Citation
Laser-Induced Breakdown Spectroscopy for the Determination of Carbon in Soil

Material Information

Title:
Laser-Induced Breakdown Spectroscopy for the Determination of Carbon in Soil
Creator:
Breckenridge, Lydia
Place of Publication:
[Gainesville, Fla.]
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (140 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Chemistry
Committee Chair:
Winefordner, James D.
Committee Members:
Yost, Richard A.
Smith, Benjamin W.
Eyler, John R.
Tebbett, Ian R.
Graduation Date:
12/14/2007

Subjects

Subjects / Keywords:
Calibration ( jstor )
Carbon ( jstor )
Graphite ( jstor )
Lasers ( jstor )
Particle size classes ( jstor )
Plasmas ( jstor )
Signals ( jstor )
Soil samples ( jstor )
Soils ( jstor )
Spectrometers ( jstor )
Chemistry -- Dissertations, Academic -- UF
libs, quantification, soil
Genre:
Electronic Thesis or Dissertation
born-digital ( sobekcm )
Chemistry thesis, Ph.D.

Notes

Abstract:
Laser-induced breakdown spectroscopy (LIBS) is a relatively young atomic emission technique that has found great utility in the elemental analyses of a variety of materials. In brief, LIBS is achieved by focusing a high-powered, short-pulse laser onto a sample surface to produce plasma that is rich in electrons, atoms and ions. The atoms and ions in the plasma emit radiation that is characteristic of the elemental composition of the sample. While LIBS has been extensively applied to soil analysis, its use as a means of determining carbon in soil has been lacking, limited to a few proof-of-principle demonstrations by research groups at Los Alamos and Oak Ridge National Laboratories. Since the sequestration of carbon in soil may offer a means of mitigating the effects of global warming, it is becoming increasingly urgent to develop a practical device that can accurately measure terrestrial carbon inventories in order to understand the status of the global carbon cycle. Due to the unique advantages of LIBS, including the potential for in-situ analysis, little-to-no sample preparation, simultaneous multi-element detection and relatively simple instrumentation, the technique has the potential to fulfill the requirements of an in-field carbon monitor. Before a field-based LIBS instrument can be designed there must be a laboratory-based demonstration of the capability to determine carbon in soil. The purpose of our research was to investigate the potential of LIBS for carbon in soil determination with an emphasis on minimal sample preparation and the potential for in-situ analysis. While the quantification of carbon in soil by LIBS was not achieved in this study, an optimal soil sampling regime was established that increased sensitivity and measurement precision. The LIBS instrumentation was optimized specifically for soil samples, including the laser energy, optical configuration and spectrometer delay time. The sample presentation was extensively studied, including the effect of soil grain size on LIBS signal and the advantages of mounting the sample on tape as a thin monolayer rather than as a pressed pellet. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2007.
Local:
Adviser: Winefordner, James D.
Statement of Responsibility:
by Lydia Breckenridge.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Breckenridge, Lydia. 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:
663110681 ( OCLC )
Classification:
LD1780 2007 ( lcc )

Downloads

This item has the following downloads:

edwards_l.pdf

edwards_l_Page_068.txt

edwards_l_Page_090.txt

edwards_l_Page_057.txt

edwards_l_Page_029.txt

edwards_l_Page_081.txt

edwards_l_Page_027.txt

edwards_l_Page_112.txt

edwards_l_Page_111.txt

edwards_l_Page_080.txt

edwards_l_Page_007.txt

edwards_l_pdf.txt

edwards_l_Page_123.txt

edwards_l_Page_055.txt

edwards_l_Page_072.txt

edwards_l_Page_033.txt

edwards_l_Page_134.txt

edwards_l_Page_019.txt

edwards_l_Page_096.txt

edwards_l_Page_076.txt

edwards_l_Page_082.txt

edwards_l_Page_097.txt

edwards_l_Page_100.txt

edwards_l_Page_073.txt

edwards_l_Page_024.txt

edwards_l_Page_133.txt

edwards_l_Page_035.txt

edwards_l_Page_075.txt

edwards_l_Page_008.txt

edwards_l_Page_124.txt

edwards_l_Page_030.txt

edwards_l_Page_136.txt

edwards_l_Page_038.txt

edwards_l_Page_117.txt

edwards_l_Page_089.txt

edwards_l_Page_103.txt

edwards_l_Page_095.txt

edwards_l_Page_115.txt

edwards_l_Page_110.txt

edwards_l_Page_048.txt

edwards_l_Page_049.txt

edwards_l_Page_002.txt

edwards_l_Page_036.txt

edwards_l_Page_050.txt

edwards_l_Page_127.txt

edwards_l_Page_128.txt

edwards_l_Page_010.txt

edwards_l_Page_093.txt

edwards_l_Page_040.txt

edwards_l_Page_118.txt

edwards_l_Page_098.txt

edwards_l_Page_066.txt

edwards_l_Page_126.txt

edwards_l_Page_045.txt

edwards_l_Page_077.txt

edwards_l_Page_114.txt

edwards_l_Page_056.txt

edwards_l_Page_120.txt

edwards_l_Page_060.txt

edwards_l_Page_069.txt

edwards_l_Page_059.txt

edwards_l_Page_104.txt

edwards_l_Page_046.txt

edwards_l_Page_025.txt

edwards_l_Page_032.txt

edwards_l_Page_064.txt

edwards_l_Page_116.txt

edwards_l_Page_091.txt

edwards_l_Page_006.txt

edwards_l_Page_105.txt

edwards_l_Page_084.txt

edwards_l_Page_101.txt

edwards_l_Page_122.txt

edwards_l_Page_028.txt

edwards_l_Page_001.txt

edwards_l_Page_031.txt

edwards_l_Page_044.txt

edwards_l_Page_034.txt

edwards_l_Page_011.txt

edwards_l_Page_129.txt

edwards_l_Page_018.txt

edwards_l_Page_067.txt

edwards_l_Page_070.txt

edwards_l_Page_039.txt

edwards_l_Page_043.txt

edwards_l_Page_061.txt

edwards_l_Page_140.txt

edwards_l_Page_113.txt

edwards_l_Page_130.txt

edwards_l_Page_083.txt

edwards_l_Page_102.txt

edwards_l_Page_138.txt

edwards_l_Page_004.txt

edwards_l_Page_022.txt

edwards_l_Page_014.txt

edwards_l_Page_042.txt

edwards_l_Page_052.txt

edwards_l_Page_005.txt

edwards_l_Page_037.txt

edwards_l_Page_131.txt

edwards_l_Page_026.txt

edwards_l_Page_063.txt

edwards_l_Page_119.txt

edwards_l_Page_086.txt

edwards_l_Page_065.txt

edwards_l_Page_106.txt

edwards_l_Page_132.txt

edwards_l_Page_109.txt

edwards_l_Page_021.txt

edwards_l_Page_139.txt

edwards_l_Page_137.txt

edwards_l_Page_012.txt

edwards_l_Page_085.txt

edwards_l_Page_094.txt

edwards_l_Page_092.txt

edwards_l_Page_003.txt

edwards_l_Page_054.txt

edwards_l_Page_078.txt

edwards_l_Page_017.txt

edwards_l_Page_023.txt

edwards_l_Page_088.txt

edwards_l_Page_135.txt

edwards_l_Page_015.txt

edwards_l_Page_079.txt

edwards_l_Page_041.txt

edwards_l_Page_016.txt

edwards_l_Page_013.txt

edwards_l_Page_051.txt

edwards_l_Page_020.txt

edwards_l_Page_058.txt

edwards_l_Page_107.txt

edwards_l_Page_121.txt

edwards_l_Page_047.txt

edwards_l_Page_108.txt

edwards_l_Page_053.txt

edwards_l_Page_099.txt

edwards_l_Page_071.txt

edwards_l_Page_009.txt

edwards_l_Page_062.txt

edwards_l_Page_087.txt

edwards_l_Page_125.txt

edwards_l_Page_074.txt


Full Text































Figure 4-1. An image of a pressed pellet and soil mounted on a microscope slide.









ACKNOWLEDGMENTS

I extend my deepest gratitude to my advisor, Dr. James D. Winefordner for his

unwavering support and guidance during my four years in his research group. His knowledge

and patience are legendary and have contributed to a research environment that fosters

cooperation and mutual respect. I am incredibly proud to have been a member of his group, and

even more honored to have been one of his last students before his retirement. I have great

appreciation for Dr. Nicolo Omenetto and Dr. Ben Smith, who have nurtured my interest in

applications of spectroscopy and provided countless insightful observations about this research.

I also extend a special thank you to Dr. Igor Gornushkin, who was so patient with me and wrote

data analysis programs that saved me from weeks, if not months, of incredibly tedious and

repetitive data processing. Additionally I thank Dr. Andy Freedman and Dr. Joda Wormhoudt of

Aerodyne Research, Inc, who initially proposed and provided funding for this research.

I thank all the past and present members of the Winefordner/Omenetto/Smith research

group, who have become like a second family to me over the years. I have learned so much from

each of them beyond chemistry and know that life-long friendships have been developed. I

directly thank Mariela Rodriguez for understanding the frustrations involved in LIB S

applications and for her many illuminating ideas, and also Benoit Lauly for his elucidation of

plasma fundamentals. Past group members whom I acknowledge for significantly influencing

my research include Dr. Galan Moore and Dr. Kirby Amponsah-Manager. No list of

acknowledgements would be complete without thanking Jeanne Karably and Lori Clark for

keeping the often forgotten paper-work in line.

I cannot begin to express that gratitude I owe to my family for their love and support

during my time in both undergraduate and graduate school. I owe so much to my Mum, Kay,

who is my inspiration and never lets me forget that anything can be accomplished. I especially










a graphite pellet using an Acton Spectra-Pro-500i spectrometer with an ICCD detector which

allowed for the integration time to be changed. Spectra from the graphite pellet were collected

for 200 ns at increasing delay times. The maximum C247 emission was achieved approximately

200 ns after plasma formation, and steadily decreased thereafter as the plasma cooled (Figure 5-

2). However, these data did not reflect the high continuum emission that was also occurring at

such early intervals in the plasma life time, and even though the C247 emission had substantially

decreased by 1 Cps, earlier research, discussed in Chapter 3, revealed that the S/B was at a

maximum.

It was initially proposed that the CN bands might be used for quantitative analysis of

carbon in soil. However, previous research has indicated that the CN molecule in the LBES

plasma is principally formed through the reaction of a carbon dimer and atmospheric nitrogen:

C2 + N2 2CN

Evidence indicates that the maj ority of the C2 available for this reaction is from the

photodissociation of compounds in the sample that contain carbon double bonds or delocalized

carbon bonds rather than the recombination of carbon atoms in the plasma.90-9 Because of this,

the use of the CN band for quantitative analysis seems relegated to unsaturated organic

compounds and not for carbon in soil, which originates from a variety of organic and inorganic

sources. Besides, inspection of the CN bands in a spectrum of soil showed that due to the high

background in the 300-400 nm spectrometer channel and the multitude of interfering species, the

CN band head at 388 nm was unresolved and barely detected, and the two lower wavelength

bands at 387 and 386 were not resolved at all (Figure 5-3).

