmospheric conditions

by an ArF excimer laser at 193 nm. Their results lacked temporal resolution of the transient

plasma; they reported temperatures in the region 13000 K varying along the plasma height.

Simeonsson et al. [151] measured time resolved temperatures in gaseous LIB plasmas in CO,

CO2, methanol and chloroform applying the Boltzmann equation to two oxygen and two

chlorine lines and the Saha equation to two carbon lines. They found spatially integrated

temporal temperature values ranging from 15000 to 20000 K. They assumed LTE based on

theoretical considerations following the Griem [152] criteria and the experimental

observations ofRadziemski [115]. The 521.82 and 510.55 nm Cu I lines were measured by

Mao and co-workers [21] and applied to a two line Boltzmann technique to determine the

axial variation of the temperature in a LIB plasma on copper targets. No temporal resolution

was measured. They found excitation temperatures ranging from 6000 to 11000 K.

Our impression is that in spite of a vast number of reports citing temperature and

electron number density measurements in LIB plasmas, there is still a need for a complete

spatial and temporal investigation of the population distributions of several different

atomic/ionic species. As part of a research project to develop a portable LIBS instrument for






83

the determination of toxic metals in the environment and in particulate samples, we

investigated the temporal and spatial development of lead atomic emission from a LIB plasma

in air, and inferred from it, the temporal and spatial distribution of the populations of the

upper levels from which the transitions originate.

Theory

Temperature Determination Using Spectral Line Ratios

The following assumptions are required:

* The source volume under observation is in local thermodynamic equilibrium (LTE),

has a homogeneous distribution of emitters, and remains in a steady state during the

time interval of interrogation.

The spectral lines measured are optically thin.

The population distribution amongst the internal energy levels is described by the

Boltzmann equation.

From the radiances, R (spectral energy per unit of time per unit of projected source area per

unit of observational spatial angle and integrated over the line profile), of two spectral lines

(indexed i andj) from different upper excitation energy levels, E, of the same element and

ionization stage, the temperature is given by [5]



T --E-E Rg J5 Ai (3-1)



with the degeneracy of the upper levels denoted by g, the emission transition probability by

A, and the wavelength by X. The two-line method has the advantage compared to a single

line radiance application that one eliminates the need of knowing the value for the effective






84

pathlength through the source, the total particle number density, the partition function, the

need for absolute radiance calibration of the detection system and absolute transition

probabilities. Practical considerations to keep in mind when selecting a line pair are to have

the wavelengths nearly identical and the difference in upper energies of the two transitions as

large as possible. The first criterion is beneficial in the sense that relative radiance calibration

is easier and more accurate and the second assures that the calculated temperature is more

reproducible and not over-sensitive to small fluctuations in the radiance ratio measurement.

Temperature Determination Using Boltzmann Plots

The two-line ratio technique can be extended to a larger number of lines from the

same element and ionization stage. This is similar to applying a least squares linear regression

to a larger number of points to determine the slope and intercept instead of determining the

linear relationship from just two points. Rearranging the spectral line radiance equation for

multiple lines, indexed i, from the same element and ionization stage gives the expression

[153]


( R 1 E, hcln,
In + hIn --0 (3-2)
g,A T k] 4xZ


where 1 is the effective pathlength through the radiation source, n, the particle number density,

Z the partition function and the other symbols have been defined above. If we now treat the

quantity In as the dependent variable and E' as the independent variable for a set of

lines from the same species and otherwise identical conditions, a linear relationship with a

slope of -- and an intercept of In ( results. The Boltzmann plot also needs

calibration of the radiances at each of the analytical wavelengths, just as with the two-line






85

ratio technique. Relative signal values proportional to the radiances and relative transition

probabilities for the measured lines instead of absolute values will not affect the calculated

temperature and will only result in a change of the intercept. As with the two-line ratio

technique, the same practical considerations apply, including the selection of closely matched

wavelengths for accurate relative radiance calibration and large differences in upper energies

from which the transitions originate to provide the best precision in the calculated

temperature.

Experimental

Lead Spectrum. Energy Level Diagram and Line Selection

The experimental apparatus, used for all measurements, was the same as that used in

part I of this chapter. In each wavelength window, a spectrum was accumulated from 10

single laser shots on a solid lead target. The target was moved horizontally between shots to

allow sampling from a fresh location and to improve the reproducibility of mass ablation.

Spectra were captured using an integration time of 15 ps after a delay of 2 ts with respect

to the plasma initiation to avoid the initial intense continuum background. Each spectrum

covered a wavelength window of~20 nm and successive sets of measurements were taken

with the center of the window shifted by 15 nm increments to cover a total spectral range

from 180-705 nm. From the energy level scheme for lead [154], we calculated the

wavelengths in air for all possible electric dipole transitions and associated the observed lines

with their respective upper and lower energy levels. We then used the following criteria in

selecting a set of lines for thermometric purposes:

S The wavelengths of the lines should all fit within the 20 nm wavelength window of

our detection system so that they could be observed simultaneously and radiance

calibration could be considered identical for all the lines.






86

* The lines should be reasonably intense to provide line to background ratios far in

excess of one in order to minimize background noise degrading the precision of the

measured line signals.

* The transition probability values of the lines should be available [145].

* The lower energy level on which the transition ends should preferably exclude the

ground state or other low lying energy levels to minimize self absorption.

* The difference in upper energy levels from which the emission lines originate should

be as large as possible to provide better precision in the temperature measurement.

From these criteria, we selected the lead atomic lines at the wavelengths 357.275, 363.958,

367.151, 368.348 and 373.994 nm as the diagnostic lines. Table 3-2 contains the

wavelengths, transition probabilities, upper and lower state's statistical weight, configuration

and energy level values, from the publication by Wiese and Martin [145]. Figure 3-11 shows

the partial Grotrian diagram for neutral lead with the transitions used in this investigation.

Data Capture and Graphical Presentation of the Raw Data

In each exposure, we simultaneously captured radiational data for the fixed spectral

window around 365 nm containing the selected set of five lead atomic lines emitted from the

LIB plasma and their intensities along the height of the plasma. Each exposure measured the

signal (as a discrete function of wavelength and height) in a time slice with a fixed integration

time of 200 ns and a delay time (with respect to the plasma initiation) that was varied in 500

ns steps for successive files. Each file was the cumulative result from 10 single laser shots on

a solid lead target. The target was moved horizontally between shots. The data files

generated by the Winspec V1.4.3.6 (Princeton Instruments) software were analyzed by means

ofin-house written Matlab 4.0 (The Mathworks Inc.) source programs. Each graph in figure







Energy
(X 103cs np
(x 10 cm')

60 Ionization limit Pb II (P (59821.0 cm')


50 7s P (49439.57 crf') ___ 3
7s P2(48188.67cm- ') 8S P (4868687 crm)

40
7s P3 (35287.24 crr) 357.275 nm
3 1 367.151 nm
30 7s Po (349990 cr1) 373994 nm


20 6p D (21457.90 cm')
363.958 nml .,
368.348 nmJ

10 3
--_ 6p P1 (781935cnr1)

0 ---- 6P3P (ocm')

Figure 3-11. Partial Grotrian diagram for neutral lead atoms and the transitions used for the thermometric diagnostics.






88

Table 3-2. Wavelengths, transition probabilities, and level specifications of the lead lines used
as thermometric species.

Lead I Transitions
Wavelength Transition Upper State Lower State
(nm) Probability
(x l s') Statistical Configuration Energy Level Configuration Energy Level
Weight (cn-1) (cn-1)
357.2748 0.99 3 7s'P, 49439.57 6p2'D, 21457.90
363.9577 0.34 3 7s'P, 35287.24 6p23P, 7819.35
367.1513 0.44 3 8s'P, 48686.87 6p2'D, 21457.90
368.3475 1.50 1 7s'P, 34959.90 6p2'Pi 7819.35
373.9944 0.73 5 7s'P, 48188.67 6p2'D, 21457.90




3-12 shows the variation in each spectrum along the plasma height for sequential delay time

increments of 500 ns. The first spectrum was captured at a delay time of 320 ns and consisted

mainly of an intense continuum with superimposed extremely broadened, barely discernable

lines. As time progressed, the continuum background decreased rapidly while the spectral

lines increased in intensity and reached a peak intensity at around 2 ts delay time after which

they continued with time to show a slower decay in intensity and with decreased line widths.

One can also see in the first frame that the plasma emission extended only about 0.5 mm

above the target surface but then gradually expanded upwards to over 1.5 mm height. From

these figures, it can be seen that the temporal and spatial development of the Pb I 367 nm line

differed significantlyfrom the other four observed lead lines.

Temporal Development of the 360 nm Series ofPb I Line Intensities

A simpler graphical overview of the temporal development of the spectrum was

achieved by accumulating in software the signals over the plasma height for each wavelength

associated column of pixels in each file and then plotting a time sequence of their temporal

development as shown in figure 3-13. The first spectrum measured at a delay time of 320 ns













320 ns
a



3,Il





05 375
Height (mm) 0 36 333 37
Wavelength (nm)



820 ns
4











075 375
e ( 375





1320 ns
5
4^
3
2 a






305 375


Figure 3-12. Time sequence of height dependent spectra for the observed lead
LIBS plasma.







1820 ns

, .


Height (mm)


Wavelength (nm)
2320 ns


2820 ns


Figure 3-12 continued.


((I'




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E5RYV46LB_M8GLYS INGEST_TIME 2013-02-07T18:24:07Z PACKAGE AA00012949_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

LASER-INDUCED BREAKDOWN SPECTROSCOPY: FUNDAMENTALS, INSTRUMENTATION, AND APPLICATIONS By BRYAN C. CASTLE 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 1998

PAGE 2

T o my wile, Cherie, and my parents, Leighton and Suzanne

PAGE 3

ACKNOWLEDG:MENTS I would like to express my sincere appreciation to Jim Winefordner for his guidance and wonderful support extended to me over the last four years. Jim has created a sound working environment that encourages experimental freedom and, therefore, increases both productivity and creativity. The Winefordner laboratories, with the wide variety of instrumentation and materials, are second to none! My gratitude also is extended to Benjamin Smith for his friendship, assistance, and always calming presence. When things began to look overwhehning, Benny was always there to straighten the curves in the road. I have been very fortunate to have two superb research advisors and friends. I am grateful to all the members of the Winefordner research group for their support and friendship. I would especially like to acknowledge those who contributed directly to this work. These people include Andrea Croslyn, Leslie King, Gretchen Potts, David Rusak, Ricardo Aucelio, Andrew Knight, Hossein Nasajpour, and Omar Hamori. Additionally, I would like to acknowledge the following visiting scientists for their input to the presented work: Kobus Visser, Piet Walters, Nico Omenetto, and Karine Ta)abardon. In particular, Kobus, is the only person, other than Jim and Ben, who has molded me into the scientist that I am today. In the eighteen months we spent performing research together he has taught me, through example, the ways of science. My family has always provided me with love and support. I am grateful to my parents for encouraging me to continue my education, and I am very fortunate to have their guidance lll

PAGE 4

and wisdom. I can not express enough how much the love, support, friendship, and the sacrifice of my wife, Cherie, means to me each and every day. I hope and pray we live a long and happy life together so I can repay her for these selfless gifts. I would also like to acknowledge my God who has always been the rock that makes up my foundation, and the calm in the storr11. Finally, I would like to acknowledge the support of the chemistry department staff including Jeanne Karably, Steve Miles, Larry Harnly, Joe Shalosky, Gary Harding, and Dailey Burch. Each of them have made a contribution to the construction of the experimental apparatus. For financial support I would like to acknowledge the National Science Foundation Engineering Research Center for Particle Science and Technology and their industrial partners, the United States Anny, and the Florida Institute of Phosphate Research IV

PAGE 5

TABLE OF CONTENTS ACKN" 0 WLEDG MENTS ................ .............................................................................. iii ABSTRACT ................................................................................................................... ix CHAPTERS 1 INTENT AND SCOPE OF DISSERTATION ...................................................... 1 2 BACKGROUND OF LASER-INDUCED BREAKDOWN SPECTROSCOPY .... 3 Introduction to Emission Spectroscopy ..................................................... 3 Introduction to Laser-Induced Breakdown Spectroscopy .......................... 5 Review of Laser-Induced Breakdown Spectroscopy Literature ................ 12 So lids ........................... .............................................................. 15 Fundamental studies ........................................................ 15 Analytical results ............................................................. 22 Liquids ........................................................................................ 3 1 Fundamental studies ........................................................ 31 Analytical results ............................................................ 3 9 Gases .......................................................................................... 46 Fundamental studies ........................................................ 46 Analytical results ............................................................ 4 9 3 FUNDAMENTAL INVESTIGATIONS OF LASER PRODUCED PLASMAS .................................................................................. 58 Part I: Spatial and Temporal Dependence of Lead Emission in LaserInduced Breakdown Spectroscopy ................................................................. 58 Introduction ........................................................ ................................... 5 8 Experimental ....................................... ................................................... 5 9 Results and Discussion ............................................................. .............. 63 Spectra ........................................................................................ 6 3 Plasllla Imagmg ........................................................................... 63 Angular Dependence ............................................ ...................... 7 5 Conclusions ............................................................................................ 77 V

PAGE 6

Part II : Level Populations in a Laser-Induced Plasma on a Lead Target ....... ..... 81 Introduction .. .. ........... ....... .. .. .... .. .. .. ..... ....... ... . ....... ........ ....... .. . .... . .. 81 Theory ..... .. .. .. .... .. .. .... ........ . ... .... ... ... .. .. ........ .... .... .. ..... ....... .... ... . 83 Temperature Determinations Using Spectral Line Ratios . ... ... ..... 83 Temperature Determination Using Boltzmann Plots ........ . . ..... .... 84 Ex perimental .. .... ....... .. .. .. ....... .. .. .. . ..... . ... ...... .. ..... ........... ....... ......... 8 5 Lead Spect~ Energy Level Diagram and Line Selection ...... .... 85 Data Capture and Graphical Presentation of the Raw Data ........... 86 Temporal Development of the 360 nm Series of Pb I Line Intensities .... ........... ... ......... .... ....... ... .......... ...... . .. .... .... .......... 8 8 Background Correction ......... ...... .. ............ .. .......... .. . .. ... ...... . .. . 96 Stripping the 367 and 368 run Line Spectral Overlap ... . ... . ... ... ... 97 Height Dependence of the 360 Series of Pb I Line Intensities ........ 99 Boltzmann Plot ............ .............. ..... . ......... ... .... .. .. ............... ... 99 Height and Temporal Development of the Boltnnann Temperature .... ...... ..... .... .... ... .. .... .. .. .............. .. .. ..... .. .... .... 10 8 Co nclusions ........................................... .. .............. ... ...... .. ... .. ...... ...... 108 4 LIBS BENCHTOP INSTRUMENT: DEVELOP~NT AND EVALUATION ..... ....... .... . ... ....... .. .. .. .. .. ........ ........ . . ............ . . .. .. ...... ... ... 110 Introduction .......................................................................................... 110 Development .................................................. ... .. .. ...... ...... ... . ... .. .... . 111 Instrument Overview . . ... ........ ....... ..... ..... ...... ...... ... . .. ........ .. . 1 l l Pierced Mirror ............ ....................................... .. ............. .. .. 115 Fundamental Evaluations ...... .. ........ .. . ... ... .. ...... .. ......... . ...... ..... ..... .. 118 Mass Remo val ........................................................................... l 1 8 Plasma Temperature ............................. .. .................. .... ...... ... . .. 120 Optimized Gate Settings ........ ......... .. ....... ............ . ... .... . .... .. 121 Analytical Evaluations ....................... .... ... .. . ....... ... ................... .. .. ..... . 126 Sampling Procedure .... ...... .. ................... .. ...... .. .. ..... ........ ........ 126 Alloys .. ..... . .... ..... .... ....... .. ....... ... . ....... ...... ..... .... ....... .. ........ 126 Soils ......... ...... . ........ ..... ...... . ...... ... .... .. ........ ... ......... .... .. .... 128 Paints ............................ .. ...... .. .. ................ .. ..... ... ... .. .... .. ... .. ... 13 3 Organics .... ............ .. ............... .... . ...... .... .......... .... ....... .... ..... 13 3 Detection Limits .............................. ... ....... .................... ...... ... .. 13 3 Day-to-Day Reproducibility ............ .. ........ ....... . ..... ..... .. ..... . .. 13 7 Depth Profiling ...... .... ..... ..... ........ ......... .......... .. . ..... ... . ..... .. ... 139 Conclusions .. ... .... ... . .. . . ............ .. ... .. ... .......... ...... .. ........ ... . .. ..... .. .. 14 5 Vl

PAGE 7

5 LIBS PORTABLE INSTRUMENT : DEV E LOPMENT AND EV AL UAT ION ... . . .... ..... ..... .... ... . .. ..... . ...... ........................ . .... .. . ..... . ... 146 Introduction .. . ... .. ... ... .. ..... .............. .. .. ... ...... .. .. .... ... .. ...... ....... ... . 146 E xperimental ............ ...... ..... . .. .. .... .. ... ... . ..... ........ ..... .. .. .. .... ...... .. ... 14 7 Instrument Design . ......... ... ... ...... ............. ...... ..... ........ .. ..... 14 7 Spectral Imaging Apparatus .............. .. .......... ...... ... ............ . .. ... 149 Sample Preparation ... .. .. ... .. ....... ... ........ .. . ... ......... ....... .... .. . 150 Results and Discussion .. .. .. .... ..... ... ...... .. ....... .. .. .. ... . ... .... .. .... .. ...... ....... 15 0 P las01a Imaging .................. ...................................................... 15 0 Lens-to-Sample Distance ........... . .............. . ...... ... .. .. ... ... ......... 151 Evaluation of Pro be Design ........ . ..... . .. .. ... ............... ...... ... ... 154 Lens-to-sample distance ..... ... ............... ........... ... ...... .. 154 Spatial filtering ..... ....... .................... ............. .. .... ....... 156 Analytical Applications ... .... ... . .... .. .... ... .. .... ........ .. ... ......... .... 156 Analysis of paint samples . ............................ ........ ... .. .... 156 Analysis of NIST steel samples ................... .. .. .... ....... 160 Analysis of NIST iron ore samples ................................ 164 Analysis of NIST organic samples ... .. ......... .... .. ............ 166 Conclusions .... .. .. .. ........... .. ............. ... ...... ......... ..... ... .. ... .. ....... ..... ... .. .. 166 6 VARIABLES INFLUENCING THE PRECISION OF LASER-INDUCED BREAKDOWN SPECTROSCOPY MEASUREMENTS .. ..... .. ...... .... .. ...... .... 169 Introduction ... .... ........... ...... .. .......................... ..... ..... ... ........ ...... ... .... 169 Experimental .............................. .. ..... ..... .... .. ... ........ ... .. . .. .......... .... ... 1 72 Results and Discussion .......................................................................... 1 73 Choice of Analytical Line . ... .............. ..... . .. ... ....... ... ..... ...... ... 173 Signal Intra-Measurement Development ... . .... .. ....... .......... ... ... 173 Dependence of Precision on Sample Movement .. ..... ... .. .... ... .... 178 Dependence of Precision on Number of Laser Shots Accumulated . .................... ... .... .. .... .. ....... .. ....... .. ... ..... .. ... .... 184 Pulse to Pulse Stability of the Laser, and its Effect on Emission as a Function of Pulse Energy ..... ......... .... .................. .... .... .. 18 5 Dependence of Precision on Gate Delay ..... ............................... 188 Dependence of Precision on Surface Roughness ............... ....... . 189 Dependence of Precision on Background Correction .. .. ... .. ....... . 191 Conclusions .... ......... .. ... .. ....... ... .. ..... .. .. . .... .. ...... .. .................... .... .. ... 193 7 MATRIX EFFECTS IN LASER-INDUCED BREAKDOWN SPECTROSCOPY ......... ......... ... ................ .. . .. ................. .... ...................... 194 Introduction ....... ............. ............. ...................................................... 194 Background ........... ............. .. ............................. . .. .. .. ......... ... ..... ......... 194 vu

PAGE 8

Ex perimental . . . .... ... .. Results and Di s cus s ion .. 1 97 199 Bulle Matrix E ffects. .1 9 9 Internal Standards. .20 5 Speciation E ffects ..... . . . ..... ..... .. .. .......... . ...... . ... . .. .... . .. .... 214 Zinc Matrix E ffect . .... .. ..... .. .. .. .... .. .. ....... .. ... ..... .. ... . . ... ...... ... 215 Co nclusions . .. . .. ... ... . .. . 219 8 ANALYSIS OF ORES : PROGRESS TOW ARDS DEVELOPMENT OF A PROC E SS MONITOR ... .... . ....... ... . .. ... .. ..... ....... ... ....... .. .... ... .. ..... . ...... 220 Introduction . 220 E xperimental ... . ... .. .. ... .... .... ........... ...... .. ................. ... .. . .. .. .... .. .... 222 Results and Discussion ... ........ ... ...... ... ..... .... ..... ..... .. ......... ... ........ 222 Phosphate Rock Analysis .... ... ..... . ...... ....... ...... ... ..... .. . ... ... 222 Iron Ore Analysis ..... .. . ...... .... . ... ..... .. ... . ..... ... .. .... ... ...... . .. .. 224 F uture Directions .. ...... ..... ... ..................... ... .... .. .......... ....... .. . .. .. . . 22 8 9 C ONCL U SIONS AND FUTIJRE WORK .. ..... .. .. . ..... ............. . ... .. .... . .. .. 231 APPENDIX . .. ... .. .. . ...... .. .. .... .. .. . ....... ...... .. ..... ... . .. . .......... . ......... .. .. ... .. .. .. . .... 233 REFERENCES . .. .. .. . . ....... .... .. ... .. ...... ............ ..... .. .. ... .. ....... .. . .. . .. ..... . . . . 234 BIOGRAPIDCAL SKETCH .... .. ........ ... .. ...... .. .. .. .... ............... . .. ... .. ....... .. .. .. . .. .. .. 244 V1ll

PAGE 9

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 LASER-INDUCED BREAKDOWN SPECTROSCOPY : FUNDAMENTALS INSTRUMENTATION AND APPLICATIONS By Bryan C Castle Chairman: Prof. James D. Winefordner Major Department: Chemistry May, 1998 In laser-induced breakdown spectroscopy (LIBS), a laser pulse of sufficiently high power is focused to produce an irradiance that exceeds the material's breakdown threshold of 10-100 MW cm 2 If the electric field in the focused laser beam is greater than the dielectric field strength of the material it encounters somewhere in the focal volume, breakdown occurs at that point and eventually results in the fonnation of a transient highly energetic plasma For analytical purposes, the emitted electromagnetic radiation is spectrally resolved, and the emitting species in the laser-induced plasma are identified and quantified by their unique spectral wavelengths and their line intensities. Fundamental investigations of laser-induced plasmas have resulted in a better understanding of their spatial and temporal development. When exploring the lead level 1X

PAGE 10

populations in a LIBS plasma, it was determined that over the first 15 s of the plasma s life time the energy distribution of the excited atoms does not follow a Boltzmann distribution. Two different types of LIBS instrumentation have been constructed and evaluated. A benchtop instrument for laboratory use and a battery-powered, portable instrument for field use. Both were evaluated for elemental deter1nioations in alloy, soil, paint, ore and organic matrices. Investigations were conducted concerning the variables that influence the precision of LIBS measurements. These included the choice of analytical line, sample movement, number of measurements laser stability detector settings, surface morphology, and data treatment. The effect of a change in matrix on LIBS measurements was explored. Five different sensitivities were observed for the analysis of zinc and chromium in aluminum, copper graphite, sancL and potassimn bromide matrices. One possible method for compensation was through the use of an internal standard. The internal standard can be selected based on the standard criteria, with the addition of the evaluation of the temporal development of the analysis and internal standard elements signals. We also report on the investigation of the usefulness of LIBS as a process monitor in a mining facility Satisfactory results have been obtained for both phosphate rock and iron ore minerals in the laboratory. Current research efforts have focused on the transferring of the LIBS technology from the laboratory to the processing site. X

PAGE 11

CHAPTER 1 INTENT AND SCOPE OF DISSERTATION It is my impression that the forces directing university research have undergone some drastic changes in recent years. At one time the majority of experiments conducted in the university setting were driven by the needs of fundamental science. More recently the focus has shifted from being fundamentally driven to being industrial driven, that is, research is accomplished with a set goal in mind. It is not my intension to criticize or praise this change, it is only mentioned to help place the research presented in this dissertation into the right context. From the beginning my research has been industrially directed. Through our involvement with the National Science Foundation s Engineering Research Center for Particle Science and Technology we have bad the opportunity to interact with s everal mining industries. These include the phosphate mining companies located in Florida, PCS Phosphate, Inc (White Springs) Cargill Fertiliz.ers (Fort Meade) and CF Industries ( Wauchula) and also one iron ore company located in Michig~ Cleveland Cliffs Mining (Marquette). Through these interactions we learned of the industry wide need for an analytical method capable of providing quantitative, elemental information, on-line, in real time. We believed that the technique of laser-induced breakdown spectroscopy, as known as LIBS could be a possible solution to their problems. It was through this interaction that the research contained in this dissertation was founded. 1

PAGE 12

2 Chapter 2 provides a brief overview of both emission spectroscopy and laser-induced breakdown spectroscopy, and concludes with a through review of the recent literature concerning the fundamentals and applications of LIBS. Chapter 3 presents research on the .......... ~:rrnental investigations of laser produced plasmas. This was a collaborative project with Kobus Visser, a visiting professor from University of Stellenbosch in the Republic of South Africa This chapter includes all of our initial investigations into LIBS completed over a one year period, including our research into the spatial and temporal development of lead emission in LIBS plasma, as well as investigations into the lead level populations in a LIBS plasma. Chapter 4 presents the research on the construction and evaluation of a LIBS benchtop instrument. Chapter 5 presents the research on the construction and evaluation of a battery -powered, portable LIBS instrument. The following three chapters report on further evaluations conducted using the LIBS benchtop instrument. Chapter 6 reports on the investigations of the variables influencing the precision of LIBS measurements. Chapter 7 presents studies on matrix effects in LIBS. Finally, chapter 8 reports on the preliminary results obtained from the analysis of mining minerals, including phosphate and iron ores. The presentations, publications, and manuscripts that have resulted from this research are listed in the Appendix.

PAGE 13

CHAPTER2 BACKGROUND OF LASER-INDUCED BREAKDOWN SPECTROSCOPY Introduction to Emission Spectroscopy Atomic spectroscopy is based on the concept that atoms, when properly energized~ may re-emit the absorbed energy as electromagnetic radiation having frequencies characteristic of the radiating species and with an intensity proportional to the number of atoms radiating. Therefore, the frequencies at which emission is observed serve to identify the radiating species, and the intensity of the emitted radiation is a measure of the number of atoms responsible for the emission. These two properties are the basis of qualitative and quantitative atomic spectroscopy. In the late 1600s, Sir Isac Newton was credited with the first investigations of visible radiation, with his discovery of the seven colors of white light [1] It wasn't until the mid1800s through the work of Kirchhoff and Bunsen, that the qualitative usefulness of analytical spectroscopy was discovered [2]. Through their work with flame emission, they demonstrated the identification power of spectroscopy by identifying several elements based on their flame emissions. The quantitative usefulness of spectroscopy was not totally realized until the l 920s with the work of Gerlach and Schweitzer [3]. Since this time, the field of spectroscopy has really blossomed, and currently analytical chemists are realizing the fruits of these and many more earlier discoveries. The three basic components of an emission spectroscopy apparatus include the excitation source, wavelength selector, and the radiant power detector The excitation source 3

PAGE 14

4 is required to produce free atoms from any sample and to transforrn these atoms into excited species. The wavelength selector is needed to provide a specific frequency (s) to the detector The radiant power detector is capable of quantifying the intensity of electromagnetic radiation emitted from the excited atoms at a given frequency deterrnined by the wavelength selector. Early excitation sources were primarily limited to flames [ 4] Since they proved too inefficient for the excitation of certain materials, in particular refractory materials arcs and sparks were developed [4]. Arcs and sparks grew in popularity primarily because of their ability to analyze most conductive samples with minimal sample preparation. The disadvantage of these sources is their poor stability. In the 1960's the development of stable high-frequency and de plasma devices offered an even more energetic excitation source [5]. These sources offered improved sensitivity which translated into lower detection limits. One other emission source which has grown in popularity is the low pressure discharge one of which is the glow discharge [6] There have been several different types of wavelength selectors used in emission spectroscopy. These devices are primarily based on the principles of absorption, interference s patial dispersion, or interferometry. The most commonly used are the spatial dispersion devices, typically consisting of a ruled grating [7]. The grating is placed in a sealed box (spectrometer) that allows light to enter and exit through the entrance and exit slits. These are adjustable to vary the resolution and light collection efficiency of the device. Collimating optics are used to direct a parallel beam of radiation onto the grating. The grating is a plane mirror that is ruled with closely spaced groves. The grating operates under the principle that when different wavelengths of light striking the grating are reflected they constructively

PAGE 15

5 interfere at different angles relative to the normal. Therefore, as the grating is r o tated different wavelengths of radiation will be directed onto the exit slit of the spectrometer The radiation at the exit slit of the spectrometer is monitored by the radiant power detector Examples of these are photomultiplier tubes (PMf) photodiodes (PD), photodiode arrays (PDA), charge-coupled devices (CCD) and charge injection devices ( CID ) [7]. These detectors convert the incident photon flux into a measurable electrical signal. The PMTs and PDs are single-channel detectors, while the PDAs and CCDs are multi-channel detectors The multi-channel detectors are more complex and expensive, but have the advantage of simultaneously monitoring several wavelengths and the background between them. Introduction to Laser-Induced Breakdown Spectroscopy Soon after the development of the ruby laser, it was realized that when the laser radiation was focused, the intense light beam was capable of vaporizing and exciting solid material into a plasma. Tue possibility of using lasers as excitation sources in atomic emission spectroscopy was first demonstrated by Breech and Cross in 1962 [8]. This technique known as laser-induced breakdown spectroscopy (LIBS), has been reviewed by several researchers [9-12]. LIBS is based on the concept that when a nanosecond laser pulse of high energy density strikes the surface of any material, the surface temperature is instantly increased beyond the vaporization temperature. The coupling of the pulse energy into the sample occurs through several ( often unknown) mechanisms including single and multi photon absorption and dielectric breakdown [13-14]. The dissipation of this energy through vaporization is slow relative to the rate at which energy is deposited. Therefore before the surface layer can vaporize the underlying material reaches critical temperatures and pressures, causing the surfuce to explode. The ablated material, in the furtn of particles free electrons

PAGE 16

6 atoms, and ionized atoms, expands at a supersonic speed and forms a shock wave in the surrounding atmosphere. After several microseconds, the plasma plume slows down via collisions with ambient gas species, and at this point the shock wave separates from the plasma front and continues propagation at a speed approaching (or exceeding) the speed of sound. Electron number densities on the order of 10 1 5 to 10 1 9 cm 3 and plasma temperatures in the range of 10 4 to 10 5 K have been reported [ 15-16]. At this stage, the pla sma begins to decay through radiative, quenching, and electron-ion recombination processes that lead to a fonnation of a high dense neutral species zone in the post-plasma plume. The decay typically ends with the for111ation of clusters, which usually occurs within hundreds of microseconds after the plasma has been ignited. For analytical purposes, the emitted radiation (integrated over the first tens of microseconds) is spectrally resolved, and the emitting species in the laser-induced plasma are identified and qt1antified by their unique spectral wavelengths and line intensities. A simplified experimental apparatus required for LIBS analysis is shown in figure 2-1. A pulsed laser source is used to vaporize and excite the material present at the focal point of the focused laser beam. The emitted radiation from the laser-produced plasma is collected and dispersed by the spectrometer and then quantified by the detector. A digitally enhanced photograph of a laser-produced plasma is shown in figure 2-2. The plasma consist of three primary regions, the high temperature core, the lower temperature middle, and the expanding outer edge known as the shock wave. The total volume occupied by a laser plasma formed under atmospheric conditions is approximately 3 mm 3 The plasma is a transient source with a typical lifetime of approximately 50 s. Figure 2-3 shows the decay of the emission spectra observed from a pure lead sample. The front most spectrum was captured at a delay of 0 8

PAGE 17

Detector Spectrometer Nd-YAG Laser Figure 2-1. Schematic of basic LIBS experimental apparatus.

PAGE 18

HOT, VAPOR PLASMA -SHOCK WAVE Figure 2-2. Digitally enhanced photograph of a laser-induced breakdown plasma. 00

PAGE 19

Pbl 357 27 nm 355 360 Pbl 363 96 nm Pbl 368 35 nm Pbl 367 15 nm 365 Wavelength (nm) 370 Pbl 373 99 nm 375 F i g ur e 2-3. Te mp ora l d eve l o p me n t of a series of l ea d e mi ss i o n l i n es o b serve d i n a L IB S p l as m a.

PAGE 20

10 s with respect to the laser pulse, and successive spectra at 0.5 s additional increments. It can be noted from this figure that over a 14 s interval (as shown in this case) the emission signal reaches a maximum and then begins to decay. This indicates that by using gated detection, the signal-to -noise ratio can be optimized. There are several attractive features to LIBS. One of the most attractive is that only optical access to the sample is required. This has allowed LIBS to be used for remote sensing and process monitoring. Also, this is what distinguishes the LIBS technique from conventional pla.~rna emission spectroscopy. The sample doesn't need to be transported to the plasma source, because the plasma is formed within the sample or at the surface of the sample. Another feature of the technique, from a spectrochemical analysis point of view, is that it provides simultaneous multi-element capability with minimal, if any, sample preparation. However, there are some severe problems to overcome with this method including the variable mass ablation from heterogeneous samples which causes elemental quaotitation to be problematic. This could account for the fact that the LIBS technique has not yet reached its full potential. An overview of the advantages and disadvantages of LIBS is listed in table 2-1. LIBS appears at first glance to be quite simple because the vaporization, atomization, and excitation processes are carried out in one step by the laser pulse. However, the mechaois111 under which these occur is not clearly defined. The reason for the complexity is based on the large number of variables that influence the ablation process. Several of these variables and their effects have been studied including: laser pulse temporal duration and shape,[17] laser wavelength,[18-20] laser energy,[21-23] physical and chemical characteristics of the target material,[24-25] composition and pressure of the surrounding environment,[2628] and effect of magnetic field [29-30].

PAGE 21

1 l Table 2-1. Advantages and disadvantages of laser-induced breakdown spectroscopy Advantages 1 Minimal (no) sample preparation 2. All states of mater can be analyzed, as well as both conductive and nonconductive samples 3. Very small amount of material is vaporized (10s of ng) 4. Easy analysis of refractory materials such as ceramics 5. Micro analysis is possible with spatial resolution of 1-10 m 6. Capability of remote analysis in harsh environments 7 Atomization and excitation are in one step 8. Capable of simultaneous multi-element analysis Disadvantages 1. Variation in the mass ablated caused by changes in the bulk matrix. 2. Difficulty in obtaining matrix matched standards. 3. Detection limits higher (poorer) than standard solution techniques (i.e. ICP OES) 4. Poor precision, typically 5 10% 5. Standard emission disadvantages such as spectral interference and self absorption 6. Possibility of optical component damage from high energy density lasers 7. Complexity

PAGE 22

12 Review of Laser-Induced Breakdown Spectroscopy Literature There have been several publications in the last decade which have re vi ewed laser induced breakdown spectroscopy either as a unique method of elemental analysis o r as a member of the family of atomic emission techniques. LIBS has also been addre s sed in reviews on laser applications, process monitoring and materials processing Many spectroscopic studies of laser plasmas have also appeared in the physics journals. In order to construct a meaningful review the literature must first be reduced to a subset. In this instance LIBS in analytical chemistry is chosen. This review will cover fundamental studies and analytical results and applications of LIBS related to the field of analytical chemistry The review is divided according to target phase. Solids, liquids and gases are treated in sections devoted to each. Articles from the physics literature are included when they are of interest to the analytical chemist Spectroscopic studies of laser-induced plasmas created in pulsed-laser deposition ( PLD ) experiments have for the most part been omitted under the assumption that these papers are more pertinent to materials science than to analytical chemistry. Fundamentals and applications of laser ablation as a sampling technique are also ignored on the grounds that, in these instances the laser plasma does not serve as the excitation source The time frame to be covered has not been specifically defined because the amount of literature dealing with each of the three phases of targets differs greatly. The literature on s olids is easily the largest, and so the review of s olids is taken almost entirely from publications within the last 5 years. Literature on laser-induced breakdown in gases is the oldest. Many of these publications are in the physics literature and deal with mechanisms of breakdown and plasma

PAGE 23

13 diagnostics. There are a smaller nwnber of papers that report upon the determination of trace metals in gas phase matrices. Most of these are intended to prove the usefulness of LIBS in hazard monitoring. In the section on gases, the analytical chemistry literature over the past l O years is reviewed and appropriate additions are made from the physics literature. The section dealing with liquids includes bulk liquids, isolated droplets, and aerosols generated from liquids. Among the chemistry journals, this literature is the smallest. The physics is of interest primarily because of medical applications of laser-induced breakdown, and publications which are judged to be of interest to the analytical chemist have been included. All publications related to the elemental analysis of liquids by LIBS during the 1ast 14 years are included. Previous reviews on this topic begin with a 1984 review by Ad.rain and Watson titled ''Laser Micro spectral Analysis: A Review of Principles and Applications'' [31]. This was followed by Cremers and Radziemski who published ''Laser-induced Breakdown Spectroscopy: Principles, Applications, and Instrumentation'' in 1990 [32]. This paper included a brief review of theory and instrumentation for LIBS followed by a few industrial applications. In 1992, Thiem et al. reviewed LIBS theory as part of their paper titled ''Lasers in Atomic Spectroscopy: Selected Applications'' [33]. The section on laser-induced plasma covered laser-material interaction, plasma production factors (wavelength, energy, and properties of target), and emission factors (temperature and electron density). Majidi and Joseph published ''Spectroscopic Applications of Laser-Induced Plasmas'' in 1992 [34]. This publication has perhaps had the most influence in terms of style and content on the review that follows. Majidi and Joseph reviewed analytical results on solids, liquids, gases, and mixed phase systems for the years 1987-1992. Emphasis was on

PAGE 24

14 applications of LIBS such as determination of hazardous elements in air toxic elements in wastewater, and elements of interest in coals and iron ores. The authors stressed the fact that LIBS required only optical access to samples. In 1993 Ibrahim and Goddard published '' An overview of Laser-Induced Breakdown Spectroscopy'' [35]. They concentrated on the topics of laser-material interactio' local thermodynamic equilibrium, and plasma diagnostics. The instrumentation was described, the use of gated detection was explained, and a few applications were presented. It was not a literature review In a 1993 review by Darke and Tyson entitled ''Interaction of Laser Radiation with Solid Materials and its Significance to Analytical Spectrometry ," applications of LIBS were reviewed in the section on laser ablation [11]. In this paper, LIBS was referred to as laser microprobe optical emission. The following year application of LIBS to process control was briefly reviewed by Noll et al [36]. The authors discussed fundamentals of plasma fonnation and laser-material interaction with respect to optical and heat penetratio' laser energy density and absorbance. The use of fiber optics and on-line sa01pling was discussed, and a periodic table was presented showing elements which bad been determined and their detection limits in iron ore. This idea was incorporated into the review that follows. Each section was concluded with a periodic table or tables with elements that have been determined and shaded according to detection limit. Most recently, LIBS was briefly reviewed as a section in the comprehensive review of atomic emission by Sharp et al published in 1995 [37].