The carbon atomic emission line at 247.8 nm was the most distinct line in the graphite

spectrum and was the obvious choice to monitor. Information concerning the energy levels,









Applications of LIBS for Soil Analysis

A great deal of research has concentrated on applying LBES to environmental analysis,

with emphasis on detecting hazardous contaminants in a variety of media. LBES has been

applied to liquids (such as waste water, river and ocean samples), 28-31 ice,33 aerosols (such as

factory effluents),33-36 plants,37-3 bactera,40 paint,41-42 and slags (the residue formed by the

smelting of metallic ores).43 Much research has also been focused on LIBS of soils and

sediments, especially for the detection of metals and quantification of toxic metals. This section

will provide a review of some of the more pertinent research concerning LIB S of soils.

Early work with LIBS of soils focused primarily on determining toxic metals and studies

evaluating the numerous matrix effects that soils present. Eppler et al.44 were able to determine

Ba and Pb in spiked soil samples with detection limits of 42 and 57 ppm respectively, well below

the Los Alamos screening levels of 5400 and 400 ppm. The soil samples were significantly

ground and homogenized before being pressed into pellets and therefore can not be considered

'natural' soil sample. Precision was improved to 4% by use of a cylindrical lens to focus the

laser beam and produce larger ablation areas which decreased the effects of sample homogeneity.

The effect of analyte speciation on the emission signal was also studied, and while it was

determined that the origin of the analyte (for example, nitrate or oxide) affected the signal, this

could not be correlated with any physical property of the originating compound. However, the

absorptivity of the bulk matrix was also found to affect the emission signal and this was well

correlated with increasing plasma electron density which caused significant changes in the

concentrations of ionized species.

Barbini et al.45 found that correcting integrated line intensities of Fe in soil by the plasma

temperature substantially improved correlation between LIBS Fe concentration and ICP Fe

concentration. Such a correction was achieved by constructing a Saha-Boltzmann plot to obtain












35000 ~


30000 Carbon

247.811


25000 I Iron

247.978 282

20000-



15000-
247.68


10000-

247.6 247.7 247.8 247.9 248.0 248.1 248.2

Waveleng th nm




Figure 6-6. A soil spectrum produced using the Acton SpectraPro spectrometer showing the
resolved carbon and iron lines.












Table 5-1. Carbon content of soil samples determined by CHN analysis
Sample Name % C (1st CHN % C (2nd CHN A %C Average (% C)
Analysis) Analysis)
Buffalo 3.271 4.58 1.3 3.9 & 0.1
River 5.029 5.090 0.06 5.1 & 0.1
Estuarine 1.624 1.758 0.13 1.7 & 0.1
Montana 1.730 1.876 0.14 1.8 & 0.1
Al 1.043 1.157 0.11 1.1 & 0.1
A4 0.143 0.128 0.01 0.1 & 0.1
A7 0.093 0.074 0.02 0.1 & 0.1
B1 0.168 0.186 0.02 0.2 & 0.1
B4 0.049 0.068 0.02 0.1 & 0.1
B7 0.056 0.058 0.002 0.1 & 0.1
Cl 1.576 1.524 0.05 1.6 & 0.1
C4 0.085 0.071 0.01 0.1 & 0.1
C7 0.034 0.033 0.001 0 + 0.1
Aerodyne 1.024 1.043 0.02 1.0 + 0.1
S 7.190 7.06 0.13 7.1 & 0.1
SE 2.901 2.741 0.16 2.8 & 0.1
NE 4.091 3.712 0.38 3.9 & 0.1
Average = 0.1








precision is undoubtedly a combination of the great sample heterogeneity and the small mass of

sample analyzed. As such, it is unlikely that dry combustion can yield a highly accurate

measurement of carbon concentration that is representative of the bulk sample. In the absence of

an alternate method, such poor accuracy was tolerated and the average carbon concentration was

used for correlation with LIBS signal. However, as a consequence, there will be some unknown

error associated with the stated carbon concentration of each soil.

LIBS Measurements of Natural Soils

All soil samples were subj ected to LBES using the previously established optimized

sampling regime. Because the soils had not been characterized in terms of texture and particle

size, the surface density of the soil on tape was determined and assumed to be proportional to the

average grain size of the soil. Surface density was determined by weighing the microscope slide

and the adhesive tape before and after the application of the soil layer. The difference between

these two measurements represented the weight of the applied soil sample, which was then

divided by the surface area of the tape (approximately 1.9 cm2). The resulting surface densities

ranged from 3.7 to 38 mg/cm2. The soils that exhibited the greatest surface densities

corresponded to the soils that were composed of the largest particle sizes. Since it had been

previously established that the LBES signal exhibits some dependence on particles size, these

surface densities were used to correct the LIBS C247 signal for soil grain size.

LBES measurements of the soil samples were obtained using the maximum laser energy of

approximately 57 mJ at a repetition rate of 1 Hz. This allowed ample time to manually

maneuver the slide to fresh sample locations. Forty separate locations were sampled, for a total

of 80 laser shots, with two shots used per spot. A gentle air stream was directed over the soil

surface to remove liberated particles. Once the data were collected, the average C247 peak area

was separately calculated for the 40 first shots and the 40 second shots. Figure 6-2 shows the









4X Magnification


10X Magnification


Figure 4-7. Microscope images of three soils to illustrate heterogeneity. The yellow circles
represent the ablation area of the focused laser beam, approximately 0.35 mm2
(Note: the 10X magnification are not the same areas as the 4X magnification areas)


F ~i~ ~d

P~"~~._ ~F J~~










However, quantitative analysis may require some minor preparation, dictated by the type of

sample and the required degree of precision and accuracy.

Micro-destructive Nature

While LIBS can not truly be considered a non-destructive technique since some amount of

sample is always permanently removed from the surface, it can be considered micro-destructive

due to the microscopic craters that are achievable under the correct laser conditions. While this

may not be so important for applications where there is an abundance of sample, it is imperative

in applications where the sample is scarce or precious, or has aesthetic quality and can not appear

damaged. Because of this advantage, LIBS has been extended to applications involving the

pigment analysis or authentication of priceless works of art20-22 and ancient artifacts.23

Portable and Stand-off LIBS

Arguably one of the greatest advantages of LIBS, and possibly the reason why limitations

such as poor precision and sensitivity are tolerated, is the ability to take the instrument out into

the field with minimal instrumental modification. There is increased momentum in the world

today to achieve analytical information about a variety of samples in a shorter time frame. In

some cases, the time it takes to retrieve a sample, transport it back to the laboratory and analyze

it may be too long in terms of preventing the threat the sample poses. In other cases, it may be

necessary to analyze hundreds or thousands of samples to get accurate results, a monumental

task if each sample is to be individually collected and sent back to the laboratory. The

underlying factor is that it is simply more convenient to take the instrument to the sample rather

transport the sample to the instrument.

The relatively simple and versatile instrumentation that LIBS employs makes it an ideal

portable system for field-work. Small, solid-state lasers are often used in portable LIBS systems

with the laser head either mounted in a small probe or laser pulses delivered to the sample via









Table 3-1. Specifications for the Big Sky Ultra Nd:YAG pulsed laser.
Repetition Frequency (Hz) Up to 20

Pulse nnergy (mJ)a Up to 50

Beam divergence (mrad)b <7

Energy stability (%)" 2

Beam diameter (mm)d 2.8

Pointing stability (Cprad)e 50

Jitter (+ ns Q-Switch)' <2

Energy drift over 8 hours <10%

a) Nominal FWHM. b) Angle containing 86.5% energy. c) Variation from mean for 99% of shots
d) Nominal value. e) Full angle, 99% of shots. f) Measured from Q-Switch sync output.









maintain the same grain sizes and ensure uniform coating, the samples were milled for 15

minutes without mixing balls. Microscope images were taken of each sample after it had been

affixed to a microscope slide using a Nikon Labophot microscope with a 4X magnification

(Figure 4-5). It is clear that graphite was much more homogeneously distributed in the quartz

samples of smaller grain size than the larger ones. In addition to this, as the concentration of

graphite in large grain size quartz matrix increased, the number of quartz particles decreased and

the graphite coated larger areas of the tape surface. This resulted in some of the laser shots

sampling 100% graphite and significantly affected the precision of the sampling.

The average integrated C247 peak of 25 separate measurements on each sample was

determined and a plot of signal vs. carbon concentration was constructed for each of the particle

fractions (Figure 4-6). The statistics for each size fraction (Table 4-1) illustrate how the

precision of the measurements significantly increased as the particle size of the matrix decreased.

This trend was expected considering the higher degree of heterogeneity associated with the

larger grain sizes. The homogeneity of the smaller grain sizes resulted in sampling that was

more consistent and representative of the bulk sample. Table 4-1 also illustrates how the slopes

of C247 signal vs. carbon concentration decreased with the particle size. This was in agreement

with the results of Wisburn et al.5o The decreasing slope was due to the decreased mass of

carbon that was sampled per shot as the particle size of the quartz matrix decreased. As

illustrated in the microscope images of the samples (Figure 4-5), the larger quartz particles were

completely coated in graphite which concentrated more graphite into the ablation area of the

laser beam. However, as the particles became smaller and closer in size to the graphite particles

almost no coating occurred, and the graphite and quartz were more independently distributed

resulting in fewer concentrated areas of carbon and a more homogenous mixture.









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 FOR THE DETERMINATION OF
CARBON IN SOIL

By

Lydia Edwards

December 2007

Chair: James D. Winefordner
Major: Chemistry

Laser-induced breakdown spectroscopy (LIBS) is a relatively young atomic emission

technique that has found great utility in the elemental analyses of a variety of materials. In brief,

LBES is achieved by focusing a high-powered, short-pulse laser onto a sample surface to produce

plasma that is rich in electrons, atoms and ions. The atoms and ions in the plasma emit radiation

that is characteristic of the elemental composition of the sample. While LIBS has been

extensively applied to soil analysis, its use as a means of determining carbon in soil has been

lacking, limited to a few proof-of-principle demonstrations by research groups at Los Alamos

and Oak Ridge National Laboratories. Since the sequestration of carbon in soil may offer a

means of mitigating the effects of global warming, it is becoming increasingly urgent to develop

a practical device that can accurately measure terrestrial carbon inventories in order to

understand the status of the global carbon cycle. Due to the unique advantages of LIBS,

including the potential for in-situ analysis, little-to-no sample preparation, simultaneous multi-

element detection and relatively simple instrumentation, the technique has the potential to fulfill

the requirements of an in-field carbon monitor. Before a field-based LBES instrument can be

designed there must be a laboratory-based demonstration of the capability to determine carbon in

soil.









can be reduced to acceptable levels. In other cases, the standards of analytical performance are

lowered in favor of the versatility and applicability of the technique to a specific analysis.

Precision

It is generally accepted that the precision of LIBS is one of its greatest disadvantages.