PAGE 25

15 Solids Fundamental studies Many papers in this area stem from the need to more fully understand the LIBS plasma in order to obtain useful quantitative results. One such paper is the 1988 publication by Chen and Yeung who used the acoustic signal generated by a laser-induced plasma as an internal standard [38] They reported that the magnitude of the acoustic wave was proportional to the emission signal for major and minor elements within the solid target Furthermore the y found that this proportionality was independent of laser power and focus spot size. In the same year Wood et al. studied the effect of laser pulse duration on soft x-ray emission from a tantalum target [39]. Using a colliding-pulse mode-locked dye laser at 620 run with pulse durations of 100 and 600 fs and a Nd:YAG laser at 1064 run with a pulse duration of 70 ps, the plasma emission in the range from 1071 nm was observed. It was found that longer pulses gave relatively shorter wavelength emission and longer emission lifetimes than short pulses. In 1989 Coche et al used laser-enhanced ionization detection in a laser plasma to study the processes of ionization and recombination [40]. They used a N 2 laser ( 337.1 run, 5 m.J 10 ns) to ablate a solid target. A dye laser was used to selectively ionize species in the plasma at different delays relative to the ablation pulse. Optogalvanic detection was used to give an indication of the number of atoms in the probe vol11me at the chosen time delay In this way ionization and recombination rates could be inferred. Also in 1989 Iida studied the atomic emission characteristics of a laser plasma in reduced pressure argon [41]. Using a ruby laser (1.5 J, 20 ns), a plasma about 10 mm in

PAGE 26

16 height, and more than 100 sin duration was formed at pressures between O and 50 Torr. The plasma at 50 Torr had greatly increased line emission and background due to confinement by the Ar atmosphere The plasma at lower pressure showed less background and less line ermss1on. Another investigation of plasma expansion was done by Balazs et al. in 1991 [20] These authors investigated ruby laser pulses on a copper target, and developed a two-part model for the interaction. The first part of the model dealt with the heating and melting of the so lid and included parameters such as thennal diffusivity pulse duration, and material density The second part of the model described the plasma expansion into a vacuum. Mason and Goldberg have characterized a laser plasma in a pulsed magnetic field [30]. The first part of the paper included spatially resolved emission studies. The authors found that when a pulsed magnetic field was oriented normal to the laser beam, it caused radial compression and axial expansion in the plasma. The emission intensities of both atoms and ions were also increased in this magnetically confined plasma In the second part of the paper the authors discussed time resolved emission and absorption studies. They concluded that the increase in emission seen in the plasma was due to Joule heating caused by the induced secondary current in the plasma. This was evident by the fact that the increase in emission was seen later in time than the maximum of the applied magnetic field. AJso, the intensity increase was attributed to increased atomization efficiency and longer residence times in the plasma. The effects of buffer gas type on the plasma produced by a Nd: Y AG laser ( 1064 nm, 100 mJ, 7 ns) on a metal target were studied by Owens and Majidi in 1991 [27] They observed an increase in the ratio of Al II / Al I intensity in heli11m gas relative to argon gas

PAGE 27

17 and air. This increase was attnbuted to the ability of excited helium atoms to transfer energy to a similar energy level in the aluminum ion. More metal target studies were evaluated by Lee et al in 1992 [25]. These authors used an ArF laser ( 193 nm, 100 m.J 10 ns) to produce plasmas on copper and lead. The lead plasma was much larger than the copper plasma (5 mm vs 2 mm) and had a slightly lower excitation temperature as determined by Boltzt11ann plots. The temperature of the copper plasma was 13200-17200 and the lead plasma was 11700-15300 K. In 1992 Marine et al studied plaqma expansion by optical time-of-flight measurements [42]. They determined that the velocity distribution of ions produced by a UV pulse of several nanoseconds duration was broad and not well defined. In contrast the ions produced by a picosecond IR pulse traveled with a velocity inversely proportional to the square root of their mass. Neutral atoms still had a poorly defined velocity distribution. These authors also noted the appearance of a bi-modal temporal profile for YO + emission and attributed this to two possible mechanisms for forniation of YO + Kuzuya et al studied the effect of laser energy and atmosphere on the emission characteristics of laser-induced plasmas [22). They used a Nd:YAG laser at 1064 run and pulse energies from 20-95 mJ in atmospheres of He, Ar and air from 1 Torr to 1 atm. The authors reported that the maximum emission intensity was observed at 95 mJ in 200 Torr of Ar. However, the maximum signal to background was obtained in helium at 40 Torr and 20 mJ of power. Images showed the different sizes and shape of these plasmas. Okana et al. studied mass removal in non-metallic inorganic solids and determined relationships between laser power and atom yields [43]. Their paper described vacancy initiated laser ablation as a process by which weakly bound atoms were released from around

PAGE 28

18 vacancies and vacancy clusters The atom yield was detennined to be an exponential function of laser fluence. Time resolved emission studies from a laser plasma on sodium chloride were reported by Yago et al in 1993 [44]. They used a Nd:YAG laser of 150 ns duration focused onto a NaCl pellet. The emission spectra showed self-reversal in air but not in a vacuum. The plasma was divided into two zones a hot core behaving as an emission zone, and a low temperature periphery behaving as a reabsorption zone. The plasma expansion rate was shown to be determined by ambient gas pressure. In 1994 Kagawa et al used a XeCl laser (308 run, 1570 mJ 20 ns) to produce plasma on a Zn target in vacuum [45] Time resolved studies showed a number density jump which represented the blast wave expansion into the observed volume. This outermost portion of the plasma was shown to be ideal for analytical measurements because the background in this area was greatly reduced. In 1994 Tamhay and Thareja studied emission in a laser plasma of Cd metal vapor fonned in a heatpipe [46]. They showed that emission from the vapor was stronger when the plasma was formed on a tungsten target than it was when the plasma was forn1ed in the gas itself The authors claimed that this was due to p11mping of the vapor by soft x-rays formed on the tungsten target. Transition probabilities of 28 Si ion lines were deter1nined using a laser-induced plasma as a source by Blanco et al in 1995 [47]. Using a Nd:YAG laser (1064 nm, 280 ml, 10 ns) to produce a plasma on pure silicon, these authors observed emission in Ar and Kr atmospheres. The plasma produced was found to have an excitation temperature of20000 Kand an electron density of 10 1 7 cm 3 Absolute transition probabilities for the Si ion lines were calculated.

PAGE 29

19 Jensen et al published mechanistic s tudies of laser-induced breakdown on model environmental samples in 1995 [24]. They used a KrF laser (248 nm, 30 ns ) to produce plasmas on Si0 2 containing Eu and Cr, which were added as the solids El 3 and K 2 Cr 2 0 7 to the sand. Detection limits of 100 ppb for Eu and 2 ppb for Cr were reported A sample was also prepared in which the source of Cr was a solution which was added to the sand and then evaporated. This method of sample preparation gave an order of magnitude less signal and a different temporal profile for the Cr emission. In 1995 Tasaka et al s tudied the emission of a laser-produced plasma on graphite [48]. They used a Nd:YAG laser in He and in air to form a plasma in which they observed a "triple phune '' composed of three distinct regions. The authors claimed that these regions appeared because of three different speeds in the expanding plasma. The fastest region was composed of carbon ions from the target and N and O ions when the experiment was done in air. The second plume was composed of the compressed neutrals in the vicinity of the shockwave. The slowest phime was then the target vapor composed of larger molecules Thareja et al also studied graphite plasmas at low pressure and found similar temporal profiles [49). Intense emission from the CN radical has been observed in plasmas produced on graphite. With the use of a high resolution spectrometer, several authors have resolved the vibrational and rotational structure in the CN emission bands. The emission from the violet band of CN around 388 nm has been used to calculate viorational and rotational temperatures in the laser plasma [50-53). In 1996 Al-Wazz.an et al studied three-dimensional number densities of species in laser produced plumes [54). They used absorption of an expanded dye laser beam to fo1m shadow images of Ba ions in a plume produced by excimer laser ablation at 248 nm. They

PAGE 30

20 also used fluorescence from planar s lices of the plasma to obtain sequential cross s ection images which could be built into a three-dimensional image. Al-Wazzan et al. also carried out an experiment in which they observed plasma in vacuum and in ambient oxygen [55]. In oxygen, the expanding plume showed increased temperature and electron number density at the shock front due to increased collisional excitation rates. In a vacuum, enhancement at the shock front was not observed. Bulatov and Liang obtained full spectra at each pixel in the image of a laser-induced plasma as depicted in figure 2-4 [56]. They used this technique to create classification maps which gave location of any species of interest within the plasma. They studied the effects of different focal length lenses and different sampling geometries on plasma for111ation and location of species within the plasma. In 1996 Mulatari et al obtained time resolved images from a laser plasma forrned on a sample at non-normal incidence [57). They varied the angle of the incident laser beam with respect to the target from perpendicular to nearly parallel. It was found that the plume was generated in the perpendicular direction regardless of the taser angle. However the maximum emission signal was obtained with normal incidence. Mulatari and Cremers also published a second study that year reporting on the use of an acousto-optic tunable fiher to capture spectrally resolved images [58). They used a series of different lenses to examine the different distributions of elements within each plasma. By collecting light from the outermost edge of the plasma, they were able to use ungated detection to obtain analytically useful spectra with low background. Nemet and Kozma studied the shape of emission lines produced on gold targets at different delays relative to the ablation laser [59}. The 406.51 and 389.79 nm lines were

PAGE 31

Figure 2-4. Spectroscopic imaging of a laser-induced plasma.

PAGE 32

2 2 observed These lines were asymmetric and shifted relative to their natural wavelength at times up to l 000 ns after the laser pulse. These lines could be descn"bed by asymmetric Lorentz-type profiles. After 1000 ns the lines appeared to be very close to their natural wavelength and were Lorentzian and symmetric In 1997 Kumiawan and Ka gawa used a long pulse Nd: Y AG laser to produce a plasma on a brass target in vacuum [60]. Toe authors were especially interested in the secondary plasma formed by compression in the vicinity of the shock wave The emission in this secondary plasma was captured and it was observed that when a wedge of ahuninum was placed very close to the target the emission signal increased in the vicinity of the wedge The authors attnouted this effect to increased co1npression provided by the shock wave interaction with the immovable ahuninum wedge. In 1997 Martin et al used a laser-induced plasma to establish spectral cahoration of their detector [61]. A Nd:YAG laser (1064 nm, 350 mJ) was focused onto targets of ceramic polyethylene ZnS and ahunin11m, all in vacuum. Line pairs in the deep UV were used with known transition probabilities to determine the relative efficiency of the detection at each wavelength. Most recently, Granse et al. modeled a laser-induced plasma and compared their model to experiments with different lasers and different materials [ 62]. The model they derived accounted for fluid dynamics of the plasma, absorption of laser energy via inverse bremsstrahlung, and the dynamics of ionization and recombination. Analytical results The number of papers describing analytical results of LIBS studies on solids is easily larger than the number of papers dealing with either liquids or gases. For this reason, only

PAGE 33

23 publications after 1992 are considered in this section. Figures of merit given for s pecific determinations differ greatly from one author to the next. In the periodic tables which follow this section, the lowest published detection limits for the elements in the matrix of interest are used to determine the shading. Carbon content in steel was determined by Aguilera et al in 1992 (63]. The authors observed the emission from the 193.l nm carbon line because the 247.9 line had a spectral interference. The experiment was carried out in a CO 2 free environment and a neighboring iron line was used as an internal standard. A Nd: Y AG laser at 1064 run was used, and it was found that the slope of the calibration curves decreased with increasing laser power Also there was a slight difference in slope for stainless and non-stainless steels. At a laser power of 100 mJ the detection limit for carbon was 65 ppm and the RSD was l.6 %. Also in 1992, Hader used LIBS for on-line quality control of rubber mixing [64). The author called the technique "remote laser microanalysis'' and used the acronym RELMA. The existing technique for quality control of rubber mixing involved a discrete sampling step in which a small portion of rubber was taken from a batch and analyzed. This led to problems when it became necessary to distinguish between bulk composition fluctuations and inhomogenieties ; with on-line sampling by LIBS in a number of randomly chosen positions in the rubber mix, the two problems could be distinguished. Elements were found in the cross-linkers and plasticizers added to the raw rubber. Lorenzen evaluated other on-line applications of LIBS in the same year [65). This time the technique was referred to as laser-induced emission spectral analysis,., and given the acronym LIESA. The authors described applications such as the determination of minor elements in liquid steel, and depth profiling of layers on metallic substrates.

PAGE 34

24 In 1993 Sabsabi et al used a KrF laser (248 nm, 100 m.J) to analyze aluminum alloys (66} They carefully optimized the delay time by using a PMT to look at the temporal evolution of several different emission lines. An aluminum line was used as an internal standard and magnesium was determined at a few ppm. The following year Thiem et al investigated LIBS of alloy targets (67). They determined Al, Cu, Fe, N~ and Zn using a Nd:YAG laser at 532 nm in a vacuum chamber. Using non-resonant lines for all the elements, the authors were able to generate linear calibration curves from a few hundred ppm to a few percent for each element. For the determination of these elements at the percent level, the authors found many potential emission lines in the spectral window from 300 to 400 nm. Thiem and Wolf used LIBS to analyze mining ores and compared the resuhs to those obtained by a digestion ICP-AES method [68}. Al, Ca, Cu, Fe,~ Mg, Mn, S~ and Ti were determined in aluminum and manganese ore The comparison showed that both methods gave roughly the same accuracy but the ICP technique had better precision The authors also pointed out the similarity in cost for the two techniques about $100,000 for either Investigations of the use of resonant wavelengths in laser ablation had been made as early as 1992 by Borthwick et al [69}. Although most of the work was carried out with time of-tlight mass spectrometry, the paper is included here becat1se of its relevance to LIBS. The authors noted that when an ablation laser was scanned through specific wavelengths, an enhanced ion yield was detected. More specifically, an enn c.err nt in Ga ions in the ablation of GaAs and an enhancement in Al ion yields in ablation of steels were noted. The ion yields were even more pronounced when grazing angles of incidence of the ablation laser were used.

PAGE 35

25 In 1995 Allen et al employed resonant wavelengths in LIBS of thin films [70J. The authors used resonant laser ablation time-of-flight mass spectrometry to investigate copper thin films on a silicon substrate Using a XeCl excimer laser pumped dye laser {5 mJ pulse energy at 463.51 nm) muhi-photon ionization of copper was observed. By using an unfocussed laser continually and watching the decay of the copper signal, the authors were able to determine the thickness of the film between 20 and 100 A. It was calculated that between 101 and 101 A per shot was removed by the laser, ahhough this was at a power below the breakdown threshold, so emission could not be collected. Anderson et al. used LIBS for depth profiling without the resonance feature of Allen s experiment [71 ). A Zn/Ni coating between 2 7 and 7.2 m thick on a steel substrate was analyzed and a cahbration curve of signal duration with thickness was found to be linear The depth resolution was far poorer than in the mass spectrometric experiment of Allen because the energy density at the target was much greater. Nonetheless, Sn coatings of less than l m, and Cr coatings on the order of a few nm on steel could be determined by this technique. Arnold and Cremers used LIBS to deter111ine metal particles on air sampling fihers [72). Tl was collected on fiher paper by passing contaminated air through a fiher or by wiping a fiher on a TI surface. The laser beam from a Nd: Y AG laser at l 064 nm was forn1ed into a line at the focus by a pair of cylindrical lenses. The detection limit for Tl was 40 ng/cm 2 offiher paper and the calibration curve was linear up to 40 g/cm 2 The 535.05 nm Tl line was used for detection. Sattmann and Stunnalso investigated the use of a muhiple Q-switch Nd : YAG laser for analysis of steel samples [73). Single, double, and muhiple pulses were used to produce plasmas on a low-alloy steel Material ablation, emission intensity electronic temperature,

PAGE 36

26 and electron number density were determined for each plasma type. All of these parameters were greater for the double and multiple pulses than for the single pulses presumably because of a snialler shielding effect in these cases. Calibration curves were correspondingly steeper. Davies and Telle used a 100 m fiber optic pair to perform remote LIBS on ferrous targets [74}. A pair of 550 m (OD) fiber optics were used to transmit laser light to the target and to return plasma emission to the spectrometer. Detection limits of 200 ppm or less were found for Cr, Cu, Mn, Mo, Ni, Si, and V. A n11mber of potential emission lines to use for calibration were given for these elements. Cremers et al. used a fiber optic probe to determine Ba and Cr in soil [75). A single 1.5 mm fiber was used both to deliver laser light and collect emission. A glass plate was used to reflect the emission to the spectrometer. The Ba ion line at 493.41 run and the Cr atom line at 425.44 were used for cah'bration which was linear over 4 orders of magnitude for both elements. The detection limit for Ba was 26 ppm, and the detection limit for Cr was 50 ppm. Cremers also determined at what distance LIBS could be performed using conventional optics rather than fiber optics [76}. By using a beam expander to increase the diameter of the laser beam and a pair of lenses with adjustable distance to focus this expanded beam onto a target, he was able to produce a plasma and collect light at a distance of 24 m. Detection limits in a simulated moon rock were at the level of a few percent due mostly to the small solid angle of collection at this distance. Bescos et al analyzed alumin11m samples in 1995 [77}. The authors simuhaneously detennined Mg, Mn, Fe, and Pb in aluminum using a spectral window from 380 to 410 om. Detection limits were around 100 ppm and calibration curves were linear up to about 1 %. Gonzales et al. also analyzed steel samples [78}. Their 1995 publication deah with sulfur determination. They reported a detection limit of 700 ppm using the 180. 73 nm atom

PAGE 37

27 line. The analysis was done in a N 2 atmosphere, and the Fe ion line at 186.47 was used as an internal standard. Precision was 7%. Hakkanen and Korppi-Tommola used LIBS to study elemental distributions of paper coatings (79]. A XeCt laser at 308 run was used to generate a plasma on paper and Ca or Si was used as an internal standard. A spatial resolution of about 250 m could be achieved with a laser pulse power of 200 mJ. It was estimated that 2 ng of paper coating were vaporized per shot Al, Si, Mg, Ca, and C were monitored quantitatively in a spectral window from 220-290 run. Sabsabi and Cielo analyzed aluminum alloys by LIBS in 1995 [80]. The authors characterized the laser plasnia on alumin11m targets using Stark broadening of Al ion lines to determine electron number densities and using Boltz111ann plots a series of Fe atom lines to determine electronic temperature. Cah"bration curves for Mg, Mn, and Si were constructed and detection limits were as low as 10 pp~ for Mg. Soil was analyzed by Ciucci et al. in 1996 [81}. They used both a Nd:YAG laser at 1064 nm and a XeCl laser at 308 run. The authors found that the background decay in the laser plasma was faster for the 308 run pulse than for the l 064 run pulse, and used a correspondingly shorter delay time with the XeCl laser. Cu, Pb, and Cr in soil were detcrinined, and an entire spectra from 350-700 run was shown with lines identified for the geological survey soil sample GXR-2. Ernst et al. used LIBS to determine Cu in A533b steel [82}. The first attempt was carried out with fiber optic delivery of the laser. This method was significant because in the hazardous environment of a nuclear reactor pressure vesset Cu concentration in the steel is an indicator of radiation embrittlement and of expected material lifetime. Because the fiber

PAGE 38

28 optic could not deliver enough power to produce sensitive detection, beam delivery was done with conventional optics. The Cu line at 324. 75 was used for calibration., and by using a second order cahbration curve CU could be determined in the range between 100 ppm and 5%. Geertsen et al revisited LIBS for aluminum samples in 1996 [83]. Using a Nd:YAG at 1064 run and a pulse power of 230 mJ the authors forrned craters ~ 5 m deep in aluminum with a single shot. The detection limits for Mg and Cu were 4 and 40 ppm, respectively with a precision of about 8 o/o. Spatial resolution was assessed by rastering the beam over a sharp Al-Cu junction. The best spatial resolution obtained was 6 m. Marql1ardt et al. determined Pb in paint using a Nd:YAG laser at 532 nm coupled into a fiber optic probe [84]. The probe consisted of two fibers, excitation and collection, terminated in a probe head with an aspheric lens to focus the laser light to a spot and focus plasma emission into the collection fiber. The common end of this fiber is shown in figure 25. Different combinations of excitation/collection fiber diameters were tried. The detection limit for Pb was 140 ppm, and precision was 5-10 % even when analysis was carried out through layers of non-lead containing paint. Miziolek also reported on a LIBS probe in 1996 [85}. This application was for determination of heavy metals in soils, and the probes were to be used in a cone penetrometer truck. In one probe, a compact laser was mounted in the probe head In the other probe, laser light was carried through a fiber optic to the probe head. The compact laser head provided the more sensitive probe with detection limits for Pb, Hg, Cr, Cd, and Zn of 1-10 ppm in sand and sih. The fiber optic delivery provided less power but a higher laser repetition rate; 15 mJ vs 28 mJ, 30 Hz vs 1/3 Hz.

PAGE 39

Collection Fiber Excitation Fiber f/2 Lens Sample Aspheric Lens Figure 2-5. Common end of fiber optic LIBS probe used by Marquardt et al for deter111ination of lead in paint

PAGE 40

30 Also in 1996, Cremers et al introduced a portable LIBS instrument and used it to determine hazardous elements in soil, in paint and in samples on filter paper [86J. A compact laser head was used as the probe. Ba, Be Pb and Sr were determined in soil with detection limits of 265 9 3 298, and 42 ppm respectively In paint the detection limit for lead was much greater because the 405.8 run line could not be used due to s pectral interference. On filter paper Be and Pb were determined with detection limits of2t ng/cm 2 and 5.6 g/cm 2 respectively. In the filter paper studies, some particle size effects were evident Palleschi e t al used a 400 mJ Nd: Y AG laser at 1064 nm for a variety of deter1ninations including Hg in air (5 ppm detection limit ), and pollutants in power plant smoke and soil [87J These authors analyzed the geographical survey GXR2 sample and determined concentrations for 18 elements. Vadillo and Laserna analyzed geological samples of vanadinite pyrite garnet and quartz [88J. These determinations were done in vacuum to increase the lifetime of the ionic s pecies and of the ion lines observed. Fe Mn, Mg and Si were determined in each of the rocks and Al was also detertnined in garnet. Differences in compo s ition were as expected. Laserna e t al used LIBS to analyze the surf.ace of solar cells [89]. A N 2 laser at 337. t nm wa s focused onto the cell and the cell was rastered under the focus providing s patial resolution of about 30 m. The C ion line at 588 9 Ag atom line at 546.5 Si atom line at 634.7, and Ti atom line at 625.9 were used in determinations of these four elements. The concentration of each element was mapped across the surface of the solar cell In another application of LIBS to alloys, Kim et al determined aluminum in a zinc alloy [90]. The authors used a Nd:YAG laser ( 1064 run, 105 mJ 3 ns ) in vacuum, air and argon. The presence of Al in the Zn alloy was important because of its detrimental effect on

PAGE 41

31 its welding properties. The delay and the distance from the surface of the target were optimized in each atmosphere. Because the Al atom line at 308.22 nm was u se d for cahoration, large delay times were used to insure no contribution from a shorter-lived Zn interference. In air, a 30 s delay was used and in Ar a 50 s delay was used. The Zn line at 307 59 was used as an internal standard. In 1997 Maravelaki et al. used LIBS to monitor the laser cleaning of marble artifacts [91]. Although lasers had been used previously for the cleaning of such treasures these authors were the first to examine the emission from the plasma. Crusts of 20-600 m consisting of gyps,1m, iron oxides soot, and calcite were ablated from the marble by a Nd:YAG laser. The plasma fortned in this ablation was analyzed spectroscopically to deter1nine the endpoint of the cleaning. Figures 2-6 and 27 show graphically the detection limits for elements that have been deter111ined using LIBS in sand, ore or soil and in steel or alloy. Li.Quids Fundamental studies Besides papers in the field of analytical chemistry, there have also been a number of papers which have dealt with the physics of laser-induced breakdown in aqueous solution. Most of these deal with the processes of cavitation and the factors which can affect the breakdown threshold. In 1984 Armstrong identified three time domains in the heating of an aerosol: the acoustic regime the internal conductive regime, and the external conductive regime [92J. He explained the phenomenon of aerosol-enhanced air breakdown as being caused by increased electron collision :frequency in the hot, dense vapor surrounding the heated aerosol particles in the beam.

PAGE 42

H He Li Be B C N 0 F N e Na S i p S C l Ar K Sc Cr Fe Co G a Ge As Se Br Kr Rb Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Te I X e Cs La Hf Ta W Re Os Ir Pt Au Hg Tl Bi P o A t Rn Fr Ra Ac Key : D < 10 ppm 10100 ppm > 100 ppm Figure 2-6. Elements deter1nined in ores rocks, and soils shaded according to detection limit.

PAGE 43

H He Li Be B N O F Ne Na Mg Si p Cl Ar K Ca Sc Ti V Cr Mn Fe Cu Zn Ga Ge Se Br Kr Rb Sr Y Zr Tc Ru Rh Pd Ag Cd In Cs Ba La Hf Ta Re Os Ir Pt Au Hg Fr Ra Ac Key: D < 50 ppm 50 200 ppm > 200 ppm Sb Te I Xe Pb Bi Po At Rn Figure 27. Elements deterrnined in metals shaded according to detection limit

PAGE 44

34 Three years later Armstrong used a CO 2 laser at 10.6 m to explode aerosols generated by a vibrating orifice aerosol generator [93}. A UV laser was used to produce shadow images of the exploding droplet and a phase Doppler particle analyzer system detennined size and velocity of the expelled particles at a distance of2 mm from the original drop. It was found that the CO 2 laser power affected the size of particles produced in the breakdown for original drop sizes between 30 and 50 m. The speed of the particles produced was independent of laser power. Chylek et al studied the effect of size and material of liquid aerosols on breakdown thresholds in 1986 [94}. The setup that they used is shown in figure 2-8. A Nd:YAG laser (532 run, 10 ns) was focused onto droplets generated by a Berglund-Liu vibrating orifice aerosol generator. Breakdown thresholds were determined on aerosols of different size generated from liquids of various refractive inde~ density, surface tension, and chemical structure. Thresholds were seen to decrease with droplet size. Refractive index had no apparent effect. Density, surface tension, and chemical structure had effects which could not be independently examined. In 19g7 Hsieh et al used a Nd:YAG laser at 532 nm to examine the exact location of breakdown initiation on water droplets generated by a Berglund-Liu vibrating orifice [95). They found that as the energy of the laser increased, the breakdown moved from outside the droplet to the inside. Theoretical calculations showed that the curved liquid gas interface of a water droplet in air focused laser energy at one point just inside the illuminated face and at another point just beyond the shadowed face.

PAGE 45

Pyroelectric Detector Acoustic Detector Microscope Viewing from the top Vibrating Orifice Frequency Generator Figure 2-8. Schematic of experimental setup used by Chylek et al to investigate LIBS of aerosol s.

PAGE 46

36 The effect of laser wavelength and irradiance on spectra from laser-induced breakdown of single levitated aerosol droplets was examined by Biswas et al. in 1988 [96]. Droplets of glycerine saturated brine solution approximately 18 m in diameter were optically levitated and probed with a Nd:YAG laser at 1064, 532, and 355 nm. Emission lines ofNa, C, and N were observed. It was shown that the energy required for breakdown, and the time of plasma emission increased with increasing wavelength. The optimum time delay for determination of these elements increased with laser wavelength and showed more variance from element to element at longer laser wavelength. Also in 1988, Zheng et al. acquired temporally and spatially resolved spectra of laserinduced breakdown of a 40 m 4 M NaCl droplet using a Nd:YAG laser at 532 nm [97]. Using fiber optic ribbons and a streak camera, these authors observed that emission began first at the shadowed face of the droplet and propagated toward the illuminated face. Hammer et al also examined thresholds with uhrashort laser pulses [9S]. Using pulses of2.4 ps, 400 fs, and 100 fs from a pulsed dye amplifier, they determined breakdown thresholds in saline solution, high purity water, and tap water. The thresholds in these media were not significantly different. Wrth the 2.4 ps pulse the breakdown threshold was found to be 5 x 1 oi 1 W cm1 with a 400 fs pulse it was 1.3 x 10'2 W cm 2 and with a I 00 fs pulse it was 5.65 X 10t2' W cm2 Kitamori et al. used the acoustic wave generated during particle-induced breakdown to detect polystyrene particles in aqueous solution [99]. A Nd:YAG laser (532 nm, -1 ml 6 ns, 10 Hz) was focused into a sohrtion. In pure water, the energy density produced by this beam was not sufficient to cause a breakdown. When a particle entered the probe volume, however a breakdown occurred and subsequently an acoustic wave was generated. By

PAGE 47

37 detecting these acoustic waves, the probability of breakdown was determined. By analyzing solutions of known particle concentration, a calibration curve of breakdown probability vs concentration was generated. This curve was used in the analysis of solutions of unknown particle concentration. Vogel et al used a Nd:YAG laser to produce and image a breakdown in water [lOOJ. The fundamental wavelength of 1064 nm was focused into the water to produce a breakdown. A portion of the beam was frequency doubled to produce 532 nm light which passed through a beam expander and through the underwater breakdown, providing a shadow of the plasma and the cavitation bubble. The size of the plasma and cavitation bubble were examined for 1064 run laser pulse durations of 30 ps and 6 ns. Sacchi discussed the mechanisms of laser breakdown in water and described the differences in threshold for short pulse and long pulse lasers in 1991 [101J. He claimed that the avalanche breakdown process or inverse bremstrahlung was responsible for breakdown when nanosecond laser pulses were used. In this case, the breakdown threshold behaved probabilistically. The probability of breakdown scaled linearly with the log of laser power for low powers, and scaled parabolically with the log of laser power at higher power. In contrast, with laser pulses of less than nanosecond duration, muhi-photon ionization appeared to be the mecbanis111 ofplas11,a fonnation, and a definite threshold above which breakdown would occur could be determined. Sacchi returned to this topic in 1996 and produced images of laser-induced breakdown on solids under water. By using a dye laser oriented parallel to the underwater target and delayed relative to the ablation laser, the authors obtained shadow graphs of the evolving plume and shock wave. A series of images followed the growth and decay over a total time of about 700s. A schematic of the setup used is shown in figure 2-9.

PAGE 48

Trigger and Delay ... "' ... Lens C :> Sample Cell Dye .......... ............. .. Beam xpander Water CCD Target Figure 2-9. Schematic of experimental setup used to obtain shadow graphs of a laser-induced plasma. w 00

PAGE 49

39 In 1992, Pinninck et al studied the effect of resident particles on laser-induced breakdown thresholds [102). It was found that even at wavelengths where the particles did not absorb laser light, there was a decrease in breakdown threshold in particle laden air versus clean air. This was taken to mean that particles were focusing laser light. Location of plasma formation (inside, shadowed side or illuminated side) was found to depend on ionization potential, gas pressure and laser wavelength. Refractive index had no effect on breakdown threshold, in agreement with the observations reported by Chylek in 1986 [94]. In 1993 Nyga and Neu created a plasma on calcite submerged in water using two fiber optics to deliver two pulses from two XeCl lasers [103}. Both pulses were 30 ns pulses at 308 nm through 600 m fibers which were brought in close proximity to the target with no focusing lense~. The pulses were separated in time by -300 s, and the second fiber was equipped with a 308 nm dielectric beam splitter and used to collect emission from the calcite. No quantitative results were reported. Feng et al also examined the effect of laser pulse duration on breakdown thresholds in water [104]. In their 1997 paper, they deterrnined the mechanism of breakdown with nanosecond pulses to be due to cascade ionization, and the mechanism of breakdown with ultra-short pulses to be muhi-photon ionization, similar to Sacchi's findings [101). These authors, however, went on to model the breakdown phenomenon in water relating the breakdown threshold to laser power, spot size, and pulse duration. They derived an extended non-linear Schrodinger equation to descnl>e the relationship. Anab'tical results Quantitative resuhs for LIBS with liquid "'lmples will be presented in two sections based on the nature of the analysis. The first section will cover determinations done in bulk

PAGE 50

40 liquid. This implies that a laser spark is generated either on the surface or beneath the surface of a relatively large liquid sample. The second section will cover determinations done in aerosols or droplets generated from liquid samples. Analysis of dry aerosols generated from solids are included in the last section on gas phase determinations. Bulk liquid. In 1984, Cremers et al reported the use of a repetitive single spark ( RSS) and repetitive spark pair (RSP) for the deter111ination of L~ Na, K, Rb Cs Be, Mg Ca, B, and Al in aqueous and organic solutions [l 05]. With the single spark, detection limits for all elements except Li were > l ppm. The detection limit for B was 1200 ppm. When a second spark which was generated from a second laser and delayed by 18 s relative to the frrst was used to forrn the plasma, the detection limit for B was reduced to &O ppm. This improvement was attnouted to formation of the plasma within a cavitation bubble when the RSP method was used. The experimental setup is shown in figure 2t 0. In 1987 Wachter and Cremers reported a detection limit of 100 ppm for uranium in solution by LIBS [106). The plasma used was formed by a Nd:YAG laser (1064 run, 260 m.J) on the surface of a 4 M nitric acid solution contained in a small glass vial. Each analysis was done by averaging 1600 laser shots. This averaging was necessary to overcome poor shot-to shot precision due mainly to small variations in lens to sample distance. In 1987, Cremers conducted an experiment to determine the maximum lens to sample distance (L TSD) that could be used in a LIBS experiment [76]. With a 250 mJ pulse from a Nd:YAG laser at 1064 run, it was found that a 2 m focal length lens could produce breakdown on molten metal. The larger problem was collection of the emitted light. A fiber optic could collect enough light at 0.5 m from the plasma for modest figures of merit to be obtained. At distances farther than this, the sensitivity of the experiment decreased rapidly.

PAGE 51

1064 nm Pulsed Laser Beam Windows 5 cm Focal Length Lenses Spark c~-----:-'> Spark Light Liquid Cavity .. . ........... ... ................ ............. .. Teflon Cell Fill Hole Figure 2-10. Experimental setup used by Cremers et al. to analyze aqueous solutions with LIBS.

PAGE 52

42 In 1993 Aragon et al used LIBS to determine carbon content in molten steel [107]. They used a Nd:YAG laser (1064 200 mJ 8ns ) focused onto molten steel in a crucible under an argon atmosphere. In order to provide a homogeneous sample, a jet of argon gas was directed downward into the crucible. This served to remove the topmost layer of molten liquid which would be enhanced in the lighter elements. Carbon was detennined by ratioing of the C 193 09 run to Fe 201 07 nm intensity and a detection limit o f250 ppm was found with an RSD of 6 %. In 1995 Stolarski e t al produced sub-surface plasmas in saline solution, triple distilled water and tap water with Nd:YAG laser pulses of < 5 mJ and pulse durations of 5 ns and 80 ps [108]. The focussing lens used was 2.54 cm. in diameter and had a focal length of 17 mm. Thi s lens was chosen to approximate the lens in the h11man eye. The authors found that an energy of 1.5 rnJ was enough to produce a breakdown at either pulse duration, and found that the plasrna temperatures and electron number densities depended little on pulse duration for the two studied These authors observed sodium emission around 590 run in a 0.9% NaCl solution, but no quantitative analysis was perfor1ned. Also in 1995, Ito et al used LIBS to determine colloidal iron in water [109] They focused a Nd:YAG laser (1064 nm, 100 mJ 10 ns) into a flowing stream of water containing FeO(OH) as subrnicron particles. The detection limit for iron was in the ppm range. This technique differed slightly from previous techniques in that it relied on particle-induced breakdown for the formation of the plasma. A plasma was formed only when a FeO(OH) particle was present in the probe volume. Because of the extremely small size of the particles however, it was essentially a detern1ination of iron in water. The following year the same authors used a second laser delayed by 1 s relative to the first in a technique similar to

PAGE 53

43 Cremer's repetitive spark pair. With this technique, they were able to improve the detection limit for iron to 16 ppb [110]. In 1996, Knopp and Scherbawn used a dye laser at 500 nm with a pulse energy of 22 mJ to produce a sub-surface breakdown in aqueous solution [111]. They reported detection limits for Cd, Pb, Ba, Ca, L~ and Na of 500 ppm, 12.5 ppm, 6.8 ppm, 130 ppb, 13 ppb and 7 .5 ppb respectively. No signal was obtained for either Hg or Er at 0.1 o/o in solution. When micron-sized particles ofErB~Cu 3 0x, rather than a soluble salt, were used as the source of Er, the detection limits decreased by a factor of 10 3 The improvement illustrated the increased sensitivity of particle-induced breakdown as compared to breakdown in particle-free solution. This experiment also suggested that 500 nm laser light had analytical advantages over 1064 nm light. In 1997, Ho and Ng used both a Nd:YAG laser (532 nm, 12 ns) and an ArF laser (193 nm, 15 ns) to determine Na in aqueous solution [112]. In order to increase absorbance of the laser light, methyl violet was added to the solution to be analyzed. It was found that the 532 nm laser light produced a more visibly intense plasma; because the continuum emission was higher in the 532 nm produced plasma, the 193 nm laser light actually gave better detection limits for Na A detection limit of230 ppb was reported for Na in the methyl violet solution with the plasma formed by the ArF laser. In 1996, a similar experiment was performed by Paksy et al. who used a Nd: Y AG (1064 nm, 15 mJ, 4 ns) focused onto the surface of a molten alloy in an argon atmosphere [113]. Because the laser power used was significantly less than that used by most authors, the delay of the detection relative to the laser was also less A 100 ns delay and 1000 ns gate width were used. Detection limits were 0.001 % for Si in Fe, 0.006o/o for Cr in Fe, 0.06% for Si in Al, and 0.007% for Mn in Al.