Common relative standard deviations (RSD) reported fall between 5-10%, though values can be

significantly higher depending on the sample. There are many different factors that contribute

to such poor precision, the most important of which is more a property of the sample than the

technique itself. The lack of sample preparation, which makes LBES so attractive, also means

that sample remains in its raw state, which most often is completely non-uniform. This,

combined with the very small mass of sample that is interrogated, ensures that the sampled mass

may not be representative of the bulk sample, and subsequent laser shots on different areas may

produce very different spectra. While increasing the number of laser shots may increase the

precision, this is still limited by the inherent inhomogeneity of the sample. There are many other

factors that will significantly affect the precision, including: choice of analytical emission line,

emission signal temporal development, sample translation velocity, detector gate delay,

background correction methods, perturbations of the plasma, changing lens-to-sample distance

(LTSD), changing fiber-to-sample distance (FTSD), and laser-pulse shot-to-shot instability.25 In

many cases computing the ratio of the element of interest to another fixed-concentration element

in the sample will alleviate the effect these parameters have on the precision.

Matrix Effects

Matrix effects are one of the greatest challenges to the LBES analysis and represent a

significant obstacle to quantitative analysis. The term 'matrix effect' refers to the physical

properties and chemical composition of the sample that affect the element signal such that

changes in concentration of one or more of the elements forming the matrix alter an element' s















1800-
A. B. 18000
1600-
1600-
1400 C247
1400-

S12001 -1200-

1000 1000-

800 -800-

600-1 600-





200 00 40 50 60 70 800 900200 300 400 500 600 700 800 900
Wavelength (nm) Wavelength, nm




Figure 3-7. Graphite pellet spectra with A) back-side emission collection and B) side emission
collection (Note: Only the C247 line is identified because it was used to determine

increase in signal intensity)









CHAPTER 3
INTSTRUMENTATION AND DATA ANALYSIS

The components of a LIBS system are very simple, as explained in chapter 1. The basic

requirements are a pulsed laser, focusing optics, some method of emission collection, a

spectrometer coupled with a detector and a computer. Naturally, there are many variations for

each of these components, and the choice of which to incorporate into the LIBS system is

dependent on the type of analysis and available resources. The instrumental components that

were employed in the maj ority of this research were selected because they were comparable to

the type of instruments that may be incorporated into a portable LIBS instrument. That is to say,

they are compact and robust, have low power requirements and require little maintenance. Each

of these components will be discussed in depth in the following section. The second half of this

chapter will describes the optimization of these instrumental components.

Instrumental Components

Laser

The laser used for these studies is manufactured by Big Sky Laser Technologies. It was a

flashlamp-pumped Nd:YAG laser operated at the fundamental wavelength of 1064 nm and with

a specified maximum pulse energy of 50 mJ. The pulse energy can be changed in increments of

5 mJ. The pulse duration is 6-8 nm and the laser can be operated at repetition rates of up to 20

Hz. Table 3-1 provides a complete summary of the laser specifications. While these

specifications state that the laser energy stability is 2%, it was assumed that this was an

underestimation of the value and that it had to be measured independently prior to use. The laser

consists of two units, the laser head where the pulse is generated, and the integrated cooler and

electronics (ICE) system from which the energy and repetition rate are controlled. For this

research, the laser head was mounted vertically with the laser pulse directed down onto the









the S/B was nearly 7 times greater for the side-collection configuration. Clearly, placing the

fiber closer to the plasma provided significantly increased sensitivity and was consequently

chosen as the best position of the fiber probe for the remainder of the study.

Optimizing Lens-to-Sample Distance

The lens-to-sample distance (LTSD) is an important parameter that must be held as

constant as possible in order to obtain reproducible spectra. This can be a challenge when

analyzing samples of different heights, or when a single sample does not have a uniform surface.

If the sample is moved farther from the focal point of the laser, the irradiance (W/cm2) Of the

sample decreases. In cases where the breakdown threshold of the sample is low, or the laser

energy is high, this can be beneficial because the laser beam samples a greater area. However, if

the spot size is too great, a breakdown will not occur. Depending on the physical and chemical

properties of the sample, there is an optimal LTSD that must be kept constant for a set of

samples.

In order to determine the best LTSD a graphite pellet was placed on the sample stage and

spectra were collected at 1 mm increments above and below the focal point of the lens (measured

as 2.5"). Ten shots at maximum laser energy were recorded and averaged. The results indicated

that the greatest C247 signal was found when the focal point was 2 mm above the sample surface

(Figure 3-8). Incidentally, this position also yielded the lowest standard deviation in

measurements. It is important to note that the measured focal point of the plano-convex lens

used in this research was measured using room light and may be slightly different for the 1064

nm laser. Regardless, the results of this experiment provide the optimum LTSD and fiber-to-

sample-to distance (FTSD) based on maximum C247 signal intensity. In order to ensure a

constant LTSD and FTSD, regardless of different sample heights, a laser positioning system was

constructed. With the graphite pellet at the height that provides the best signal, a small helium-












Establishing an optimized sampling regime for soils is much more difficult than for most

solid samples encountered in LIBS. Due to the highly heterogeneous physical and chemical

nature of all soils, combined with the micro-sampling property of LIBS, single shot

measurements were not representative of the bulk sample. As such, it was found that a minimum

of 20 sample spots must be measured to achieve the minimum, if not always acceptable, RSD.

In reality, it might take thousands of sampling shots to produce a precise measurement that is

representative of the bulk sample. The particulate nature of the soil samples also greatly affects

the LIBS measurements. It was confirmed that the LIBS signal was greatly dependent on the

matrix particle size and that the increased homogeneity imparted by samples composed of

smaller particles yielded much higher measurement precision. It was also determined that when

a single, preparatory laser shot was used to modify the sample surface, the subsequent second

laser shot produced increased emission intensity across the entire spectrum. Microscopic images

of the soil samples before and after the sampling indicated that surface modification resulted

from the removal or break-up of large particles which most likely allowed more sample to

interact with the second laser pulse, thus increasing the mass ablated. Under these sampling

conditions, the precision of the C247 signal for all 38 soil sample did not exceed 18%.

The optimized sampling regime was applied to natural, dried soil samples to assess the

potential of LIBS as a quantitative technique. However, the results revealed no correlation

between the uncorrected carbon peak area of all soil samples and their reported carbon

concentrations. There are many factors that must be considered before quantification can be

ruled out. For instance, all of these soils suffered from very different matrix effects and therefore

could not be all be compared together or used to produce one single calibration curve. In














30-

28 a

26-

0_ 24-








12 -





















m 87











4-5 Images of different quartz size fractions spiked with graphite (4X magnification). .........90

4-6 C247 peak area as a function of carbon concentration (wt %) for different particle
sizes of quartz matrix. .............. ...............91....

4-7 Microscope images of three soils to illustrate heterogeneity ................. .....................93

4-8 RSD of the C247 signal intensity as a function of the number of shots on separate
spots. ............. ...............94.....

4-9 C247 and Ca422 signal intensities as a function of shot number on the same spot. .........96

4-10 Microscope images of soil after the indicated number of laser shots (4X
m agnification). ............. ...............97.....

4-11 Proposed mechanism for the second shot signal enhancement on soil ................... ...........98

5-1 Full spectrum of carbon taken on a graphite pellet with prominent lines and bands
identified. ............. ...............106....

5-2 C247 peak area as a function of spectrometer delay time. The integration time is
fixed at 200 ns. ........._.._.. ...._... ...............107...

5-3 The CN molecule, with a band head at 388 nm, is barely detected in a soil sample.......108

5-4 Theoretical curves of growth for carbon in a silicon dioxide matrix at three different
plasma temperatures ................. ...............109................

5-5 Calibration curves produced from graphite in ................. ...............110........... ..

5-6 Results of LBES measurements on finely ground soil samples ................. ................. .1 12

6-1 Microscope images of natural soil samples (3.5X magnified). ............. .....................121

6-2 A histogram of the RSD's from LBES C247 measurements on natural soils using the
first shot and the second shot. .............. ...............124....

6-3 LIBS C247 peak area as a function of % carbon. ........................... ........123

6-4 Normalizing the C247 peak area by soil surface density increases the correlation
between LBES signal and carbon concentration ................. ...............125........... ..

6-5 Extent of the iron spectral interference. ............. ...............126....

6-6 A soil spectrum produced using the Acton SpectraPro showing the resolved carbon
and iron lines. ................. ...............127___ .....

6-7 LIBS C247 peak area as a function of carbon concentration for measurements
produced using a high resolution spectrometer. ............. ...............128....


































hi 1


Soil Samples
Graphite in Silica
Graphite in
Aluminum Silicate


340
320
300
280
26-
260
22-
240
18-
220
14-
1200
10-
180
160
140
120
-_


V


0 1 2 3 4 5 6 891011

% Carbon


)S+River


02 04 06 08 10 12
% Carbon


14 16 18 0


12 3 4
% Carbon


56 7


Figure 5-6. Results of LBES measurements on finely ground soil samples A) With graphite in
silica and aluminum silicate calibration curves B) The two separate, distinct slopes
formed by the soils (error bars are + 1 standard deviation).





112












400-





300-


o


v, 200-
C




i0 100-






247.2 247.6 248.0 248.4

Wavelength, nm




Figure 3-9. Calculation of intensity (I), background (B) and area (above the red line).










2). The accuracy of each curve to predict the carbon content varied greatly, with a range of 5.2 -

155 %. If the soil that produced an accuracy of 155 % is omitted, the range was reduced to 5.2 -

64.5 % with averages of 34.9 % for the aluminum silicate curve and 23.3 % for the silica curve.

Still, one problem with these results is that they were obtained with advance knowledge of which

curve to apply to which soil (based on Figure 5-6). It is still unknown what properties of the soil

matrix caused such separate and distinct slopes, though particle size can be dismissed as a cause

because all soils were extensively ground to a fine powder. Without firm evidence as to why one

soil aligned with the silica curve and another with the aluminum silicate curve, it would be

impossible to know which to apply to an unknown soil. Furthermore, the results of this research

indicate that even in the absence of particle size effects, it would be impossible to produce a

universal carbon calibration curve that is applicable to all soil samples. This includes calibration

curves produced from simulated soils or from a range of standard soils.









more of the incoming laser light, thus reducing the portion of the laser pulse that reaches the

solid sample, effectively decreasing the signal. It was found that the optimal repetition rate

varied as a function of the sample mean particle size and for a soil sample was 1 Hz.

The crater formed during laser interrogation of the sample spot with multiple pulses was

also studied. It was found that there was a slight increase in the emission signal from the second

laser pulse that could not be attributed to persistent aerosol due to the much lower repetition rate.

Instead it was theorized that the sampling site was enriched with larger particles after the

shockwave from the first laser pulse swept away the smaller particles. After the second laser

pulse, a sharp decrease in signal was observed which was explained as a crater effect: the deeper

the crater became the greater the plasma-to-fiber distance and enhanced shielding of the plasma

by the crater walls from the fiber.

Another very important aspect of LIBS soil analysis that was investigated by Wisburn was

the effect of soil grain size. The importance of this parameter on any analysis is rather dependent

on the analyte of interest and how it is distributed in the sample, for instance if it is a surface

contaminant or if it is bound within the particles. To investigate this, soil samples were

separated into four different grain sizes by sieving (0.38-1 mm range). Each was contaminated

with Cd, which adheres to the surface of the particles. Good linear correlation between signal

intensity and Cd concentration for each grain size was found, as well as a decrease in slope as the

particle size decreased. While these results are significant in that they show a great dependence

of the LIBS signal on the particle size, two important factors must be considered for practical

soil analysis. First, most raw soil samples are composed of a broad distribution of particle sizes

that exceeds the range studied in this research. In addition to this, these results were based on the

assumption of only surface contamination. For many different analytes of interest this is not the








Normalizing the C247 signal to surface density proved to be the most promising method.