PAGE 54

44 Arca et al deter1nined Cr Pb and Cu in aqueous solution with a single laser shot on the surface of the water [114] Detection limits were 100, 100, and 50 ppm for the three elements, respectively. Calibration curves were linear over l order of magnitude. The authors also analyzed aqueous solutions by placing a drop of water to be analyzed on a KBr pellet. When this technique was used, detection limits were improved by almost an order of magnitude. Aerosols and droplets. In 1983 Radziemski et al used a nebulizer/heat chamber system to produce an aerosol from a liquid solution and analyzed the aerosol with LIBS [115]. Detection limits for Na, P As, and Hg were reported to be 6 ppb, 1.2 ppm, 0.5 ppm, and O. 5 ppm, respectively. The authors detert11ined that local thertnodynamic equilibri,1m existed in the aerosol spark for time delays greater than 1 s relative to the laser. In 1987, Eickmans et al spatially resolved the emission from laser-induced breakdown of an aerosol [116]. Approximately 45 m diameter droplets were formed by a vibrating orifice aerosol generator which was synchronized with the Nd:YAG laser beam at 532 nm. Emission ofL~ Na and H was observed in 5 M salt solutions of either NaCl or LiCl. Spatial resolution of the e~ion revealed that in the plasma plume in the vicinity of the illuminated face of the droplet, emission lines showed Stark broadening and self-reversal. Near the shadowed face of the droplet, emission lines were narrower and less intense. In 1988, Essien et al used LIBS to detertnine Cd, Pb, and Zn in an aerosol [117]. A Nd:YAG laser (1064 nm, 100 mJ, 5 os) was focused onto an aerosol generated by a nebulizer/heat chamber. Detection limits were 19 ppb, 210 ppb, and 240 ppb for Cd, Pb, and Zn, respectively

PAGE 55

4 5 Also in 1988 Archontaki and Crouch used an isolated droplet generator to pr o duce equally spaced, uniformly sized droplets [118] The experimental setup was such that a s ingle drop of known s ize was always in the probe volume of the laser for a given repetition rate. A Nd:YAG laser ( 1064 nm, 100 mJ) was used to determine detection limits for Li, Na, Mg, Ca, Mn and Al of 0.3 2 2 1.9 0 4 7 2, and 5 2 ppm, respectively. Droplet diameter had little effect on the detection limit for sizes between 58 and 75 m. In 1992 Ng et al used an ArF laser ( 193 nm, 150 ml) to analyze an aerosol generated by a concentric glass nebulizer / spray chamber traditionally used for ICP-AES [119). The authors determined the optim11m delay for detection to be 6 s relative to the laser pulse Detection limits for N~ Li, In, AL Ga, Ca, Mg, K, and Sr were determined to be 0.9 0.3, 10, 3 3 8 3 2, and 20 ppm, respectively. The following year Parigger and Lewis used a picosecond XeCl laser (308 nm, 4 mJ) to produce LIB in 66 m water droplets [120]. Plasma diagnostics showed a temperature of ~ 10 000 Kand an electron number density of 10 18 cm 3 The Ha line at 656 nm was used as an internal standard and the detection limit for sodium was ~ 1 ppm. Poulain and Alexander perfor1ned a very similar experiment in 1995 using a KrF laser ( 248 nm, 200 m.J). The detection limit for Na was found to be 165 ppm [121). Raisch and Paine reported on the characterization of colloidal particles by LIBS in 1996 [122]. Field flow fractionation was used to sort particles by size and deposit them onto filter paper using partitioned pumping to retain time resolution. A cylindrical lens was used to focus the laser onto the paper to avoid ablation of the paper which could occur if the energy density became too high. Detection limits for Si were in the ppb range.

PAGE 56

46 Figures 2-11 and 2-12 s how graphically the detection limit s for elements that have been determined using LIBS in bulk liquids and in aero s ols generated from liquids Gases Fundamental studies In 1988 Kumar and Thareja studied laser-induced gas breakdown in the presence of an electric field [123] They used a XeCl laser (308 nm, 60 mJ 8ns) focused between two electrodes separated by 6.5 mm. A transverse static electric field was produced by applying different potentials to the electrodes. In order to study the effect of the field on breakdown thresholds the current on the electrodes was adjusted s o that no breakdown occurred using only the electric field or onl y the laser The authors were then able to determine that pre breakdown electron densities of 10 10 cm 3 created by the field were sufficient to lower the laser energy density needed to produce breakdown. Parigger and Lewis used a Nd : YAG laser ( 1064 nm, 300 mJ 7.5 ns ) to pr o duce a plasma in CO [124] They observed the emission of the C 2 Swan band around 565 nm to determine VIbrational temperatures in the decaying p1asrna. The observed s pectra was fit with a model in which te eratllf"'" was a parameter. At a time delay of 30 s relative to the laser pulse a temperature of 6745 K was obtained This temperature was confirmed by use of the CN violet band around 388 nm to calculate the vibrational temperature in a s imilar way. Yagi and Huo investigated breakdown in H 2 gas at different pressures initiated by a KrF laser ( 248 nm, 250 mT 20 ns) [125]. This work was related to Raman spectroscopy in that the breakdown threshold detern1ined the maximum energy with which the Raman transitions could be pumped The authors found that below 600 Torr breakdown thresholds were pressure independent. Thre s holds decreased between 600 and 3000 Torr and increased s lightly above 3 0 00 Torr.

PAGE 57

H He Li C N 0 F Ne Na Si S Cl Ar Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag In Sn Sb Te I Xe Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Fr Ra Ac Key: D < 1 ppm l 10 ppm > 10 ppm Bi Po At Rn Figure 2-11 Elements dete11nined in bulk wa ter shaded according to detection limit

PAGE 58

H He Li B C N 0 Ne Na Mg Al Si Ar K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge Se Br Kr Rb Y Zr Nb Mo Tc Ru Rh Pd Ag Cd Cs Ba La Hf Ta W Re Os Ir Pt Au Fr Ra Ac Key: D < 1 0 ppm l O 100 ppm > 100 ppm Sn Sb Te I Xe Tl Pb Bi Po At Rn Figure 21 2. E l eme n ts d e t e r1r1 ine d in liqui d a e r oso l s s h a d e d accor din g t o d e t ec ti o n limit

PAGE 59

49 Yalcin et al investigated the sensitivity of a laser spark produced by a Nd:YAG at 532 nm to ambient conditions such as variation in background gas, presence of particles, and humidity [126]. This was done in order to assess the potential usefulness of LIBS in toxic metal monitoring. The authors found that excitation temperature and electron number density in the plasma remained essentially constant with variations in the ambient conditions. Analytical results Cremers and Radziemski again carried out, pioneering work on gasses. In their 1983 publication, they detected chlorine and fluorine in air [127]. Detection limits for these two gases were 8 and 38 ppm by weight, respectively. These values corresponded to absolute detection limits of 80 and 2000 ng. In a helium atmosphere, the absolute detection limits improved to 3 ng for both. Experiments were carried out in which molecular gases containing both chlorine and fluorine were introduced to the sampling chamber. The relative intensities of the chlorine and fluorine signal were found to be an indication of the number of each of these atoms in the molecular gas. All experiments were carried out using a Nd:YAG laser (1064 nm, 100 mJ, 15 ns) and detecting Cl at 837.6 nm and Fat 685.6 run. These authors also detected Be in air [128]. The limit of detection in this experiment was 0.5 ppb by weight. The Be was introduced into a chamber by laser ablation and diluted by air. A second laser produced a breakdown in the chamber from which the emission was measured. It was found that the emission from the Be II doublet at 313 .1 om was stronger than any atomic emission even at delay times of 20 s. This doublet was used for calibration which was linear over 4 orders of magnitude. Cremers et al. used LIBS to deter,nine sodium and potassi11m in a coal gassifier stream as early as 1983 [127]. These elements were present as particulates in the stream and

PAGE 60

50 so agai~ the question of whether or not the target was a gas or solid arises Regardless, the authors were able to determine Na with a detection limit of 4 ppb by weight using the doublet at 589 run. Figures of merit for potassium were not given. In conclusion, the authors noted that sulfur could also be monitored. Sneddon extended this technique to the determination of P in air and lowered the absolute detection limit of Cl in air. In a 1988 paper, Sneddon detected phosphorous at 15 ppm and Cl at 60 ppm by weight in air [129]. A calculated plasma volume of 0 010 cm 3 was used to deter111ine an absolute detection limit of 60 pg for Cl. Ottenson et al reported on the application of LIBS to analyze particulates in a coal combustion vent in a 1989 publication [130]. A Nd:YAG laser was triggered on the scattering from a HeNe laser to explode particles detected in the probe volume. The experiment used on-line t~ .. .L .... e correction to account for different excitation conditions in plasmas formed on different size particles. Moreover, the size of particles to be sampled could be adjusted by varying the laser power. With lower laser power, only larger particles were sampled; the smaller particles were not sufficient to provide plasma formation. The on line temperature correction automatically accounted for differences in excitation conditions with varying laser power. In 1990 Morris et al used laser-induced breakdown as a detector for gas chromatography [131]. The UV laser light was focussed into the effluent from a gas chromatograph at a power below the breakdown threshold of the carrier gas. When eluting carbon containing molecules were present in the effluent, the threshold for breakdown was lowered and a plasma forn1ed. The organic analytes were indirectly detected by optically detecting plasma formation. A schematic is shown in figure 2-13. No emission from the plasma was measured, and so the potential of this technique was only as a universal detector.

PAGE 61

Focused Laser Beam 193 run Screen Gas Sheath Optogalvanic Probe Microplasma Capillary Figure 2-13. Schematic of setup used by Morris et al for use of laser-induced breakdown as a detector in gas chromatography.

PAGE 62

52 In 1991, Cheng et al studied polyatomic molecular impurities in helium gas at ppm levels [132]. Using ungated detectio~ they determined B 2 ~ PH 3 and AsH 3 with detection limits of 1, 3 and 1 ppm, respectively The ionic phosphorous lines at 602.4, 603.4, 604.3, and 605.5 nm were used. As lines were observed at 228.8, 235.0 278.0 and 286.0 nm. B lines were observed at 336.0 and 434 5 om The plasmas were produced by a Nd: Y AG laser at 532 nm. Also in 1991, He and co-workers focused a pulsed laser onto a metal rod in an atmosphere of helium or argon seeded with a reaction gas to produce ion-molecule complexes and observe their emission [133]. Ion lenses were used to preferentially draw ions into the He or Ar gas for reaction with the seeded gas. Emission from species such as AlO, AIO-CO Al-Ar and Al + H 2 was observed. Joseph and Majidi used an electrothennal atomizer to create a gas phase sample from a liquid for analysis by LIBS [134]. A few L sample was deposited in the furnace and was dried, ashed, and atomized. The setup is shown in figure 2-14 At the onset of atomization, a Nd:YAG laser ( 1064 mn, 100 m.J) was focused into the furnace and fired at a repetition rate of 10 Hz. Emission from the plasma was collected through the dosing hole of the tube. Experiments with cobalt and cadmium gave absolute detection limits of 5 and 50 pg. Casini et al. also determined several elements of interest in air using LIBS [135] Their experiment used a Nd:YAG laser (1064 nm, 40 mJ, 7 ns) to produce a plasma in an atmospheric pressure chamber as shown in figure 2-15. Using the saturated concentrations of air over chosen liquids at the given temperature, and then successively diluting these concentrations in the chamber yielded calibration curves for a number of elements. C l was detected at 449 0 nm with a detection limit of 60 ppm, S at 415.3 nm with a DL of 200 ppm,

PAGE 63

Computer Dichroic Mirror Pulse Atgplifier Carbon Rod ---A I rap 1 e Lens Tube Lens High Energy Mirror Photodiode Array Spectrometer Figure 2-14. Schematic of setup used by Joseph and Majidi for LIBS in a graphite furnace.

PAGE 64

Parabolic Mirror Aspheric Lens < > Nd:YAG Laser Laser Controller Lens Monochromator PMT-------~ Stepper Moto~.......,.. Oscilloscope 0 0 0 0 Computer Figure 2-15 Schematic of setup used by Cassini et al. for air analysis using LIBS

PAGE 65

55 Pat 442.1 nm with a DL of 200 pptl\ Na at 371.1 run with a DL of 110 ppm, Hg at 404. 7 run with aDL of50 ppm, Be at 381.4 nm with a DL of 130 ppm, and As at 454.4 nm with a DL of 130 ppm. In 1994, Flower et al. used LIBS to monitor metal aerosol emission generated by a pneumatic nebulizer [136]. A chromium salt solution was passed through the nebulizer and a heating pipe to produce a gas with known concentration of Cr. The detection limit was 200 ng Cr / standard cubic meter of air ( scm). The laser used was a Nd: Y AG ( 1064 run, 180 mJ) and detection was at 3 12 run. In 1994, I azzari et al. detected mercury in air using LIBS with a Nd:glass laser ( 400 mJ, 8 ns) to produce plasma in a chamber filled with saturated Hg vapor [ 13 7]. With successive dilution of the mercury vapor, they obtained a detection limit of 10 ppb by observation of the 253.7 line. Zhang et al. performed LIBS in a particle loaded methane / air flame and in an oil fueled combustor vent [138]. In their 1995 publication, they described the addition of a flange with four movable windows to a pre-existing combustor vent. LIBS was performed through these windows which were kept clean by a flow ofN 2 gas and could be switched out periodically for cleaning. In 1995, Nordstrom studied the laser-induced plasma emission spectra ofN 2 0 2 and ambient air from 350 to 950 nm [139]. This range was covered with approximately 0.5 nm resolution and atomic, ionic, and molecular emission from a plasma created by a CO 2 laser (10.6 m, 320 mJ, 200 ns) was observed. Spectra were computed for N, 0, and a few molecular species, and a comparison between experiment and calculation was done

PAGE 66

56 The same year, Parigger et al published a fundamental paper on laser-induced plasma in the gas phase (140]. The breakdown was produced by a Nd:YAG (1064 run, 220 mJ 6 ns) in hydrogen at 150 and 810 Torr. The H a line at 656 nm was used to calculate electron number densities in the decaying plasma with temporal resolution as high as 6 ns early in the plasma and 1 sat later times. Spatial resolution was also achieved by focusing different portions of the 1 :I plasma image onto the spectrometer entrance slit. Spatial resolution was approximately 50 m. Surface plots for electron number densities resolved in time and space were then created Electron number densities varied between I 0 16 and 10 19 cm 3 In 1996 Raisch et al used LIBS to detect chlorine in the gas phase [ 141]. Two experimental setups were described. The first was a benchtop experiment using a 320 mJ Nd:YAG laser. The second setup used a miniature Nd:YAG laser built into a sensor head. This miniature laser provided 18 mJ of pulse energy. Detection of Cl in chlorinated hydrocarbons was accomplished by focusing the laser onto a copper target in the presence of the gas to be analyzed. This provided for fonnation of CuCl in the plasma. The luminescence of the D-system of CuCl was then detected around 440 nm. Detection limits were a few ppm. Figure 2-16 shows the elements determined in a gas phase matrix and their relative detection limits. The t1nit scm refers to a standard cubic meter of gas.

PAGE 67

H He Li Be B C N 0 F Ne Na Mg Al Si K Ca Sc Ti V Cr Mn Fe Co Cu Zn Ga Ge Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag In Sn Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Fr Ra Ac Key: D < 1 0 g/scm 1 0 1 00 g/scm > 100 g/scm p S Cl Ar Br Kr Te I Xe Bi Po At Rn Figure 2-16 E l ements determined in gas phase matrices s had ed according to detection limit T he unit sc m refers to a standa rd c ubi c meter of gas.

PAGE 68

CHAPTER3 FUNDAMENTAL INVESTIGATIONS OF LASER PRODUCED PLASMAS Part I: Spatial and Temporal Dependence of Lead Emission in Laser-Induced Breakdown Spectroscopy Introduction Studies investigating the spatial and temporal development of the LIBS plasma will provided an insight into the complexity of these pulsed sources. Through the use oftime resolved spectral imaging a better fundamental understanding of the plasmas behavior is attainable. The results of these studies have the potential to considerably improve the analytical usefulness of LIBS results. Kuzuya e t al. [22] reported on the change in spatial distribution of the LIB plasma spectral radiation at 400 nm caused by gaseous atmospheres of Ar air, and He at different pressures. Their results indicated that breakdown in the gaseous atmosphere above the target occurred at excessively high laser power densities leading to decreased ablation and emission from the target material because a major portion of the laser radiation was absorbed by the opaque gaseous plasrna in front of the target material. Sattmann et al. [73] studied the effect of double laser pulses on the spatial structure of the plasma integrated over time and wavelength using photographic detection Temporal and spatial resolution of the LIB plasma structure was reported by Palau et al.[142] for three different laser wavelengths, without wavelength resolution. Multari and Cremers[58] reported on the spatial and temporal 58

PAGE 69

59 distribution of the Cr 425 run line emitted by the LIB plasma when a flat target was tilted at various angles with respect to the incident laser beam. They concluded that after the plasma formed on the solid target, it expanded upward along the path of the incident laser pulse but once the chromium line emission became dominant over the continuum background the emission appeared symmetric about the normal to the surface rather than along the axis of the incident laser pulse. Vadillo et al. [143] showed non-dispersed time resolved LIB images captured by means of a CCD and monochromator in mirror mode ( zero o rder ). They incorrectly inferred the width of the plasma from the digital images failing to recognize that the measured image actually represented an image of the entrance slit with non-uniforrn irradiance caused by the plasma image projected upon the entrance slit. Bulatov et al [56] reported 2D spectral images from the LIB plasma created by means of Fourier transform visible spectrometry The wavelength resolution was in 4 nm spectral windows over the range 400-800 run which was insufficient for spectral classification of single emission lines. The results also lacked time resolution. A s part of a research project to develop a portable LIBS instrument fo r the determination of toxic metals in the environment and in particulate samples we have investigated the temporal and spatial development of lead ionic and atomic emission from a LIB pla~roa. Experimental A schematic of the LIBS system which was used is depicted in figure 3-1. It consists of a laser motorized X Y Z translational stage carrying the sample spectrometer ICCD detector, detector gating and control electronics and a computer for control and data acquisition. The Nd:YAG was a Quante! YG580 which delivered 1 J at the fundamental

PAGE 70

,,....,. ___ ....j Nd-YAG Laser S pe .,,,c ~ tr -:-,;:-, o ;"""".""':"" m ~ e :-:--te r __,,, 1 CC 0 g ~~Cl = .,. 45 t Controller Pulser =~_-,L l l 7, 7[ l 7, If L 7, If X Y,Z Stage Stage Control Computer Figure 3 1 Experimental apparatus for capturing time resolved LIBS spectra. 7 7 7 0

PAGE 71

61 wavelength of l 064 nm and produced a 9 ns duration pulse at a maximum of 30 Hz repetition frequency. A 80 / 20 beam spliter was used to attenuate the laser output so that only 20% of the laser power was used in the generation of the plasma. The laser beam was focused on the target using a 2.5 cm diameter, 10 cm focal length lens. The resulting plasma created a 200 m diameter crater with an average incident laser irradiance of 18 GW / cm 2 The initial design directed the laser beam perpendicular to the target material. In a later modification, the laser beam was incident at 45 with respect to the target surface to allow for the desired movement of the fiber optical system used for radiation collection. Pressed sample pellets of 1.4 cm diameter and varying thickness were made with a pellet press at 4500 psi pressure. In pressing the pellets pure lead flakes (Fisher Scientific, Catalog# L-24) and NIST soil standards (SRM 2711) were used as sample material. A positioning system consisting of a helium-neon laser, lenses and a position sensing detector was used to monitor the position of the srunple surf.ace with respect to the laser beam focus and the light collecting system's focal point; control was accomplished with a motorized X, Y, Z stage ( Orie4 Model# 18011) For all measurements, the target was moved horizontally between shots to allow sampling from a fresh location and to improve the reproducibility of mass ablation. A 2.5 cm diameter fused silica lens with a focal length of 5 cm was used as the collection optic to produce a one-to-one image of the plasma as viewed from the side on the entrance slit of a 0.5 m focal length spectrometer (Acton Research, Model# SP-500) with a grating of 1200 grooves / mm. The reciprocal linear dispersion was 1.6 nm / mm at 250 nm and an entrance slit width of 25 m was used. The intensified charge coupled device (ICCD), ( Princeton Instruments, Model# ICCD-576S) being used, had pixel dimensions of 22.5 x 22.5 m and a viewing area of 12.9 mm by 8.6 mm. It was oriented in such a way that the

PAGE 72

62 longer dimension corresponded with the wavelength dispersion axis and the shorter dimension with the entrance slit height The ICCD was operated by a controller (Princeton Instruments Model# ST-130). A programmable pulse delay generator (Princeton Instruments, Model# PG-200) was used to gate the ICCD. The PG-200 was computer programmable, and was capable of gating the ICCD with integration times as short as 20 ns. The Q-switch trigger signal from the laser was connected to the "Ext trig in" input of the programmable pulser (PG-200) and served as the master control of the delay gate. This trigger signal provided 250 ns pretriggering with respect to the onset of the laser pulse and enabled the system to open the ICCD gate before the plasma actually started emitting. The PG-200 high voltage gate pulse was connected to the ICCD while its "Aux dly'd trig out" signal was used to trigger the ICCD Controller ST-130 "Ext sync" input. The PG-200 was programmed to activate the ST-130 by means of this signal to perform the data read-out from the ICCD and to transfer data to the control computer 1 s after the completion of the gate pulse. The ST130 "not scan" output was connected to the PG-200 "Inlnbit in" connector to prevent the PG200 from being pretriggered before the most resent data capture, transfer and real time processing was completed. The system was thus operated in a synchronous mode and the detector delay and integration time settings and selection of active pixels were completely controlled by the software. The entire experimental apparatus was under the control of a Gateway 2000 Pentium 133 :Mliz computer running the Winspec Vl.4.3 6 (Princeton Instruments) software. A miniature spectrometer ( Ocean Optics, Model# S 1000) consisting of collection fiber, optical components, modular grating disperser, and linear CCD detector with control electronics was used in the angular dependence measurements.

PAGE 73

63 Results and Discussion Spectra Since existing elemental spectral intensity tables [ 144-14 7] are based mainly on sources such as electric arcs sparks and gas discharges, one often finds that they are not directly applicable to the observed laser-induced plasma intensities. We therefore captured and analyzed the spectrum of our element of interest, lead Spectra were accumulated from 10 single laser shots on a solid lead target. Spectra were captured after a delay of2 s using an integration time of 15 Each spectrum covered a wavelength window of 20 nm and successive sets of measurements were taken with the center of the window shifted by 15 nm increments to cover a total spectral range from 200-450 nm. The resulting spectra are not shown due to space limitations. However a classification of the observed lead spectral lines are given in table 3-1 The relative intensities presented in the table are not corrected for the detector's varying spectral efficiency due to a lack of a reliable radiance standard for wavelengths below 250 nm. Plasma lmaiini Time resolved images of the laser produced plasma were captured by using the modified experimental setup shown in figure 3-2. The collecting lens was placed at its focal length ( 250 01rn) from the laser-induced p)asn:ia to provide a nearly parallel beam of radiation incident on the interference filter. The focusing lens (f=250 mm) focused this transmitted parallel beam of radiation in its focal plane where the ICCD was placed to capture the spectrally filtered plasma image. The beam combiner (a neutral density filter of 0.3 optical density ) allowed radiation from the plasma as viewed from the front via one of the plane

PAGE 74

64 Table 3-1. Table of lead lines with their relative intensities as observed in the LIB plasma and their associated states. U pper State Lowe r State Wavelength Intensity ( nm ) ( Relative ) Term Energy Te rm E nergy ( cm ') ( cm ') 205.328 15236 8s 3 P I 48686 87 6p2 ) Po 0 .0 0 217 000 83017 6d 3 D I 46068 57 6p2 3 P 0 0 00 220 353 4* 159839 7 s2 S 1 12 59448 00 6p1P 312 14081 00 223 7425 1 6023 7d 3 D 1 52499.53 6p2 3 P 1 7819 37 22 4 689 39766 7d 3 D 2 523 11 37 6p2 3 P 1 7819 35 239 3792 529 41 7d 3 F 1 52412 30 6p2 l p 2 10650 47 2 440 1940 3 1515 7s 1 P 49439 .73 6p2 3 P 1 7819 35 241 1734 145 52 7d 3 F 2 52101 77 6p2 l p 2 10650 47 2 44 3829 22780 8 s3 P 0 48726 16 6p2 1 p 7819 35 2 44 6181 35 871 8s 3 P 1 48686 87 6p2 3 P 1 7819 35 247 6378 59174 7s 3 P 1 48 188 67 6p2 3 P 1 7819 .3 5 257.7 260 63 035 7s 1 P 1 49439 57 9p21p 2 10650 47 261 4175 1516 3 6 6d 3 D 2 46060 90 6p2 3 P 1 7819 .3 5 262. 8262 1 3 882 8 s P 1 48686 87 6 p2 1 p 2 1 0650 47 266 3 157 115415 7s 3 P 2 48188 67 6p2 3 P 2 1 0650. 47 269 7541 9837 l4 s' P 1 58517 67 6p2 1 0 2 21457 90 280 1995 161677 6d 3 F 3 46328 81 6p2 3 p 2 10650 47 282 3189 90279 6d 3 02 46060 90 6p2 3 P 2 10650 47 283 .3 053 140343 7s 3 P I 35287 24 6p2 3 P o 0 00 287 33 11 109403 6d 3 F 2 45443 .2 6 6p21P 2 10650 47 357.2 729 62390 7s 1 P 1 49439 57 6p2 1 0 2 21457 90 363. 9568 64527 7 s3 P 1 35287.24 6p2 3 P 1 7819 35 36 7 1491 13181 8 s 3 P 1 48686 87 6p2 1 D 2 21457 90 36 8 3 46 2 72813 7s3P 0 3 4959 90 6p2 3 P 1 7819 35 373.9935 55761 7s 3 P 2 48188.67 6p1D 2 1457 90 401 9632 10236 6d 3 F J 46328 81 6p2 1 0 2 21457 90 405 7807 116066 7s3P 1 35287.24 6p2 3 P 2 1 0 650 47 406 2136 23673 6d 3 0 I 46068 57 6p2 1 0 2 21457 90

PAGE 75

Plane mirror for side-on view Beam combiner I I I I I I I I I ......... I I 1 I I I I I I I I I I I I I I I I I I I I / I I I I I I I I I I f s Plasma I I I I I I I I I I I I I I I I I I I I Plane mirror for frontal view Collecting lens Spectral filter Focusing lens ICCD Figure 3-2. Schematic of the n1odified optica l system for capturing time resolved 2D spectral images of the LIB plasma simultaneously from two orthogonal directions

PAGE 76

66 mirrors and a side view via the other plane mirror to reach the detector simultaneously. By slight rotation of the plane mirrors around vertical axes, the two images f and s ( representing the frontal and side-on images respectively) could be laterally offset from one another. The enlargement of the image was given by the ratio of the focal length of the focusing lens to that of the collecting lens This setup allowed simultaneous registering of two perpendic,1lar views of the plasma during a programmable time interval. A solid lead pellet was used as the sample. Interference filters were used sequentially to produce narrow wavelength band images of the plasma. With a proper choice of peak wavelength and spectral bandwidth, the interference filter isolated and transmitted a single transition, providing a means of measuring the temporal and spatial dependence of the wavelength-associated population distribution. An interference fiher with a spectral FWHM of 8 nm was used to observe the ionic line emission from the transition Pb II at 220.3534 nm. The filter was tilted at 17 with respect to the optical axis to have the filter's transmission peak correspond with the ionic line wavelength. This filter could not completely isolate the ionic line since its full-width at half-maximum perrnitted a minimal detection of the Pb I 217.000 nm atomic line transition. An interference filter with a peak transmission at 280 run and a FWHM of 8 nm was used to detect the lead atomic line This filter permitted radiation from the lead atomic transitions at the wavelengths 280.1995, 282.3189, 283.3053 and 287.3311 nm to be observed simultaneously, with the major contribution coming from the 280 nm wavelength. It is important to note that during the initial 500 ns of plasma existence, the observed radiation was mainly due to broadband continuum radiation at the filtered wavelengths and not due to the elemental line emission. This was verified by analysis of a sample which did not contain lead ( or any spectral emitters

PAGE 77

67 at 220 or 280 nm). To obtain a wavelength integrated image, a neutral density filter ( optical density 1) was used to attenuate the radiative flux sufficiently in order to prevent saturation of the ICCD The top two images in figure 3-3 show a mm scale as viewed by the ICCD from the front (left band picture) and from the side (right band picture), when the ruler was placed at the position where the laser-induced plasma would have forrned, using identical imaging. The physical orientation for the left and right sides of the frontal view and the front and rear of the side-on view is also shown. The ruler images were obtained from a double exposure blocking each of the orthogonal paths successively and rotating the ruler through 90 between the exposures. The clear plastic ruler was back illuminated with a mercury pen lamp The lower five frames in figure 3-3 show the wavelength integrated temporal development of the LIB pla<;ma using a neutral density filter with optical density 1 for radiative flux attenuation. In each frame the plasma image on the left is the front view, and the side view is on the right The laser beam is entering from the right on the side view, striking the sample at about a 45 angle. During these delay times, only continuum radiation was emitted. These images were generated by subtracting two successive image files captured with a pulse width of200 ns, separated by a 10 ns delay. Even though the PG-200 could gate the ICCD with pulses as short as 20 ns, it was not practical for full area ICCD imaging, due to the "iris effect"[ 148] which caused the central pixels in the ICCD to have almost zero sensitivity at such short pulse durations. For figures 3-4 thru 3-6, a pulse width of 200 ns was used. Each image at a specific delay time is from a different laser produced plasma. The data shown here are from sequentially captured laser shots. No attempt was made to use the "best matching" images. A no11nallzed intensity scale was used for each figure, 3-4 thru 3-6. This allows comparison

PAGE 78

Left Right Front Rear 10 ns 20ns 40ns 60ns 80 ns Figure 3-3. Wavelength integrated lead plasma images. Top two images show a back illun1inated mm ruler for scale. The lower five images s how the temporal development of the wavelength integrated plasma with increasing delay time

PAGE 79

: r 170 ns 370 ns 570 ns 770 ns 970 ns 1170 ns 1370 ns 1570 ns 1770 ns Figure 3-4. Wavelength integrated lead plasma images on a longer time scale than shown in figure 3-3.

PAGE 80

3.77 s 4.27 s 4.77 s 5 77 s 6 77 s 7.77 s Figure 3-5. Wavelength integrated lead pla s ma images on a longer time scale than s hown in figure 3-4 -.J 0

PAGE 81

"' 1 ii' l ,, . ' 1.J ' 170 ns 370 ns 570 ns .... -. .. . .. .. ~, } ~ .. .. ~.. . . ,. . 770 ns 970 ns 1170 ns 1370 ns 1570 ns 1770 ns Figure 3-6. Images showing temporal and spatial development of the ionic emission from the LIB plasma transmitted by a 220 nm filter. --..J

PAGE 82

72 of images within one figure, but not comparison between figures. Figure 3-3 shows that the plasma starts from the spot where the laser strikes the target and then continues to grow along a preferential axis coinciding with the incident laser beruIL Figure 3-4 and figure 3-5 shows the further development of the wavelength integrated plasma on a longer time scale and with a different neutral density filter to show the temporal development of the combined line emission Lines have been drawn on the pictures to indicate the surface of the lead target. Since the top and side of the pellet are highly reflective, anything below this line should be ignored, since it represents reflected radiation. The reflection from the side of the pellet is due to back reflection from the neutral density filter. These images show that at ~ 1 s after the initiation of the plasma, it bas reached a maximum size and brightness, and bas begun to decay. It also shows that the development of the LIB plasma on a homogeneous solid sample is very reproducible. Finally non-normal incidence of the laser beam onto the target material leads to non-symmetrical ( cylindrical) pla.~ma for1nation. The protrusion seen in the side-on view and starting to develop at delay times of > 500 ns occurs long after the termination of the laser pulse, but along its' incident path. Figure 3-6 shows the 220 nm ionic line emission. The first two plasma images at delay times of 170, and 370 ns show primarily the continuum background emission. Note that the lead ions are confined to a smaller volume, and that they do not extend out into the plascna tail seen in the images in figure 3-4, thus resulting in a ''cut off" shape at the rear of the side-on view. The ionic emission is also shorter lived. This becomes more obvious when comparing the ionic emission in figure 3-6 to the atomic emission shown in figure 3-7. In figure 37 and figure 3-8 the atomic emis.~ion from the 280 run atomic line and also the contribution of emission from a few other minor lines is shown . Initially, during the

PAGE 83

' ' 170 ns 370 ns 570 ns 770 ns 970 ns 1170 ns 1370 ns 1570 ns 1770 ns Figure 37. Images showing temporal and s patial development of the atomic emission from th e LIB plasma transmitted by a 280 run filter

PAGE 84

3.77 s 4.27 s 4.77 s 5.77 s 6 77 s 7 77 s Figure 3-8 Continuation of the 280 nm filtered lead plasma images on a longer time scale than s hown in figure 37

PAGE 85

7 5 continuum emission, the plasma images, have a "cut off' appearance similar to the corresponding images for the lead ionic images. Then, as the plasma decays, i t develops a more cylindrically symmetric shape. These images around 280 run do not show the protrusion in the side-on images for the wavelength integrated images of figure 3-4. This would mean that the emitters responsible for the protrusion in figure 3-4 must be lead atomic or ionic excitation states other than the ones associated with the 280 or 220 nm lines. Because of a lack of appropriate narrow band wavelength interference filters, we could not identify the wavelength( s) responsible for the protrusion. From these measurements it is obvious that the elemental spectral line emission in the LIB plasma is a complicated three dimensional spatial and temporal function, which is different for ionic and atomic lines and most likely different even for different wavelength lines from the same element and ionization stage. This indicated that the assumption of local thermodynamic equilibrium might not be valid [149]. An~ular Dependence Since the plasma images observed in the previous section are not spherically symmetrical, the spectral brightness will be an anisotropic function. The line of sight along which the spectral brightness is highest will also be the one providing the optimum sensitivity in a single parameter optimization. Experiments were carried out to determine if there was a dependence of the signal intensities on the polar angle of radiation collection when the whole volume of the LIB plasma was observed. Since it was necessary in these measurements to align the horizontal axis of rotation of the radiation collecting system with the pla~ma the experirnental apparatus shown in figure 3-9 was used. It was necessary to keep the lens-to plasma distance constant while the collection optic was rotated along a horizontal axis,

PAGE 86

Optical fiber with focusing lens on polar rotational stage ., 'I. / r -' Laser beam incident at 45 1-Target on X, Y,Z stage Figure 3-9 Schematic diagram of the modified radiation collection system used for the angular dependence measurements.

PAGE 87

77 otherwise the focus would be changing. For these measurements the Ocean Optics spectrometer was used. Since the spectrometer we used didn't have the capability of gating, no temporal resolution was available. Therefore, the time integrated emitted radiation from the plasma (including background) was used for the measurement. Six successive laser produced plac;rnas from a solid lead target were averaged to produce the spectra shown in figure 3-10( a). The plot shows the l O background-corrected spectra collected at 10 intervals as the collection optic was rotated from 90 to ff with respect to the horizon. Figure 3-10( c) shows the wavelength integrated net signals over the observed line profile for each lead line at the corresponding viewing angle. As can be seen in Figure 3-lO(c), the signal intensities are roughly constant from 90 to 40, with a slight increase at 30 before they slope off. The decrease in signal was due to blocking of the solid angle of collection by the target. We also observed a greater enhancement at about 40-30 when this experiment was repeated with a soil sample. This more or less constant behavior with a peak at about 30 with respect to the horizon is consistent with a cylindrical or mushroom shaped structure. Conclusions We found that the spatial and temporal behavior of the transient emission generated by a laser-induced breakdown plasma is rather complicated. The different spatial and temporal distributions for the ionic and atomic lines is to be expected during the existence of the p1asrna considering the dynamic shift in ionization equilibrium as the plasma initially heats up and then cools down. What is quite surprising though is that different atomic lead lines also seem to have vastly different but uniquely reproducible time and space dependencies. Measured time resolved spectra, to establish the most appropriate time interval producing an optimum signal to background ratio ( or what ever parameter is optimized), will be dependent

PAGE 88

(a) 1400 12 00 1000 ,...._ !1 800 600 400 200 0 ~ 80 ._ __ 60 .__ __ ~40 ~(>J ____ __ 20 Jo -o o ,:r276 278 280 282 284 286 288 Wavelength (nm) 290 Figure 3-10. Angular dependence plots. 8 .._, 1 en (a) Background corrected lead spectra plot ; (b) Background plot. (b) 4 Viewing Angle 15000 10000 5000 0 280 run 283 nm i ~l:;, 287 nm ~ 6 282 run .!f ....., -..l 00

PAGE 89

(c) 4 Viewing Ang l e Figure 3 10 conti nu ed 20000 15000 V> ..... 0 0 ,._, 1 1 0000 ... en 1s tis 61) 5000 4.) ... .El 0 280 nm 2 83 nm ~t:.t? 'q} 287 nm 282 nm -~ 4...,1 4 View i ng Ang l e ( c) The wavelength integrated net signal p l ot; ( d) Signal to background ratio plot. ( d ) ' ~ t ~ 1 .6 'O 1 2 0 ... 00 0 ..!Id a 0.8 a 00 0 4 ... en 0 0 280 nm 283 nm 'l.J~~ 'q} 287 nm 'l.J 282 nm .:.._;~

PAGE 90

80 on the specific volume of plasma being observed and the line of sight of the detector. These additional variables affecting the LIB emission should be considered together with the well documented independent variables such as the laser pulse duration and shape in time, [ 17] the laser wavelength [ 18] and energy, [21] the physical and chemical characteristics of the target material, [24] and the composition and pressure of the surrounding environment [26]. Multi-variable optimization techniques for increased sensitivity and decreased limits of detection will consequently also be complex procedures, especially considering that the physical dimension of the plasma is of the order of millimeters in size and microseconds life time with steep spatial and temporal gradients in spectral brightness. With homogeneous material in the solid phase as the target and constant laser and imaging parameters, the LIB plasma is a very reproducible emission source which is a promising characteristic for analytical applications. We conclude that time resolved, spectral imaging has the potential to considerably improve LIBS results and lead to a better fundamental understanding of the LIB plasma.