Figure 6-4 shows such normalization transforms the raw measurements and produces a much

more linear correlation plot. Based on these results, it would seem that correcting the LBES

signal for the physical rather than then chemical matrix effects is the most promising method to

relate C247 signal to carbon content. This is advantageous because the physical properties of

any given soil are easier to assess during in-field analysis than the chemical composition. In

order to correct the LIBS signal for chemical matrix effects, the chemical composition of the

soils would have to be determined prior to LBES measurements. In some cases, this may be as

simple as determining high iron content based on color; however, even subtle differences in

chemical compositions will contribute to significant chemical matrix effects and reliable

analytical techniques would be necessary to classify soils into groups of similar composition for

LB S measurements.

High-Resolution LIBS Measurements on Soils

One potential problem with the C247 line was the possibility of an interference from the

weak iron ionic line at 247.8 nm and atomic line at 247.9 nm. To study the extent of this

interference, a pure iron powder was analyzed, as well as a 1:1 mixture of iron and graphite

powders (Figure 6-5). The spectrum of iron powder revealed that the iron atomic line is detected

at 247.99 nm. A small shoulder on the left of the peak corresponded to the ionic iron line. The

spectrum of the graphite/iron mixture revealed that the Fe247.98 line was not resolved from the

C247.83 line. While the extent of this interference was expected to be minimal due to the low

iron content of these soil samples, it was still necessary to ensure that the poor correlation

between the LIBS carbon signal and the carbon content was not due to the unresolved iron peak

contributing to the carbon signal.










The purpose of our research was to investigate the potential of LIB S for carbon in soil

determination with an emphasis on minimal sample preparation and the potential for in-situ

analysis. While the quantification of carbon in soil by LBES was not achieved in this study, an

optimal soil sampling regime was established that increased sensitivity and measurement

precision. The LIBS instrumentation was optimized specifically for soil samples, including the

laser energy, optical configuration and spectrometer delay time. The sample presentation was

extensively studied, including the effect of soil grain size on LIBS signal and the advantages of

mounting the sample on tape as a thin monolayer rather than as a pressed pellet.










vegetation, such as plants, trees, crops and grasses. Due to poor land management and

agricultural practices during the first half of the 20th century, over half of the maximum soil

organic carbon (SOC) has been lost. However this has resulted in a greater capacity of many

soils to sequester carbon.

The maj ority of carbon is introduced into soil as decaying biomass. Plants and other

vegetative material withdraw CO2 fTOm the atmosphere during photosynthesis and convert it to

sugars, starch and cellulose. Upon death, soil microorganisms cause decomposition and respire

CO2 in the process. However, an organic residue remains after decay and is referred to as

humus. In the absence of oxidation, this organic matter will remain in the soil for hundreds,

possibly thousands of years.65 Carbon sequestration in soil can be significantly increased by up

to 0.1% per year by converting marginal arable land to forest and grassland, planting high yield

crops, and not tilling crop lands, especially in warm, moist climates. Even such a seemingly

small increase in carbon sequestration can have a huge impact on the global (atmospheric)

carbon balance. Current estimates indicate that successful sequestration of carbon in soil can

potentially offset 30,000-60,000 million metric tons of carbon that is released by fossil fuel

combustion over the next 50 years.66

A direct result of the Kyoto Protocol is the establishment of a so-called 'carbon stock

market'. In brief, it allows countries that have large carbon sinks, such as forests or vegetation to

deduct certain amounts from their CO2 emiSsion targets, thus making it easier for them to

achieve the desired net emission level. Countries that have already met their emission targets,

and have additional CO2 Sinks may wish to sell the benefit that these sinks provide (an offset of

the target emission levels) to countries that are not on target to meet required emission

reductions. Such a "cap-and-trade" system has also been proposed for the national level, where









neon laser was directed over the sample surface toward a photodiode connected to a voltmeter.

Half of the laser beam was blocked by the sample and the voltmeter measurement for the

blocked beam was recorded. When samples of different heights were placed on the sample

stage, they were moved either up or down until they blocked the same portion of light as the

graphite pellet in its optimal position. While using this positioning system allowed for each

sample to be analyzed under the same conditions, little could be done to account for the

variations in height within the sample itself, such as the case for particulate soil matter.

Data Analysis

Once the required number of spectra had been collected for a sample, it was necessary to

garner the required information by processing the data. As previously mentioned, the OOILIBS

software had a mode which allowed the user to collect statistics during the analysis; however the

information provided was limited to simple statistics, with no regard to the trends or patterns

which may appear within a sample. Instead, the data collected during this research was

processed using Microcal OriginTM and a Matlab program written specifically for the analysis of

LBES data. Both the peak area and the background subtracted peak intensity were calculated

(Figure 3-9). A comparison between the statistics (standard deviation, standard error) produced

from calculating either the intensity or integral revealed that they were both very similar,

however in general, computing the integral yielded slightly lower RSD's than the intensity, most

likely due to the dependence on more data points than the one pixel used to calculate intensity.

The Matlab program also calculated the intensity and integral of the background (B), the noise

(N), and RSDB.









after concentrations of carbon in excess of 6% both curves reach a plateau. When points after

6% were removed from the curve, a linear fit could be applied that produced a correlation

coefficient, R, of 0.9880 for silica and 0.9932 for aluminum silicate. It should also be noted that

both curves exhibit a non-zero intercept. Samples of silica and aluminum silicate were submitted

for dry combustion CHN analysis, which revealed that both contained carbon at concentrations

of approximately 0. 1 %. However, such a concentration can not explain such a large signal. It

was determined that the same particle-adhesive effects were contributing greater carbon from the

tape. However, the linearity of these curves up to 6% provide further evidence that this

contribution is likely constant for all sampling shots.

When constructing these potential calibration curves, it was realized that their use for

natural soil samples would be limited. The effect of the two different matrices on the LIB S

signal is significant, and such a matrix effect was expected to be more pronounced for soil

samples, which are much more varied in chemical and physical composition. Based on this

knowledge, the possibility of constructing a universal LBES calibration curve for all soil samples

is not reasonable. Furthermore, it was considered highly unlikely that a calibration curve

produced from graphite spiked silica or aluminum silicate powders would be highly accurate

predictors of % carbon in natural soils. However, the level of accuracy required for carbon in

soil measurements is still in question, and these results provided a good foundation for studying

carbon in a soil-like environment.

LIBS Measurements of Powdered Soil Samples

Before any measurements of natural soils were undertaken, soil standards were obtained

from NIST. These four soils, Buffalo River Sediment (SRM 2704), Estuarine (SRM 1646),

River Sediment (SRM 1645) and Montana Soil (SRM 2710) were very fine powders and

considered completely homogenous. Only two of the soils, Buffalo River and Montana Soil




















200

-- 160


120
100





1 2 4 6 8 10 12 14 16 18 20 22
% Carbon



Figure 6-7. The LIBS C247 peak area as a function of carbon concentration for measurements
produced using a high resolution spectrometer (error bars are + 1 standard deviation).











Table 4-1. Statistics for the analysis of quartz size fractions spiked with graphite
Quartz particle Average RSD for Correlation Slope (C247 Carbon
size all carbon coefficient (R) area/% C) concentration
concentrations (wt %)a
(%)
>500 Cpm 40 & 20 0.9143 2.74 0.01

250-500 Cpm 26 & 21 0.7554 1.83 0.001

>30 Cpm 12 & 21 0.8691 1.52 0.016

18-30 Cpm 10 &2 0.9657 1.25 0.020

<18 Cpm 7.2 & 1.2 0.9898 1.30 0.034

0.5-1 Cpma 0.118

a) Determined by dry combustion. b) Not used in study due to limited sample









area of LBES has seen remarkable progress, including plasma characterization, instrumentation,

and applicability. Indeed, it is the broad range of the latter that is fueling the increased interest

in LIBS, which has been successfully applied to an extraordinary variety of samples, from

priceless works of art to sewage; from environmental aerosols to forensic analysis. The revival

of LBES is reflected by the numbers of published j journal articles on the subj ect, which began to

increase in the late 1980's and continues to do so each year thereafter, as illustrated in Figure 1-



Today the potential of LIBS as a qualitative and potentially quantitative elemental

analysis technique is well documented in virtually every scientific field. The unique advantages

of LBES that have contributed to widespread interest include remote sensing capabilities, in-situ

analysis, little-to-no sample preparation, micro-destructive nature, applicability to all media,

simultaneous multi-element detection capability and relatively simple instrumentation. While

LBES may currently be limited by its poor precision (typically 5-10%), higher detection limits

(relative to conventional techniques), and difficulty in preparing matrix-matched standards, the

preponderance of applications reported in the literature may indicate that the advantages

overcome the limitations.

Characteristics of LIBS

The principles of laser ablation and plasma formation on a solid target have been

extensively covered in previous literature.4-1 As such, only a brief description of the

mechanisms that contribute to atomic emission during the LBES technique will be covered.

LBES is an atomic emission spectroscopic (AES) technique that shares similar fundamental

principles to ICP-OES, microwave induced plasma (MIP-OES), direct current plasma (DCP-

OES), flame-OES, arc-OES and spark-OES. All atomic emission techniques exhibit the

following basic steps:
































O 2007 Lydia Edwards










CHAPTER 5

THE SPECTROSCOPY OF CARBON AND CALIBRATION CURVES WITH SIMULATED
SOILS

The emission properties of carbon have been well studied, especially for the detection and

potential quantification of molecular fragments such as C2 and CN.84-90 Much of the previous

work concerning atomic carbon emission has been achieved by monitoring the 247.8 nm line.

This chapter will discuss the characteristics of this line and why it was chosen for this research,

as well as alternate aspects of the carbon spectrum which were considered for quantitative

analysis. The second half of this chapter describes the results of calibration curves that were

produced from graphite spiked silica and aluminum silicate powders and subsequent validation

with natural soil samples.

The Spectroscopy of Carbon

Choice of Analytical Line

To determine the best emission line of carbon for quantitative purposes using the Ocean

Optics LBES2000+ spectrometer, emission from the ablation of a graphite pellet was studied.

Initially, spectra were taken with no acquisition delay. This was because the ionization energies

of the carbon atoms are relatively high, and are therefore expected to be more prominent early in

the plasma lifetime, where the temperatures are high enough to sustain emission. When no delay

was used, the graphite spectrum showed four emission lines in the UV region: three ions that

produce emission lines at 229.6 nm, 251.1 nm and 283.8 nm and have ionization energies of 53.5

eV, 27.5 eV and 27.5 eV respectively and one stronger atomic line that emits at 247.8 nm and

has an ionization potential of 7.7 eV (Figure 5-1). When an acquisition delay of 1 Cps was

applied, the three ionic lines were not detected and the emission intensity of the atomic line

decreased. To further study the temporal properties of the C247 emission, spectra were taken on









sample stage. This type of laser is classified as a Class 4 laser which means that the output

beam, by definition, is a safety and fire hazard. As a result, protective eyewear was worn at all

times to prevent accidental exposure to both direct and reflected beams.