PAGE 91

Part II: Level Populations In A Laser-Induced Plasma On A Lead Target Introduction 81 In laser-induced breakdown spectroscopy the mass ablation rate and characteristic elemental emission intensity vary non-linearly on a pulse-to-pulse basis (apart from the latter's temporal and spatial variation during a single shot) because of variations in the laser energy, the temporal pulse shape and the spatial profile of the laser beam on the irradiated material. The chemical composition and homogeneity in the targeted volume, the physical properties (phase, pressure, temperature) and the mechanical properties (smoothness of the surface, crystalline orientation) of the material and the atmospheric conditions (gas composition, pressure) play a role, suggesting the likelihood for strong matrix effects if not properly controlled. Objective interlaboratory comparisons are further obscured by the use of different (and often tioreported) laser focusing conditions (affecting the energy and power density ), the angle of laser beam incidence, the volume o r subvolume of plasma actually being observed, the a ogle of light collection, and the presence of previously ablated particulates in the breakdown volume creating a memory effect for subseq uent laser shots. An understanding of the characteristics of the processes involved will enable better control and lead to improved analytical performance. Temporal and spatial resolved observations of laser-induced plasmas have been reported [45 150]. In an often cited publication [115], Radziemski et al. concluded that local thermodynamic equilibrium (L TE) was established in the LIB plasma at delay times greater than I s. Their conclusion was based on the agreement between calculations from a theoretical one-dimensional hydrodynamic model and the intensity ratios ofC II / CI; NII / NI; Be II / Be I applying the Saha equation and a Boltzmann plot for Be I lines in an air plasma at 580 torr. They found

PAGE 92

82 time-resolved te,.....,.h,...,,s ranging from 22000 to 8000 K during the plasma decay. Spatial resolution using Abel integral inversion was applied only to the Be II / Be I line pair. Spatially resolved temperatures changed less than 5% compared to the line-of-sight integrated values, leading Radziemski et al to conclude that the uncertainties introduced by unfolding a small source of only approximate cylindrical symmetry would not improve the temperature or electron density values. Lee et al [25] measured spatially resolved intensities from lead lines at 357.27, 363.96, 368.35, 373.99 and 405.78 nm and applied the Bohmiann plot technique to deter111ine temperatures for the plasma induced in air at atmospheric conditions by an ArF excimer laser at 193 nm. Their results lacked temporal resolution of the transient plasma; they reported temperatures in the region 13000 K varying along the plasma height. Simeonsson et al [151] measured time resolved te ~UfA s in gaseous LIB plasmas in CO, CO 2 methanol and chloroforrn applying the Boltzmann equation to two oxygen and two chlorine lines and the Saha equation to two carbon lines. They found spatially integrated temporal temperature values ranging from 15000 to 20000 K. They assumed L TE based on theoretical considerations following the Griem [152] criteria and the experimental observations ofRadziemski [115]. The 521.82 and 510.55 nm Cu I lines were measured by Mao and co-workers [21] and applied to a two line Boltzr,,ann technique to determine the axial variation of the terr 1f~T-3ttll'in a LIB p)aSina on copper targets. No temporal resolution was measured. They found excitation temperatures ranging from 6000 to 11000 K. Our impression is that in spite of a vast n11mber of reports citing temperature and electron number density measurements in LIB plasmas, there is still a need for a complete spatial and temporal investigation of the population distnoutions of several different atomic / ionic species. As part of a research project to develop a portable LIBS instrument for

PAGE 93

83 the determination of toxic metals in the environment and in particulate s ample s, we investigated the temporal and spatial development of lead atomic emission from a LIB plasma in air and inferred from it, the temporal and spatial distnbution of the populations of the upper levels from which the transitions originate. Theory Temperature Determination Usin~ Spectral Line Ratios The following assumptions are required: The source volume under observation is in local thermodynamic equilibrium (L TE), has a homogeneous distribution of emitters, and remains in a steady state during the time interval of interrogation. The spectral lines measured are optically thin. The population distribution amongst the internal energy levels is described by the Boltzmann equation From the radiances R (spectral energy per unit of time per 11nit of projected source area per unit of observational spatial angle and integrated over the line profile) of two s pectral lines ( indexed i and} ) from different upper excitation energy levels E, of the same element and ionization stage the temperature is given by [5] E E T = j I In k R g A A J I I j R g A A I '} J I (3 -1) with the degeneracy of the upper levels denoted by g, the emission transition probability by A, and the wavelength by A. The two-line method has the advantage compared to a single line radiance application that one eliminates the need of knowing the value for the effective

PAGE 94

84 pathlength through the source the total particle number density the partition function, the need for absolute radiance calibration of the detection system and absolute transition probabilities. Practical considerations to keep in mind when selecting a line pair are to have the wavelengths nearly identical and the difference in upper energies of the two transitions as large as possible. The first criterion is beneficial in the sense that relative radiance calibration is easier and more accurate and the second assures that the calculated temperature is more reproducible and not over-sensitive to small fluctuations in the radiance ratio measurement. Temperature Determination Usin~ Boltzrnann Plots The two-line ratio technique can be extended to a larger number of lines from the same element and ionization stage. This is similar to applying a least squares linear regression to a larger number of points to determine the slope and intercept instead of deter1nining the linear relationship from just two points. Rearranging the spectral line radiance equation for multiple lines, indexed i from the same element and ionization stage gives the expression (153] I g A l l T + 1n hcln, 41tZ ( 3 2) where l is the effective pathlength through the radiation source, Di the particle number density Z the partition function and the other symbols have been defined above. If we now treat the as the dependent variable and as the independent variable for a set of k lines from the same species and otherwise identical conditions, a linear relationship with a L h c ln slope of and an mtercept of In results. The Boltzmann plot also needs T 4n Z calibration of the radiances at each of the analytical wavelengths, just as with the two-line

PAGE 95

8 5 ratio technique. Relative signal values proportional to the radiances and relative tran s ition probabilities for the measured lines instead of absolute values will not affect the calculated temperature and will only result in a change of the intercept As with the two-line ratio technique the same practical considerations apply including the selection of closely matched wavelengths for accurate relative radiance calibration and large differences in upper energies from which the transitions originate to provide the best precision in the calculated temperature. Experimental Lead Spectnims Ener'' Level Dia~am and Line Selection The experimental apparatus used for all measurements, was the same as that used in part I of this chapter. In each wavelength window, a spectrum was accumulated from 10 single laser shots on a solid lead target. The target was moved horizontally between shots to allow sampling from a fresh location and to improve the reproducibility of mass ablation. Spectra were captured using an integration time of 15 s after a delay of 2 s with respect to the plasma initiation to avoid the initial intense continuum background. Each spectrum covered a wavelength window of ~ 20 nm and successive sets of measurements were taken with the center of the window shifted by 15 nm increments to cover a total spectral range from 180-705 nm. From the energy level scheme for lead [ 154] we calculated the wavelengths in air for all possible electric dipole transitions and associated the observed lines with their respective upper and lower energy levels. We then used the following criteria in selecting a set of lines for ther1nometric purposes: The wavelengths of the lines should all fit within the 20 nm wavelength window of our detection system so that they could be observed simultaneously and radiance calibration could be considered identical for all the lines.

PAGE 96

86 The lines should be reasonably intense to provide line to background ratios far in excess of one in order to minimize background noise degrading the precision of the measured line signals. The transition probability values of the lines should be available [145]. The lower energy level on which the transition ends should preferably exclude the ground state or other low lying energy levels to minimize self absorption. The difference in upper energy levels from which the emission lines originate should be as large as possible to provide better precision in the temperature measurement From these criteria, we selected the lead atomic lines at the wavelengths 357.275, 363 958, 367.151 368.348 and 373.994 nm as the diagnostic lines. Table 3-2 contains the wavelengths, transition probabilities, upper and lower state's statistical weight, configuration and energy level values, from the publication by Wiese and Martin [145]. Figure 3-11 shows the partial Grotrian diagram for neutral lead with the transitions used in this investigation. Data Capture and Graphical Presentation of the Raw Data In each exposure, we simultaneously captured radiational data for the fixed spectral window around 365 run containing the selected set of five lead atomic lines emitted from the LIB plasma and their intensities along the height of the plasma. Each exposure measured the signal (as a discrete function of wavelength and height) in a time slice with a fixed integration time of200 ns and a delay time (with respect to the plasma initiation) that was varied in 500 ns steps for successive files. Each file was the cumulative result from 10 single laser shots on a solid lead target. The target was moved horizontally between shots. The data files generated by the Wmspec Vl .4.3.6 (Princeton Instruments) software were analyzed by means of in-house written Matlab 4.0 (The Mathworks Inc.) source programs. Each graph in figure l

PAGE 97

60 50 40 30 20 10 0 ns np __ l i ~ iJ _________ _ __ ___ _____ -----Pb 11 ( 2 ~J ( 59821 0 cni 1 ) 3 7s 3p1 (35287 24 crrr 1 ) = ::::;;:: _ ., == 7 S P O ( 34959 90 crrf 1 ) ,;. .. .. .. .. .. ., ,, .... .. . . ,,, .. , ~~. :: . .. ' .. ~ ;;, .. ,, I>;~ .... ~ .,. .. ' ,, 363 958 nm} 368 348 nm 357 2 7 5 nm 367 151 nm 373 994 nm ., \, (21457 90 cm 1 ) (7819 35 c rrr 1 ) Figure 3-11. Partial Grotrian diagram for neutral lead atom s and the tran s ition s used for the therrnometric diagnostics 00 -.l

PAGE 98

88 Table 3-2. Wavelengths transition probabilities, and level specifications o f the lead line s u s ed a s thertnometric spec ies Lead I Transi t io n s W ave l e n gt h Trans iti on U pper S tate Lower State (run ) P roba b i lit y (x I 0 8 s 1 ) Statis ti cal Co nfigur a t io n E n e r gy Le v e l Co nfigur a t io n Ene rgy Level Weigh t ( cm1) ( cm -1 ) 357 27 48 0 .99 3 7s 1 P 1 4 9 4 3 9 57 6p 2 1 D 2 2 14 57 90 363 9577 0 3 4 3 7s 3 P 1 3 52 87 2 4 6p2 3 P 1 78 1 9 35 36 7 1 513 0. 44 3 8s 3 P I 48 6 86 87 6p2 1 D 2 2 14 57 90 36 8 .3 475 1 50 I 7 s 3 P 0 3 4 95 9 9 0 6 p2 3 P 1 78 1 9 3 5 373 99 44 0 73 5 7s 3 P 2 48188 67 6 p 2 1 D 2 21 4 57.90 3-12 shows the variation in each spectrwn along the plasma height for sequential delay time increments of 500 ns. The first spectrum was captured at a delay time o f 320 ns and consisted mainly of an intense continuum with superimposed extremely broadened, barely discernable lines A s time progressed the continuum background decreased rapidly while the s pectral lines increased in intensity and reached a peak intensity at around 2 s dela y time after which they continued with time to show a slower decay in intensity and with decreased line widths. One can also s ee in the first frame that the pla~ma emission extended only about 0 5 mm above the target surface but then gradually expanded upwards to over 1.5 mm height. F rom these figures it can be seen that the temporal and spatial development of the Pb I 367 run line differed significantlyfrom the other four observed lead lines Temporal Development of the 360 run Series of Pb I Line Intensities A s impler gr aphical overview of the temporal development of the spectrum was achieved by acc1unulating in software the signals over the plasma height for each wavelength as s ociated col11mn o f pixels in each file and then plotting a time sequence of their temporal development as s hown in figure 3-13. The first s pectrum measured at a dela y time of 3 2 0 ns

PAGE 99

6 ,-.. 1 5 Cl) 4 8 3 2 00 Cl) 1 0 2 Height(m m ) 6 5 4 3 2 1 0 2 6 6 4 3 2 1 0 2 320 ns ~---~--~36:6:---""'3;;70;-------"375 Wavelength ( nm ) 820ns ~---~------,;::;:;;--i37;0-----"'375 355 1320 ns ~ -~ --\;:;;----~3; 70;----"37 5 355 Figure 3-12. Time sequence of height dependent spectra for the observed lead LIBS plasma 89

PAGE 100

0 2 6 1820 ns 05 "'t'"---'==---~;---~;;;-----,375 Heigh t (m m ) o :a, :E5 370 6 5 4 3 2 1 0 2 6 5 3 2 1 0 2 I 5 I Figure 3-12 continued W av elength ( nm ) 2320 ns o~---~=---~;---~;;370:;;----375 2820 ns 05 0 ~---~-:---"';;------"l-;;37;0----...375 90

PAGE 101

6 l s ;!1 3 4 0 3 aJ C 2 0() 0 2 Height (mm) 6 5 2 1 0 2 6 5 4 3 2 1 0 2 I 0 5 0 Figure 3-12 continued. I 3320 ns ~----r:-----~3;5;;----~37;0----175 :Ji) l55 W av elength ( run ) 3820 ns ~----r---~;--~;;37;0--"-'"'375 a 4320 ns ~----r---~;--~:;;3~;0--"--1375 l55 9 1

PAGE 102

Pbl 357 27 nm 355 360 Pbl 363 96 nm Pbl 368 35 nm Pbl 367.15 nm 365 Wavelength (nm) 370 Pbl 373 99 nm 375 F i g ur e 3 -1 3. Tem p o r a l deve l opme nt of t h e h eig h t int eg r a t e d l ea d s p ectra.

PAGE 103

93 was omitted from this graph, since the high continuum background would ob s truct the view of the successive spectra. Therefore the front-most spectrum in figure 3-13 represented the plasma height integrated spectrum captured at a delay time of 820 ns and each successive one behind it was captured 500 ns later. From figure 3-13 one can see the line emi s sion integrated over the plasma height increased initially with time, reached a peak at about 2 s and then gradually decayed over the next 14 s. Note the difference in the behavior of the 367 run Pb line which peaked at approximately 4 compared to the other observed lead lines. It is important to keep in mind that the graphs at the different delay times are indeed captured sequentially in time from different laser shots altogether and variations in plasma conditions from s hot to shot may have obscured the true temporal development of a s ingle LIB plasma. Figure 3-14 s hows the temporal development of the line profiles from figure 3-13 in a smaller wavelength interval around each line and in a smaller time range To avoid cluttering in figure 3-14 only every second graph from figure 3-13 is shown. Interesting features to note are that the lines initially suffer severe Stark broadening and peak s hift towards longer wavelengths ( caused by the high electron number density in the L IB plasma ). Also note the asymmetric profiles at the shortest delay time. The Stark effect gradually decreased as the plasma decayed. The difference in the behavior of the 367 nm Pb line which peaked at approximately 4 compared to the other observed lead lines peaking at about 2 sis also obvious. Table 3-3 contains the halfwidths (FWHM) and peak shifts of the Pb I lines from figure 3 -13 at different delay times. Since the measured halfwidths exceeded the instrument halfwidth by a factor of five or more the contribution caused by instrumental broadening was negligible

PAGE 104

' J!3 C :, 8 co C C) en 1 4 -,----------------------------, 1 2 1 0 0 8 0 6 0 4 0 2 Pb I 357 nm : ,l )J ., 1 .. 1 I ,., :, : I i I I ; I : If i I I 1 I 1 1 I I . It I I ' . I ,, . ~ i : I I I . I \ I ... ---0 82 1 82 2 82 3 82 s s s s . ... . ~:-:~....._ _______ J ' .. . ... 411, .. . . ----:.._ .. -.;..- .. ---0 0 ~~-~:!::;""':: 1 ::.:::.::.. __,,----...---..----.__ .:-:.::.: :;.::-:.:::.:~:.:::~::.~:.:::~.!J'UallDll-~-lill-lJil-"""-w,-"'-illi--1,,li;.a.W..a 357 0 357 5 358 0 358 5 359 0 Wavelength (nm) 1 6 Pb I 364 nm 1 4 1 2 1 0 0 8 0 6 0 4 0 2 .. :1.~ ,'lA .:i,,.j~ 1,,l, 363 5 I I # I : I : I ) I 1 I :. : !f I .. ... : ), I '.'I' \ \ : I t I I : I I I I : I I I \ : I I I I I I .. ' \ \ ' ... . ~~';':~::-:-:------_J . . .. . . ,1T.--., ,. 364 0 364 5 365 0 Figure 3-14. Initial temporal development of the Stark broadened and wavelength shifted lead line profiles.

PAGE 105

2 0 1 8 1 6 1 4 co G> fl) ... 1 2 C ::::, 0 0 1 0 ffl C 0 8 0) Cl) 0 6 0 4 0 2 0 0 Pb I 367 & 368 nm lines '\. I .. -"' .. . ,. 367 0 . ,., ..... ,., ... -.:.,-368 0 I ,, I I I I t f'' I ,:, I I ; I I t ,I ) : I l . I .-, :, .. I I. I I I I I I I I ; .. I I I I I ' I I . ' I I . ' . --0 82 1 82 2 82 3 82 . .. . .. . -... . . ........ ~,A ... _._.. 369 0 Wavelength (nm) Figure 3-14 continued. s s s s 370 0 1 4 -------------------------------, 1 2 1 0 0 8 0 6 0 4 0 2 Pb I 374 nm line ~ .. ., \ I ' : I : ,' ., j \\ I I : ' : ' ' ' !, 'i : ' ' .. L' \ ;, t,.,. ..... -.! I ' I J I ' I .. .. . I .. .. --4 I 11 --.. --... .. --------.. . .... ., . . --'-~ 1,l '-'I 0 0 ..Jil:~~~=:;.:.:.:.:.::.:~,---~---r--r----r----,--~---.-~=p 373 6 373 8 374 0 374 2 37 4 4 37 4 6

PAGE 106

96 Table 3-3. Full width half maximum values for the Stark broadened lead I line profiles and their peak wavelength shifts as a function of delay time. The profiles are representative of the height integrated values. Lead I Spectral Lines Spectral Line 357 run 364 run 367 run 368nm 37 4 run Dela y Time Width Shift Width Shift Width Shift Width Shift Width Shift ( s) ( nm ) ( nm) { run ) ( run ) ( run ) ( nm) (nm) ( nm ) ( run ) (nm) 0.82 0.740 0 269 0 402 0.129 0.402 0.155 0 367 0 125 1.32 0.43 8 0.135 0 268 0.095 0.502 0.134 0.301 0.08 8 0.267 0.09 1 1 82 0 .3 01 0.067 0. 201 0 062 0. 402 0.067 0.234 0.055 0.200 0 .0 58 2.32 0 236 0.03 4 0.235 0.02 8 0 301 0.033 0.234 0.055 0.234 0.02 5 2.82 0.202 0.034 0.235 0.028 0.23 4 0.033 0.234 0.0 55 0.200 0 .0 25 3.32 0.236 0.034 0.20 I 0.028 0 201 0.033 0.234 0.021 0.200 0.02 5 3 82 0 202 0.000 0.235 0 .02 8 0.23 4 0.00 0 0.234 0.021 0.200 0 025 4 32 0.202 0 000 0.20 l 0.02 8 0. 20 l 0.000 0.23 4 0.0 21 0.167 0.0 25 4 82 0 868 0.000 0.20 1 -0 .00 5 0.201 0.000 0.234 -0.012 0 167 -0 009 BackfUound Correction In order to compare the relative intensities of the lines, the underlying background upon which they are superimposed had to be subtracted. Since this background contribution was also a time spatial and wavelength dependent function, we estimated it from values at wavelengths remote from the five lead lines of interest, but within the 20 run spectral window that was captured in each measurement. For each delay time and each plasma height, a second order polynomial function of wavelength was fitted to the signal values at these selected background monitoring wavelength pixels. From the polynomial fit, the estimated background at all wavelength pixels in the 20 run window was calculated and subtracted from the gross measured spectrum to yield the net signal at each wavelength. A typical example of the gross spectrum, the signal at the wavelength pixels selected for background calculation,

PAGE 107

97 the estimated background and the net spectrum at an arbitrary height and delay time coordinate are shown in figure 3-15 Strippini the 367 and 368 run Line Spectral Overlap As was already mentioned and can been seen from figures 3-12 thru 3-14 there is spectral overlap of the wings from the Stark broadened Pb I lines at 367 and 368 nm during the initial development of the LIB plasma. A complicating factor in stripping such spectral overlap (apart from the large number of spectra involved for all heights and delay times) is the shift in peak wavelength of the two lines and the change in profiles with both delay time and with height of the plasma We avoided theoretical modeling of the line profiles because the observed profiles are the convolution of profiles along the inhomogeneous line of sight and fitting parameters for a single Voigt profile will be inappropriate. We tried empirical stripping of the two lines, by mirroring the long wavelength wing of the 368 run line around its peak, reasoning that this wing will be the least affected by spectral interference from the 367 nm line of lesser intensity. Subtracting the 368 nm mirrored spectrum from the net signal provided a difference spectrum representing the stripped 367 nm line, as well as the deviation from symmetry in the 368 run line profile closer to its peak wavelength. To calculate the signal associated with the intensity 11nder the spectral line profile for the 367 nm line, the signal values of the difference spectrum were accumulated over a wavelength interval which included its peak and was bordered by the first zero crossings to shorter and longer wavelengths. The signal under the line profile for the 368 run line was calculated by accumulating the signal of the net spectrum in the interval around the 368 nm peak extending from the upper wavelength limit of the 367 nm line to the longer wavelength first z.ero crossing, and adding the 368 nm-mirrored spectrum underneath the 367

PAGE 108

en I C: :::J 8 C'O C: C) (/') x10 4 3.5~---------------------------1 -Gross spectrum 3.0 Estimated background Net spectrum ~. . .. .... .. . I 2.5 2.0 1.5 1.0 0 5 0.0 o Background pixels ' . . -. . .. J I . . . . . . . -. I I . . . . . . ' . . . ' . ' . . -,. -0.5-+----------------.~---------'--------------~ 355 360 365 Wavelength (nm) 370 375 Figure 3-15. A typical example of the gross spectrum the wavelength pixels selected for estimating the background by fitting a second order polynomial function the calculated background estimate and the net spectrum at an arbitrary height pixel and delay time

PAGE 109

99 run line to it. A typical example of the net spect~ the mirrored 368 nm line profile and the difference spectrum is shown in figure 3-16 Hei~t Dependence of the 360 nm Series of Pb I Line Intensities The wavelength integrated signal under the line profile for each of the five Pb I lines was calculated at each pixel height and for each delay time (see figure 3-17) An interesting feature, worth pointing out is that at the first delay time of 0.82 s, there was hardly any 367 nm emission; the 367 nm emission appeared later in time. At all delay times the 3 7 4 nm line reached a maxim11m value higher in the plasma compared to the height where the other lines maximized and decayed more gradually with height. Boltzmann Plot The points for a typical Boltzmann plot at an arbitrary height and delay time coordinate from our data is shown in figure 3-18. The excitation energy E is used as the independent variable and the parameter 1n R,).., as the dependent variable. For a g A physically acceptable temperature from the Boltzrnann plot the dependent variable should always be inversely proportional to the excitation energy. In an L TE model, only small deviations from this inverse proportionality, due to random statistical fluctuations in the measurement for closely spaced energy levels are acceptable. The steeper the slope of the line, the lower the temperature and vice versa Just presenting the calculated Boltzr11ann temperature as a function of height and delay time, as is the customary practice, does not indicate how the actual points change with time and height. In order to retain this information we plot the dependent variable 1n R,1.. for the five spectral lines, but since K 1 A 1 their independent variable values (ie. their excitation energies) are fixed we shifted the points horizontally to the same horizontal coordinate value on the graph, which we take to be

PAGE 110

en I t C: :::, 8 ca C: C) Cl) x10 4 3.5-,--------------------------3 0 2.5 2.0 1 5 1.0 0.5 ' -Net profiles .. .... ... ..... .. ............................. L ..... ... ....... .. .. Difference spectrum o Mirrored 368 nm line ' ' ' .. .. .. . .. .. .. .. .. .. .. . . .. ... .. ' ' I I I I -, ..... r r "' ' ' .. .. .. ... .. .... ... ... ... .. .. ..... .. .. ... .. .... ... .. ........ .. ' .......... .. ....... ............ ........ .... ........... ...... l .. ... .. ......... ............................... L .... ................... .. ' ' ' .. .. .. .. ' ' . ' .. . .. .. .. .. .. . .. .. .. .. .. . .. . .. .. ""-' V'Jo... -0.5-+__ _,__ ___ __ _,__ _____ ....._ ____ __._ _____ ~------1 366 367 368 369 370 371 Wavelength (nm) Figure 3-16. A typical example of the net s pectrum s howing the spectral overlap of the 367 and 368 run Pb I lines the mirrored 368 run line profile resulting from mirroring the long wavelength side of the 368 run profile and the difference spectrum resulting from s ubtracting the mirrored spectrum from the net s pectrum. 0 0

PAGE 111

7,--------------,------------------. 6 5 I() 4 Q) .2? C ::, 8 3 a, C 0) i:i5 2 1 0 82 s 0 0 I ... .. I .. ' I I '\ I \ \ ' t I 1\ I ,. l I I .... .. : I I 0 5 l 1 0 Height (mm} --368nm ----364nrn 37 4 nm 367 nm 357 nm 1 5 2 0 7 -,-----------------------------1 82 s 6 5 4 3 2 1 o~ 0 0 I I I I I I , I I ' ' ,. I \ I .., \, ,, . . . . . . \ ... ' .. ., ' \ ' . . ' .. .~ . . . -' ' . . .. . ......... --... ' . . . . :. --' -.:, 0 5 1 0 1 5 2 0 Figure 3-17. Temporal and height variation of the wavelength integrated signals for the ob s erved Pb I line s 0

PAGE 112

LO Q) C :::, 0 0 co C C) Cl) 7 6 5 4 3 2 1 0 0 0 2 82 s , I ... .. ... .. ... .---. .... ,--, ' .,. . ,, . "' .. --368nm ---364nm 37 4 nm '36 7 rvn 357 nm .. ... .. . -.. -' __ ... . --, .. . . . .. . . .--.. ..--s..: --- -.-:_, 0 5 1 0 1 5 He ig ht (mm) Figure 3-17 continued. 7 3 82 s 6 5 4 3 2 1 ,-.. , ... ... .,..4, c .. "\, . . . . . . . . '> . . . . '"'' .. .. .... ... -. -, ..._ ... --.. 0 2 0 0 0 0 5 1 0 1 5 2 0 0 N

PAGE 113

16.8 -,----,,----r---r-----,------.,.------------------0 368 nm 16.4 364 nm 16.0 ~' C) 0:: t= 15.6 15.2 374 n ......... + 367 nm 357 nm 14 8 -+-----------------------------t 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 Excitation Energy (cm1 e4) Figure 3 18 A typi c al Boltzmann plot at an arbitrary height pixel and delay time ....... 0 w

PAGE 114

104 representative of the height and delay time coordinate Different symbols are used to identify the five individual points The further apart the five points are vertically the lower the Boltzmann temperature and vice versa. The vertical ordering of the associated symbols, which is inversely proportional to the excitation energy, should also be consistent apart from small random fluctuations between the points associated with closely s paced excitation energies. If this vertical ordering would be systematically different from the expected in v erse proportional relationship with excitation energy it would indicate that s omething is s ystematically wrong. The temporal and height dependence of this 1n R ).., for the five g A s pectral lines is shown in figure 3-19. The graph for the 820 ns delay time indicates that the 3 67 nm line initially is too low in intensity while the values for the 357 nm line are too high at almost all delay times and in the lower half of the plasma. The intensity of the 374 nm line is too high in the upper part of the plasn,a. The almost negligible emission of the 367 run line in the initial phase of the LIB plasma can only be explained in ter11s of a very low population o f its excitation energy level in s pite of it being bracketed in between the excitation energy levels associated with the emitting 3 5 7 and 3 7 4 run lines At all delay times the height dependence of the 357 and 374 nm line intensities are also systematically incompatible with the latter always extending higher up in the plasma and the former decaying faster with increasing plas11ia height. Since these two lines have a common lower energy level at 21457 9 cm 1 and their signals do not differ by more than a factor of 2 absorption can be ruled out as a possible cause for the observation. The only plausible explanation for their dis s imilar behavior is that their populating rates were selectively different indicating non-equilibrium conditions

PAGE 115

, 1 5 1 4 1 3 1 2 l l 1 0. 9 0. 8 0. 7 0 1 5 1 4 1 3 l .2 1 1 1 0 9 0. 8 0 7 0 1 5 1 4 1 3 1 2 1 1 1 0 9 0. 8 0. 7 0 ---~ ----------"1 --820 ns .... .. ... + ---~---.... .......... ,._ . . .. . .... . . . .... . -------~ -----~----~-----~---- 1 -, ' :-::t ... r r ... : : .. .. ; 0.2 0 4 0 6 ----~ -' _. _____ ...J __ 0 8 l 1 2 1 4 Height (mm) 1 6 1 8 2 1320 ns ' j t'"1 .... .. .. 0 I I I I V I I I I "' e,..... .. .... ... 1 .. ... T o, ' _J _____ t r -- ~ -' I I I I I I I I --r-----,-----~--1 --r---r--~----~-0 I -----r' 0 2 0 4 0 6 0 8 1 1 2 1 4 1 6 1 8 2 1820 ns .. I -1=-:-+ + ' I . --' --r--~-----,-1 --0 2 0 4 0 6 0 8 I I ' ~ -~---,-----1-.. r... -.' r r : ' ----. -----1I 1 1.2 1 4 1 6 1 8 ' 2 0 X + 3 6 8 364 374 367 3 57 105 Figure 3-19. Modified Boltz111ann graphs showing the variation of ln(IU/gA ) for the five observed lead lines with height at different delay times.

PAGE 116

l .5 1 4 1 3 1 2 1 1 l 0.9 0.8 0.7 l .5 1.4 1 3 1 2 1 1 1 0.9 0. 8 0.7 1 5 1 4 1 3 1.2 1.1 1 0.9 0. 8 0.7 2320 ns o I -.. -~----.... ~ ~-.., ~ ~ .. . ---... --' + ' I -..,... ., -' .J J I I ..., .. ...... .. I I I -: -. .; ' a. ____ .., ___ .., ' : ~----L-.. ~t---, .. ' I I I 1+ + 0 I o 0 ---?-' w. : +' -1-_.., __ ,_ ...i ... ... :.:II J---t ... .l ... + ' -~ --r-+ ,. .. .. ..... ..,.. .. o I I I ' .. r -,-.I ,-----y0 0.2 0.4 0.6 0 8 1 1 2 1 4 1 6 1 8 2 Height (mm) 2820 ns ,-,' ,I ' -" ----l ---, -~------,-----, ~ .. -~. ... -1 ..... ., .. .. . .. ...... L ............ .......... J ..... I + I 1-----'1'' I I r' I I I I It I t I I ----~----~-----,----,-----r---r----,-----~-----,-----,-' I 0 0.2 0 4 0 6 0 8 1 1 2 1 4 1 .6 1 8 2 0 ---,' --,fir ----.:+ I I + 3320 ns L __ ,. _____ .. __ I I I ., ............. ~ ---r ---r----,----~-1 ........... J ...... ___ ,1 _ _ -..... '-.. j_ ' -----~----~-----~----'I'---------~-----~----~-' ' .. ----,. -' I I ----1-I 1 --I ---. .. -..,l ..... ;-___ i ____ ,-,-, ---.-I I I I I 0 2 0 4 0.6 0.8 1 1 2 1 4 1 6 1 8 2 Figure 3-19 continued. 0 X + 368 364 374 367 357 106

PAGE 117

I 5 1 4 l 3 1 2 I l 1 0. 9 0. 8 0. 7 1 5 1 4 1 3 1 2 1 1 l 0 .9 0 8 0 7 1 5 1 4 1 3 1 2 1 1 l 0 9 0 8 0 7 .. -. 3820 ns r I .... .. ........ ' I -. .. r -~ J ---~' ' ' ' ---r---,---,--' i ............ L ...... ... ..... _ --------.. -----' ; ' -' -1----1 ---, r ~ : .,.,._ -+ I J---' --+ +! 1+ --1.., .... I I I ----------' 0 0 2 0 4 0 6 0 8 1 1 2 1 4 1 6 1 8 2 Height ( mm ) .. -.. r... -.4320 ns :..I,. :.. L ..... _,_ : q,' ' --.,--I o .. .....__ + -+ .. J I I + + I I ' I I I ----.. -' .. .... L.. I -~ .. -~ I --, ----~ ... -, --.' ' ' -----r-~ --~ ----,I I -r----i, I ... -'., .. J .............. .. ......... I J --~-----,-----,-0 0 2 0. 4 0. 6 0 8 I 1 2 1 4 1 6 1 8 2 0 4820 ns ____ ,_ ... .............. .. I I --r --,-' 1 0 2 \ . . . ' --I .. .. I I ~ --, J. ---1 I I ------------' 0 4 0 6 I I I I .. J ........... -1 ..... .. .. .. L ........ .. .... ... .. .. .......... ~. -"' -' I -t ----,. ,... ..... .... -----. ---.. I I I l 0 8 1 I I I I J ~--' I I I I 1 2 1 4 1 6 I I I I -i--,-1 8 2 F i gure 3 19 co ntinu e d 0 X + 3 68 364 374 36 7 3 5 7 1 0 7

PAGE 118

108 Hei2ht and Temporal Development of the Boltzmann Temperature The calculated value for the Boltzmann temperature as a function of height and delay time is given in figure 3-20. The general impression is that it does not change significantly with either plasma height or delay time, but its' general validity should be strongly questioned when comparing it with the non-equilibrium results in figure 3-19. Conclusions From the log intensity values in figure 3-19, we conclude that the populations of the upper levels from which the five Pb I lines originate, are mostly not in a state of thennal equilibrium with respect to one another even at delay times as long as 14 s in certain regions of the plasma. The observed effect cannot be explained in terms of absorption, poor background correction and/or detector cahbration. Although the measurements were not truly spatially resolved along the spectrometer's line of sight so as to interrogate a small enough localized homogeneous vol11me of the plasma, doing so will not remove the inconsistencies regarding the non-emission of the 367 run line initially and the larger than expected emission of the 374 run line higher up in the plasrna when the other lines have almost vanished. Theoretical models [115] have suggested that the relaxation time needed for a transient plasma under such conditions of pressure and energy density to reach thermal equilibrium is of the order of 1 s, while our observations suggest that some parts of the plasma have not reached thermal equilibrium even at delay times as long as 14 s. A possible explanation for the discrepancy is that the LIB plasma acts like an explosion in which the highly energetic particles are mostly moving radially directed outward in space away from the target material. Therefore, a longer relaxation time will be needed for sufficient collisional exchange to occur before the equilibration of energy has taken place compared to a similar source with random motion of the energetic particles

PAGE 119

15000 g 10000 ::J co 'Q) a. 5000 I0 2 1 5 Height (mm) 1 0 5 0 0 2000 4000 6000 8000 10000 12000 Delay Time (ns) Figure 3-20 Graphical display of th e temporal and height dependence of the calculated value representing the Boltzmann plot temperature ..... 0

PAGE 120

CHAPTER4 LIBS BENCHTOP INSTRUMENT: DEVELOPMENT AND EVALUATION Introduction The reason for developing a benchtop LIBS instrument which was outlined in chapter 1 was to move in the direction of satisfying our goal of placing a LIBS instrument on-line in a mining facility for process monitoring In every LIBS publication various forms of experimental instrumentation have been used. At the center of each instrument was some type laser system based on solid-state, gas, or dye principles Each laser system has several advantages and disadvantages including wavelength, beam quality, reproducibility ease of operation, size and cost. The most common solid-state laser was the Nd:Y AG (A= 1064 532,355 and 266 run). For the gas lasers the N 2 ("= 337nm), the CO 2 ( A = 10.6 ), and the excimer (ArF KrF and XeCl with A = 193, 248 and 308 nm, respectively) were the most common. In the majority of the instruments reported in the literature conventional optics were used to focus the laser radiation onto the sample and to collect the emitted pla5ma light. In the remaining fraction, optical fibers were used to perfor1r1 all or some of the light collection and transportation duties. The final constituent of the literature instrumentation was the dispersion and detection of the emitted radiation. The spectrometer of choice has been the Czerny-Turner configuration with focal lengths varying from 0.33 1 m. Several different detection devices have been used including: photomultiplier tubes, photodiode arrays, intensified photodiode arrays, charge coupled devices, and intensified charge couple devices. By adopting the most useful aspects of these various possibilities, we have 110

PAGE 121

111 constructed a complete stand alone user-friendly instrument for analytical measurements Some of our major design considerations were cost size, weight flexibility and ruggedness The last few considerations were necessary to ensure the development of a field transportable unit. Our plan was to design the instrument on paper construct the prototype, evaluate the prototype, make any structural or parameter changes, and finally field evaluate the unit. At the time of this writing we are in the third stage of this plan. This chapter will discuss the development, evaluations (both fundamental and analytical) and possible applications Development Instrument Overview A schematic of the completed LIBS system is depicted in figure 4-1 It consisted of a laser motorized X Y Z translational stage, spectrometer ICCD detector detector gating and control electronics and a computer for control and data acquisition. The Nd: YAG laser (Model# YQL-102 + Laser Photonics) delivered up to 220 mT at the fundamental wavelength of 1064 nm and produced a 15 ns duration pulse at a maximum of 20 Hz repetition frequency The 6 mm laser beam was focused to a 0.9 mm diameter spot on the target using a 2 5 cm diameter 30 cm focal length lens. For the majority of the research in this chapter a pulse energy of 180 mJ was used, and with the above optical arrangement, these conditions provided a typical incident laser irradiance of 2 GW / cm2. A positioning system consisting of a diode laser and photodiode detector was used to monitor the placement of the sample The positioning system was located in a horizontal plane perpendicular to the incident laser beam. It was optimized so that the half-value of the maximum intensity readout, from the diode electronics, indicated the correct positioning of the sample. Using this method, a sample could be placed at the same vertical position with

PAGE 122

PC I . . . . t I . I I I I I .. .... .. ... ICCD Pulser DL ,, rJ::: .. r 4" ,( , ------/ ,, cco 1 .... r<:i ."' ,.. 0 ,.. .. "-..I: ~ .._ ...J, :. :. :: : &&P O>-.. ... .. I ~ ; I D L Spectrometer Controller .... Nd: YAG Laser ~ ) r . I I . Spectrometer . .. ' ' '. .. .. . ' .. .. OO 000 0 0000 .......... ,o U o0 000 ... 00.... 00-HOO .. 00o0o00 0 ,0 .. .. 0 00 0000o0 00" Laser Beam I I I I I I I i I I ..... ~ ....... ..... i j t I I I I ! \ l t \ l fJ ............. ., ... ..... i \ \I : ... ... i \ 1{ (:.::: .. : : Spectrometer .. I ... .......... ...... ........ ............... .... ...... ... .... ......................................... I\ \ Figure 4-1 Schematic of the LIBS benchtop instrument [\ \ ICCD Controller r--..... Laser Power Supply I a a 00 I 0 0 C 0 O @ == DL = Diode Laser ....... N

PAGE 123

113 a reproducibility limited by the s tep size of the stepping motor ( 5 ) Movement of the s ample was accomplished with a computer controlled motorized X Y Z stage ( Model Unislide, Velmex). A Qbasic program was written to move the sample during measurement For all measurements the target was moved horizontally between measurements. Emitted plasma light was reflected from a 5 cm planar pierced mirror and was collected by a 2.5 cm diameter fused silica lens with a focal length of 10 cm. A pair of2.5 cm planar mirrors were used to lower the plane of light and reflect it into the spectrometer. The lens was positioned in such way as to fill the first collimator optic inside the spectrometer instead offonning an image on the entrance slit. Homogeneous illumination of the entrance slit ensured a homogeneous intensity on the detector along the height of the spectral images of the entrance slit. Another reason to fill the collimator and therefore the grating, was to obtain the highest possible resolution (narrower instrument line profiles). The 0.5 m focal length spectrometer (Model# SP-500 Acton Research) was equipped with three automatically interchangeable gratings with 1200, 2400, and 3600 grooves / mm. The adjustable entrance slit was set to a fixed width of 20 m for all measurements unless noted. An intensified charge coupled device (ICCD), (Model# ICCD-576S Princeton Instruments) was used as the detector with its controller (Model# ST-130 Princeton Instruments). A programmable pulse delay generator ( Model# PG-200 Princeton Instruments) was used to gate the ICCD The firing of the laser and data collection was under the control of the computer. This was made possible by connecting the ''NotScan'' output of the ICCD controller (ST-130) to the flash lamp input of the laser. The Q switch output signal from the laser was connected to the "Ext trig in" input of the programmable pulser (PG-200). This trigger signal provided ....,0. 8 s pretriggering with respect to the onset of the laser pulse and enabled the system to open the

PAGE 124

114 ICCD gate before plasma emission actually started. The PG-200 high voltage gate pulse was connected to the ICCD, while its "Aux delayed trig out" signal was used to trigger the ST130 "Ext sync" input. The PG-200 was programmed to activate the ST-130 by means of this signal to perfor1n the data read-out from the ICCD and to transfer data to the control computer 1 s after the completion of the gate pulse. Since the ST-130 "Not scan" output was used to fire the laser and initiate data capture, the sequence was not repeated until the most recent data capture, transfer, and real time processing was completed. In our case this was limited to 500 ms, or a sampling repetition rate of2 Hz, by the control/data collection computer. The ~ system was thus operated in a synchronous mode where the firing of the laser, the detector delay, integration time settings and selection of active pixels were completely controlled by the software. The entire experimental apparatus was under the control of a CyberMax Pentium 200 :MHz computer nmning the Wmspec Vl.4 3.6 (Princeton Instruments) software. Multiple laser shot spectra could be stored in one of two ways by the Winspec software, either as "spectra" or as "accumulations". When multiple LIBS spectra were stored as "spectra", each individual spectrum, from a single shot, was stored within a file. When "accumulations" was selected, the multiple LIBS spectra were summed, and stored as one spectrum. In this mode, all of the shot-to-shot data was lost. For the majority of the data presented in this chapter the data was stored as ''acct1m1tlations'' by the Winspec software. The captured data files were analyzed for background correction and peak height dete1111ioation by MatLab programs which were written in-house. (These are the same type ofMatLab files described in detail in chapter 3.)

PAGE 125

115 Pierced Mirror As mentioned above, a pierced mirror was used to image the plasma emission into the spectrometer. In the original design, a side-on collection method was chosen instead of the end-on collection achieved by the pierced mirror approach. To understand the difference between the two collection methods, see figure 4-2. There were several reasons why this cha nge was necessary. One was that the sample roughness can change the lens to sample distance, and therefore influence the plasma characteristics and also affect the collection of the emitted light due to a changing plasma focus. With side-on collection, the light path may be blocked by the surface structure or distorted by lateral movement of the plasma on the surface with respect to the entrance slit of the spectrometer. Structural blockage becomes a major concern when perfor111ing depth profiling measurements. As the laser drills a channel in the sarnple part of the emitted light will be blocked and no longer detected. Another reason for using the end-on collection was that eventually we were planning on incorporating a remote probe into the instrument. We had designed the remote probe to incorporate an end on optical design for convenient monitoring of material on a conveyor belt. To explore the advantages / disadvantages of both designs (sideand end-on) we evaluated an aluminum ~ample (SRM 603 NIST) with both optical collection methods. We designed the experiment to investigate the variation in emission signal when the sample height was changed over a limited range. The lens to sample distance measurement of sample displacement from collection focal point was varied by moving the sample stage up and down in the vertical direction. The results of these studies are graphically presented in figure 4-3 With the side on collection, it was observed that as the lens to sample distance changed over a 10 mm range, the Al signal drastically varied however with the end-on collection there was a

PAGE 126

Plasma Pierced Mirror Plasma (a) Laser ~ Lens I I Lens (b) Laser Lens Lens ;/ 1 r~-------I Spectrometer I Spectrometer ==Figure 4-2. Schematics of the two types of light collection. (a) Side-on collection; (b) End-on collection. 116

PAGE 127

(a) (b) 6 6 ~5 ~5 Q) Q) (/) en c4 c4 ::J ::J 8 8 ~3 ~3 Ii I (/) (/) C C ~2 ~ 2 C C -,-. Cl) I_!_ Cl) C 1 C 1 -.I .....J _J ... <{ 0 <{ 0 ..... 80 100 280 285 290 295 300 305 Lens to Sample Distance (mm) Lens to Sample Distance (mm) Figure 4-3 Results of using the two types of light collection (a) Result s using side-on collection ~ (b) Re sults using end-on collection ......J

PAGE 128

118 negligible change in the Al signal The end-on 180 optical detection approach was therefore adopted Fundamental Evaluations Mass Removal By quantifying the mass removed per laser shot, a value was provided that can be used in dete11r1ining the absolute limits of detection. Another reason for investigating the mass removal was to dete1r11ine how the removal rate varied with laser pulse energy the main variable that controls the amount of material ablated. For these measurements pure copper discs 5 mm in diameter and 0 004" thick were used ( Alfa AESAR Catalog# 11394, 99.9985% Cu) The sample preparation entailed cleaning of the copper foil surface with soap, water, and then a final rinse in acetone. Discs were then cut from the foil using a standard hole ptmch. Each disc was transferred to a clean plastic vial after weighing ( Perkin Elmer, Model# AD-2Z). The stationary sample was then subjected to 1200 laser shots and re-weighed. The mass removal was determined by the difference between the mass before and after the 1200 laser shots Triplicate measurements were made at 8 different laser pulse energies. The pulse energy was varied by adjusting the power to the laser flash lamps using the front panel control. The pulse energy was monitored using a volume absorbing energy meter (Scientec~ Catalog# 38-0101 and 362). The results of the mass removal study are plotted in figure 4-4( a). The y-axis and x-axis represent the average mass removed per laser shot and the laser pulse energy respectively. The plot shows that over a pulse energy range of 89 156 mJ the mass removal had a linear response. The slope of the line indicated that in this region approximately 0 5 ng was removed per mJ of pulse energy As the energy was increased further the mass removed no longer continued to increase. This trend has been

PAGE 129

g> 40 +J 0 .c. Cl) '30 8. u Q) > 0 20 E Q) er: en 10 (a) I Slope= 0 4674 ng/mJ (b) 16 -14 en +J 12 8 10 J C O> 8 Cl) C 0 6 en en E 4 w .. :::J 2 0 O-+-----.-----r-----.---.-...---.---------------------------~ 0 50 100 150 200 80 100 120 140 160 180 200 220 240 Pulse Energy (mJ) Pulse Energy (mJ) Figure 4-4 Results indicating how the laser pulse energy effects the mass removed and the emission signal (a) Mass removal per laser shot results ; (b) Copper emission signal results 2 50

PAGE 130

120 observed by others and they attnbuted it to a change in plasma conditions where the increased energy is absorbed by the growing plasma, and therefore not coupled into the sample for mass removal [155]. Simultaneously while the sample was being subjected to the 1200 laser shots, the spectrum ( t d = 2 t g = 15 s) of each plasma was recorded. The average ( from the 3 measurements) emission signal from the copper 427.51 run atomic line is plotted in figure 44(a). The emission signal shows a trend similar to that of the mass removal plot. A plot of the signal to background ratio ( SBR) vs. the pulse energy showed that the highest three measurements ( 222 200, and 178 mJ) had approximately the same SBR. A laser pulse energy of 180 mJ was therefore chosen as the optimum. It is important to emphasize that this optimum was determined for a pure metallic copper sample and it is unlikely to be optimum for other matrices We also investigated redeposition during the mass removal studies. We repeated the above set of measurements, incorporating a perpendicular flow of argon. A 6 mm diameter tube was placed l cm from the laser spot/sample interface. An argon flow of 30 L / min was used to sweep away any material that might be redeposited. The results from this trial were the same as those without the argon flow. This does not e1iminate the possibility of redeposition, but it does indicate that redeposition did not significantly distort the mass removal measurements. Plasma Temperature As already discussed in chapter 3 the plasn1a excitation temperature can be estimated by preparing a Boltzmann plot, and is a useful diagnostic of the energy in the plasma. The plasma temperature as a function of laser pulse energy and detector time delay was determined. The data was the same as that used in the mass removal studies. For the

PAGE 131

121 temperature as a function of detector delay, the same sample was used as well as a fixed laser pulse energy of 180 mJ, a detector gate of 1 s, and 25 laser shots were accumulated for each measurement. Three copper ion lines were used in the temperature determinations. Spectroscopic data for these lines can be found in table 4-1 [156-157]. Using the two data sets, BoltZinann plots were prepared and excitation temperatures were calculated from the slope of the best fit line. The resulting t"'a ... ...,. ..... L-14 es are plotted in figure 4-5 with (a) and (b) distinguishing the laser pulse energy and the time delay plots, respectively. The trend in the pulse energy plot is very similar to that observed in the mass removal plot, where above 186 mJ, the plasma temperature and the mass removed remained approximately constant. The trend shown in the time delay plot is what would be predicted for a decaying plac;:ma source. Optimized Gate Settin~s For the analytical evaluation of the benchtop instrument, it was necessary to deterr11ine if one set of detector parameters could be used for several different samples/matrices. Chromium (425.41 run) was detected in three NIST SRM standards, one each of steel, aluminum, and zinc base, as well as lead ( 405. 78 nm) in NIST soil, paint, and spinach leaves. For all analysis, 10 laser shots were accumulated for each measurement, and the measurement was repeated in triplicate for each detector setting. For the detector delay studies, a fixed gate of 15 s was shifted at 1 s intervals over a range of 1 10 s. Once the optimum delay was determined, the samples were re-analyzed to determine the optim11m gate width. This was accomplished by fixing the detector delay at the optimum, and then lengthening the gate in additional 10 s intervals over the range from IO 180 s. All of these results are not shown here to save space; however, figure 4-6 shows the results of the analysis of the steel SRM where (a) and (b) represent the delay time and gate width results, respectively. The top

PAGE 132

82008000 :::, ... Q) a. E Q) t7800. I (a) I I 7600-+---~---,---.---,---.---,---.---4 50 100 150 200 250 Pulse Energy (mJ) (b) 1600n-r-------------.---.----------------1400 ..-. 1200 1000 :::, ... 8000 E Cl) t6000 4000 2000-+----------------------------1 0 5 10 15 20 Time Delay (s) Figure 4-5 Plasma excitation temperature as a function of laser pulse energy and detection time delay (a) Laser pulse energy results ; (b) Detection time delay results

PAGE 133

(a) 3 1 ?-T---,----.-------.-------..------------1 I lO (l) 2 "C C "' ::J 0 I C L. 8 ::J 0) 8 I .:. 0 -2 a, CXl a, C -... 6 0) I a, I C (f) 0) L. (.) 1. (f) I 1 rH---,----.------,---,----r----w---...----..---~ 2-t----------------------"""" 0 2 4 6 8 10 0 2 4 6 8 10 Delay Time (s) Delay Time (s) Figure 4-6 Determination of optimum detection settings for chromium in steel deter11unations (a) Left : Chromium signal as a function of dela y time, Right : Signal to background ratio as a function of dela y time

PAGE 134

(b) 3 2 3.0 16 III I Q) 2 8 I "C en C ...., 2 6 5 14 I I 8 "C) II -2 4 0
PAGE 135

125 Table 4-1 Wavelengths ionization stage, level specifications, and transition probabilities of the copper lines used as thermometric s pecies Wavelength ( nm ) 212 2 979 2 1 2 6 045 214 8948 l o niz.ation S tage n II II Lower Level ( cm I ) 2 6 2 64 568 22 84 7 131 2 19 2 8 754 U pper Level ( cm I ) 7 3353 292 69867 983 68447 741 Stati s tical Weight 5 5 7 T ran s ition Probability ( xl 0 8 s 1 ) 2.52 1 64 0 64 plot in both ( a ) and ( b ) shows the development of the Cr signal and the lower plot in both shows the development of the signal to background ratio ( SBR) The background signal was determined by rnonitorir1g the signal of a pixel on the I CCD that was assigned to background ( no signal due to line emission ) In figure 4-6 (a) top, it is evident that the Cr s ignal was initially the most intense and then decays over the remainder of the observation period. A similar trend was observed for the other saJlples for both elements. In figure 4-6 ( a ) bottom, the SBR was initially the lowest increased up to a delay of 5 s before leveling off f o r the remainder of the observation period. Each sample exhibited a similar trend however the point at which a maxim11m was reached and the SBR began to level off varied slightly over the range from 3-6 s. 1be non-conductive samples reached a of the range (3-4 while the conductive samples reached a maximum over the latter part of the range (5-6 ) An optimized delay for all samples was therefore chosen as the middle of the range at 4.5 s. Figure 4-6 (b ) shows the gate width results obtained at the fixed delay time of 4 5 The trend for both the signal and the SBR for all the samples was that they both increase until about 50 sand then remained approximately constant for the remainder of the measurements. It is also known from other studies that the emission signals typically only last about 50 s. Therefore 50 s was chosen as the optim11m gate width

PAGE 136

Analytical Evaluations Samplin~ Procedure 126 The same sampling procedure was followed for all the following analytical evaluations. The laser pulse energy was set to 180 mJ the optimum detector settings discussed in the previous section were used ( td = 4 5 ms and t = 50 ms ), a measurement consisted of 50 laser shots, and each measurement was repeated four times. The sample was also transJated both during and between measurements The reason for translation during a measurement was to try and obtain representative sampling from what might be a very heterogeneous sample The reason for translation between measurements was to introduce a fresh area for analysis. Figure 47 shows a photo of the resulting sampling track on an aluminum sample. The plot in figure 4-7 shows the chromi11m line intensity ( 425 44 nm ) for each laser shot. The shot-to-shot RSD of the 500 shots was 6.34o/o however when those 500 s hots are accumulated and the measurement is repeated 4 times RSDs on the order of I % are attainable. Chapter 5 will explore the question of precision further. The translational velocity was the only parameter that was changed when different samples were analyzed. For the alloys and paints a translational velocity of l 00 mis was used and for the soils and organics a translational velocity of 200 mis was used. All o f the samples analyzed were NIST SRM materials with the exception of three sand samples obtained from High Purity Standards (Charleston, SC ) Alloys Adhering to the guidelines discussed in the sampling procedure section the following NIST alloys were analyzed for the elements listed in parentheses with respective lines: alumin,1m (Cr 425.44 run ) steel (Cr 425.44 nm, Cu 327.40 run, Mo 386.41 run) and zinc ( Cr

PAGE 137

1mm 2 4~--,,---..---,-----,.---,---..----,.----r--....---.--...-------, 12 2 (/) C :, 2 0 8 ...__ :?;-1 8 (/) C Q) c 1 e Q) C :.:J 1 4 \,,. 0 1 2 0 0 0 0 0 0 O 0 goo 0 RSD : 6 34 /o 1 o-+---,,---..---,-----,.---,---..----,-----r------r---r-~ 0 100 200 300 400 500 Laser Shot 127 Figure 47. Plot shows the cbromilim signal obtained for each of 500 laser shots. The picture is of the resulting laser track.

PAGE 138

128 425.44 run, Cu 327 40 nm). Table 4-2 provides a comprehensive listing of the samples analyzed. There are two ways to present the analysis results, and the method chosen is dependent on the number of standards present in the series One way is to use the results to detennine a limit of detection (LOO), and the other, when there are enough standards in the series, three or more standards can be used to dete1111ine a cahbration curve and the remaining standards can be treated as unknowns and their concentrations determined. LODs were determined for each element in the different matrices and the grouped results are presented at the end of this section in table 4-8. Q1Jantitative results are presented for the copper in steel determinations. The resulting cahbration plot is shown in figure 4-8, along with the qt1antitative results in table 4-3. The values in table 4-3 indicate that we were able to obtain good accuracy and precision. Soils Several different soil samples obtained from NIST and High Purity Standards were analyzed for both Pb (405.78 nm) and Cu (327 40 nm) all of which are listed in table 4-4. Before analysis, these sa11111Ies needed to be prepared to improve sample rigidity. To do this a portion of each SRM was mixed 90 % w / w with cellulose binder (Spex CertiPrep, Cataolg#3642). Proper mixing was accomplished by ball milling each sample for 15 minutes. Then the sample was transferred to a pellet press, and pressed into pellets using a pressure of 4500 psi. Ag~ only one cahbration plot is shown in figure 4-9, and the quantitative results are listed in table 4-5. These results indicate reasonably good accuracy and precision; this is especially obvious when considering that the ''soil'' samples were river sediments soils, sands, and loams of varying matrices and particle sizes

PAGE 139

129 Table 4-2. NIST alloys analyzed by the LIBS benchtop instrument. Aluminum Zi nc S t ee l C r C r C u C r Cu Mo SRM Certified S RM Certified SRM Certified ( ppm ) (p pm) ( ppm) 601 200 10 626 395 1 9 560 30 1 26 1a 6930 50 42 0 1 0 19 00 100 602 70 10 627 38 7 1320 70 1 263a 13100 1 00 980 50 300 1 0 603 2400 1 00 631 l 1 1 3 l 1264a 660 50 2500 50 4900 1 00 1 265a 72 5 58 l 50 1 0 805 1800 1 200 1 00 807 11 500 900 350 808 6400 1200 809a 250 11 00 818 9600 2200 Table 4-3. Quantitative results obtained with the LIBS benchtop instrument for the analysis of copper in NIST steel STND indicates the samples treated as standards in deter1nining the others concentration. NIST Steels NIST SRM Type 1261a, AISI 4340 1263a, AISI 94Bl 7 1265a, Electolytic Fe 805, Medi,1m Mn 807, Cr-V 808 Cr-Ni 809a, Ni Cu Concentration (ppm) Certified Experimental 420 10 STND 980 50 933 36 58 1 STND 1200 STND 900 STND 1200 1157 47 1100 1123 46

PAGE 140

......-.. co Q) 7 6 en 5 I C ::J 0 (.) 4 '-""' 2 130 R 2 = 0.99802 1.....,__ ____________________ ______. 0 200 400 600 800 1000 1200 Cu Concentration (ppm) Figure 4-8. Cahoration plot for copper in NIST steel samples.

PAGE 141

3 5 3 0 r (0 a> 2 5 en I C: ::::, 2. 0 0 (.) rt"' co 1 5 C: 0) en 1.0 0 5 131 R 2 = 0.98684 0.0-+----------------.......---.----,,----,---4 0 20 40 60 80 100 Cu Concentration (ppm) Figure 4-9. Calibration plot for copper in NIST soil samples.

PAGE 142

132 Table 4-4 NIST and High Purity Standard soils analyzed by the LIBS benchtop instrument Soil Pb Cu SRM Certified 1645 1646 2704 2709 2710 2711 Sand A SandB Loam A (ppm) 714 28 28.2 1.8 161 17 109 19 18 3 98.6 5 18.9 0.5 34.6 0.7 5532 80 2950 130 1162 31 114 2 15.2 2.0 3.2 0.3 129 6 62 1.5 27 6 1.8 12.4 1.4 Table 4-5. Quantitative results obtained with the LIBS benchtop instrument for the analysis of copper in NIST and High Purity Standard soils. NIST Soils Cu Concentration (ppm) NIST SRM, Type Certified Experimental 1646, Estuarine Sediment 18 3 17. 3 3. 7 2704, Buffalo River 98.6 5 STND Sediment 2709, San Joaquin Soil 34.6 0.7 STND 2711, Montana Soil 114 2 90.7 5.9 Sand A, SC Soil 3.2 0.3 STND Sand B SC Soil 62 1.5 70.2 5.6 Loam A, KT Soil 12.4 1 4 15.2 3.4

PAGE 143

133 Paints Five leaded paint samples were prepared by adding known amounts of lead oxide ( Aldrich Chemical Company Inc., Catalog# 24 1547) to standard white latex paint ( ACE Seven Star Model# 16492) Samples were mixed for 60 minutes prior to application onto flat 25 mm square aluminum plates. Three coats of paint were applied over a three day period. The resulting layer of paint was 730 m thick which would indicate each dried coating was approximately 243 m thick. Analysis of these samples resulted in the calibration plot shown in figure 4-10. By treating three of the samples as standards and the other two as unknowns we can estimate the concentrations of the unknowns. Table 4-6 shows the results. The experimentally determined values are not in total agreement with the calculated ones. This was attributed to the fact that these samples were not standardized and that the calculated concentration should only be considered an estimate due to the possibility of losses of lead oxide in mixing (i.e. deposition on plastic mixing vial). Or~anics The organic samples which are provided by NIST in a JX>Wdered form, were prepared for analysis in the same way as the soil samples. The samples are listed in table 47 and the resulting cahbration curve is shown in figure 4-11. Due to the distnbution of the points on the plot it was not possible to use some standards to deter111ine the concentration of others. Detection Limits Limits of detection (LOD) were calculated for all measured samples when at least three samples were analyzed resulting in a linear calibration curve. The LOD is defined as LOD = 30 m ( 4-1 )

PAGE 144

12 10 I'(].) 8 en I C ::J O 6 u ,,,,, = 4 0) en 2 0 134 R = 0.99723 0 1000 2000 3000 4000 5000 Pb Concentration (ppm) Figure 4-10. Calibration plot for lead in prepared paint samples.

PAGE 145

135 Table 4-6. Quantitative results obtained with the LIBS benchtop instrument for the analysis of lead in prepared paint samples. Prepared Paints Pb Concentration (ppm) Calculated Experimental PPl PP2 PP3 PP4 PP5 5000 1472 1072 118 0 STND 1700 107 1240 190 STND STND Table 4-7. NIST organic samples analyzed by the LIBS benchtop instrument. NIST Organics 1566a, Oyster Tissue 1567 Wheat Flour 1570 Spinach 15 71, Orchard Leaves 1573 Tomato Leaves 1575, Pine Needles 15 77 Bovine Liver Mn Concentration (ppm) 12.3 1.2 8.5 0 5 165 6 91 238 675 15 10 3 1.0

PAGE 146

13 6 20 15 ,,.." (0 Q) en I C: :::J 10 0 (.) ,,,,, m C: C) R = 0.98679 5 en 0 0 50 1 00 150 2 00 250 Mn Concentration (ppm) Figure 4-11 Calibration plot for manganese in organic samples.

PAGE 147

13 7 where a is the standard deviation of the blank and m is the slope of the calibration curve. Since it is usually not possible to find a matrix-matched blank a was estimated as the noise in the background (pixel(s) where no line emission occurs) of the lowest concentration s ample. The LODs o f the measured samples can be found in table 4-8. The absolute LOO for chromit1m in zinc matrix can be estimated at 0.2 pg by assuming that approximately 2 g of material was analyzed per measurement Day-to-Day Reproducibility Over a six day period five steel samples ( 3 standards and 2 unknowns ) were repeatedly analyzed. The following procedure was followed each day: ( I ) the instrument was turned on and allowed to warm up for 30 minutes ( 2) the surface of the steel samples were prepared by sanding with 150 and 400 grit sand papers ( 3 ) the five steel samples were analyzed for 4 measurements of 50 shots each ( 4) a calibration curve was constructed and the unknown concentrations were determined Figure 4-12 s hows the results of these studies where (a) is the calibration plot of molybdenliro in NIST steel samples obtained on day one and (b) shows the results of unknown concentration detetninations for the six day period. The higher concentration unknown (unknown 2) had the best precision with an RSD of 11 o/o, and unknown I had the worst with 13%. These values are encouraging considering the current crudeness of the instrument and the sample preparation.

PAGE 148

12 10 co Q) 8 u, .... C ::, 6 8 :S C 4 O> (/) 2 0 0 (a) (b) ,:. E RSD = 11/o 0. B 1000' C 0 .... .. .... C Q) 0 C 0 Standard (.) Unknown 1 0 .. Unknown 2 RSD = 13/o 100 -. I I I 1000 2000 3000 4000 5000 1 2 3 4 Mo Concentration (ppm) Day Figure 4-12 Results of the day-to-day reproducibility data (a) Calibration plot of molybdenum in NIST steel samples .; (b) Day to-day reproducibility results Unknown 1 Unknown 2 I I 5 6 ..... v,) 00

PAGE 149

139 Table 4-8. Limits of detection (LOD) obtained with the LIBS benchtop instrument Sample Element LOD (ppm) Aluminum Cr 1 Cr 0.1 Zinc Cu 0.6 Cr 0.9 Steel Cu 3 Mo 11 Paint Pb 10 Pb 15 Soil Cu 2 Organic Mn 1 Depth Profilin~ The LIBS benchtop prototype was further evaluated for it's usefulness in depth profiling. For these measurements, the sample was not translated during a single measurement. This allowed the laser to continuously sample in one spot by drilling a channel into the material. The current method of choice for depth profiling of conductive samples with coating thickness of 0.1 to 200 mis by either glow discharge emission spectrometry (ODES) or glow discharge mass spectrometry (GDMS) [158]. The glow discharge techniques are very useful, but they are time consuming for thick layers (10 m2 hours), and are not very applicable to on-line measurements since the sample must be placed in an evacuated chamber. The LIBS technique, as will be demonstrated, overcomes the two disadvantages found in the glow discharge techniques.

PAGE 150

140 The samples were 50 mm diameter steel disks that had been electrolytically plated with varying thicknesses of zinc. Five samples were obtained with a zinc layer thickness of 1.22, 6.15 12.45 18.67 and 25.17 m (Kocour, Catalog# zn-20452 zn-20451 zn-20443 zn-20446 and zn-20445). The experimental parameters were the same as those used in previous sections of this chapter with the exception of the laser pulse energy The pulse energy was optimiz.ed to obtain a balance between analysis time and resolution. Over the size range of interest it was found that a pulse energy of 30 mJ satisfied these two criteria. If thicker layers were attempted the laser energy could be increased to decrease the analysis time. Conversely, for coatings thinner than 1 ~ a lower laser energy could be used to improve the spatial resolution. Figure 4-13 shows the overlaid spectra obtained for the first and thousandth Jaser shot on the 25 m thick layered sample. The spectra obtained from the frrst and thousandth laser shots are identical to spectra obtained from pure zinc and iron, respectively. To obtain depth profiling information the peak height of the 308.32 nm zinc line and that of the 306.72 run iron line were monitored as a function of laser shot This data is shown in figure 4-14. To be able to use this data to analyze unknowns, it is necessary to calibrate the instrument by finding the relationship between the experimental data and the known layer thickness. The two criteria that are typically used are the cross-over point and the integrated signal. The cross-over point refers to the point where the zinc signal crosses the iron signal on the depth profiling plot, and the integrated signal refers to the integrated signal under the zinc curve in the depth profiling plot. By plotting both of these criteria versus the known zinc layer thickness the plots in figure 4-15 were obtained. From the good correlation values obtained for both plots it is obvious that either criteria can be used for cahbration, and that a linear relation will be obtained. One other interesting point observed

PAGE 151

5 -..----r---...---.........------.------.-------.--....-----.--4 4> 3 C: :::s 0 0 .._ 2 "' C: C) Cl) 1 .. .... . ' . . . . . ' ' . . . . . ' . . . ' ' . ' .. I ' \ \ \ ' . .. ' . . .. ' ' . '' . .. . . . . .. . . ' ' ' . I I : . . . .. . . . . . ' ' . I Shot 1 -Shot 1000 ' . ' ' ' . ' . . . . ' ' . . . . \.w/1,J~ v,.,...;,.I. ~ ... 0 -+----r---...---.........------.-------..---, ---.-------.--~ 306 307 308 Wavelength (nm) 309 310 Figure 4-13. Typical depth profiling spectra obtained from a the analysis of a zinc plated steel. 141

PAGE 152

7 .---.---.-----......--...---.----r----r---....----.---, 6 .-.5 4) 4 :::, 8 -3 CV C: C) u5 2 1 200 25.17 m zn --Fe 400 600 800 1000 Laser Shot 7 ---.------.---......--------------. 6 18 67 m 5 \ 4 3 2 1 0 100 200 300 400 500 6 --------------.--......-.......-----..---------, 5 12.45 m 4 3 2 0 50 100 150 200 250 300 Figure 4-14. Depth profiling plots for the results of the analysis of the five standards. 142

PAGE 153

7 ,----,----,---.----,-..--.....-----.---,.-..---,----,--.----,,--.....-.....-----.-6 ,-. 5 "d" 0 rn ... 4 0 0 ] 3 00 t;/) 2 1 0 20 6 15 m 40 60 80 100 120 140 160 Laser Shot 7 ....----.----.---...---.------.------.----.--....----,--, 6 1.22 m 5 4 3 2 1 0 10 20 30 40 50 Figure 4-14 continued. 143

PAGE 154

(a) (b) 350-.-----,,--,---,r------r-----r------------. 300 250 a 200 .c CJ) 150 (/) J ....J 100 50 0 0 5 R=0 99811 14 < "' 12 C: ::J 0 O 10 J C: Q> 8 CJ) u 6 i C: 4 R=0 99588 2 -+---r---r-----r---r-----r----,------------~ 10 15 20 25 5 10 15 20 2 5 Zn Layer thickness (m) Zn Thickness (m) Figure 4-15 Depth profiling calibration plots (a) Zinc and iron cross-over plot ; (b) Zinc integrated signal plot

PAGE 155

145 during the depth profiling studies becomes evident upon closer examination of the plot s in figw-e 4-14. As the laser begins to break through the zinc layer and sample the iron base the iron signal begins to increase. The iron signal then reaches a maximum and decays to ~ 50o/o of the maximum value where it remains for the remainder of the analysis It was intri g uing that the iron signal would actually reach a peak We found that the iron signal would always decay to a constant value at the same point that the zinc signal reached a constant minimum. This leads us to the conclusion that the presence of the zinc was somehow enhancing the iron signal This was confirmed based on the observations we made with other zinc-containing compounds discussed in chapter 6 Conclusions A benchtop LIBS instrument has been constructed and evaluated. The figures of merit accuracy precision, limits of detection, and day-to-day reproduc i bility of the measurements are satisfactory for performing analytical analysis The usefulness o f this instrument bas been exlubited on several different types of materials for a handful of elements. With more refinement and the incorporation of a fiber optic probe this instrument will be capable of routine onand off-line analysis.

PAGE 156

CHAPTERS LIBS PORTABLE INSTRUMENT: DEVELOP:MENT AND EVALUATION Introduction In recent years, there has been a resurgence of interest in the LIBS technique. One focus has been upon moving LIBS from the laboratory to the field, for at-site analysis. This has been accomplished through the use of fiber optic probes, and portable instruments [84,159). Miniaturization has also played a big role, for example, where microchip lasers occupying a volume of less than 3 cm 3 have been used for LIBS analysis on soils [160]. Yamamoto et al have developed such a portable instrument, packaged in a suitcase, consisting of a hand-held probe for sample analysis (159]. The probe contained a compact laser which produced the LIB plasma. The emitted plasrna light was collected with a fiber optic and transported to a 1 / 8 m spectrometer for analysis. They evaluated the instrument for the analysis of metals in soils (Ba, Be, Pb, and Sr), paint (Pb), and particles collected on filters (Be and Pb). Detection limits were best for Be in soils (9 .3 ppm), and worst for Pb in paint (8000 ppm). Relative standard deviations for these measurements varied from 3 4 7% This paper by Yamamoto et al. [159] is the first example in the literature of a portable instrument ( operating on 120V AC) for LIBS analysis. In this chapter, we report the development of a compact, battery powered, portable LIBS spectrometer, and its application to the analysis of paint, alloy, ore and organic samples. The application of LIBS to the analysis of paints, [84,159] steels, (82,107, 161-164] 146

PAGE 157

147 and ores [165-166] has been widely investigated and was therefore used as the ba sis for evaluation of this new instrument Since the use of a compact laser has only once been reported in the literature [159] we have studied the spatial development of the laser-produced plasma using conventional techniques [167 58]. These measurements have led to the development of a more reliable probe design. Experimental Instrument Desi~ A schematic of the LIBS instnurent is depicted in figure 5-1. It consisted of a laser s pectrometer / detector computer electronics optics, and a re-chargeable battery The individual components were placed in a suitcase (48.3 x 33 x 17.8 cm), and had a combined weight of 13.8 kg. The Nd : YAG laser was a IGgre MK-367 (Kigre Inc., Hilton Head Island, SC) which delivered a 3.6 ns duration pulse of2 l mJ at the fundamental wavelength of 1064 run at a maximum of 1 / 3 Hz repetition frequency. A repetition rate of 1 Hz was available if the laser head was mounted in an optional water-cooled mount. The advantage of using the Kigre was its small size (10.11 x 2.06 x 2.21 cm ), which allowed incorporation into a compact sampling probe. When the laser was operated at maximum power, a sequence of five laser pulses were produced each separated by ~ 16 s. The layout of the probe optical bench is shown in figure 5-2. The laser beam was focused on the target using a coated 12.7 mm diameter, 20 mm focal length lens. The resuhing plasma created a 450 m diameter crater with an average incident laser irradiance of 0.92 GW / cm 2 The emitted plasma light was collected back through the focusing lens and reflected off a beam splitter (Corion, Model# HT-500-F) and mirror, before it was focused with a 12.7 mm diameter 25 4 mm focal length lens onto a 400 m optical fiber (Ocean Optics, Dunedin, FL). The 2 m fiber optic was

PAGE 158

.. ,,.,,1 111 ,. ...... .. . "' . ....... . . .. .. .... ....... .. !'~ 1! ... ........ _.-------~-... =~~"% "'. ,: ,,r """" _, --i .... ,.,. ,.. ... ~"'~JY J! ~ ,.. ,. ,. ... "e., ,." ." .;: = --1:"~ ": ( ulll:., .. 4w, .. "'ii;,.~ ~qj.~ -~ , .L IJ',;~ t. t :;. ,.,er ~.~ ...... ... 1 1.1o ........ .. .-. ... -;i._..: ,_ ,.... .. Probe Storage Spectrometer Laser Power Supply 148 Laser Control .... ., -;;-,_,... ,"" "ffi"i~ ; ,:, .._,_ ..... -.. -' ei>"';,O .. ~~ L ,._~~ .. :- ,.. A;~-.... ..... .. t,;; .M.,tc r .. . ff~""J : It ..... .............. .. ....... .. ~.t_i, '"~.~~ -. .... j!-J: ;.;-: ,,,: -~.: tH--. .:: Figure 5-1. Schematic of the battery powered LIBS portable instrument. Beam Splitter ~~---------------~ --------------~~---------------~ Kigre Laser Fiber Optic Mirror -1cm I I Figure 5-2. Schematic of the optical bench part of the LIBS portable probe

PAGE 159

149 connected to a miniature spectrometer (Model# S2000, Ocean Optics) consisting of optical components, a modular grating disperser and a linear CCD (2046 pixels) detector with control electronics. The small dimensions (13.8 x 10.4 x 3.9 cm) of the S2000 allowed its incorporation into the suitcase design. The S2000 had a spectral dispersion of 0.060 run per pixel over a fixed spectral coverage from 339 462 nm. The spectral resolution was determined from the FWHM of the Hg 435.8 run line to be 0.4 nm. The S2000 CCD detector was a non-gateable device; however it did have the capability of external triggering. A timing circuit was designed to control laser firing and data collection. For data collection, the laser was operated at 0.25 Hz and the S2000 was set to an integration time of 8 ms. Data was collected using a laptop computer (IBM ThinkPad, Model# 365XD) running the Ocean Optics Inc. Basic Acquisition Software (OOIBASE) version I .OB. For all measurements, the probe was moved horizontally between shots to allow sampling from a fresh location and to improve the reproducibility of mass ablation. Ten laser shots were accumulated per sampling site, with four sites being analyzed per sample. Spectral Imaiini Apparatus The same detection system, described in detail in chapter 3, was used for capturing spectral images. A collecting lens was placed at twice its focal length (25 cm) from the LIB plasma and an ICCD (Princeton Instruments, Model# ICCD-576S). A beam combiner ( 0.3 neutral density filter) allowed radiation from the plasma as viewed from the front via one plane mirror and the side view via another plane mirror to reach the detector simultaneously. By slight rotation of the plane mirrors around vertical axes, the two images could be laterally offset from one another. This setup allowed for simultaneous capturing of two perpendicular views of the plasma during a programmable integration time.