Optics

A quartz plano-convex focusing lens with a diameter of 1" and a focal length of 2.5" was

used to focus the laser. The lens was housed in a 1" lens tube which was mounted in a steel cage

system (Figure 3-1). This allowed for the lens to be easily moved up or down while still

remaining on axis with the laser. The lens was located approximately 11" from the laser

aperture. The diameter of the ablation area produced on a copper surface, with a pulse energy of

50 mJ was approximately 0.35 mm (Figure 3-2).

Fiber Optics

The fiber optic used to collect the plasma emission was produced by Ocean Optics, Inc.

especially for the LIBS2000+ spectrometer unit. The fiber bundle actually consists of seven

individual fibers, each of which collects a certain portion of the plasma radiation and transmits it

to an individual spectrometer. The fiber bundle has a nylon jacket and PVC sheathing. The fiber

cores are pure fused silica and the cladding is doped fused silica. The core diameter of each of

the individual fibers is 600 pm. The numerical aperture of the optical fibers and the probe is

0.22, which yields an acceptance angle of 24.80 in air. The probe houses a collimating lens

which has a focal length of approximately 1".

Spectrometer

The spectrometer used for this research was the LIBS2000+ and was also manufactured by

Ocean Optics (Figure 3-1). As the name implies, it is a commercial spectrometer that is uniquely

designed for LBES applications. It is in fact not one spectrometer but seven individual miniature

spectrometers which each cover a specific spectral range (approximately 100 nm). Each of the










CHAPTER 7
CONCLUSION AND FUTURE WORK

Natural soils are arguably the most challenging of all samples for quantitative LBES

measurements. There are many variables that must be accounted for that each contributes to

detrimental chemical and physical matrix effects. Chemical matrix effects can include spectral

interference from other soil components, speciation of the analyte of interest, water content and

the overall chemical composition, which may alter the sample's breakdown threshold. Physical

matrix effects include the broad range of different particle sizes, the distribution of the analyte

throughout the sample and extensive differences in physical properties such as reflectivity and

heat of vaporization on both the macro and micro scale. The initial purpose of this research was

to assess the potential of LBES for quantitative carbon measurements in soil; however, it became

readily apparent that a more fundamental approach was necessary before any conclusion

concerning quantitative results could be achieved. As such, the majority of this work was

devoted to establishing optimal instrumental and sampling parameters exclusively for LIBS

measurements on soils. While the ultimate goal of this research was for carbon determination in

soils, most of these studies can be extended to any other analyte of interest.

A number of modifications were made to the LIB S instrumental components for the

purpose of soil analysis, including the use of a positioning laser to ensure constant LTSD and

FTSD for soils of variable heights. A gentle air stream directed over the sample was found to

eliminate the interaction of the laser beam with liberated, air-borne particles along the laser axis,

thus increasing measurement reproducibility. Careful alignment of the fiber-optic probe,

positioned close to the plasma, was found to provide superior emission collection relative to the

popular back-side emission collection. A spectrometer delay time of 1 Cps was found to produce

the best S/B for the C247 line, as well as the lowest limit of detection for carbon in soil.












742190


Oceans

Atmosphere

0 Terrestrial
Biosphere






39120






Figure 2-1. Magnitudes of the reservoirs of actively cycling carbon in gigatons (Gt). Produced
with data from ref. 61.








Particle Size


>500 pm


250-500 pm


>30 am


18-30 am


<18 ram


% Graphi



2


4


6


9


14


19


11111 r


Figure 4-5. Images of different quartz fractions spiked with graphite (4X magnification).


;"









addition to the different matrix effects, other factors independent of the LIBS measurements may

have contributed to such poor correlation. For instance, the significant variability in the CHN

analyses introduced significant error to the carbon concentrations, which in turn will influence

the correlation. In order to minimize this error, for future work, soil samples should be

extensively ground prior to dry combustion and analyzed multiple times. Another consideration

is whether the mounting tape contributes to the LIBS carbon signal of soil samples, and whether

this contribution is relatively constant. Other sample mounting methods that do not involve a

carbon based substrate should be explored. For instance, future work could produce LBES

measurements on loose soil samples.

Correlation is likely to increase when soils of very similar chemical and physical

compositions are analyzed together, rather than as suite of very different soils. Additional

research must focus on identifying the chemical or physical properties that contributed most to

the LIBS matrix effects. For instance, this research exploited the particle size dependence of

LBES signal and found an increase in correlation by normalizing the carbon signal to surface

density, which is related to the particle size. In addition to particle size other variables may

influence the LBES signal and could be used to categorize soils or to correct the LBES signal.

Examples include the chemical form of carbon, chemical composition, and mineral content.

The use of these soil properties would require extensive prior knowledge about each soil sample

and collaboration with soil scientists, who often work with fully characterized soils, would be a

valuable resource.

Future work should also consider the use of a multivariate analysis of the entire soil

spectrum. In the absence of prior knowledge about a specific soil, the LIBS spectrum can

provide valuable information concerning the relative concentrations of elemental components.









the original physical properties of the soil, and as such allowed for further studies of the effects

of grain size on LBES signal.

Analysis of Bare Mounting Tape

The tape that was used to mount the sample to the microscope slide was produced by

Scotch 3M and consisted of a foam center coated on both sides with an adhesive. The foam

backing and the adhesive coating are composed of acrylic and vinyl polymers which contain high

concentrations of carbon. For this reason, the tape was extensively analyzed to determine the

extent of carbon interference. First, nine separate measurements on different spots of the bare

tape were collected. The C247 line was not detected in any of them (Figure 4-3a). Following

this, 12 shots were collected on the same spot of the tape. Each of the resulting spectra were

compared, and it was determined that carbon was not detected after the first shot; however, it

was detected after subsequent shots (Figure 4-3b). A plot of the shot number directed on the

same spot and the resulting C247 intensity showed that the maximum amount of carbon

(approximately 55 counts under these specific instrumental conditions) was detected for the

second laser shot, after which time the signal decreased, until a plateau was reached after 6 shots

(figure 4-3c). To investigate the reproducibility of this phenomenon, 10 pairs of shots were

taken on the tape. In each instance, the first shot yielded no carbon peak and the second shot

produced a carbon peak of with a net intensity of approximately 60 counts (Figure 4-3d).

Images taken of the ablation spots after a number of different laser shots illustrated that the

first laser shot did not produce a discernible crater and the surface of the tape merely exhibited

signs of localized melting. It is possible that the breakdown threshold of the acrylic adhesive is

not reached under these sampling parameters; however the first shot modified the surface of the

tape, allowing for the second shot to produce plasma and for carbon detection to occur.

Regardless of the mechanism, this occurrence suited the purposes of this research well. During










* atomization/vaporization of the sample to produce free atomic species consisting of atoms
and ions,
* excitation of these atoms,
* detection of the radiation emitted by these atomic species,
* calibration of the intensity to concentration or mass relationship,
* determination of concentration, masses, or other information.4

In LBES, the first step, atomization and vaporization of the sample, as well as the second

step, atomic excitation are both achieved by focusing a short-pulsed laser of adequate power

density onto the surface of a sample. Laser pulses of 5-10 ns are most commonly employed.

Increasing research is being performed on LIBS achieved with picosecond or femtosecond

lasers." Once the laser pulse strikes the target, single- or multi-photon absorption, dielectric

breakdown and other, often undefined mechanisms contribute to the surface temperature

instantly increasing beyond the vaporization temperature of the material. The subsequent

dissipation of energy through vaporization is slow relative to the rate of energy deposition from

the laser pulse. This results in vaporization of the surface layer and causes the underlying

material to reach critical temperature and pressure, finally causing the surface to explode.

Particles, free electrons and highly ionized atoms are released into the surrounding atmosphere

and begin to expand at a velocity much faster than the speed of sound, forming a shockwave in

the process. If the laser pulse is of long enough duration, as is the case for nanosecond lasers, the

expanding plume and the material within will continue to absorb energy from the laser and

intense plasma up to a few millimeters in height is visually observed. After several

microseconds, collisions with ambient gas species cause the plasma plume expansion to slow

down while the shockwave detaches from the plasma front and continues propagating at a rate of

approximately 105 m/s. As the plasma begins to decay through radiative, quenching, and

electron-ion recombination processes, high-density neutral species and clusters of dimers or

trimers are formed via condensation and three-body collisions and with the thermal and








concluded that the lack of correlation between LIB S signal and carbon observed with the Ocean

Optics measurements must be attributed to other matrix factors and is not caused by an iron

interference.

One other important aspect to note concerning these results is the poor precision. As

previously mentioned, this was likely due to the optical arrangement of the emission collection.

For many soils, the RSD was double that achieved using fiber optic emission collection. It is

likely that the precision can be increased by further grinding the soils to Eine powders or

analyzing pressed pellets; the flatter surfaces will minimize the plasma translation with respect to

the spectrometer entrance slit. However, such additional samples preparation steps are

cumbersome, and fiber optics is likely the best method of emission collection when analyzing

natural soils of varying textures.












Table 4-2. The C247 statistics for intra-sample and inter-sample measurements.
Slide Number of spots C247 Area RSD (%)
sampled
A 272 28 15 19

B 293 28 16 19

C 291 27 15 18

Average of the shots 856 28 & 5 19
taken on all slides









m2 to a depth of 50cm.70 It is apparent that in order for quantitative results to be meaningful,

both bulk soil density and particle density must be assessed at the time of analysis.

Clearly, for accurate and precise soil analysis a carefully considered analytical scheme

must be developed. Consideration must be given to the inherent variables in soil and the effect

these will have on the analysis method. LIBS may be a suitable technique for practical soil

analysis due to its portable nature (which precludes costly sample transport) and high-throughput

(which allows for the orders of magnitudes of samples necessary to gain representative data).

However, as also discussed, LBES itself may suffer from poor precision due to a variety of

instrumental and physical variables. As such, any analysis of soils by LIBS requires an

experimental plan that reduces the effects of these variables, and balances the required accuracy

and precision with the limits of the technique and sample. The research described in the

following chapters describes an experimental scheme that optimizes these parameters and

assesses the feasibility of using LIBS for in-situ carbon determination









BIOGRAPHICAL SKETCH

Lydia Edwards was born in Shrewsbury, England, in 1981 and is the oldest of four

siblings. In 1994 she moved to Woodbridge, Virginia, with her family and attended C.D. Hylton

High School from which she graduated summa cum laude in 1999. After attending Northern

Virginia Community College and working full-time for one year, she moved to Blacksburg,

Virginia, to attend Virginia Tech, where she studied chemistry. Lydia graduated from Virginia

Tech in May 2003 with a Bachelors of Science in Chemistry. In August 2003 she enrolled in the

chemistry graduate program at the University of Florida in Gainesville, Florida and j oined the

research group of Dr. James D. Winefordner in the analytical chemistry department. In addition

to performing graduate research, she also took classes in Forensic Drug Chemistry and Forensic

Toxicology. In August 2007 Lydia completed her doctoral research as well as a Master of

Science in Forensic Drug Chemistry. In 2008 she will receive a Master of Science in Forensic

Toxicology from the UF Department of Veterinary Sciences. Lydia has been married to Daniel

Breckenridge since 2006. She has accepted a position in atomic spectroscopy with Bristol-

Myers Squibb in New Brunswick, New Jersey.










6-8 A plot of the C247 peak area normalized by surface density vs. carbon content for
measurements produced using a high resolution spectrometer ................. ................. .129









relative intensities and exact wavelength for carbon were obtained from the NIST atomic line

database.93 In Order to assess the accuracy of the LIBS2000+ spectrometer, the actual measured

wavelength of 247.837 nm was compared with the true wavelength of 247.856 nm. A difference

of 0.019 nm corresponded to less than one pixel and indicated high accuracy in the UV channel.

The precision of the spectrometer was also assessed by collecting 100 spectra on a pure,

homogenous graphite pellet. The RSD of these measurements was an acceptable 2.4%.

Curves of Growth

To determine the linear concentration limits of calibration linearity, theoretical curves of

growth (COG) were constructed using a software program developed in our group. At high

concentrations of analyte, a calibration curve is expected to exhibit non-linear behavior, although

the actual concentration at which this will occur varies as a function of plasma temperature,

electron number density, emission line monitored and matrix composition. The theoretical

COG' s for the C247 line were calculated assuming a matrix of SiO2 and plasma temperatures of

8000, 1200 and 16000 K. While the use of such a simplified matrix undoubtedly introduced

some error in the results, it would be impossible to account for all components of the soil matrix

as well as their relative concentrations. From the calculated theoretical COG' s shown in Figure

5-4, it is clear to see that the theoretical integrated C247 peak varies greatly with plasma

temperature. For instance, the relative C247 peak areas are 0.0053:0.26:1 for plasma

temperatures of 8000 K, 12000 K and 16000 K respectively. Despite the signal disparities for

different plasma temperatures, all three COG's indicated a departure from linearity for carbon

concentrations in excess of 8 %.

While these COG's provide useful information concerning the theoretical LIBS response,

it is important to realize that the inhomogeneity of soil samples precludes the direct use of these

calibration curves. As previously discussed, and is evident from Figure 4-7, each plasma that is










instruments are commercially available for the determination of total carbon content. Most

analyze hydrogen and nitrogen as well as carbon and are termed CHN analyzers.

More recently, the potential of visible, near infrared, mid infrared and diffuse reflectance

spectroscopy have been investigated as a means of analyzing soil, specifically for carbon

content.767

The Challenges of Soil Analysis

The greatest challenge to the analysis of soil, regardless of analytical technique or analyte

of interest, is the difficulty in obtaining a truly representative sample of soil. There are several

factors that contribute to this difficulty, including sampling errors, soil variability and soil

homogeneity. Soil variability refers to different types of soils located within a relatively small

sampling region. Many environmental and anthropogenic factors can affect the spatial

variability of soil, including localized decomposition events and farming practices. Variability

can occur in both the horizontal and vertical directions, and must be compensated for with a

careful sampling regime. The simplest way to account for it is to take multiple samples;

however the hundreds, or possibly thousands of samples that may be needed to obtain

representative data can be cumbersome. Due to the importance of determining carbon pools on a

regional scale, much attention has been directed toward the spatial distribution of carbon in a

variety of different lands.79-81

Soils are also highly heterogeneous even within the same soil type and therefore create a

so-called resolution problem. Soils are composed of many different constituents that are

unevenly distributed throughout a given sample. Inorganic constituents such as quartz are

present in three distinct particle size distributions referred to as sand, silt and clay, each of which

exhibits different physical and chemical properties related to the different surface area. Table 2-

2 shows the physical size ranges of these particle fractions, as well as the sizes of sub-groups









Table 3-2. Mean pulse energies of the Big Sky laser measured with a power meter.
Expected pulse energy Mean pulse energy of Mean pulse energy Mean pulse energy
(mJ) 32 laser pulses (mJ) without first ten of first 10 laser
laser pulses (mJ) pulse (mJ)

25 29 & 1.7 (5.9%) 30 & 2.2 (7.2%) 29 & 1.4 (4.8%)
30 35 A 1.9 (5.3%) 36 & 2.4 (6.7%) 34 11.3 (3.8%)
35 40 & 1.5 (3.8%) 40 & 1.7 (4.3%) 40 & 1.4 (3.6%)
40 46 & 2.3 (5.1%) 47 & 2.9 (6.2%) 45 A 1.7 (3.7%)
45 51 & 3.7 (7.2%) 54 & 2.7 (5.0%) 50 & 3.7 (7.4%)
50 57 & 1.8 (3.2%) 59 & 2.2 (2.3%) 57 & 1.3 (2.3%)














200


180


160


140


S120





6 00


4 0







20






Year




Figure 1-1. Number of LBES publications per year since 1960.









Spectrometer


Pulsed
Laser


Fiber
Optic


~c~


lens


Laser-Induced
Plasma
Emission
Collection
Sample





Figure 1-4. Typical schematic of a LIBS set-up


460.829


410 420 430 440 450 460
















Ions


Neutral s


Strong continuum emission


Molecules


Laser


Continuum


1 ns 10 ns 100 ns 1 CLs 10 CLs 100 CLs


Figure 1-3. Temporal evolution of the LIBS plasma. Td is the detector delay time and t, is the
gate pulse width, also referred to as the integration time. Constructed with data from
ref 4.













850-






700-

650-

600-

550-

500-

450-

400-

-6 -4 -2 0 2 4 6


Focal Distance Offset (mm)




Figure 3-8. C247 signal intensity (the average of 10 shots) as a function of focal point offset
(error bars are + 1 standard deviation).









to be valid for in-field analysis, samples must also be pressed into pellets. Obviously, this

additional sample preparation defeats the high throughput capability desired for the portable

instrument.

An alternate method of sample presentation is to mount the sample on a microscope slide

using double-sided foam mounting tape. This is achieved by attaching the desired area of tape

onto a standard microscope slide and removing the backing to expose the adhesive. The tape is

then firmly pressed onto a thin layer of the sample and the excess, unstuck sample is gently

tapped away. The tape is then fully covered with a monolayer of practically undisturbed soil

particles. The so-called mounting-tape sample preparation method is very quick, cheap,

introduces few contaminants, and minimizes sample modification. Due to these very attractive

qualities, it was the method of choice for sample preparation during this research.

In order to study the effects of the using the mounting tape sample preparation, LBES

analysis was performed on a pressed soil pellet and the same soil mounted on a microscope slide

(Figure 4-1). The soil pellet was prepared by applying in excess of 16,000 psi for 30 seconds.

One drop of water was added to the pellet before pressing to make the sample more cohesive. In

general, a binder would be used for this task; however common binders such as cellulose and

wax contain carbon and would therefore interfere with the LBES carbon signal. The resulting

pellet was 13 mm in diameter with a height of approximately 4 mm. Both samples were

analyzed using the same laser and spectrometer parameters (pulse energies of 57 mJ, a delay

time of 1 Cps and repetition rate of 1 Hz). For the slide sample, two laser shots were used for each

of 50 sample spots and data from each shot were collected (termed first-shot and second-shot).

The purpose for this double-shot analysis will be discussed later in the chapter. The resulting 50

spectra were collected, and the average integrated C247 peak was determined for both the first
















Soll Spectrum Varying Q~-Switch (2-2 L 20 38% C) -*-+4
Positive No delay ---+3
Negative Delayed (1000ns Increments) -*-+2
247.837 --1

A : -


C247 Signal Intensity vs Q~-Switch Delay


350

300-

250-

S200-

S150 -






o-


4 3 2 1 0 -1 -2 -3 -4 -5
Q~ Swtch Delay (us Increments)


Wavelength, nm


S/B vs Q~-Switch Delay


006-


m






a


4 3 2 1 0 -1 -2 -3 -4 -5
Q~ Switch Delay (us Increments)


,004-



S003-



0 002


4 3 2 1 0 -1 -2
Q-Switch Delay (us Increments)


-3 -4 -5


Figure 3-6. Results of changing the Q-switch delay times on soil spectra. A) The C247 line

drastically changes with spectrometer delay. B) C247 signal intensity vs. delay. C)

C247 S/B vs. delay. D) CL VS. delay.









At lower pressures, the plasma readily expands into the lower pressure regions increasing in

volume relative to those formed in atmospheric pressures. The signal enhancement at lower

pressures is the result of increased mass ablation due to decreased shielding of the sample

surface from the laser pulse by the more diffuse plasma. It was also determined that pressures of

less than 10 torr were required to detect O(II) in the soil due to the increased electron density

which causes the ionic oxygen atom to convert to the neutral oxygen atom. Arp et al 53 TepOrted

on LIBS of ice/soil mixtures which were synthetically produced to mimic the polar surfaces of

Mars. They determined detection limits for elements in these mixtures (at a pressure of 7 torr),

which were typically 10% soil, were not significantly changed from those obtained by analyzing

dry soil.

The determination of carbon in soil is arguably one of the most pressing issues concerning

soil analysis due to the current attention to global warming. However, very few papers have

been published regarding the potential of LBES to achieve quantitative data on carbon in soil.

Ebinger et al. first reported on the analysis of carbon in dried soil samples from farms in

Colorado and Hields surrounding Los Alamos, New Mexico. Each sample was sieved to <2mm

and placed in a quartz tube. A 50-mJ, 10 ns, Nd:YAG laser was focused into the quartz tube and

the resulting emission collected via fiber optics and delivered to a 0.5-m spectrograph with an

intensified photodiode array detector. Shot-to-shot variations were normalized by taking the

ratio of the 247.8 nm C line to a Si line. Verifieation of the C content was achieved by dry

combustion. The results showed good correlation between LIBS signal and the carbon

concentration, with an R2 Value Of 0.966. In turn, the resulting calibration curve predicted C

concentration in other soil samples with an accuracy of 3-14%. The detection limit was reported

as 300 ppm.54,55 This research certainly demonstrated the feasibility of LBS as a method for








To investigate this, LBES measurements were performed on the same soil samples from the

previous study and emission radiation was spectrally resolved using an Acton SpectraPro-500i

spectrometer with a 3600 grooves/mm grating which provided a resolution of approximately

0.02 nm. This resolution was high enough to fully resolve the carbon and iron lines (Figure 6-6).

Most other experimental conditions were maintained, including a laser pulse energy of

approximately 60 mJ, spectrometer delay time of 1 Cps and an integration time of 100 Cps.

Samples were mounted vertically and shifted manually after every two sampling shots to ensure

a fresh sample location. Emission was collected by a 2" lens and focused onto the 35 Cpm slit of

the spectrometer. The focusing lens produced 1:1 inverted image of the plasma onto the

entrance slit of the spectrometer, and alignment was achieved using a flat graphite pellet.

However the soil samples used in these studies were far from flat, and the positioning of the laser

induced plasma could be horizontally translated by up to 1 mm depending on the size of the

particle where the plasma was formed. As such, this optical configuration was highly sensitive

to the coarse textures of soils and emission collection efficiency was not always optimal or

reproducible.