PAGE 160

150 Sample Preparation Leaded paint samples were prepared by adding known amounts of lead oxide ( Aldrich Chemical Company, Inc., Catalog# 24 154-7) to standard white latex paint ( ACE Seven Star Model# 16492 ) Samples were mixed for 60 minutes prior to application onto flat 25 mm square aluminum plates Three coats of paint were applied over a three day period The resulting layer o f paint was 730 m thick which would indicate each dried coating was approximately 243 m thick. Validation of the prepared samples was accomplished by the use of a LA-ICP-MS system ( Finigan MAT SOLA, San Jose CA ). Other s amples were pressed into 31 mm diameter pellets of v arying thickness made with a pellet pre ss at 4500 psi pressure. In pressing the pellets pure lead flakes (F isher Scientific Catalog# L-24) and NIST soa iron ores, and several organic standards were used as sample materials. The NIST soa iron ore, and organic standards were mixed 90 % ( w / w ) with a cellulose binder ( Spex CertiPrep C atalog# 3642 ) to improve sample rigidity The alloys were also NIST standards and these were polished with 600 grit sand paper prior t o analy s is. Results and Discussion Plasma lmaiini It was necessary to capture images of the plasma produced by the Kigre laser to determine if the lower pulse energy laser produced a plasma different in size / shape compared to that produced by the higher pulse energy laser used for the LIBS measurements described in earlier chapter s Measurements were made to determine if the plasma structure varied with sample matrix. For these measurements a pressed lead pellet pressed lead-spiked soil, and lead-spiked paint samples were used The emission was accumulated for 15 s with a

PAGE 161

151 delay of 1 s from the beginning of the plasma to eliminate the initial intense continuum emission. No spectral filtering was employed so these images represented the wavelength integrated emission from the plasma. The images observed for the three different samples, captured under identical conditions are shown in figure 5-3. In each frame, the image on the left represents the plasma as viewed from the front and the image on the right represents the plasma as viewed from the side by the ICCD. From these images, we can see that the plac;ma has a distinctly different shape depending on sample matrix. The plasma produced on the pressed lead pellet was cylindrical in shape (figure 5-3(a)). However, the plasmas produced on the soil and paint samples were more flat, and had a dome-shaped appearance ( figures 53 (b) and 5-3(c) respectively). This flattened appearance of the plasma indicated that the best method to view the plasma would be slightly from the side in order to increase the line of sight. However, collecting the emitted light perpendicular to the incident laser is not usually practical with field measurements where the probe to sample distance is not always constant The next section will point out that the perpendicular method of light collection will result in a totally different volume of the plasma being viewed as the probe-to-sample distance changes. The overall solution for variable matrix samples, where the plasma shape changes, would be to view a larger vol11me than the plasma would act\Jally occupy. Lens-to-Sample Distance Further imaging of the plasma was performed to determine if the plasma strucrure changed when the distance between the laser focusing lens and the sample was varied. For these measurements, the 10 % lead-spiked paint sample was used, and the resuhs are shown in figure 5-4. Images captured with the focusing lens positioned closer to the target are indicated by negative distances and ones further from the target as positive distances. As the

PAGE 162

152 (a) (b) (c) Figure 5-3. Wavelength integrated images of the LIB plasma on three different samples. Each individual image shows orthogonal views of the plasma. In each picture the left image is the frontal view and the right image the side view. The images were captured after a 1 s delay and with a 15 s integration time. (a) pure lead; (b) 25 % lead added to soil; (c) 10 o/o lead added to paint.

PAGE 163

153 -3 mm -2mm -1 mm 0mm +2mm +3 mrn Figure 5-4. The images present the effect of changing the lens-to-sample distance (L TSD). The sample is a paint sample spiked with 10 % lead. A distance of O mm represents the laser lens at its focal length from the sample, and negative distances indicate a focus further into the sample and vise versa for positive distances.

PAGE 164

154 laser radiation is focused deeper into the sample, the plasma retained its shape. To the other extreme, as the laser radiation is focused above the target the plasma becomes elongated due to the occurrence of a breakdown in the air above the sample. From these results, we determined it was best to place the laser focusing lens at its focal length from the target or slightly closer. Evaluation of Probe Desi~ Lens-to-sample distance We have found in previous studies that sampling reproducibility is improved using an axial view of the plasma to collect the emitted light. The reason is that small fluctuations in the lens-to-sample distance (L TSD) do not affect the signal or signal-to-background ratio. If the plasma is viewed from the side, a slight change in the sample height can cause a different region of the plasma to be detected (see figure 5-4). In designing the LIBS probe, this idea was incorporated To verify the suitability of this choice we made one 10-shot measurement on a NIST steel sample containing 0.67 o/ o Mn by weight ( SRM 1261 a ) The peak height of the Mn 403.08 nm line was used as the signal, and a pixel where no line emission was observed was used as a measurement of the background. These results are shown in figure 5-5. The signal data is presented as a signal-to-background ratio (S BR ) where the background acts as an internal standard. The data indicates that at a LTSD of 18 5 mm (lens bas a nominal focal length of20 mm), a .5 mm change in the LTSD resulted in a RSD of the SBR of 4.32 %. Because the LTSD of 18.5 mm was the middle of the range over which the SBR was the most constant, this was chosen as the optimum L TSD to minimize changes in L TSD effects.

PAGE 165

0.8 0 I t 0 6 '"C C: :::, 0 'C> u ca cc 0 I ca C: C) 0.4 en 0.2 ... 0.0-+-----.---.-----.---.-----.---.----.---.-----.---.-----.---.----1 -6 -4 -2 0 2 4 6 Lens to Sample Distance (mm) 155 Figure 5-5. Plot of the Mn signal to background ratio as a function ofLTSD. A distance of O mm represents the laser lens at its focal length from the sample, and negative distances indicate a focus further into the sample and vise versa for positive distances.

PAGE 166

15 6 spatial filterini With axial viewing the hot center part of the plasma is directly viewed. Other researchers have found that by spatially filtering out this center emission, there is a great impro v ement in the signal-to-background ratio [57] Because the detector c o uld not be temporally gated, spatial fihering was used to improve the signal-to-background ratio. F igure 5-6 s hows the observed improvement. This was achieved by moving the fiber o ptic o ff-axis to view the outer edge of the plasma (see figure 5-7). The on-axis viewing doe s increase the s ignal, but also proportionally increases the background. For comparison, the off-axis view ha s a decreased s ignal and more than proportionally decreased background. The s ignals backgrounds s ignal-to-background and s ignal-to-noise ratios are shown in table 5-1 Note that the precision o f the measurements and the signal-to-noise ratio improve by a factor of 2 and 25 respectiv e ly when the off-axis collection method was used. All values are the average of 10 single s hot measurements of the Mn 403.08 nm line emitted from an analyzed s teel sample. The noise was estimated by the standard devi ation of the background pixel signal for the 10 measurements. Because of the enhancement in the signal-to-noise ratio of more than an order of magnitude off-axis collection was used in all analytical measurements. Analytical Applications Analysis of paint samples Both NIST and the prepared paint samples were analyzed for lead. Figure 5-8 shows ten-shot accumulated spectra obtained for samples of pure Pb 10 % Pb in paint and blank paint. The spectra have been vertically offset for clarity but have nearly equivalent background levels Two series of paint standards were analyzed for lead contamination, those prepared in-house and a series of standards from NIST The two sets of standards resulted

PAGE 167

(/) ... C :, 20000 15000 0 10000 0 co C 0) Cl) 5000 0 340 18000 16000 14000 12000 (/) C 10000 :::, 0 0 8000 cu C 0) 6000 (/) 4000 2000 400 157 (a) \ Off-Axis 360 380 400 420 440 460 Wavelength (nm) (b) Mn Line \\ /\ \ I I _r \ ____ ; -----' ----------..... ----J On-Axis Off-Axis \ I 402 404 406 408 Wavelength (nm) Figure 5-6. Spectra obtained from the analysis of a steel sample both onand off-axis. The top and bottom spectra in each plot are the result of onand off-axis light collection. ( a) signal over the wavelength range from 340-460 nm; (b) zooms in to show the Mn line

PAGE 168

I \ \ C>n-Axis LIB Plasma Cdlection Zone Off-Axis Figure 5-7 A visual representation of spatial gating indicating the difference between onand off-axis light collection. Table 5-1. Table of Mn signal, background, signal-to-background, and signal-to-noise ratios for both onand off-axis collection. Measurement On-Axis Mn Signal 198 21 Background 1150 140 Signal/Background 0.174 0.020 Signal/Noise 1.5 Off-Axis 167 14 120 4 1.40 10 37.6 158

PAGE 169

(a) 4r----....------------------------3 Pure Pb c..s C:: I 00 'q" Cl.) ..... c:: :::s 0 (.) ,.__,,, c..s c:: 00 00 10% Pb in Pa int 0 Blank Pa int 320 340 360 380 400 420 440 460 480 Wavelength ( nm ) (b) 4 0 -------------......... -----...-------3.5 3.0 2. 5 2 0 I 5 I 0 0 5 0.0 39 5 400 405 Wavelength (nm) Pure Pb 10% Pb in Paint Blank Paint 410 415 1 59 Figure 5-8. Single shot spectra captured for samples of blank paint, 10 o/o lead added to paint, and pure lead. The spectra have been offset vertically for clarity

PAGE 170

160 in calibration curve s with different s lopes. Examination of the laser sampling spots indicated that there was a difference in mass removal for the two s ets of samples based o n the size / depth of the craters. The reason for this is the difference in the way the s ample s were prepared. The NI ST sample s were pre s sed paint particles and the prepared paint s were layered coats of paint Because of this difference we were not able to show both sets of data on the same plot and because of the limited number ofNIST samples only the prepared paint s tandards were used for q11antitation. Figure 5-9 shows the linear calibration curve obtained for the prepared paints At a concentration of ,..,, 5 %, the slope of the calibration curve began to decrease At this point a weaker lead line at 368 35 run could be used. The points represent the average of three measurements where each measurement consisting often laser s hot s Table 5-2 lists the results obtained when three of the samples are used to establish a calibration curve and the other two are treated as unknowns. A 3a-based limit of detection ( LOO ) was detern1ined using the lead peak height and the standard de vi ation of a background pixel. The LOO for lead in paint was 0.12 o/ o by weight for a 1 0 shot measurement. T he preci s ion for these measurements varied from 4.0 44 1 %. As previously mentioned, the prepared paint samples were also analyzed by LA-ICP-MS. There was a good correlation ( 0.9995 ), shown in figure 5-10 between the results obtained by the LIBS instrument and the LA-ICP-MS results. Analysis of NIST steel samples The instrument was also evaluated for the detection of manganese in steels The sample s were from two different series of NIST steel samples. The 403.08 nm manganese line was used for these measurements. Figw-e 5-11 shows the calibration curve obtained. An iron line at 404.58 run was used as an internal standard. The decrease in slope of the curve

PAGE 171

1000-----------------------800 r' 600 Cl) I C :::, 0 (.) -,,,,, m 400 C C) en .c a.. 200 0 I R 2 = 0 98886 0.0 0 5 1 0 1.5 2.0 2 5 Pb in Paint (w/w % ) 161 Figure 5-9. LIBS calibration curve (linear region) of the peak intensity of the 405.78 nm Pb line versus the amount of lead added to paint standards.

PAGE 172

1 62 0.5 0 3 0.4 .,, C 0 I r m 0.3 'I t C Q) 0 C 0 (.) 0.2 ..c a. en R 2 = 0.99953 I 0.1 a. (.) I ::i 0.0 0.0 0.1 0.2 0.3 0.4 0.5 LI BS Pb Concentration (w/w % ) Figure 5-10 This plot shows the correlation of the LIBS measurements to standard LA-ICP-MS measurements.

PAGE 173

0.50 0.45 0 40 0 0 35 ro a:: ro 0.30 C: C) en 0 25 (1) u. ..., ..., 0 20 0.15 0 10 0.05 I R 2 = 0 99687 -0.2 0.0 0.2 0 4 0.6 0 8 1.0 1.2 1.4 1.6 Mn Conctration (w/w 0 /o) 163 Figure 5-11. Plot of the ratio of the peak intensity of the 403 08 run Mn line to that of the 405.58 run Fe line versus the concentration of Mn in the NIST steel standards

PAGE 174

164 Table 5-2. List of the experimental results obtained when three of the paint samples (STND) were used to establish a calibration curve and the other two are treated as unknowns The actual values were calculated from the amount of lead oxide added to the paint. Prepared Paints PPl PP2 PP3 PP4 PP5 Pb Concentration (w / w o/o) Actual Experimental 2.10 1 29 0.10 0.500 STND 0.147 0.156 0.036 0.103 STND 0.0 STND around 0. 7o/o is reasonable for the onset of self-absorption, considering that a 0. 7% Mn sample corresponds to a Mn number density of ...., 10 14 atoms / cm 3 [168]. Table 5-3 lists the results obtained when three of the samples are used to establish a calibration curve and the other three are treated as unknowns. Using the linear section of the calibration curve, the 3o LOD was calculated to be 0.016% Mn by (w/w) in steel. The precision for these measurements varied from 2.6 4 7 o/o. Analysis of NIST iron ore samples NIST iron ore sarnples were analyzed to determine the manganese concentration using the 403.08 nm emission line. An iron line at 404.58 nm was used as an internal standard. In this case the concentration of Fe also varied. The cahbration curve from these measurements is shown in figure 5-12. Large departure from linearity was observed at concentrations above about 0.2o/o. The precision for these measurements varied from 1 9 4.1 %.

PAGE 175

1 65 0 25 0 0 20 I m 0:: m C C) en Q) 0.15 LL -C 0.10 0.0 0 1 0.2 0.3 0 4 Mn Concentration (w/w 0 /o) Figure 5-12 Plot of the ratio of the peak intensity of the 403.08 nm Mn line to that of the 405.58 nm Fe line versus the concentration of Mn in the NIST iron ore standards

PAGE 176

166 Table 5-3. List of the experimental results obtained when three of the steel samples ( STND ) were used to establish a calibration curve and the other three are treated as unknowns. The actual values were provided by NIST. NIST Steels Mn Concentration ( w / w % ) Actual Experimental 1261a 0.67 0.01 0 635 0.046 1264a 0.258 0.005 STND 1265a 0.0057 0.0005 STND 808 0.62 0.595 0.058 809a 0.70 STND 818 0.52 0.515 0.048 Analysis ofNIST or~anic samples Several NIST powdered organic samples (see table 5-4) were analyzed to determine the calcium concentration. Figure 5-13 shows the resulting calibration curve for the calcium 422.01 run line. A reason for the poorer results is that we were \1nahle to find a suitable internal standard for Ca. Table 5-4 lists the results obtained when three of the samples are used to establish a calibration curve and the other three are treated as unknowns. From the slope of the calibration curve the 3o LOD was deterrnjned to be 0.13% Ca ( w / w ). The precision for these measurements varied from 0.4 4.9 %. Conclusions A battery-powered, portable LIBS instrument bas been constructed and evaluated for a variety of sample types. Despite the low pulse energy laser and the use of a poor resolution, non-gated spectrometer, useful qt1antitative results were obtained. Limits of detection ranged

PAGE 177

500 400 ,., en I t C: 5 300 (.) ca C: C) en ca 200 0 100 R 2 = 0.98815 .... --o....,_ ____________________ -....-----. 0.0 0.5 1.0 1.5 2.0 2.5 Ca Concentration (w/w % ) 3.0 Figure 5-13. Plot of the peak intensity of the 422.01 nm Ca line versus the concentration of Ca in the NIST organic standards. 1 67

PAGE 178

168 Table 5-4. List of the experimental results obtained when three of the organic samples (STND) were used to establish a calibration curve and the other three are treated as unknowns. The actual values were provided by NIST. NIST Organics Ca Concentration (w / w %) Actual Experimental 1570, Spinach 1.35 0.03 1.48 0.18 1571 Orchard Leaves 2.09 0.03 1.72 0.40 1573, Tomato Leaves 3.00 0.03 STND 1575, Pine Needles 0.41 0.03 STND 1577, Bovine Liver [0.0123) N.A. 1566a, Oyster Tissue 0.196 0.03 0.355 0.089 1567, Wheat Flour 0.019 0.001 STND from 0.016 % for Mn in NIST SRM steel to 0.13% for Ca in NIST SRM organic samples. Acceptable precision (0.4 4.9 %) was obtained for steel, ore and organic SRM samples. The poorer precision (4.0 44.1 %) obtained for the detection of Pb in paint can be attributed to the heterogeneity of the samples. Such a portable instrument may be of use for rapid field determinations of Pb in paint, or for the rapid identification of alloys. Future studies will investigate other elements/samples, transition to unprepared (field) samples, spectrometer drift, and the repeatability of measurements.

PAGE 179

CHAPTER6 VARIABLES INFLUENCING THE PRECISION OF LASER-INDUCED BREAKDOWN SPECTROSCOPY MEASUREMENTS Introduction The main attributes of laser-induced breakdown spectroscopy, non-invasiveness, requiring small amounts of sample, and a need for minimal sample preparation, are also partially responsible for the unsatisfactory figures of merit of the technique. Since only a small amount of material is removed and analyzed, the accuracy and precision of the measurement can be dependent on the homogeneity of the sample. With only minimal sample preparation, there is the possibility that the mechanical properties (smoothness and contamination) of the samples surfuce can also affect the accuracy and precision. The surface roughness can change the lens to sample distance, and therefore influence the plasma characteristics and also affect the collection of the emitted light due to a changing plasma focus. For right-angle (with respect to incident laser beam) collection of the emitted light, light collection may be blocked by the sample structure or distorted by lateral movement of the plasma on the surface with respect to the entrance slit of the spectrometer. Other conditions that affect the figures of merit are physical parameters (sample state, pressure, temperature), atmospheric conditions (gas composition, pressure) and sample matrix composition. Several more 'controllable'' variables include choice of analytical line, laser shot-to-shot variance, speed of sample movement, and detector settings ( time delay and gate 169

PAGE 180

170 width). A better understanding of the effects of these variables should lead to improved analytical perfor,nance. The focus of the chapter is to identify and quantify the variables that affect the precision ofLIBS measurements. At presen 4 no similar systematic study has been reported. Wachter et al used LIBS to deterrnine uranium concentration in solution and examined the effects of laser repetition rate, detector gating parameters, and the number of averaged laser shots on the precision [169]. They fo11nd an increase in precision (relative standard deviation) with increasing laser repetition rate, which was attributed to the increase in and stabilization of the concentration of airborne material above the liquid surface at higher laser repetition rates. They also deterrnined the relative standard deviation (RSD) as a function of detector parameters including time delay after the laser pulse (tJ and the gate width (!w). Over a range oftd = 0.1 10 sand tw = 0.3 1000 s, they found that the RSD was independent of the timing parameters. They also studied the RSD as a function of the total number of laser pulses averaged, for net and ratioed signals. For the net signals, the RSD decreased from 13.3% for 50 laser shots to l.8o/o for 1600 laser shots; they also saw a similar, but less dramatic, decrease for the ratioed signals. W1Sburn et al. also studied the RSD of LIBS measurements on particulate samples as a function of laser repetition rate, freshness of sampling site, and the number of spectra binned for lines of various intensity [ 1 70]. They found that the RSD was low up to a laser repetition rate of 8 Hz after which the RSD increased dramatically. Investigations of the differences in RSD between fresh and same site measurements indicated that when more than two shots were collected from the same site, the RSD increased. The lowest RSD was obtained when each shot sampled totally new material. They also found that there was an improvement in the RSD by increasing the

PAGE 181

171 nwnber of spectra binned up to a limit, and that the amount of improvement was dependent on the signal to noise ratio (SIN). Eppler et al when studying matrix effects in the detection of Pb and Ba in soils, investigated precision as a function of the laser focusing lens for both spherical and cylindrical lenses [171] They found a reduction in the RSD when using the cylindrical lens instead of the spherical lens, when the particle size of the ratioed elements varied significantly. However, when the particle sizes of the ratioed elements were similar, the RSD was unaffected by the choice of focusing lens. The authors attributed the reduction in the RSD to the increase in material sampled by the cylindrical lens and therefore sampling of a more statistically homogeneous sample. Pakhomov et al looked into precision effects when using LIBS to determine Pb concentration in concrete (172]. They determined precision as a function of td, and found that the standard deviation of the background increased with increasing time delay over the range of O 18 However over the same time span, the background signal was also increasing, which would not be expected. In this chapter we have attempted to identify and q11antify the controllable factors that affect the precision of LIBS measurements. We have investigated the effect of the choice of analytical line, emission signal temporal development, sample translational velocity number of spectra acct1mulated, laser pulse stability, detector gate delay surface roughness, and background correction, on LIBS precision. All of these measurements were made using a pure copper sample to eliminate the effect of sample homogeneity on precision. Our reason for using a pure metal was for the application of the LIBS technique to the determination of the purity of precious metals. For the reader to understand our results it is necessary to name the goal, that is whether inter-measurement (between measurements consisting of the swnmation of spectra from several laser shots) precision, or intra-measurement (within one

PAGE 182

172 measurement comparing individual shot-to-shot spectra) precision is to be optimized. As we will show both are not optimized under the same set of conditions. This study was aimed at a conventional set-up where no internal standardization, photoacoustic or special temporal normalization procedures were used. Experimental The same experimental apparatus described in chapter 4 was used for these measurements. It is important to restate how the data can be collected and stored. As was mentioned in chapter 4 the multiple laser shot spectra could be stored in one of two ways by the Wmspec software, either as "spectra" or as "acc1rrnulations". When multiple LIBS s pectra were stored as "spectra", each individual spectrum, from a single shot, was stored within a file. When "accumulations" was selected, the multiple LIBS spectra were summed, and stored as one spectrum. In this mode all of the shot-to-shot data was lost. For all of the data presented in this chapter the data was stored as '' spectra '' by the Winspec software. To evaluate intra-measurements precision the data was used as is. To evaluate inter measurement precision on the same data sets, the spectra from an individual measurement (file) were sununed into one spectrum. The captured data files were analyzed for background correction and peak height determination by MatLab programs which were written in-house. The samples were either pressed pellets or pure copper discs. The pressed sample pellets of 31 mm diameter and varying thickness, were made with a pellet press at 4500 psi pressure Copper powder (stock# 13990, 99 % pure, 325 mes~ Alfa AESAR) and graphite (stock# 014145-03, 99.999 % pure, 100 mesh, Carbone of America) were used as sample materials for the pressed pellets. The pure copper discs were 25 mm diameter 5 mm thick disk made from stock copper rod( ~ 99 % pure). For the surf.ace roughness study, the surface

PAGE 183

173 of the copper discs were modified by using silicon carbide sand paper of 600 and 400 grits and aluminum oxide sand paper of 220, 150 100, 60 and 40 grit. Results and Discussion Choice of Analytical Line Since a pure copper sample was used in these measurements to minimize the effect of heterogeneity on the precision, it was necessary to verify that the analytical lines used were still linear for the high atomic vapor concentrations produced (no self-absorption). Investigations showed that the 427.511 nm atomic and the 213.598 run (observed in the s econd order) ionic lines were good choices Table 6-1 lists the line wavelengths order of observation(* denotes second order) ionization stages, energy levels for the transitions, and relative intensities of the lines in the observed spectral window [173]. To verify that line emission intensities produced a linear response with concentration, for high concentrations 8 standards were prepared of ptrre copper diluted in graphite over the concentration range of 5 to 90 %. The resulting linear calibration curves for these lines are shown in figure 6-1. It is also interesting to point out that the pure copper target resulted in signals comparable to the signals resulting from the 20 % Cu in graphite sample. The lower signal observed for the pure copper is at least partially attnbutable to the removal of less material in the case of the pure copper sample. For a pressed particulate sample, more material is ablated than for a solid metallic srutlple. It is clear that these two lines will exhibit a linear intensity dependence with copper number density for pure copper targets. Si~al Intra-Measurement Development For the majority of the measurements, the individual spectrum from each laser shot was recorded. This was necessary to identify any variations from shot-to-shot caused by a

PAGE 184

1 7 4 Table 6-1. Table of line wavelengths, order of observation, ionization stages energy levels for the transitions and the observed relative intensities of the lines present in the observed spectral window. The* denotes the lines observed in the second order Wavelength ( run) 211.7300* 212.2979* 424 8956 212 6045* 425.9401 213.008* 213 4341 213.5982* 427.5107 214 8948* 215.1801 Ionization Stage II II I II I II II II I II II Lower Level (eV) 8.49 3.26 5.08 2.83 4.97 8 52 8.52 2.72 4.84 2.72 8.66 Upper Level (eV) 14 34 9.09 7.99 8.66 7.88 14.34 14.33 8.52 7.74 8 49 14.42 Relative Intensity 2 14 3 14 I 2 4 24 18 13 l

PAGE 185

(a) 10-r---r---r----'T'---r---r--.....--------.---,------.-----. ,, I ,/ I 8 ,-... 6 Q) VJ C: :, 8 4 C'O R = 0.990 C: O> ci5 2 / 0 7 6 ,:::5 Q) VJ c 4 :, 8 -::::3 C'O C: O> (/) 2 1 (b) R = 0 992 0 4----r---.....-----T---.---,---.....--~---.----.----,-------1 0 20 40 60 80 100 0 20 40 60 80 100 Cu Concentration (wt%) Cu Concentration (wt 0 /o) Figure 6-1. Calibration plots obtained from the analysis of copper in graphite standards (a) Plot obtained using the 427.511 nm atomic line; (b) Plot obtained using the 213.598 nm ionic line ( observed in the second order).

PAGE 186

176 variety of sources. All measurements were made at a fixed delay of 2 s, and fixed gate of 15 s, parameters which we have found to be good compromise for a wide range of samples. Upon captwing the spectra from 900 individual laser shots, and plotting the peak height of the 427.511 nm line as a function of laser shot (i.e. intra-measurement development of the Cu emission signal), a non-random behavior was identified. As can be seen in figure 6-2, the Cu signal steadily increased from the first laser shot up to about the 50 th laser shot where the signal began to level off and became more random in nature Repetition rates up to the maxim11m JX>ssible rate of2 Hz produced similar results to those in figure 6-2. Another way to evaluate the sar11pling rate dependency was to fire any number of laser shots, stop and wait for several minutes and then start sampling again. These measurements resulted in the same temporal development of the signal. A possible cause was the velocity at which the sample was translated during the measurement. To collect the data in figure 6-2 the sample was translated at 30 m s 1 as the laser fired. The sample translational velocity was of interest because at slower speeds sequential laser shots overlapped. For a laser spot size of ~ 900 m and sampling rate of2 Hz, the sample needs to be moved at ,..., 1800 m s 1 for sequential laser shots to sample entirely new material. If the sample is translated at slower speeds, only a fraction of new material will be sampled for sequential laser shots. The idea that the translation speed caused the non-random development was investigated by capturing individual spectra from 500 laser shots at sample movement rates of Oto 50 m s 1 These measurements showed practically no variation in the initial intra-measurement signal development. In a subsequent study 100 laser shots were captured at sample movement rates of Oto 400 m 1 Only 100 shots were captured due to the limitations of sample size and the total distance which the sample stage could travel. These results several of which are

PAGE 187

,> Cl) Q) Cl) I C :::l 0 (.) -,,,, ca C C) en :::l (.) 10-,-----,---,---,----y--,----T-----.--------8 6 4 2 ... ~I ,..,~ I -.. .. .. _,, o_..,. __________ __________________ __,. 0 200 400 600 800 1000 Laser Shot Figure 6-2 Tl1e intra-measurement development of the 427.511 nm line peak height. -.J -..l

PAGE 188

178 shown in figure 6-3, indicate that the velocity at which the sample was moved or the percent of overlap of sequential shots, does cause the growth in emission signal observed over the first ~ 50 shots. The plots in figure 6-3 show that as the translational velocity was increased, the 011mber of shots required to reach a steady-state ( or random) signal was reduced from -45 shots for O m s 1 to ~O shots for 300 m s1 At faster translational speeds of 500 4000 m s 1 10 laser shots were captured. This data continued the same trend as in figure 6-3, in that over the ten shots there was practically no change in the emission signal. This sampling phenomenon has also been observed for other samples such as alumintim alloys for both major and minor components. Dependence of Precision on Sample Movement From the results of the previous section, a dependence of the emission signal on the sample translational velocity became evident. The following set of studies were perfortned to deter1nine the effect of sample movement on precision. The same data presented in the previous section was used. Figure 6-4(a) shows the average Cu emission signal obtained as a function of sample translational velocity, (b) shows the RSD as a function of sample translational velocity for inter-measmements, and ( c) shows the RSD as a function of sample translational velocity for intra-measurements. The top set of plots in figure 6-4 are the result of l 00 laser shots, and the bottom set of plots in figure 6-4 are the result of 10 laser shots. The plot in figure 6-4(a) shows that the lowest average signal was observed at a translational velocity of 100 m s1 and the highest average signal was observed at 2000 m s 1 The high average signal at 2000 m s 1 can be explained by the lack of sampling overlap which is discussed in section entitled, signal intra-measurement development. The decreased signal at 100 m s 1 might have been due to the influence of redeposited material. Ablated material

PAGE 189

Cl) ::, (.) I 10 8 2 . I I I I I . ... . ..... . .... : ............ . ..., ... . ,,, .. .. ... T r anslationa l Veloc i ty : 0 m / s Ov erlap Between Shots : 100 /4 0 4-., --.. -, ,.---......,----,, --.. -..-, -,.--..., ..... -, ..----1 10 8 6 4 2 0 20 4 0 6 0 8 0 I 00 Lase r S h o t ff, . ...... . . . ...... ... . ,u_!.--.""'. ., .... . 1: ,. T r ans l at i onal Velocity : 2 0 m i s O verlap Between Shots : 98 9 /4 0 +--.---.----,-......----,,-,.--...,........--, .--,-.....--..-----1 0 10 8 6 4 2 20 4 0 60 8 0 Translational Velocity : 40 / s Overlap Between Shots : 9 7 8 /4 100 .... ...--:.,, -... .. ........... .,..... . ..... 0 -+-..--,..-.....-......-----,,........... --r----,-..........-----1 0 20 4 0 60 80 100 1 79 Fig ure 6 3 The intra-m eas ur e m e n t d eve l o pment of the 4 2 7 511 nm lin e peak heig ht as a fun c ti on of tr an s lat ion al ve l oc i ty

PAGE 190

10 8 M 0 8 6 ..._,, rn 4 ::s u 2 Translational Velocity : 100 mis Overlap Between Shots : 94 4 /o . .. ... . ,,,. ,, :.-..,. .... _,_ .,,.,, .. .. .. .... : ..... -la 0 -+---~-------.---------....--..-----..----------' 0 10 8 6 4 2 20 40 60 80 Laser Shot Translational Velocity : 200 m/s Overlap Between Shots : 88 9 /o 100 . .... ... :. -... __ ., _Jll"-8\ .... .. .. ............ ., ...... .. .... . .. . ... 0 4---.----.----....-----------------~----------1 0 20 40 60 80 1 00 Figure 6-3 continued. 180

PAGE 191

M 0 !!l 8 .._ en u 10 8 6 4 2 ... 1 ,,, ,11. .. ,-.. ,,.~ .-:._',.. .._-:. . . ... .. . .. Translational Velocity : 300 mis Overlap Between Shots : 83 3 /o 0 -+--....-----.-----....-----.--...,.....-....-----.---.---....-----4 10 8 6 4 2 0 20 40 60 Lase r Shot . .. . ...... 80 JOO .. ~--,. . ... ._1. ...... ... .. . . . : ...... 0 Y "" Translational Velocity : 400 m/s Overlap Between Shots : 77.8 /o 20 40 60 80 100 Figure 6-3 continued. 181

PAGE 192

182 (a) OJer1ap Be-tv.een S hos ( 0 k) 100 94 4 88 9 llJJ 77 8 8 I l. ....... 7 I 0 6 -l C) .L CIO i Cl.l 5 l C: I 0 _. en en I ~ _._ s w 1 3 0 100 200 3 00 400 Sample Movement ( is ) (b) Over1ap Between Shots(%) 100 9 4 4 88 9 8 3 3 77 8 5 4 i ....... o J o 'Cl Cl.l ex: 2 J O -l---~ -~--,--~--.,---J 0 100 200 300 400 Sample Movement ( mis) (c) 100 Owr1ap Between Soots (%) 94 4 88 9 83 3 I 77 8 28 . Cl 20 I ~ I I -~ 16 -I T T 0 100 200 JOO 400 S ample M ove ment ( i s) Figure 6-4. Results of the dependence of precision on sample roughness study. This set of figures are the result of 100 laser shots, and figures on the following page are the result of 10 laser shots. (a) Plots of the average Cu emission signal as a function of sample translational velocity; (b) Plots of the RSD as a function of sample translational velocity for inter-measurements; ( c) Plots of the RSD as a function of sample translational velocity for intra-measurements.

PAGE 193

183 ( a ) Overlap Between Shots(%) 100 44.4 0 0 0 16 ... I ,,...., !;J 14 j j 0 12 1 0 '-' "' I 5b 1 10 Cl) I ,.. i I 0 I j "' 8 "' e 1 l w I 6 7 I 0 1000 2000 3000 4000 Sample Movement ( is) ( b ) C>J ertap Betweel S hots (%) 100 44 4 0 0 0 I 30 -I c 20 0 Cl) 0:: 10 I I I () I 0 1000 2000 3000 4 000 Sam p l e Movement ( / s) ( c ) O wrtap Betwee1 S hots (%) 100 44 4 0 0 0 20 T IS I "$. T ._, 10 0 tll 0:: 5 0 0 1000 2000 3000 Sam p le Movement ( is) F igure 6-4 c o ntinued

PAGE 194

184 could be carried by the shock wave and then deposited in the region surrounding the original crater. At certain translational velocities, only this redeposited material might be sampled. However the reason why a translational velocity of 100 m s 1 where sequential shots are 94% overlapping (6o/o new material), should be different than a velocity of 350 m s 1 with an overlap of 81 o/o (19o/o new material) and with double the emission signal, is not known. Figure 6-4(b) shows the inter-measurement RSD. Here, we see that the RSD does not fluctuate much over a range of0 -400 m s 1 in translational velocity, which indicates that the observed initial growth of the emission signal at lower velocities was quite reproducible The lower plot in figure 6-4(b) shows that the best obtainable precision (0.03 %) occured when each laser shot sampled 1 OOo/o new material at translational velocities of greater than 1800 m s 1 In figure 6-4( c ), the intra-measurement variance is plotted. The precision was poor due to the observed initial growth in the emission signal at lower translational speeds. However once again the best precision occurs when each shot begins to sample entirely new material at translational velocities greater than 1000 m s1 Precision worsened at translational velocities above 3000 m s 1 The reason for this trend is not presently known. Dependence of Precision on Number of Laser Shots Accumulated Triplicate, 900 shot measurements were made on a stationary sample and on the sample as it was moved at several translational velocities (15, 30 and 60 m s 1 ). The first 100 shots of each m~nt were discarded since the emis.~ion signal continually increased during this interval (as discussed in the titled, signal intra-measurement development). Within one measurement, the individual spectra were accumulated for each 10 shots plus the previous# of shots (i.e. 1-10, 1-20 1-30, etc.). The RSD was then calculated using the accumulated signals from the three sequential measurements. The resulting RSD was plotted

PAGE 195

185 versus the nwnber of accumulated laser shots. Figure 6-5 shows the results for the stationary sample. Up to ~ 325 laser shots, the RSD decreased with the number of shots, and then began to increase indicating two different sources of noise. For the earlier laser shots, a fundamental noise response was observed (N 1 1 2 dependence). The cause for the increasing RSD after 325 laser shots has not been deter111ined When the sample was moved at any of the three velocities the general trend was the same except that the point of minimum RSD occurred at varying points ranging from 50 to 450 shots. In all cases, the RSD eventually began to worsen after that point as shown in figure 6-5. PlJlse to Pulse Stability of the Laser, and its Effect on Emission as a Function of Pulse Ener~ To investigate the precision of the laser pulse and its relation to LIBS measurements the Cu emission signal and the laser pulse energy were simultaneously monitored. A signal proportional to the pulse energy was obtained using a photodiode (UDT, PIN 100) connected to a current to voltage converter (Stanford Research, SR560) and then to a boxcar (Stanford Research, SR250 ) from which the data was captured using an AID interface and Stanford software (v. 3.3). Measurements of 200 shots were captured, with only the last 100 shots being used for data analysis. The sample was only moved between measurements to sample new material. It is important to point out that the average signal intensity decreased from a maximum of ~ 6500 counts at high pulse energies (224 mJ) to a low of ~ I 000 counts at the lowest pulse energy ( 40 m.J). Because the signal decreased, this could have had an effect on the measured precision; however, the average signal values varied over less than an order of magnitude Again, we looked at the data in two different ways to present the precision as a function of laser energy for both intraand inter-measurements. Figure 6-6(a ) presents the RSD for intra-measurement determinations. The circles and triangles represent the RSD of

PAGE 196

7-r---r---~--~----y----r---r-----.--------6 5 .. .... r 4 0 ,,,, 2 1 .. "' 0-+-----------------------------0 200 400 Laser Shot 600 Figure 6-5 RSD vs the number of accumulated laser shots. 800

PAGE 197

(a) 24 20 16 0 O 12 (/) 0::: 8 4 I I Emission Sig 'v Diode Sig ~ I I I ~ (b) 10-,-------r-----,.--~--.-------.----,---.-----. 8 6 0 0 (/) 4 0::: 2 Emission Sig 'v Diode Sig 'v 'v 'v -VV-V-=9'V' VV'V' -V 0-+-------r-----,.------.--...;.__------,.------,---------1 0 -+------,_;___...,..._;lit...-~-----..X-.:!1.---------l 50 100 150 200 250 50 100 150 200 Pulse Energy ( mJ) Pulse Energy (mJ) Figure 6-6. RSD as a fucntion of laser pulse energy. The solid circles are the Cu emission signals (427.511 run line ), and the triangles are the diode signals (proportional to the pulse energy) in both (a) and (b ). (a) intra-measurements results ; (b) inter-measurements results. 250 ,__. 00 -...J

PAGE 198

188 the Cu emission s ignal and the photodiode signal (which is proportional to the laser pulse energy) respectively. The RSD of the Cu emission signal decreased with increasing pulse energy. The RSD of the photodiode signal remained approximately constant (l o/ o) over the range of pulse energies, with a slight increase at lower energy Figure 6-6(b) presents the RSD for inter-measurement determinations. Again the circles and triangles represent the RSD of the Cu emission signal and the photodiode signal, respectively The top curve s hows the Cu emission signal, and the lower curve the photodiode signal. Note that the RSD of the emission signal is larger at high pulse energy, and smaller at low pulse energy, which is opposite to the relationship between RSD and pulse energy noted in the intra-measurements. A reason for this was not evident. Also the photodiode RSD varried over a greater range for the inter-measurements compared to the intra-measurements. However the RSD was still less than 2 %. Another comparison was made between laser pulse variance and emission variance, by comparing the shot-to-shot Cu emission signal to the photodiode signal. No significant correlation was observed, indicating that the laser pulse variance played a small role if any in the overall variance of LIBS measurements. One may conclude that even though the intra measurements bad anRSD of 10 % or greater, the inter-measurements of the same data have RSDs of less than 10 % at all pulse energies, and less than 2 o/o at lower energies. Dependence of Precision on Gate Delay Normally r,ms measurements are gated and delayed with respect to the laser pulse to obtain an optimum signal-to-background noise ratio. This is because the background signal ( and the noise associated with it) decays at a faster rate than the emission signal. Measurements were made to deterrnine the dependence of LIBS signal precision on gate delay (tJ. For all measurements, a fixed gate width (lw) of 15 s was used and the sample

PAGE 199

189 was only moved between measurements. Three measurements of200 individual s pectra were acquired at each delay time and the spectra from the first 100 shots were discarded. Fig ure 6-7 shows the resuhs of these measurements with 6-7(a) representing the inter-measurement and 6-7 ( b ) representing the intra-measurement results. The data in figure 67(a) do not show any trend of RSD with td. Figure 6-7(b) shows the RSD for intra-measurements where a definite trend was evident. The intra"""' mmie:nt data indicated that the precision decreased with increasing t d. Therefore, a spectrum captured at td = 0 (highest background) was the most reproducible. It is also important to note how the signal to background noise ratio (S/BN) changes since the background noise is typically the limiting noise in LIBS measurements. Figure 6-7(c) shows the relationship of the S/BN to t d for inter measurements. The background noise was determined as the standard deviation of a single background pixel between the three measurements. Except for a few outliers the S/BN ratio increased with increasing td up to td = 6 and at longer delays, the S/BN ratio remained practically constant. This indicates that a td of -6 s should be chosen to maximize the signal to background noise ratio. Dependence of Precision on Surface Rouihness As mentioned in the experimental section, the surface of the copper disc was altered using varying grits of sand paper. The surface was prepared with 600 grit sand paper to achieve the smoothest surface to analyze initially. The last few strokes on the sand paper were in only one direction; and then the laser sampling track was generated in a direction perpendicular to the etched marks on the disc surface. After analysis the surface was prepared in the same way with the 400 grit sand paper. This entire procedure was repeated for all 7 different grades of sand paper The 600 grit sand paper produced the smoothest

PAGE 200

(a) 12 10 8 '#. 0 6 en a::: 4 2 0 --+-.....-.....-.....--.--....-...--...--...--..--..--..--...-,;--...-...--,--,---f (b) 22 20 1 8 '#1 6 0 en a::: 1 4 1 2 10 (c) 8 000 z ID 00 4000 0 0 I 0 0 2 T 1 2 4 2 4 6 8 10 12 1 4 1 6 Gate Delay ( ) II I 1 6 8 1 0 12 14 16 Gate Delay ( ) 6 8 10 12 14 16 Gate Delay () 1 90 Figure 67 Precision and the signal to background noise ratio as a function of gate delay. ( a) shows the RSD for inter-measurement; (b) shows the RSD for intra-measurement; ( c) shows the signal to background noise ratio (S/BN) for the inter-measurement.