The resulting LBES measurements produced with the high resolution spectrometer are

shown in Figure 6-7. It is clear that there is no increase in correlation between the LBES carbon

signal and the carbon content; indeed, the trend is very similar to the results achieved using the

lower resolution Ocean Optics spectrometer. Normalizing these results by the surface density of

the samples (Figure 6-8) did not significantly improve the correlation as it did for the previous

data. The C247 peak produced using the high resolution spectrometer were completely resolved,

and therefore wholly due to carbon alone. If the poor correlation observed with the Ocean

Optics spectrometer were due to the iron spectral interference, the fully resolved carbon peak

would significantly improve the correlation. This does not occur, and therefore it can be









At the very least, inspection of the entire spectrum can be used to group soils based on maj or

constituents. More advanced multivariate algorithms may be applied to entire spectral regions

instead of simply one carbon line and will account for varying peak intensities of other elemental

constituents which may be producing specific matrix effects.

The quantification of carbon in natural soils by LIBS was not achieved during this

research. While it is possible that this will be accomplished in future work, it is important to

reflect on the original motivation of this research: to assess LBES as a means of in-situ carbon

determination in natural soils. During this research, every attempt was made to minimize

sample preparation and reduce cumbersome instrumentation in order to mimic the LIBS

measurements that would be produced using a portable LIBS system. Future work may prove

that quantification is achievable on finely ground, pelletized samples and precision drastically

increased by taking thousands of measurements. However, once the sample preparation negates

in-situ analysis and requires extensive laboratory based work, the advantages of LBES become

redundant and more conventional techniques such as dry combustion or wet oxidation may

provide more accurate results.

The inability of LBES to determine carbon in soils at this point does not diminish the

importance of the sampling and instrumental studies that were performed using for the purpose

of soil measurements. For instance, studies concerning the dependence of LIBS signal on matrix

particle size are applicable to nearly all types of particulate samples. In addition to this,

qualitative analysis of soils is successfully achieved using the soil sampling characteristics

studied here. Indeed, the increase in emission intensity that is observed after a single laser

preparatory shot will increase LIBS sensitivity for trace toxic metals such as Hg, As or Pb. A

variety of important elements in soil, including Mg, Mn, Na, Fe, Cu can be readily detected in












LIST OF REFERENCES

(1) Brech, F.; Cross, L. Appl. Spectrosc. 1962, 16, 59.

(2) Rusak, D.A.; Castle, B.C.; Smith, B.W.; Winefordner, J.D. Trends Anal. Chent. 1998,
17, 453-461.

(3) SciFinder, version 2006; Chemical Abstracts Service: Columbus, OH, 2006; Accessed
June 3, 2006); calculated using ACD/Labs software, version 8.14; ACD/Labs 1994-2006.

(4) Laser-hIduced Breckdown Spectroscopy (LIBS): Fundamentalsd~~~dd~~~ddd~~ and Applications;
Miziolek, A. W.; Palleschi, V.; Schechter, I., Ed.; Cambridge University Press: New
York, 2006.

(5) Weyl, G.M.; Root, R.G.; Baldis, H.A, Hauser, A.A. In La;ser-hiducedPla;snza;s and
Applications; Radziemski, L.J.; Cremers, D.A., Eds.; Marcel Dekker, Inc.: New York and
Basel, 1989.

(6) Lee, Yong-Ill, Song, Kyuseok, Sneddon, Joseph. Laser hIducedBreakdown
Spectroscopy; Nova Science Publishing, Inc.: New York, 2000.

(7) Sneddon, J.; Lee, Y.; Curr. Top. Anal. Chent. 2004, 4, 111-117.

(8) Winefordner, J. D.; Gornushkin, I. B.; Correll, T.; Gibb, E.; Smith, B. W.; Omenetto, N.
J. Anal. At. Spectrons. 2003, 19, 1061-1083.

(9) Radziemski, L. J. Spectrochint Acta, Part B. 2002, 57B, 1 109-1 113.

(10) Lee, Y.; Sneddon, J Adv. At. Spectrosc. 2002, 5, 235-288.

(11) Radziemski, L.J.; Loree, T. R.; Cremers, D. A. Springer Series in Optical Sciences,
1983, 39, 303-307.

(12) Gunratne, T.; Kangas, M.; Singh, S.; Gross, A.; Dantus, M.; Chens. Phys. Lett. 2006, 423,
197-201.

(13) Le Drogoff, B.; Margot, J.; Vidal, F.; Laville, S.; Chaker, M.; Sabsabi, M.; Johnston, T.
W.; Barthelemy, O. Plasma Sources Sci. Technol. 2004, 13, 223-230.

(14) Le Drogoff, B.; Chaker, M.; Margot, J.; Sabsabi, M.; Barthelemy, O.; Johnston, T. W.;
Laville, S.; Vidal, F. App. Spectrosc. 2004, 58, 122-129.

(15) Angel, S. M.; Stratis, D. N.; Eland, K. L.; Lai, T.; Berg, M. A.; Gold, D. M. Fresenius'J.
Anal. Chent. 2001, 369, 320-327.

(16) Lenk, A.; Witke, T.; Granse, G. Appl. SurJ: Sci. 1996, 195, 96-98.









Data from CO2 trapped in ice cores and long term CO2 mOnitoring has indicated that the

current CO2 leVOIS are the highest they have been in 650,000 years, at 390 ppm in 2007.62 Such

elevated levels can be attributed to increased land clearing (destruction of vast rain forests,

depleting woodlands) and burning of fossil fuels (oil, coal, and natural gas), which contributes

over 5 Gt of CO2 into the atmosphere each year. Indeed, the atmospheric concentration of CO2

was 270 ppm prior to the industrial revolution of the 18th century, after which it exponentially

increased to the current value of 390 ppm. After the complete expenditure of all fossil fuels in

approximately 300-400 years, the atmospheric concentrations of CO2 are predicted to be

1200ppm.63

Increased CO2 leVOIS in the atmosphere contribute to a phenomenon referred to as global

warming. Carbon dioxide, along with other gases such as water vapor, methane (CH4), and

nitrogen oxide (N20), are referred to as 'greenhouse gases' because they trap heat from the sun

in the earth's atmosphere, which results in increased global temperatures, a process referred to as

global warming.64 The ramifications of global warming are ominous, and include decreased

vegetation production and accelerated polar ice-cap melting.

There is little argument that atmospheric CO2 leVOIS need to be reduced; however there is

much debate on how to achieve this goal. The obvious recourse is to reduce fossil fuel

consumption however current economics depend on fossil fuels and the stunted development of

alternate energy sources is reflective of this.63 An alternative to stemming the CO2 Source is to

enhance natural carbon sinks (where carbon is stored for long periods of time). As illustrated in

Figure 2-1, the largest carbon sink by far is the ocean; however the ability to manipulate such a

vast sink is limited and may take hundreds of years to produce appreciable changes. The next

largest carbon sink is terrestrial soil, the carbon content of which is double that of carbon in









emission signal even though the element concentration remains constant. For example, the

signal intensity for carbon in silica is very different from that of carbon in alumina, despite

having the same carbon concentration. This is an example of a chemical matrix effect and is

especially problematic for samples that appear the same (such as soils) but may exhibit small

unknown variations in composition, such as iron or aluminum content. This makes it very

difficult to produce a universal calibration curve that will encompass a broad range of soils.

Some success in correcting matrix effects has been found by normalizing signal intensities to

excitation temperatures and electron densities using the Saha-Boltzmann equilibrium

relationships.26

Physical matrix effects are also of concern in a LIBS analysis. Examples of physical

matrix effects include differences in specific heat, latent heat of vaporization, thermal

conductivity, and absorption between matrices. Each of these factors can affect the mass of

sample ablated and therefore influence quantitative results. Compensation for these physical

matrix effects can be achieved by producing a ratio between the element of interest and an

internal standard, if it can be assumed that the relative ablated masses of both remain constant.

Another method of compensation that has been reported involves acoustic monitoring of the

spark sound.27 An additional example of an important physical matrix effect is grain size, or

particle size. It has been well documented that the LIBS signal is dependant on the particle size

and arrangement (pressed pellet, free powder, etc.) of the sample.

Matrix effects are either the primary focus or at least an integral part of much of the current

research concerning LIB S. While there is still much work to be done, great progress has been

made in identifying the source of different matrix interference and conceiving different ways to

prevent or correct for them.

















E DI


Figure 1-2. Significant events during plasma formation. A) The laser pulse strikes the sample
surface. B) Energy is dissipated through the sample. C) Vaporization of the
underlying material causes the sample surface to explode. D) Plasma expands while
still absorbing energy from the residual laser pulse and a shockwave is formed. E)
Radiation is emitted. F) Plasma cools and slows. G) A crater remains after the
ablation event.


I
I
r\


1! '









case; rather they may be found tightly bound within the particle, a property that is not often

known before the analysis.so

The issue of soil grain size was also raised by Theriault et al.,51 who performed LIBS of

soil with a cone penetrometer system which probes deep into the soil. They found that the

detection limits of Pb, Cd, and Cr increased as the particle size was decreased to less than 106

pm. This variation was explained by considering the soil surface area over which the

contaminant was spread. Because a fine-grained material has a larger total surface area per unit

mass than a coarse-grained material, the total surface area over which equal contaminant

concentrations by mass are dispersed is larger for the fine-grained material and smaller for the

coarse grain material, resulting in a thicker surface layer of contamination on coarse-grain

material. If the laser spot size is fixed and ablates the contamination on the surface of the grains,

there is an overall larger volume of contaminant in the plasma for coarse-grained material,

resulting in lower detection limits.

One of the most interesting applications of LBES on soils has been for the purpose of space

exploration. It has been proposed that stand-off LIBS can be used for the exploration of

planetary surfaces that are not accessible for analysis by other techniques such as X-ray

fluorescence and alpha-proton X-ray spectrometry (APXS), which both require close proximity

to the sample, and in the case of APXS, exhibit very slow data acquisition. Because of the

similarity between the geological properties of the Martian surface and some Earth soil samples,

preliminary research regarding the feasibility of incorporating a LIBS system into a space

mission have been conducted using soil samples. Knight et al 52 Studied the effect of decreasing

pressure on both the plasma formed on the soil and the emission intensities. It was determined

that down to pressures of 1 torr there is an increase in plasma dimensions and emission signal.















Plasma Temperature
-m- 8000 K
0 -*-1200 K
16000K


% C In Quartz


r) 10 20 30 40 50 60
% C in Quartz2

12000K


70 80 90


16000K


8000 K




H-


% C in Quartz


%C in Quartz


Figure 5-4. Theoretical curves of growth for carbon in a silicon dioxide matrix at three different
plasma temperatures.






























109









concentration diffusion of species into the ambient gas. This usually occurs within hundreds of

microseconds after the plasma has been ignited.2 The temperature of the plasma and the electron

number densities are time dependant and range from 104 to 10' K and on the order of 1015 to 1019

cm-3, TOSpectively (Figure 1-2).16

The temporal properties of the LIBS plasma are reflected in the emission radiation.

During the first 1 Cps of the plasma, emission is dominated by broad background continuum

emission that is caused by bremsstrahlung and recombination radiation resulting from the

interaction between free electrons and ions as the plasma cools. While the atoms and ions in the

plasma do emit their characteristic radiation during this interval, weaker lines can not be

differentiated from the continuum. However, after approximately 1 Cps, the continuum begins to

subside while the atomic and ionic species in the plasma continue to emit. Consequently,

collection of the plasma emission at time intervals after the decay of the background continuum

results in superior signal-to-background (S/B) and narrower peaks (Figure 1-3).