PAGE 201

191 surface, and the 40 grit sand paper produced the roughest surface. Three measurements of 200 shots were recorded as the sample was translated at 30 To evaluate the precision, only the last 100 shots were used Figure 6-8 shows the results of these measurements with 6-8(a ) and 6-8(b) representing the intraand inter-measurement precision respectively The intra-measurement results indicate that the surface roughness had no effect on the shot to shot precision. This was evident because the RSD only varied over the range from 11 6 13.7 %. The inter-measurement RSD results varied over a much wider range, from 0.4 5 .7 %. However these results did not demonstrate any trend ofRSD with surface roughness, and therefore indicate that the surf.ace roughness has practically no effect on the intraand inter measurement precision. Dependence of Precision on Back~ound Correction Spectral background correction was done using an in-house written MatLab program. Pixels were chosen where there was only background emission (no line emission ) and were assigned to a background set A second order polynomial was then fit to the background set, and the resulting polynomial was subtracted from the raw spectrum to create a background corrected spectrum. The purpose of this study was to determine the effect of this background subtraction method on precision. This was investigated by using one data set and processing it with and without background subtraction for both interand intra-measurements. It would be expected that by performing a background subtraction, the RSD for inter-measurements would be improved (assuming the background is fluctuating). A greater improvement would be expected for intra-measurements than inter-measurements, due to the shot-to-shot fluctuations. For the inter-measurement evaluations the RSD was 8.68% and 8.34% with and with out background subtraction, respectively, and for the intra-measurement evaluations, the

PAGE 202

14 12 -10 a en a:: 8 6 0 (a) (b) 65. 4-0 0 3en a:: 21 0 -.--,--.--,---. ----i--- --. --.--,--.--,-.--,-~ 100 200 300 400 500 600 0 100 200 300 400 500 600 Sand Paper (grit) Sand Paper (grit) Figure 6-8. RSD as a function of sand paper grit (used to roughen the surface). The 600 grit sand paper produced the smoothest surface, and the 40 grit sand paper produced the roughest. ( a) intra-measurements; (b) inter-measurements ......

PAGE 203

193 RSD was 13.7 + 1.5% and 13.1 + 1.5% with and without background subtraction, respectively It was evident that there was no improvement in either the intraor inter measurement precision by using background correction. The reason for this was because very high signal to background ratios were attained with pure materials and therefore the fluctuations in the background had an insignificant effect on the precision of the measurement. Conclusions To optimize measurement precision, it is first necessary to name the goal, that is whether interor intra-measurement precision is to be optimized. We have shown in this work that both are not optimized under the same operating conditions. For inter measurement precision, the following variables should be considered (% RSD ranges indicated inside the brackets): translational velocity (9.5 0.03), laser pulse energy (8.5 1 2), andnumberofspectraaveraged (6.3 1.0). Gate delay, surface roughness, and background correction only have a minimal effect on the inter-measurement precision. For intra measurement precision, translational velocity (29 0.29), laser pulse energy (23 11 ), and gate delay (21 11) should be considered. Surface roughness and background correction also only have a minimal effect on the intra-measurement precision. In the case of pure samples, it will be possible to improve the sampling precision by adjusting some of the variables discussed; but for a heterogenous sample, such as a soil sample, the sampling reproducibility is typically limited by the homogeneity of the sample itself. The best precision reported here is 0 03%

PAGE 204

CHAPTER 7 MATRIX EFFECTS IN LASER-INDUCED BREAKDOWN SPECTROSCOPY Introduction Qt1antitation can be difficult with LIBS because of the variation in the mass removed by each laser shot. If the mass removed is not constant then the total number of atoms in the plasr:na can fluctuate. This fluctuation can have several adverse effects including the addition of another variable into the concentration expression, as well as changing the distribution of atoms in excited states due to the variation in the number of atoms ''sharing'' the plasma energy. The variation in mass removed can be caused by one or more of the following: laser pulse stability, change in laser focusing lens to sample distance, sample compositio' sample phase, and sample rigidity. With the new diode pumped, Q-switched Nd:YAG lasers variation due to pulse stability can be eliminated. Also there are ways to keep the lens to sample distance constant, as discussed in chapter 4. This leaves the last three possible causes of variation in mass removal to be addressed. In this chapter, two of these three, sample composition and rigidity will be investigated along with possible solutions. (Sample phase was not investigated because the scope of this dissertation only includes solids.) Backiround Matrix effects in laser ablation have been reported in the literature several times [ 1741 78]. Our focus in this background section is summarize what has been published in the area of matrix effects concerning LIBS. Eppler et al reported on matrix and chemical speciation effects in LIBS on sand and soil [171]. In the bulk matrix, studies they investigated the 194

PAGE 205

195 matrix effects observed when analyzing sand and soil matrices for both barium and lead. They found that the absolute emission signal varied from one matrix to the others, as well as the slopes of the respective calibration curves. One way to correct for this was by using an internal standard. It was found that Ba(I) / Si(I) Pb(I) / Cr(I) and Pb(I)/Fe(I) provided matrix independent results whereas the ratio of Ba(II) / C(I) did not. The authors explained the lack of improvement with the Ba(II) / C(I) by associating the ion/atom ratio with the varying plasma electron density. Possible causes of the matrix effects were investigated including aerosol production, sample absorptivity, and sample composition. The only factor that could be related to the change in electron density was the sample composition since the soil contained appreciable concentrations of more easily ionized elements such as Al, Fe, Na, and K. However, the exact results of the matrix effects eluded the authors. In their speciation studies, they investigated the effect of change in the chemical composition on the emission signal for oxides, carbonates, sulfates, chlorides, and nitrates of Ba and Pb. The final result was that speciation effects were observed with carbonate and oxide giving the highest signals for Ba and Pb respectively and the nitrate providing the lowest. The authors failed to correlate the observed signals, for the different compounds, to any of the following physical properties: enthalpy of formation, enthalpy of vaporization, Gibbs enthalpy of formation, entropy, enthalpy of fusion, heat capacity, density and molar volume. Ko et al. reported on the use of interrial standardization as a way to correct for matrix effects in LIBS [179]. Their studies included binary matrices of Fe / Cr and Zn/Cu, and the LIBS measurements were perfortned in a chamber at reduced pressure ( 140 mbar) with argon as a buffer gas. They found in their investigations into the Fe / Cr matrix that the plasma temperature was dependent on concentration ratio. The sample with the highest amount of iron (Fe / Cr= 100 / 0) produced

PAGE 206

196 the highest temperature, as well as the lowest amount of material ablated (30 ng per shot), and that the sample with the lowest amount of iron (Fe / Cr= 10 / 90) produced the lowest temperature, as well as the highest amount of material ablated (45 ng per shot) This makes sense considering that if there is more material in the plasma sharing a fixed amount of energy, then the temperature will be lower. They found that by using the ratio of a Fe / Cr pair of emission lines (449.46 nm/450.03 run), they were able to obtain a linear calibration curve at any delay time. The resuhs for the analysis of the Zn/Cu matrix were not the same. From the plot of the temporal development of the zinc (468.01 nm) and the copper (453.08 run ) lines, it was deter1nined that the responses of the two elements were different. The zinc and copper signals reached maxima at about 2 and 5 s, respectively. It wasn't until about 16 s that both signals exhibited the same decay. Based on these findings, they repeated their analysis of Zn/Cu alloys at a delay time of24 and obtained a linear calibration curve. The authors attributed the variation in temporal developments to the differences in vapor pressures between zinc and copper. Since zinc has the lower vapor pressure, it vaporized more quickly than the copper and thus fractional evaporation occurred. Chaleard et al also investigated LIBS matrix effects and reported on a possible correction scheme consisting of acoustic and temperature sensors [180]. The authors claimed that there were two primary parameters responsible for the observed matrix effects in LIBS vaporized mass and plasma excitation temperature, and that by monitoring both of these a correction function could be deterr11ined. The amount of vaporized mass was found to be proportional to an acoustic signal from a microphone placed near the laser plasma. The resulting correlation plot showed favorable linearity, but very poor sensitivity which ultimately would limit the measurement precision. The plasma excitation temperature was monitored through the use of a ''temperature sensor'',

PAGE 207

197 which was copper for all of the studies presented. The ratio of two copper emission line s ( 5 I 0.55 and 515 32 nm) was used to determine the temperature according to a Boltzrnann distribution. For the detennination of manganese ( 414.14 nm) in steel, aluminum, and nickel matrices matrix effects were not observed; however for the determination of copper (51 0 55 nm) in aluminum, brass, and steel, severe matrix effects were observed. They also found that with the increasing concentration of zinc the copper signal greatly increased. By using their correction function (inclusion of the copper signal, acoustic signal, and temperature ), they were able to produce a linear cahbration plot (R 2 = 0 .9 996). The advantages of their method over the use of an internal standard, is that the concentration of the '' temperature sensor'' doesn't need to be known; however the same '' temperature sensor'' must be present in all s amples. The focus of this chapter will be to examine matrix effects observed with the LIBS benchtop instrument for several different types of matrices including altiminum, copper, graphite, potassium bromide and sand. We also report on the investigations into the use of internal standards (IS) in complex matrices and the determination of the criteria to consider when selecting an element as the IS The last section will cover the results of our studies into the zinc matrix effects already report by Ko et al and Chaleard et al E xperimental The same experimental apparatus described in chapter 4 was used for all measurements reported in this chapter. The laser pulse energy was fixed at 180 mJ and the detector and stage settings were variables that will be indicated in each section. The s amples investigated were either prepared, Apex standards, or NIST standards The prepared samples consisted of two series of samples containing zinc oxide (Mallinckrodt, cat# 8832) or

PAGE 208

198 chromium metal (Alfa AESAR, cat# 10148) at concentrations of approximately 50, 100, 500, and 1000 ppm in aluminum (Alfa AESAR, cat# 10576A) copper (Alfa AESAR, cat# 13990), graphite (Carbone of America, cat# 014145-003), potassium bromide (Alfa AESAR, cat# 39794), and sand (Mallickrodt, cat# 7062) matrices. Proper mixing was accomplished by ball milling each sample for 15 minutes. The sample was then transferred to a pellet press, and pressed into pellets using a pressure of 4500 psi. In the case of the sand and potasium bromide, they were ground in an alumina grinding vial (Spex CertiPrep, Model 8003) for 20 minutes to produce a particle size similar to that of the other matrices ( -50 m). The sand samples also included lOo/o by weight cellulose binder (Spex CertiPrep, cat#3642), which was added before mixing and was required to improve sample rigidity. For the speciation effects studies, two sets of samples were prepared containing various compounds (i.e. carbonates, silicates, oxides, ect.) of Ba and Zn in a sand matrix (Mallickrodt, cat# 7062 ) at a concentration of -750 ppm. In addition, Ni (as sulfate) was added at the same concentration ( ...... 500 ppm) in each sample as the internal standard. For the Ba samples nitrate, carbonate, oxide, chloride, and sulfate compounds were used, and the same were used for the Zn samples with the replacement of the carbonate by a sulfide. {barium oxide (Aldric~ cat# 288497), barium carbonate (J.T Baker, cat# 0950), barium chloride (Fisher, cat# B-31), barium sulfate, (J.T. Baker, cat# 1030), barium nitrate (J.T. Baker, cat# 101B), zinc nitrate (Aldric~ cat# 22873-7), zinc sulfide (Matherson Coleman & Bell, cat# CB844), zinc oxide (Malinckrodt, cat# 8832), zinc chloride (Fisher, cat# 233-500), zinc sulfate (Fisher, cat# Z-68)} In all cases, cellulose binder was added at a concentration of I Oo/o by weight, and pellets were prepared by the method mentioned in the previous paragraph. Similar samples were also prepared in aluminum and copper matrices, however

PAGE 209

199 a homogeneous sample was not attainable, and therefore these results are excluded from the discussion section. The standards that were obtained from Apex and NIST are listed in table 7-1 and 7-2 respectively Results and Discussion Bulle Matrix Effects The two series of prepared samples described in the experimental section consisting of zinc oxide and chromium metal in alwninum, copper graphite, potassium bromide, and sand matrices were analyzed. The settings on the detection system were 2 s and 15 s for the gate delay and gate width, respectively The emission line of 334 50 nm was chosen for the detection of zinc and 425.44 run for the detection of chromium. Each measurement consisted of 50 laser shots, and was repeated in triplicate on each sample. All samples were translated at 100 m 1 during the analysis. The calibration plots shown in figure 7-1 were constructed after analysis of the samples, parts (a) and (b) represent the zinc and chromium results respectively. The zinc results exclude the potassium bromide curve due to a spectral interference prohibiting the detection of the zinc line. It can be seen from figure 7-l ( a) that the zinc sensitivities obtained have the following rank: sand > copper > aluminum > graphite. The sensitivity obtained from the sand matrix was almost an order of magnitude higher than that obtained from the graphite matrix. After analysis of the chromium samples the calibration plot in figure 7-l(b) were prepared. The ranking of the chromium sensitivities was the following: potassium bromide > sand > alwninum > graphite > copper. These results are exactly the same as for the zinc samples, with the exception of the copper. A possible explanation for the change in copper ranking eludes us.

PAGE 210

SRM 1102 1104 1107 1108 1109 1110 1 1 l 1 1112 1113 1114 1115 1116 1117 Cl 123 Table 71. Listing of NIST bass standards. Type Cartridge Brass C Free-Cutting Brass Naval Brass B Naval Brass C Red Brass A Red Brass B Red Brass C G tiding Metal A Gliding Metal B Gliding Metal C Commercial Bronze A Commercial Bronze B Commercial Bronze C Beryllium-Copper Cu Cone. (wto/o) 72.85 61 33 61.21 64.95 82.2 84.59 87.14 93.38 95.03 96.45 87.96 90.37 93.01 97.10 Zn Cone. (wt%) 27.10 35.31 37.34 34.42 17.4 15 20 12.81 6.30 4.80 3.47 11.73 9 44 6.87 0.01 200

PAGE 211

Table 72. Listing of Apex aluminum standards. 8-8 D-33 M-7 R-14 V-14 SM-1 Z-8 AA-3 TZ-3 D-28 SM-9 Z10 SM-10 AA-I S-11 S-4 S-5 Mg 0.076 0.038 0.060 0.87 0.025 1.00 1 .27 0.20 0.035 0.004 0.43 1.03 1 .08 Zn 0.52 0.59 0.51 0.48 0.42 0.50 0.79 3.20 3.52 3 60 3.70 5.39 5.45 Mn 0.40 0.40 0.34 0.92 0.58 0.11 0.26 0.21 0.35 0.59 0.76 0.49 0.295 Si 2.33 8 .5 4 0.52 14 .00 6.20 0.39 0.84 17 .00 3.20 9.66 1 .69 0.69 2.92 Cu 6.95 2.89 11 12 2.05 4 .05 0.97 16 .05 8 .00 3.50 1 .76 3.00 16 41 2.80 Fe 0.80 1.15 1 .2 8 0.63 0.90 0.41 1 .09 1 .77 1 .00 0.98 3.70 0.472 1 .96 Ni 0.20 0.50 0.205 0.97 0.33 0.05 0.53 0.106 0.45 0.43 0.20 0. 191 0.065 Ti 0. I 6 0 055 0.065 0. 16 0.17 0.03 0.17 0.078 0.054 0.033 0.07 0.12 0.055 Cr 0.17 0.047 0.050 0.11 0.18 0.012 0.15 0.10 0.062 0.21 0.38 0.27 0.20 Sn 0.155 0.048 0. I 05 0. J 2 0.28 0.023 0.026 0.12 0. JO 0.30 0.31 0.025 0.26 Pb 0.165 0. J 4 0.11 0.10 0.045 0.02 0.027 0.080 0.30 0.34 0.32 0.023 0.245 Bi Zr Cd 0.099 0.68 0.17 0.043 0.035 0.025 0.055 0.54 0. 123 0.021 0.17 1 11 0.35 0.09 5.90 6.85 10 .90 14 .9 4 0.54 0.50 0 .38 0.55 14 .6 0.45 l .03 2.24 5.70 0.98 2.64 5.75 l .73 0.57 0.119 0.76 0.60 0.10 0.18 0.21 0.03 0.065 0.12 0.11 0.28 0.115 0. 13 0.1 I 0.50 0.026 0.15 0.11 0.38 0.022 0.13 0.12 0.08 0.06 0.015 N 0

PAGE 212

30 25 ,!1 20 C: ::::, 8 1ij 15 C: 0) en CtS Q) a.. rG (/) +J C: ::::, 0 u 10 5 o_._;: 0 40 35 30 25 co 20 C: 0) en 15 CtS Q) a.. 10 '(.) 5 X I 0-1-----= 0 (a) Aluminum Copper Graphite Sand w/Binder I 200 400 600 800 1000 Zn Concentration (ppm) (b) Aluminum Copper Graphite Sand w/Binder KBr 200 400 600 800 1000 Cr Concentration (ppm) Figure 7-1. Calibration plots for the zinc and chromium samples (a) Zn calibration plot; (b) Cr calibration plot. 202

PAGE 213

203 Based on the results presented in the above paragraph, it is evident that matrix effects are observed in laser-induced breakdown spectroscopy. Our next step was to identify the cause of the matrix effects. A s was stated in the introduction, these effects are pr o bably cau s ed by a change in the mass removed or the plasma energy The optimum method for determining a variation in the mass removed from the different matrice s would be t o weigh the samples before and after each analysis to determine if the same amount o f material is analyzed This was impossible to accomplish with our samples for several reasons. F ir s t the s amples were to heavy (> 20 mg ) to weigh on a balance sensitive enough to quantify the s mall change in ma ss. Second, because of the particulate nature of the sample it would be impos s ible to distinguish between material that is actually analyzed by the plasma, and the other fraction of the material that is removed by the shock wave generated by the plasma. Hence the only option remaining was to qualitatively determine the mass removed Two options were available to achieve this, one was to use a profilometer to obtain the s hape and depth o f the laser created sampling track or by evaluating the tracks visually under magnification. Attempts to use the profilometer failed due the large depth of the laser track, s o we were left with the visual option. Figure 7-2 shows the photographs of the laser sampling tracks generated on the 500 ppm Zn in ( a) aluminum, ( b) copper, (c ) graphite ( d ) potasshim bromide and ( e ) sand samples. Visually, it can be seen that the order of sampling track depth was the following: sand > potassium bromide > copper > alumintim graphite. This ordering is in agreement with the ordering found for the sensitivities of Zn in the various matrices. It is important to note that by visual observation we still have not eliminated the possibility of material being removed by the plasrna shock wave. However the fact that there is an agreement between laser track depth and cah'bration sensitivity provides some insight

PAGE 214

204 (a) (b) (c) (d) (e) Figure 7-2. Photographs of the laser sampling tracks on the different matrices. (a) Aluminum; (b) Copper; (c) Graphite; (d) Potassium Bromide; (e) Sand.

PAGE 215

205 The possibility that the plasma energy varied was also investigated as a possible cause for the observed matrix effects. The plasma energy was determined by using lead as the temperature species The same scheme discussed in chapter 3 was used to construct a Boltzmann plot and the data was captured under the same conditions used for the above analysis. The approximately 100 ppm Zn samples were used. It was not possible to determine a temperature for the graphite plasma due to a spectral interference from the CN molecular emission bands The resulting temperatures calculated for the other matrices are listed in table 7-3 The order of potassi11m bromide < sand < copper < aluminum is practically the opposite order of what was observed for the calibration sensitivities, and the mass removed This is what would be expected considering that if more material is present in the plasma the temperature would be lower We also included the inverse log of the Boltzrr1ann intercept for each matrix in table 7-3. Looking back at equation 3-2 in chapter 3 it is evident that the Boltzmann intercept is equal to ln h ct n where n t is the total number density of lead 4 1tZ atoms. Therefore the Boltz1r1ann intercept is proportional to lead number density From table 7-3, we can note this value for the studied matrices followed the order of potassium bromide > sand > copper > alumin11m. Again, this is in agreement with what has been described. Internal Standards As we reported in the introduction section, the use of internal standards in LIBS has been throughly investigated for binary matrices. Our attempt in this section is to investigate the use of internal standards for complex matrices. We approach this from the beginning by looking at the classical characteristics to consider when selecting an internal standard. In Ahrens and Taylor s book they list the following criteria [181]:

PAGE 216

Table 7-3. Temperature and Boltzmann intercepts for the different matrices. Matrix Aluminum Copper Graphite Sand Potassium Bromide Major Considerations: Temperature (K) 11000 10000 -8000 7000 Boltzmann Intercept ( e x) 2.64 e 12 1.59 e 13 -2 90 e 13 3.91 e 14 206 1. If the internal standard is added its concentration in the samples should be low. 2. The IS and analysis element (AE) should have similar rates of volatilization. 3. The IS and AE lines should have similar excited state energies. 4. The IS line should be free of self absorption. 5. Both lines should be present in the same spectral window to be captured simultaneously. Minor Considerations: 6. Both should have similar ionization potentials. 7. If the IS is added, it should be of high purity. 8. When dealing with lighter elements, the IS and AE should have similar atomic weights. At this point our research focused on deterrnining if these standard criteria for choosing an internal standard were a valid guideline to follow when selecting an internal standard for LIBS measurements. To verify this, we attempted to use an IS to correct for matrix effects in complex matrices. The samples chosen were a series of aluminum standards

PAGE 217

207 ( listed in table 7-2 ), that varied greatly in silicon (17.0 0.4 o/ o) copper ( 16 0 1.0 %), zinc ( 14 9 0 4 %), and several other elements. The settings on the detection s ystem were 4 s and 150 s for the gate delay and width, respectively. Each measurement consisted of 50 laser s hots and was repeated four times on each sample. All samples were translated at 100 m s 1 during the analysis Several emission lines from multiple elements were analyzed and cahbration curves were generated. A representative selection is shown in the following pages and the further justification of the IS criteria are explored. Figure 7-3 shows the calibration plots for the c o pper 2 2 9 364 run line with and without the use of the internal st andard aluminum line at 226 92 om. It is obvious that with the IS a better calibration plot was o btained ( correlation coefficient improved from 0 844 to 0.989 ) Figure 7-4 s h o ws the calibration plot for a silicon line 230.31 run, present in the same spectral window with and without the use of the same IS Here we can see that the al\1minum signal was a poor choice as an internal standard (R 2 was degraded from 0.831 to 0. 703). Looking at the criteria of our choices listed in table 7-4 we can see that the copper / aluminum was a g ood selection ( based o n Ahrens and T a y lor's criteria for choosing an IS ); however, the silicon/aluminum was not valid due to the differences in boiling points and excited state energies In another s pectral window, we observed the copper 510.52 om line and figure 7-5 shows the calibration curves obtained for the analysis with and without the use of the IS aluminum 256.80 run line ( observed in the second order ) In this case the aluminum line was a good choice as an internal standard (R 2 was improved from 0 801 to 0 991 ) Figure 7-6 shows the calibration curves resulting from the analysis of a chromium at 257 17 nm with and without the use of the same IS. The data indicates that the al\1minum line was not a good choice as the internal standard in this case (R 2 was only slightly improved from 0.563 to 0.595 ) L o oking at the

PAGE 218

(a) 1 0 _a a I i (/) .-J I 0 6 8 m c: 0 4 I 0) en 0 2 :a: I Cu 229 36 nm 0 0 0 5 10 15 20 Concentration (o/o) (b) 25---------------------(/) :t= 20 C: ::::, ..a ..... ~15 m C: 0) Cl) 10 ::::, <.) E O 5 z Cu229 36 / Al226 92 o_.,__ _________________ ---1 0 5 10 15 20 Concentration ( 0 /o) Figure 7-3. Calibration plots for the internal standardization investigations. ( a) Plot of Cu raw signal; (b) Plot of Cu ratioed to internal standard 208

PAGE 219

0 5 0 4 ...< (J) _. 0 3 0 (.) co c 0 2 0) en 0 1 (a) I I Si 230.31 nm O Q -&.-----------------------t 7 ..-. ti) ~ 6 C :::, 0 ~4 .Q> I en 3 en E2 I '0 ...__ z 1 5 10 15 Concentration ( 0 /o) (b) I I ft=0.703 Si 230.31 / Al 226.92 20 o ....,_ __________________ -----4 0 5 10 Concentration ( 0 /o) 15 20 Figure 7-4. Calibration plots for the internal standardization investigation s. ( a ) Plot o f Si raw signal ; (b) Plot of Si ratioed to internal standard. 20 9

PAGE 220

r-Q) "' C :J 0 C) ro C 0) en (a) 1 2 1 0 0 8 I 0 6 0 4 0 2 R2=0 801 Cu 510 52 nm o o -+---------.....---------------4 0 5 10 15 20 Concentration (o/o) (b) 200 -------------------------"' .... C 150 :J .c L.. co 1 00 0) en :J (.) E 5 0 0 z I R 2 =0 991 Cu 510 52 / Al 256 80 0 -1------------------------t 0 5 10 Concentration ( 0 /o) 15 2 0 Figure 7 -5 C alibration plots f o r the internal standardization in v est ig ation s ( a ) Plot of C u raw s ignal ; (b ) Plot o f C u ratioed to internal standard 2 1 0

PAGE 221

(a) 1 4---,----,-----,----,---~----,------1 2 ,...._ 1 0 Q) "' +J 0 8 8 j 0 6 C C) ci5 0.4 I 0.2 11: ft=0.563 I Cr 257.17 nm 0 0-1----------------,-----.--------t "' 80 :t:= C ::::s .a s...
PAGE 222

212 Table 7-4. Criteria of elements and their respective emission lines from the aluminum studies. E lement A ( run) Cu 229 38 Si 230.3 1 Al* 226 92 Ion S tage I I I 11 203 15394 11 2 5 4 7 84 58802 44166 Boiling Point { C) 2562 3265 2519 Ionization Pot ( eV) 7.726 8 152 5.986 Atomic Weight 63.55 2 8 .09 26.92 Table 7-5. Criteria of elements and their respective emission lines from the alumintim studies. Element C u C r Al* A ( run ) 5 10 52 257. 17 256. 80 Ion Stage I I I 11203 8095 0 307 84 46968 38929 Boiling Point (C) 25 6 2 2671 2519 Ionization Pot. ( eV) 7 726 6 766 5.986 Atomic Weight 63.55 52.00 26.92 specifications of the selected lines listed in table 7-5, it is evident that based on the criteria given by Ahrens and Taylor, the aluminum line should have been a good choice for an internal standard. This points out that there might be another criteria to be considered when selecting an IS in LIBS measurements. To investigate this further we measured the temporal development of the Cu, C r and Al emission signals. To do this we used a fixed gate width of 1 sand delayed the gate over the range of 1 to 80 s with respect to the laser pulse. The results of this investigation are shown in figure 7-7. It is evident from this study that Cu and Al signals have almost exactly the same temporal behavior, whereas the Cr signal decays at a much faster rate than the IS Al. This indicates that temporal development is another criteria that should be considered when choosing the IS Unfortunately, this can only be deter1nined experimentally.

PAGE 223

1 ...-... I'Q) 0 1 en ...., C 0 (.) C) 0 01 0 ,,,,, m C C) (I) 1 E-3 A .& A 0 A 20 40 Delay {s) Cu 510 518 nm Al 256 798 nm A Cr 257 174 nm 60 80 Figure 7-7. Temporal development of the Cu, Cr, and Al signals from Apex standard AA-1. 213

PAGE 224

214 Speciation Effects Since metals exist as a variety of compounds in soil and organic samples, the effects of chemical speciation on LIBS measurements were examined. Another reason for investigating speciation effects was to determine if it was necessary to match the chemical speciation of the prepared standards to that of the 11nknown. The settings on the detection system were 2 s and 15 s for the gate delay and gate widt~ respectively. Each measurement consisted of 50 laser shots, and was repeated in triplicate on each sample. All samples were translated at 200 m s 1 during the analysis. The same samples were also analyzed by a Finnigan-MAT SOLA laser ablation inductively coupled mass spectrometer (LA -I CP -MS ). The barium 551 91 nm line was detected, and the results from the analysis are listed in table 7-6. The table contains the sensitivity value whic~ in the case of LIBS, is the emission signal divided by the barium concentration, and in the case ofLA-ICP-MS is the ratio of the bari11m sensitivity divided by nickel sensitivity. From the LIBS data, it is evident, that within our range of precision, speciation effects were not observed. However in the case of the LA-ICP-MS results, the oxide sample shows a drastic reduction in the sensitivity. For the zinc sample, the 334.56 run line was chosen, and the results are listed in table 7-7. Again the LIBS results do not indicate any speciation effects, and the LA-ICP-MS results do show a variation in the sensitivity with a change in chemical species. The overall LIBS results indicate no speciation effects, and therefore any chemical species of an element can be used in preparing standards. However, this conclusion might only be valid for sand matrices, since other matrices were not investigated.

PAGE 225

Table 7-6. Barium chemical speciation results (sand matrix). LIBS LA-ICP-MS Compound Ba(I) Ba/Ni Sensitivity ( e6) Sensitivity ( e-2) Ba(N0 3 ) 2 11.1 0.3 101 3 BaC0 3 10.0 0.2 88 4 BaO 11.3 0 6 49 4 BaC1 2 11.9 0.2 90 5 9 60 + 0.15 90 Table 7-7. Zinc chemical speciation results (sand matrix). Zinc Matrix Effect Compound Zn(N0 3 ) 2 ZnS ZnO ZnC1 2 ZnS0 4 LIBS LA-ICP-MS Zn(I) Sensitivity ( e6) 15.4 1.2 15.1 1.5 15.1 0.9 15.2 0.5 14.8 0.1 Zn/Ni Sensitivity ( e-2) 24.1.7 29.1 0.8 23.4 3 19.8 0.8 28.9 0.9 215 Zinc matrix effects have already been reported by Ko et al. and Chaleard et al Our attempt in this section is investigate a larger number of brass samples (14) that cover a wider concentration range ofCu/Zn(97.1 / 0.01 -61.2 / 37.3). We also report on the temperature and number density investigations of the plasnl8 generated on the brass samples. The settings for

PAGE 226

216 the detection system were 4 s and 150 s for the gate delay and gate width, respectively. Each measurement consisted of 50 laser shots, and was repeated four times on each sample. All samples were translated at 100 m s 1 during the analysis. A copper ion emission line at 212.60 nm was measured and plotted as a function of concentration in figure 7-8(a), and a zinc atom line at 213.86 nm was plotted in figure 7-8(b). Figure 7-8(a) shows a very interesting trend where the signals from the lower copper concentration (high zinc concentration) samples anomalously increased. This was also observed for other copper ion and atom lines and is similar to the trend that has been reported by others [ 179-180]. However, the zinc data produces a normal cahbration curve with a good linear fit. To gain an insight into the cause of the increased copper signal, we investigated the plac;ma temperature and number density. This was accomplished by using a series of copper lines (the same series used in chapter 4) to construct a Boltzt11ann plot. Once again the copper number density was assumed to be proportional to the Boltzmann intercept. The resulting temperature and Bohzmann intercept data are plotted in figures 7-9 (a) and (b) respectively as a function of zinc concentration Based on the scatter of the points, and the fact that both the temperature and Boltzrnann intercept only vary over a limited range, it was determined that these results are inconclusive. Our future plans will include the investigation of mass removal as soon as small, thin chips of each brass sample can be machined. These results will prove or disprove the theory that the presence of the zinc (low boiling point substance) enhance s the ablation of copper. We are also investigating other binary matrices of copper and low boiling point metals such as cadmium and rubidium.

PAGE 227

6 0 5 5 (0 <1> 5 0 U) ... 4 5 0 0 -4 0 cu 3 5 en (a) ::, 3 0 2 u R =0.61339 2 5 2 0 Cu 212.60 nm 1 6 1 4 (1) U) c 1 2 ::, 0 ~ 1 0 cu C .Q> 0 8 en C N 0 6 0 4 60 0 ... 70 80 90 100 Cu Concentration (wt 0 /o) (b) I :t: ft= 0.98685 Zn 213 86 nm 10 20 30 40 Zn Concentration (wt 0 /o) Figure 7-8. Ca libration plots from the analysis of the NIST brass standards. (a) Copper calibration plot ; (b) Zinc calibration plot. 217

PAGE 228

(a) 76507600--7550Q) .... ... ro 7500.... Q) a. E Q) 7450 t7400ti) ... C 1 40E+014 .ri 1 20E+014 .... ro ... a. Q) u .... Q) ... C C 1 00E+0148 00E+013 E 0 t 6 00E+013 0 10 20 30 40 Zn Concentration (wt 0 /4) (b) I I T 1 I I 0 10 20 30 40 Zn Concentration (wt 0 /4) Figure 7-9. Plasma temperature and Boltm1ann intercept plots as a function of Zn concentration. ( a) Plasma temperature plot; (b) Boltzrnann intercept plot. 218

PAGE 229

219 Conclusions Matrix effects are present in LIBS measurements, and for that reason there is difficulty in using one set of standards for all matrices. The way to compensate for the matrix effects is to use matrix matched samples ( as descnbed in chapter 4 ), or through the use of an internal standard. An internal standard can be s elected base upon standard criteria, h o wever experimental verification is required to assure that the obtained results will be free of matrix effects This is required due to the differences in temporal decay of the different element emission lines. Finally the proof of the zinc matrix effect is presented and a theory for its cause is being investigated.

PAGE 230

CHAPTERS ANALYSIS OF ORES: PROGRESS TOW ARDS DEVELOPMENT OF A PROCESS MONITOR Introduction In this chapter, we report on the progress towards meeting our ultimate goal of using a LIBS instrument for on-line monitoring of a process stream at a mining facility The LIBS literature contains only two examples of the analysis of ore minerals. Grant et al. published two papers in 1990 and 1991 concerning their work with Australian iron ores [165-166]. In their first paper they dete1tnined the suitable detection parameters for Fe, Si, Mg, Ca, and Ti in an iron ore sample. This was accomplished by measuring the temporal development and decay of the individual emission lines. They found that in order to achieve the best signal-to no ise ratio, a gate delay of 2 -3 s with a detection width of 1 s was optimum. In their second publication, they reported on the qtJantitative results obtained for Si (390.55 nm), Mg ( 518.36 run), Ca ( 431.86 nm), Ti ( 498.17 nm), and Al (396.15 nm) in a series of iron ore samples. They acquired data according to the optimum detection parameters described in their first publication. In all cases, an iron line at 389.56 nm was used as an internal standard. Because their detection system was only single channe~ each element (including the internal standard) was detected sequentially in time. This lead to their somewhat poorer precision (3 26 % RSD). Linear calibration curves were obtained for CaO and MgO over the concentration range of 0.01 -10.8 %, and 0.005 1.47 % respectively. However the 220

PAGE 231

221 cahbration curves for Si0 2 Al 2 0 3 and Ti0 2 were found to be curved due to self-absorption, following a second order fit. In this chapter, we report our initial results from the analysis of phosphate rock and iron ore minerals Even though our initial results are promising, much research is still required to place a LIBS instrument on-line. Both an iron ore mining facility and a phosphate rock mining facility were visited so that we could gain a better understanding of how and where a LIBS instrument would fit into their process plan The processing of the two minerals are very similar, and for that reason only a brief overview of the iron ore processing scheme is described here. The raw iron ore containing approximately 35o/o iron oxide is removed from the ground in open pit mines. The -9 inch diameter chunks of rron ore are then transported to the main processing facility. At this point the size of the ore chunks must be reduced. This is accomplished through a series of grinding steps. After grinding, the ore is a fine powder of ~ 100 m diameter particles The iron particles must then be separated from the waste or tailings. The process by which this occurs is dependent on the type of ore. If magnetite is being processed, then a magnetic separation is used however for hematite a froth flotation scheme must be used. After separation, the concentrated ore slurry must be filtered and dried. It is then mixed with a binder and rolled into 0.5 inch diameters balls. These ''green'' balls are fired in a furnace to create a pelletized product that is suitable for shipping to a steel mill. Since the profit margin is so smaii the most efficient separation steps are required. Because the characteristics of the ore are always changing, the separation steps need to be constantly adjusted

PAGE 232

222 Experimental The same experimental apparatus described in chapter 4 was used for all measurements reported in this chapter. The laser pulse energy was fixed at 180 mJ, and the detector and stage settings were variables that will be indicated in each section. The phosphate rock samples were obtained from NIST and CF Industries (Wauchula, FL) and the iron ore samples were obtained from the Cleveland Cliffs Mining Company ( Marquette Ml). In all cases the samples were used as received, which was as a fine powder with a typical particle size of 100 m. They were mixed with 10 % (by weight) cellulose binder and pressed into pellets following the same procedure described in detail in the proceeding chapters. Standard samples of phosphate minerals were prepared by dilution of NIST Florida phosphate rock standard ( SRM 120) with calcium oxide (Alfa AESAR, Cat# 10684). Calcium oxide was used as the dilutant because it was the constituent of highest concentration in this reference material, therefore resuhing in the least modification of the matrix ( i.e. matrix matched standards ). The final concentration ratios were: (Binder/FL phosphate rock/calcium oxide) 10 / 90 / 0 10 / 72 / 18, 10 / 59 / 31 10 / 45 / 45 10 / 22 / 68. Results and Discussion Phosphate Rock Analysis For the analysis of the phosphate rock samples, the settings on the detection system were 2 sand 15 s for the gate delay and width, respectively. Each measurement consisted of 50 laser shots and was repeated in triplicate on each sample. All samples were translated at 200 m s 1 during the measurement. After analysis of the NIST Florida phosphate rock over the spectral range from 200 to 600 nm, the spectral window shown in figure 8-1 was selected. Within this window the emission lines of Si, Fe, P, Al, and Mn could be

PAGE 233

5 ,-----,i------,,----r---.-----,---------------------p 4 Si Si ..r" 0 3 p
PAGE 234

224 simultaneously monitored. All five of the prepared standards and four unknowns obtained from CF Industries were analyzed in this window Calibration curves were prepared using the 254 60 nm line for iron, 255.49 nm line for phosphorus, 256 80 nm line for alumin~ and 252.41 nm line for silicon, and are shown in figure 8-2. Good signal-to-noise was obtained for all concentration levels encountered which is reflected in the good precision and near unity correlation coefficients for the linear regression fits. These calibration curves were used to determine the concentrations of the four elements in the CF Industries unknown samples. Table 8-1 shows the results of the analysis of the four samples ~ identified as #5 #17, A, and B. The tabulated 'Act.' is the concentration value ( w / w o/o) reported by CF Industries ( obtained from dissolution followed by laboratory ICP emission spectroscopy measurements). In general, there is good agreement between these results and those obtained from the LIBS measurements. In the case of the Si (Si0 2 ) measurements, we compare our results to the '' Insolubles '' determined by CF Industries. It is to be expected that some fraction of the insoluble component of these samples would be composed of material other than Si0 2 Iron Ore Analysis For the analysis of the iron ore samples the settings on the detection system were 4 s and 100 s for the gate delay and width, respectively. Each measurement consisted of 50 laser shots, and was repeated in triplicate on each sample. All samples were translated at 200 m s 1 during the measurement. After analysis of one of the Cleveland Cliffs Mining Company ( CCMC) samples over the spectral range from 200 to 500 mn, the spectral window shown in figure 8-3 was selected. This window was selected because it contained elemental lines from both iron and manganese, the two elements that were of major interest to CCMC. Calibration curves were prepared using the 406.80 nm line for iron and the 404.88 nm line

PAGE 235

16 1 4 12 T (0 Q) l J!? C 10 :, 0 0 8 co C R 2 =0 99593 ,,,. 0) ~ / (f) 6 4 -t--,----,,------.----r---r---.----r----r-......---~ 5 10 15 20 25 30 P 2 0 5 Cone ( wt. %) 14.---------------------, 12 / cg (/) -, 1 0 C ~ 1 :, 8 / co C 8 I 0) R 2 =0 99785 (f) 6 2 4 6 8 10 Si0 2 Co n e (wt 0 /o) 7-.----------------6 (0 5 Q) (/) C :, 8 4 co C 0) (f) 3 R 2 =0 99324 t 2 ---r----r-----,-----,.---..----.-----,----,-----,.--J 0 4 0 8 1 2 1 6 2 0 8._ (0 6 Q) (/) C .! :, 0 0 -. 4t co C 0) (f) 2l 0 4 I 0 8 .-!. I 1 2 Al 2 0 3 Co n e (wt. %) R 2 =0 97713 I 1 6 2 0 Figure 8 -2 Calibration plots generated from the analysis of the pho s phate rock standards.

PAGE 236

Table 8-1. Quantitative results obtained from the analysis of the CF Industries phosphate rock samples. Sample #5 #17 A B Act. 25 2 16.2 29.2 29 4 Exp. 31.1 1 9 16.8 1.4 36 2 6.7 35 0 5.5 Act. 1.5 2.4 1.2 0.85 Concentration (w / w %) Exp. Act. Exp lnsol Exp. 1 .18 0. 09 3. 2 1 95 0.07 22.5 9 51 0 64 1 10 0. 07 3 2 1.65 0.07 27 .6 8 24 0 63 1.38 0.18 1 4 1 40 0.14 12 1 8 28 1 38 1 05 0.13 1 7 1 41.17 12 6 7.95 1 01

PAGE 237

,-, (0 Q) I C: ::J 0 0 ,,,, co C: C) en 50 Fe Fe 40 Fe 30 Mn 20 Mn Mn Mn 10 Fe 0 ______ _,._ ___ ..._ __ _____.i.....________________ ____. 404 405 406 407 Wavelength (nm) Figure 8-3. LIBS s p ectrum obtained from the analysis of a Cleve land Cliffs Mining Company iron ore sample. N N -.:a

PAGE 238

228 for manganese, and are shown in figure 8-4. In both cases, the fit to the experimental data exhibited curvature. This is most likely due to self-absorption caused by the large number density of atoms in the plasma for such high iron and manganese samples. Our next step was to investigate several different spectral windows in an attempt to generate linear calibration curves. Figure 8-5 shows the best (from a linearity perspective) results obtained for manganese (222.18 nm line) and iron (284.40 nm line) We are now in the process of obtaining a larger set of samples and standards from CCMC which will be analyzed at these optimized wavelengths. Future Directions At this point we have determined the optimtim detection parameters for analyzing both phosphate rock and iron ore minerals for several elements. These parameters will be used as a basis for the analysis of various types of ''real word'' samples, including loose particulate samples, damp powders, and slurries. For each sample type, optimum analysis conditions will be determined in the laboratory. Once all of these studies are complete, then one of the benchtop LIBS instruments will placed at a mining facility where it will be evaluated for process control

PAGE 239

(a) (b) 4 0 1 6 Fe 406 7977 nm .... Mn 404 876 nm 3 5 _,,..,.1 4 ,.. ------! 3 0 ~ 1 2 ,.. (0 ,._ Cl) 2 5 Cl) 1 0 !. "' / "' C: ..t: C: 0 8 :::J 2 0 :::, 0 0 0 0 1 5 R 2 = 0.99979 0 6 -! CV J: CV C: C: .. R 2 = 0 99269 C) 1 0 ,. C) 0 4 en en 0 5 0 2 0 0 0 0 l I I I I I 0 10 20 30 40 50 60 0 0 0 5 1 0 1 5 2 0 2 5 3 0 Concentration ( 0 /o) Concentration (o/o) Figure 8-4 Calibration plots generated from the analysis of the Cleveland Cliffs Mining Company iron ore samples

PAGE 240

4 o Mn 222 184 nm 3 5 3 0 en c 2 5 ::, 8 2 0 a, 6, 1 5 (J) 1 0 0 5 (a) R2 = 0 9932 0 0 ___.__....._____._ _._____,__......____._-'-_....---L.--'--___J 0 5 1 0 1 5 2 0 2 5 3 0 0 0 Mn Concentration (wt. o/o) 2 0 Fe 284 3977 nm 1 8 ;::::-1 6 Q) en 1 4 C 5 1 2 0 00 1 0 C Q> 0 8 (J) 0 6 0 4 (b) I R2 = 0 99849 20 30 40 50 0 2._____.,_ _,____._ _.___,__..,_____.__..J......-.&.-...._____,_____, 60 0 10 Fe Concentration (wt 0 /o) Figure 8-5. Calibration plots generated from the analysis of the Cleveland Cliffs Mining Company iron ore samples

PAGE 241

CHAPTER9 CONCLUSIONS AND LIBS has proven to be a very useful technique with a few minor flaws. Examples have been presented where quantitative results were obtained for the detection of several elements in alloy s oil paint ore and organic samples. It has also been shown that LIBS is u s eful in perfor1ning depth profiling analysis of layered samples. However the presence of matrix effects and at times poor precision, indicate that there is a need for further research If laser-induced breakdown spectro s copy is to become a valuable tool for elemental analy s is both in the laboratory and the field then the focus of future research should be to develop efficient multi-component data analysis routines which take advantage o f the wealth of s pectral information acquired from multichannel detection systems In atomic emi s sion spectrochemical analysis the enormous fundamental informing power o f the emission technique is typi c ally wasted by the selection of a single elemental atomic or ionic transition for quantitative measurement. Traditionally this approach, largely dictated by limited data acquisition capabilities and single channel detection system, has been adequate for reliable elemental detenninations in most conventional, continuous plasma sources Nevertheless for most elements especially in highly energetic plasmas such as the laser-induced plasma, numerous atomic and ionic transitions can be observed. For some elements thousands of s pectroscopic line s can be observed over the entire uv-visible region. These lines can be used as qualitative and qliantitative indicators of specific elements; however as a group they carry infonnation about the total number density of plasn,a species and total plasma energy content 231

PAGE 242

23 2 When taken as a whole the complete elemental composition of the plasma ( and thus the s ample which was analyzed ) can be detennined ( i e s tandardless analysis ) In the past the acquisition and manipulation of such large quantities of data was liniroaginable e s pecially on a time s cale o f a few s which is required for LIBS systems acquiring data at repetition frequencies of hundreds ofHz. Modem echelle spectrometers coupled with ICCD detector s can provide simultaneously complete spectral coverage over the entire uv-visible region with high spectral resolution ( ~ 20 pm). For effective application to LIBS the resulting spectra ( ~ 1 MB file size ) must be captured on a s hot-tos hot basis and rapidly evaluated to provide the necessary spectroscopic information to achieve correction for plasma influences which are caused by variati o ns in the matrix With the currently available computing power data files of this size can be acquired and manipulated at the necessary speed. The technology now exists to develop such a LIBS instrument which will take full advantage of the enormous informing power of the emission technique and more importantly use this information t o s olve calibration and matrix-related problems which have up to now plagued the LIBS technique.

PAGE 243

APPENDIX LIST OF PRESENTATIONS AND PUBLICATIONS Presentations ''Variables that Effect the Sampling of Solids by Laser-Induced Breakdown Spectroscopy," The 24 r d Annual Conference of the Federation of Analytical Chemistry and Spectroscopy Societies, Providence, RI, October, 1997. ''Benchtop Laser-Induced Breakdown Spectrometer for Elemental Dett:rtninations," The 1997 Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Atlanta, GA, March, 1997. ''Laser Induced Breakdown Spectroscopy for Environmental and Process Monitoring," The 23 rd Annual Conference of the Federation of Analytical Chemistry and Spectroscopy Societies, Kansas City, MO, October, 1996. Publications and Manuscripts K.L. Riter, 0.1. Matveev, B.C. Castle, B. W. Smith, and J.D. Winefordner, "A Novel Two Step Excitation Scheme for the Analysis of Lead by Laser Enhanced Ionization in a Flame," Proceedings of RIS 96, AIP Press, 431-434, 1997. B.C. Castle, K. Visser, B.W Smith, and J.D. Wmefordner, ''Spatial and Temporal Dependence of Lead Emission in Laser-Induced Breakdown Spectroscopy," Applied Spectroscopy 51 7 1017-1024, 1997 D.A. Rusak, B.C. Castle, B.W. Smith, and J.D. Winefordner, 'Fundamentals and Applications of Laser-Induced Breakdown Spectroscopy," Critical Reviews in Analytical Chemistry, 27, 4, 257-290, 1997. B.C. Castle, K. Visser, B. W. Smith, and J.D. Winefordner, ''Level Populations in a Laser Induced Plasma On a Lead Target," Spectrochimica Acta Part B, 52, 1995, 1997 D.A. Rusak, B.C. Castle, B.W. Smith, and J.D. Winefordner, "Excitational, Vibrational, and Rotational T erm: er:ati~s in Nd:YAG and XeCl Laser-Induced Plasmas," Spectrochimica Acta Part B 52, 1929, 1997. B.C. Castle, K. Talabardon, B.W. Smith, and J.D. Winefordner, "Variables Influencing the Precision of Laser-Induced Breakdown Spectroscopy Measurements," Applied Spectroscopy, m press. B.C. Castle AK. Knight, K. Visser, B. W. Smith, J.D. Wmefordner, ''Battery Powered Laser Induced Breakdown Spectrometer for Elemental Detern1inations," Submitted to Journal of Analytical Atomic Spectrometry 233

PAGE 244

REFERENCES 1 Newton, I. '' Opticks, or a Treatise of the Reflections, Refractions, Inflections and Colours of Light," Dover Publications New York 1952 (1 st ed. 1704). 2. Kirchhoff, G and Bunsen, R. Phil. Mag ., 20, 89 1860. 3 Gerlach, W and Schweitzer E., ''Foundations and Methods of Chemical Analysis by the Emission Spectrum," Adam Hilger London, 1931. 4. Mavrodineanu, R. and Boite~ H., ''Flame Spectroscopy," John Wiley & Sons, New York, 1965 5 Boumans P W ., ''Theory of Spectrochemical Excitation,'' Asam Hilger Bristol, England, 1966. 6. Marcus, K., '' Glow Discharge Spectroscopies," Plenum Press, New York 1993. 7 Ingle J.D and Crouch, S.R. ''Spectrochemical Analysis," Prentice Hall, New Jersey 1988. 8. Brech, F. Appl. Spectrosc ., 16, 59, 1962. 9. Radziemski, L.J. and Cremers, D.A., Eds., "Laser-Induced Plasma and Applications ," Marcel Dekker, New York, 1989. 10. Moenke-Blankenberg, L., ''Laser Microanalysis," John Wiley New York 1989 11. Darke, S. and Tyson, J.F., J.Anal .A t Spectrom ., 8, 145 1993. 12. Radziemski L.J., Microchem J., 50,218, 1994. 13. Minko, L.Y. and Chivel, Y.A., J. Opt Technol., 63, 154, 1996. 14. Boardman, A.D., Cresswell, B., and Anderson, J. Appl Surf Sci. 96-98, 55 1996. 15. Lenk, A ., Witke, T., and Granse, G. Appl. Surf Sci ., 96-98, 195 1996. 16. Zapka, W and Tam, A.C., J. Opt. Soc. Am., 71, 1585, 1981. 17. Autin, M., Briand, A. and Mauchien, P., Spectrochim. Acta, 48B, 851 1993. 234

PAGE 245

235 18 Pinnick, R.G., Chyle~ P. Jarzembski, M., Creegan, E. Srivastava, V Fernandez G. Pendleton, J.D., and Biswas, A., Appl Optics 27 987 1988. 19. Simeonsson, J.B., and Miziolek A.W. Appl. Phys B 59 1 1994 20. Balazs L. Gijbels, R. and Vertes A. Anal C hem ., 63 314, 1991. 21. Mao, X.L ., Shannon, M.A ., Fernandez, A.J. and Russo R.E ., Appl Spectros c., 49 l 054 1995 22. Kuzuya, M. Mats11moto H., Takechi, H., and Mikami, 0., Appl Spectrosc ., 47 1659 1993 23 Hwang Z.W. Teng, Y.Y. Li, K.P. and Sneddon, J., Appl Spectrosc. 45 435 1991 24 Jensen, L. C., Langford, S.C. Dickinson, J.T., Addleman, R.S ., Spectrochim. Acta, SOB 1501 1995. 25 Lee Y.I ., Sawan, S.P. Thiem, T.L., Teng, Y.Y. and Sneddon, J. Appl Spectrosc ., 46 436 1992. 26. Singh, J.P ., Zhang H. Yueh, F. and Carney K.P ., Appl Spectrosc. 50 764 1996. 27. Owens M. and Majidi, V. Appl Spectrosc ., 45 1463 1991 28. Autin, M. Briand A. and Mauchien, P. Spectrochim Acta 48B 851 1993. 29 Mason, K.J and Goldberg, J M. Anal C hem. 59 1250, 1987. 30. Mason, K.J and Goldberg, J.M. Appl Spectrosc ., 45 1444 1991. 31. Adrain, R. S and Watson, J. J. Appl Phys. D 17 1915, 1984. 32. Radziemsll L. J. and Cremers, D. A., SPIE, 1318 71 1990. 33 Thiem, T. L. Lee, Y. and Sneddon J. Microchem J. 45, 1 1992 34. Majidi, V. and Joseph, M. R., C ritical Reviews in Analytical C hemistry, 23 143 1992. 35. Ibr~ A and Goddard B. J. Jurnal Fizik Malaysia, 14 43 1993. 36. Noll R. Sattmann, R. and Sturm, V. SPIE, 2248, 50, 1994. 37 Sharp B. L ., Chenery S ., Jowitt R., Fisher A. and Sparkes S. T ., J. Anal. A t Spectrom ., 10 139 1995.

PAGE 246

236 38. Chen, G. and Yeung E S ., A nal C hem ., 60, 2258 1988 39. Wood, 0. R. Silfvast W. T., Tom, H. W. K. Knox, H. Fork, R. L ., Brito-Cruz C. H. Downer M C ., and Maloney P. J ., Appl Phys Lett. 53, 654 1988 40. Coche M. Berthoud, T ., Mauchien, P., and Camus P Appl. Spectrosc. 43 646 1989 41 Iida, Y ., Appl Spectrosc ., 43 229 1989. 42. Marine, W. Gerri, M. d'Aniello, J. M. S., Sentis, M. Delaporte, P. Forestier B. and Fontaine, B. Appl Surf Sci ., 54, 264 1992. 43. Okano A ., Matsuura, A. Y. Hattori, K. and Itoh, N. J. Appl Phys ., 73 3158 1993. 44. Yago H. Furuta, K. Ishikawa, K. and Komura, H. Phys Stat Sol ., 179 223 1993. 45. Kagawa, K ., Kawai, K., M Tani and Kobayashi, T. Appl Spectrosc 48 198 1994. 46. Tambay R. and Tbareja, R. K. Laser C hemistry, 14 225, 1994. 47. Blanco F. Botho B ., and Campos J. Physica Scripta 52, 1995 48 Tasaka, Y. Tanaka, M and U sami, S. Jap J. of Appl Phys ., 34, 1673 1995 49 Thareja R. K. and Dwivedi, R. K., Laser and Particle Beams 13 481, 1995. 50. Hatem, G. Colon, C. and Campos J. Spectrochim Acta, 49A 509 1993. 51. Dreyfus R. W. Kelly R., and Walkup R. E., Nuclear Instruments and Methods in Physics Re s earch, B23 557 1987. 52. Koren, G. and Yeh, J. T. C J. Appl Phys 7 2120 1984 53. Deshmukh, S. and Rothe E.W. J. Appl. Phys 66 1370, 1989. 54. Al-Wazzan, R. A. Lewis C L. S. and Morrow, T. Rev Sci Instrum., 67 85 1996 55. Al-Wazzan, R. A., Hendron, J M., and Morrow T. Appl Surf Sci ., 96-98 170 1996 56. Bulatov, V Xu L. and Schechter I. Anal Chem 68, 2966 1996. 57 Multari, RA., Foster L. E. Cremers, D. A and Ferris M. J. Appl Spectro s c ., 50, 1483 1996 58. Muhari, RA. and Cremers D A IEEE Transactions on Plasma Science, 24 39 1996.

PAGE 247

237 59. Nemet, B. and Kozma, L., Fresenius J Anal C hem., 355, 904, 1996 60 Kurniawan, H. and Kagaw~ K., Appl. Spectosc., 51 304, 1997 61. Mart~ P., Campos, J. and Santander J L. G., J Quant Spectrosc Radial. Transfer 57, 459 1997 62. Granse G. S. Vollmar Lenk A., Rupp, A. and Rohr K. Appl Surf Sci., 96-98, 97 1996. 63 Aguiler~ J. A. Aragon, C., and Campos, J., Appl Spectrosc ., 46, 1382, 1992 64. Hader, W., Tech mitt Krupp Engl. Ed 2, 97 1992. 65. Lorenzen, C. J. Carlhofl: C., Hahn, U., and Jogwich, M., J Anal At. Spectrom 7, l 029, 1992 66. Sabsab~ M ., Cielo, P. G. Boily, S. and Chaker, M., SPIE, 2069, 191, 1993. 67 Thiem, T L. Salter, RH. Gardner, J. A Lee, Y. I., and Sneddon, J ., Appl Spectrosc ., 48, 58 1994. 68 Thiem, T. L. and Wolf, P J. Microchem J, 50, 244, 1994 69 Borthwick, I. S., Ledingham, K. W. D ., and Singbal, R. P ., Spectrochim Acta 47B, 1259, 1992. 70 Allen, T. M ., Kelly, P B., Anderson, J.E. Taylor, T N., and Nogar, S. S ., J A ppl. Phys ., 61 221, 1995. 71 Anderson, D R., McLeod, C. W ., English, T. and Smith, A. T., Appl Spectrosc ., 49 691, 1995 72. Arnold, S. D. and Cremers D. A. AIHA Journal 56, 1180, 1995. 73. Sattmano, R Sturm, V., and Noll, R. J Appl. Phys D, 28 2181, 1995 74. Davies C. M. Telle H. H., and Williams, A. W. Fresenius J Anal Chem., 355 895 1996. 75. Cremers, D A. II, J.E B., and Koskelo, A. C., Appl. Spectrosc ., 49, 857 1995. 76 Cremers, D. A., Appl Spectrosc ., 41 572, 1987. 77. Bescos, B. Castano, J., and Urena, A.G., Laser Chemistry, 16, 75, 1995

PAGE 248

238 78. Gonzalez, A., Oritz, M., and Campos, J., Appl. Spectrosc., 49, 1632, 1995. 79. Hakkanen, H.J. and Korppi-Tommola, J.E. I., Appl. Spectrosc ., 49, 1721, 1995. 80. Sabsabi, M. and Cielo, P., Appl Spectrosc ., 49, 499, 1995. 81. Ciucci, A., Palleschi, V ., Rastelli, S., Barbini, R., Colao, F., Fantoni, R., Palucci, A., Ribezzo, S ., and Steen, H. J. L. v. d., Appl. Phys. B, 63, 185, 1996. 82. Ernst, W. E., Farson, D. F ., and Sames, D. J., Appl. Spectrosc., 50, 306, 1996. 83. Geertsen, C., Lacour, J. L., Mauchien, P., and Pierrard, L., Spectrochim. Acta, 51B, 1403, 1996. 84. Marquardt, B. J., Goode, S. R., and Ange~ S. M., Anal. Chem., 68, 977 1996. 85. Miziolek A. W., Optics and Photonics News, 39, 1996. 86. Radziemski, L. J., Loree, T. R., and Cremers, D. A., Optical Sciences, 39, 1983. 87. Palleschi, V., Arca, G., Ciucci, A., Rastelli, S., and Tognoni, E., IEEE-International Geoscience and Remote Sensing Symposium, 2, 854, 1996. 88. Vadillo, J.M. and Laserna, J. J., Talanta, 43, 1149, 1996. 89. Vadi11o, J. M, Palanco, S., Romero, M. D., and Laserna, J. J., Fresenius J Anal. Chem., 355, 909, 1996. 90. Kim, D. E., Yoo, K. J., Park, H.K., Oh, K. J., and Kim, D. W., Appl. Spectrosc., 51, 22, 1997. 91. Maravelaki, P. V., Zafiropulos, V., Kilikoglou, V., Kalaitzaki, M., and Fotakis, C., Spectrochim. Acta, 52B, 41, 1997. 92. Armstrong, R. L., J Appl. Phys ., 56, 2142, 1984. 93. Alexander, D. Rand Armstrong, J. G., Appl Optics, 26, 533, 1987. 94. Chylek, P. Jarzembski, A., and Chou, N. Y., Appl. Phys. Lett., 49, 1475, 1986. 95. Hsieh, W. F., Eick:maos, J. H., and Chang, R K., J Optical Society of America, 4, 1816, 1987. 96. Biswas, A., Latifi, H., and Radziemski, L. J., Appl. Optics, 27, 2386, 1988.

PAGE 249

239 97. Zheng, J. Hsieh, W. F ., Chen S. and Chang R. K ., Optics L e tters 13 559 1 9 88. 98. Hammer D X. Thomas R J. Noojin, G. D. Rockwell B. A ., Kennedy P. K. and Roach, W P ., IEEE Journal o f Quantum Electronic s, 32 670 1996. 99 Kitamori, T. Yokose K. Suzuki, K. Sawada, T ., and Gohshi, Y ., Japan. J. o f Appl Phy s., 27 L983 1988 100. Vogel A. Busch, S. and Parlitz U. J. Acou s t Soc Am ., 100 148 1996 101. Sacchi C A. J. Opt.teal So c iety of America 8 337 1991. 102 Pinnick R. G ., Biswas, A ., Pendleton, J D. and Armstrong, R. L. Appl Opti cs, 31 No 3 311 1992. 103 Nyga, R. and Neu W ., Optics Letters, 18 747 1993. 104. Feng Q Moloney J V. Newell, A. C. Wright, E. M. Coo~ K. Kennedy P. K. Hammer D X. Rockwell, B A. and T hompson, C. R. IEEE Journal of Quantum Electroni cs, 33 127 1997 105. Cremers, D A. Radziemski, L. J. and Loree T R. Appl Spectro s c ., 38 72 1, 1984. 106. Wachter J R. and Cremers D. A. Appl. Spectrosc ., 41, 1042 1987. 107. Aragon, C., Aguilera, J. A. and Campos J. Appl Spectrosc ., 47 606 1993 108. Stolarski, D J. Hardman, J ., Bramlette C. M. Noojin, G. D. Thomas, R. J ., Rockwell B A., and Roach, W. P ., SPIE 2391, 100 1995. 109. Ito Y. U eki 0. and Nakamura, S. Anal y tica C himica Acta 299 401 1995 110 Nakamura, S. Ito Y ., Sone K., Hiraga, H. and Kaneko K., Anal C hem. 68 2981 1996. 111. Knopp, R Scherbaum, F. J. and Kim, J. I. Fresenius J. Anal Chem ., 355 16 1996. 112. Ho, W F. Ng C. W. and Cheung N. H. Appl. Spectrosc. 51 87 1997. 113. Paksy L. Nemet B. Lengyel, A. Kozma, L., and Czekkel, J. Spectrochim Acta 51B 279 1996. 114 Arca, G. C iucci, A, Palleschi V. Rastelli, S., and Togononi, E. IEEE-International Geoscienc e and Remote Sensing Symposium 2, 856 1996.

PAGE 250

240 115. Radziemski, L. J., Loree, T. R, Cremers, D. A., and Hoffinan, N. M., Anal. Chem 55, 1246, 1983. 116. Eickmans, J H., Hsieh, W. F., and Chang, R. K., Appl. Optics, 26, 3721, 1987. 117. Essien, M., Radziemski, L. J., and Sneddon, J., J. Anal. At Spectrom 3, 985, 1988. 118. Archontaki, H. A. and Crouch, S. R, Appl. Spectrosc., 42, 741, 1988. 119. Ng, K. C., Ayala, N. L., Simeonsson, J.B., and Winefordner, J. D., Analytica Chimica Acta, 269, 123, 1992. 120. Parigger, C. and Lewis, J. W. L., Appl. Phys. Com 12, 163, 1993 121. Poulain, D. E. and Alexander, D.R., Appl. Spectrosc., 49,569, 1995. 122. Raisch, C., Paine, U., and Niessner, R., Technical DigestEuropean Quantum Electronics Conference, 25, 1996. 123. Kumar, V. and Thareja, R K., J. Appl. Phys., 64, 5269, 1988. 124. Parigger, C., Plemmons, D. H., Horn.kohl, J. 0., and Lewis, J. W. L., J. Quant Spectrosc Radiat. Transfer, 52, 707, 1994. 125. Yagi, T. andHuo, Y.,Appl Optics, 35, 3183, 1996. 126. Yale~ S Crosley, D.R., Smith, G. P., and Faris, G. W., Hazardous Waste and Hazardous Materials, 13, 51, 1996. 127. Cremers, D. A. and Radziemski, L. J., Anal. Chem., 55, 1252, 1983. 128. Cremers, D. A. and Radziemski, L. J., Appl Spectrosc., 39, 57, 1985. 129. Sneddon, J., Trends in Anal Chem 7, 222, 1988. 130. Ottesen, D. K., Wang, J.C. F., and Radziemski, L. J., Appl. Spectrosc., 43, 967, 1989. 131. Morris, J.B., Forsch, B. E., and Miziolek, A. W., Appl. Spectrosc., 44, 1040, 1990. 132. Cheng, E. A. P., Fraser, R. D., and Eden, J. G., Appl. Spectrosc., 45, 949, 1991. 133. He, K. X., Hammond, T. D., Wmstead, C. B., Gole, J. L., and Dixon, D. A., J. Chem Phys, 10, 7183, 1991. 134. Joseph, M. R. and Majid~ V., J. Trace and Mictroprobe Techniques, 10,207, 1992.

PAGE 251

241 135. Casini, M. Harith, M.A ., Palleschi, V., Salvetti, A., Singh, D. P., and Vaselli, M., Laser and Particle Beams, 9, 633 1991. 136. Flower, W L., Peng, L. W., Bonin, M. P., French, N. B., Johnsen, H. A. Ottesen, D. K., Renzi, R. F., and Westbrook, L. V., Fuel Processing Technology, 39 277 1994. 137 Lazzari, C., Ro~ M. D., Rastelli, S., Ciucc~ A., Palleschi, V., and Salvett~ A. Laser and Particle Beams 12, 525, 1994. 138 Zhang, H. Singh, J.P ., Yueh, F. Y ., and Cook, R. L., Appl. Spectrosc. 49, 1617, 1995. 139. Nordstrom, R. J., Appl. Spectrosc ., 49, 1490, 1995. 140. Parigger, C., Plemmons D. H., and Lewis, J. W., Appl. Optics, 34, 3325, 1995. 141. Raisch, C ., Niessner, R, Matveev, 0., Panne, U., and Omenetto N. Fresenius J. Anal Chem, 356 21, 1996. 142. Palau, J., Sowinska, M., Varela, M. Summ, P., Esteve, J., Serra, P., Morenza, J.L., and J.A. Miehe Appl Surf Sci., 86, 59, 1995. 143. Vadillo, J.M., Milan, M., and Laserna, J.J., Fresenius J. Anal. Chem., 355 10, 1996. 144. Reader J. Corliss, C.H., Wiese, W.L., and Martin, G.A., '' Wavelengths and transition probabilities for Atoms and Atomic Ions Part I. Wavelengths, Part II Transition probabilities'', Natl. Stand. Ref. Data Ser., Natl. Bur. Stand. (U.S.), 68, 1980. 145. Lide, D.R. Ed., "C RC Handbook of Chemistry and Physics '', CRC Press, Boca Raton, Florida, 1995. 146. Harrison, G.R., ''M.I.T. Wavelength Tables'' The M.I.T. Press, Cambridge, MA, 1969. 147. Boumaos, P.W.J.M., ''Line Coincidence tables for Inductively Coupled Plasma Atomic Emission Spectrometry'' Pergamon Press, Ox.ford, 1980. 148. Sweedler J.V., Ratzlaff, K.L., and Denton, M.B., Eds., ''Charge-Transfer Devices in Spectroscopy'', VCH Publishers, Inc., New York, 1994. 149. Castle, B.C., Visser K., Smith, B.W., and Wmefordner, J.D., Spectrochim. Acta., 52B 1995 1997. 150 Sdorra, W ., and Niemax, K., Spectrochim Acta, 45B, 917, 1990.

PAGE 252

242 151. Simeonsson, J.B. and Miziolek, A.W., Appl. Opt ., 32,939, 1993. 152. Griem,H.R., Phys. Rev ., 131, 1170, 1963. 153. Thome, A.P., ''Spectrophysics," Chapman and Hall, London, 1974. 154. Moore, C.E., Atomic Energy Levels, Vol. ill, Natl. Stand. Ref. Data. Ser. Natl. Bur. Stand. (U.S.), 35, 1958. 155. Russo, R.E. Appl. Spectrosc., 49 14A, 1995. 156. NIST Internet Database, 1997. 157. Kono, A., Hattori, S., Opt Soc. Am ., 72, 5, 1982. 158. Raith, A., Hutton, R.C. and Huneke, J.C., JAnal.At.Spectrom., 8, 867 1993. 159. Yamamoto K. Y., Cremers, D. A., Ferris, M. J., and Foster, L. E., Appl. Spectrosc. 50, 222, 1996. 160. Zayhowski, J.J. Laser Focus World, April, 73, 1996. 161. Belliveau, J., Cadwell, L., Colleman, K., Huwe!, L., and Griffin, H., Appl Spectrosc ., 39, 727, 1985. 162. Lorenzen, C.J., Carlhoff, C. Hahn, U., and Jogwich, M., J Anal. At Spectrometry, 7 1029, 1992. 163. Gonzalez A., Ortiz, M., and Campos, J., Appl Spectrosc., 49, 1632, 1995. 164. Sattmaon, R. and Sturm, R.N.V., J Appl. Physics, 28, 2181, 1995. 165. Grant, K.J., Paul, G.L., and O'Neill, J.A., Appl Spectrosc 44, 1711, 1990. 166. Grant, K.J., Paul, G.L., and O'Neill, J.A., Appl. Spectrosc., 45, 701, 1991. 167. Castle, B.C., Visser, K., Smith, B.W., and Winefordner, J.D., Appl. Spectrosc ., 51, 1017, 1997. 168. Alkemade, C. Th. J., Hollander, Tj., Snelleman, W., and Zeegers, P .J. Th., ''Metal Vapours inflames'' Pergamon Press Inc. New York, 1982, ch. 2. 169. Wachter, J.R., and Cremers, D.A. Appl. Spectrosc., 41, 1042, 1987.

PAGE 253

243 170. WISbum, R Schechter, I. Niessner, R., Schroder, H., and Kompa, K.L., Anal. C hem ., 66, 2964, 1994. 171. Eppler, AS ., Cremers, D.A. Hickmott, D.D., Ferris, M.J., and Koskelo A .C., Appl. Spectrosc. 50, 1175, 1996. 172. Pakhomov, A. V ., Nichols, W., and Borysow, J., Appl. Spectrosc. 50, 880, 1996 173. Striganov A.R. and Sventitskii, N.S., '' Tables of Spectral Lines: Neutral and Ionized Atoms," IFI/Plenum Data Corporation, New York, 1968. 174. Sneddon, J ., Spectroscopy Letters 20, 423, 1987. 175. Chan, W.T. Mao, X.L., and Russo, R.E., Appl Spectrosc ., 46, 1025, 1992. 176. Mitchell, P.O., Sneddon, J ., and Radziemski, L.J. Appl. Spectrosc. 41, 141 1987. 177. Cromwell, E.F. and Arrowsmith, P.,Anal Chem., 67, 131, 1995. 178. Mao, X.L. Chan, W.T., Shannon, M.A., and Russo R.E., J Appl Phys., 74, 4915, 1993. 179. Ko, J.B., Sdorra, W., and Niemax, K., Fresenius Z Anal. Chem., 355, 648, 1989. 180. Cbaleard, C., Mauchien, P. Andre, N. Uebbing, J., Lacour, J.L., and Geertsen, C., J Anal. At. Spectrom., 12, 183 1997. 181. Ahrens, L.H. and Taylor, S.R., ''Spectrochemical Analysis'', Addison-Wesley Publishing, Reading, MA, 1961.

PAGE 254

BIOGRAPHICAL SKETCH Bryan C Castle was born in West Palm Beach, Florida, on February 28, 1971, and was raised in Lake Worth, Florida He is the oldest child in a family that includes his parents, Leighton and S1Jzaone, and brother Brody. Bryan received his Associate of Science degree from Palm Beach Commlmity College in December of 1991. He received his Bachelor of Science degree with Honors from Florida Atlantic University in May of 1994. Bryan entered the University of Florida in the fall of 1994 and joined Graduate Research Professor Jim Wmefordner' s group to work on his doctoral degree in analytical chemistry. 244

PAGE 255

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality, as a dissertation for the degree of Doctor of Philosophy. /JJilrnes D. Winefor~er Chairman raduate Research Professor of Chemistry I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. '""l...::'. CJ .( Ii ~ I <.S Z, Willard W. Harrison Professor of Chemistry I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Robert T. Kennedy Associate Professor of Chemistry I certify that I have read this study and that in my opinion it conforrns to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Samuel 0. Colgate Professor of Chemistry

PAGE 256

I certify that I have read this study and that in my opinion it confor1ns to acceptable standards of scholarly presentation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy HugbA oye Professor of Food Science and Human Nutrition This dissertation was submitted to the Graduate Faculty of the Department of Chemistry in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fu1fi11ment of the requirements for the degree of Doctor of Philosophy. May 1998 Dean, Graduate School

PAGE 257

LD 1780 199g C -S'-.)-.3