Instrumentation

One of the greatest appeals of the LIB S technique is the simplicity of the instrumentation

involved in the analysis. The most basic LIBS analysis requires a pulsed laser, focusing and

collection optics, a spectral selection device and a detector. The nature of these components can

vary greatly depending on the purpose of the analysis (Figure 1-4).

Laser

The choice of laser for LBES is often dependant on the purpose of the analysis. Several

considerations must be made regarding the laser parameters, including: (1) size and portability

potential, (2) maximum pulse energy and repetition rate (3) power source and cooling

components, (4) beam quality and stability and, (5) safety. Pulsed CO2 laSers (with a wavelength

of 10.6 Cpm) and excimer lasers (typical wavelengths of 193, 248, and 308 nm) have all been









spot. This was surprising since to the naked eye, it looked as though there was no soil left to

sample after the first shot, least of all the second shot. Initially it was thought that the carbon

signal increased because the second shot was measuring only the carbon from the mounting tape,

and indeed, based on the adhesive-particle carbon enhancement reported previously, this may be

a contributing factor. However, this significant increase in signal occurred across the entire

spectrum and not just for carbon. A plot of signal intensity vs. shot number for carbon and

calcium (which was certainly not present in the tape) revealed the same trend (Figure 4-9 a,b).

The carbon signal was prevalent until around the 175th shot on the same spot and the calcium

signal was prevalent until the 80th shot. One explanation was that airborne particles that had

been liberated from the sample by the force of the plasma shockwave were either entering into

the plasma or interacting with the laser beam along its axis and forming mini-plasmas. When

watching the sampling in the dark, it was easy to ascertain these small plasma events occurring

up to an inch above the sample surface. The emission from these small independent plasmas was

likely contributing to the long lifetime of the calcium and carbon signals.

To reduce this effect, which decreased the amount of laser energy available for the next

sampling spot, a stream of air was directed over the target surface to sweep any loose particles

from the sampling areas. This drastically decreased the number of laser shots that still produced

carbon or calcium emission, however, even with the air stream carbon and calcium could be

detected until approximately the 20th shot (Figure 4-9 c,d). The greatest difference was that

without the air, the emission intensity for both elements was very random, increasing or

decreasing with no discernable pattern. Yet with the air stream, the emission intensity reached a

single maximum, usually after the first or second shot, and exponentially decreased to zero after

10- 20 shots (it takes longer for the carbon signal to disappear due to the carbon being measured




























Dedicated to my Mum and my husband, Daniel










(3 6) Radzeimski, L.J.; Loree, T.R.; Cremers, D.A.; Hoffman, N.M. Anal. Chem. 1987, 997-
981.

(37) Kaiser, J.; Samek, O.; Reale, L.; Liska, M.; Malina, R.; Ritucci, A.; Poma, A.; Tucci, A.;
Flora, F.; Lai, A.; Mancini, L.; Tromba, G.; Zanini, F.; Faenov, A.; Pikuz, T.; Cinque, G.
M~icrosc. Res. Tech. 2007, 70, 147-153.

(38) Samek, O.; Lambert, J.; Hergenroeder, R.; Liska, M.; Kaiser, J.; Novotny, K.;
Kukhlevsky, S. Laser Phys. Lett. 2006, 3, 21-25.

(39) Sun, Q.; Tran, M.; Smith, B. W.; Winefordner, J. D. Can. J. Anal. Sci. Spectros. 1999,
44, 164-170.

(40) Morel, S.; Leone, N.; Adam, P.; Amouroux, J. Appl. Opt. 2003, 42, 6184-6191.

(41) Haisch, C.; Liermann, J.; Panne, U.; Niessner, R.; Anal. Chem. Acta, 1997, 346, 23.

(42) Marquardt, B.J.; Goode, S.R.; Angel, S.M. Anal. Chem. 1996, 68, 977-981.

(43) Kraushaar, M.; Noll, R.; Schmitz, H.-U. Appl. Spectrosc. 2003, 57, 1282-1287.

(44) Eppler, A. S.; Cremers, D. A.; Hickmott, D.; Ferris, M.; Koskelo, A. Appl. Spectrosc.
1998, 50, 1175-1181.

(45) Barbini, R.; Colao, F.; Fantoni, R.; Palucci, A.; Capitelli, F. Appl. Phys. A 1999, 69, S175
S178.

(46) Bublitz, J.; Dolle, C.; Schade, W.; Hartmann, A.;Horn R. Euro. J. Soil Sci. 2001, 52, 505-
312.

(47) Capitelli, F.; Colao, F.; Provenzano, M.R; Fantoni, R.; Brunetti, G.; Senesi, N. Gerderma
2002, 106, 42-65.

(48) Gornushkin, I.B.; Kim, J.E.; Smith, B.W.; Baker, S.A.; Winefordner, J.D. Appl.
Spectrosc. 1997, 51, 1055-1059.

(49) Hilbk-Kortenbruck, F.; Noll, R.; Wintj ens, P.; Falk, H.; Becker, C. Spectrochim. Acta:
Part B 2001, 56, 933-945.

(50) Wisbrun, R.; Schechter, I.; Niessner, N.; Schroder, H.; Kompa, K. Anal. Chem. 1994, 66,
2964-2975.

(51) Therlault, G.; Bodensteiner, S.; Lieberman, S. FieldAAnal. Chem. Tech. 1998, 2, 117-125.

(52) Knight, A.; Scherbarth, N.; Cremers, D.; Ferris, M. Appl. Spectrosc. 2000, 54, 331- 340.

(53) Arp, Z.; Cremers, D.; Wiens, R.;Wayne, D.; Salle, B.; Maurice, S. Appl. Spectrosc. 2004,
58, 897 909.












4000 -


3000-



O

> 2000-




1000-






200 300 400 500 600 700 800 900

Wavelength (nm)



Figure 3-3. A soil spectrum that clearly shows the discontinuities between different spectrometer
channels in the LIBS2000+ unit.









One troubling detail evident in Figure 4-6 is the large C247 signal that was produced with

the blank quartz samples, especially for the three smaller particle size fractions. The initial

assumption was that the bulk quartz sample contained some carbon contamination. To measure

this, each of the five size fractions was analyzed by dry combustion. The results indicated that

trace amounts of carbon were present, with the concentration increasing as the particle size

decreased (Table 4-1). Indeed, the 0.5-1 Cpm quartz fraction, which was not used for the LIBS

analysis due to limited sample, had a carbon concentration of 0. 118%, 10 times greater than the

carbon concentration of the 500 Cpm fraction. In spite of this, such a contamination is still very

small, less than 0.1% for the quartz samples used in this study, and could not account for such a

large carbon signal.

To further study this, the smallest grain size fraction was pressed into a pellet and

measured by LIBS. Several cleaning shots were used to remove surface contaminants introduced

by the pellet press, and a total of 20 measurements were taken. Interestingly, the carbon peak

was not detected. Based on these results, it can be concluded that the quartz sample was not

contributing to the large carbon signal in the blank particle size fractions. The only available

source of carbon was the mounting tape, yet previous studies had exhaustively shown that no

carbon was detected on the tape after the first shot, and only a small peak was detected for

subsequent shots, the magnitude of which did not account for the large carbon signal. In spite of

this, LBES measurements were taken on bare tape and then immediately on fresh quartz covered

tape to monitor the difference in carbon measurements. As expected, no carbon was detected on

the fresh tape, however on the quartz covered tape a large carbon peak appeared.

The mechanism behind this amplified carbon detection on the tape is unknown; however, it

seems that small particles adhere to the viscous adhesive resulting in a physical or chemical









In order to determine the best Q-switch delay time for the C247.8 nm line, 15

measurements were taken on a soil sample and averaged for delays of +4 Cps to -4 Cps. From this

data, the minimum detectable concentration of carbon, CL WAS determined for each delay time

(Equation 3-1).


CL = kC(RSD))n (3-1)



S is the intensity of the 247.8 nm signal after the background, B, is subtracted, RSDB is the

relative standard deviation of the background, C is the concentration of carbon and k is a

statistical factor equal to 3. For positive Q-switch delay times, S and B remain relatively

constant (Figure 3-6). This is expected because there is no delay (emission acquisition begins at

different time periods before the plasma if formed) and as such radiation from the entire plasma

is collected each time. However, once the Q-switch delay time becomes negative and delays

occur at intervals of 500 ns, both S and B decrease and S/B increases. A Q-switch delay time of

-1 Cps was found to yield the lowest minimum detectable concentration of C and employed for

subsequent studies. It should be noted that this corresponds to a concentration of 1.4% C;

however for practical purposes C was still detected at concentrations lower than this.

Optimization of Fiber Probe Position

The placement of the fiber optic probe is a crucial parameter in LIBS analysis. The fiber

must be placed in a position that renders it as insensitive as possible to the spatial inhomogeneity

and shot-to-shot instability of the plasma. Commonly the fiber is either mounted close to the

plasma at a 30-450 angle, so called side-collection, or plasma emission is focused on to the fiber

by the same lens that focuses the laser and a dichroic mirror, termed back-side collection. Both

orientations provide unique advantages. The back-side viewing is less sensitive to spatial








correlation plots ofLIBS signal vs. % carbon for both the first and second shot data. Previous

studies on soils indicated that the C247 signal from the second shot on the same spot is

proportional to that of the first shot (by a factor of between 1.6 and 2.0) and as such, besides an

increase in signal, the correlation plots were not expected to be significantly different. This is

supported by Figure 6-2, which clearly shows a very similar trend between first and second shot

measurements. One important advantage of using the two shot sampling method is the increase

in precision that it provided relative to first shot measurements. Figure 6-3 is a histogram that

compares the range of both first and second shot RSD's associated with the average C247 peak

areas for each of the natural soil samples. From this histogram, it can be seen that the maj ority

of second shot soil measurements provide lower RSD's and thus higher precision than the same

measurements collected from only the first shot. This higher precision for the second shot

measurements is most likely due the homogenizing action of the first preparatory shot.

There are several important features to note about the plots in Figure 6-2. First, there was

no evident correlation between LIBS C247 signal and the average % carbon determined from

CHN analysis across all different soil samples. This was expected based on previous results that

established the dependence of LIBS on properties such as matrix particle size and chemical

composition. Due to this, it would not be feasible to expect one correlation plot to produce a

single calibration curve applicable to all soils, regardless of texture or chemical composition.

Second, the plots showed that there was an upper concentration limit of approximately 6%, after

which the C247 signal plateaus. This would indicate that under these measurement conditions

LBES would not be applicable to high carbon soil samples. The third point of interest from these

plots was the high carbon signal that the soils containing less than 1% carbon produced. There

are two possible explanations for this. Unlike the quartz particle size fractions, which also

produced a large signal for trace carbon, these soil samples contain iron, which has a weak










800000-

C 700000-

0 600000-

(n 500000-



4Q 00000 -

Y 200000-

100000 -


U 5 10 15 20
% Carbon


Figure 6-8. A plot of the C247 peak area normalized by surface density vs. carbon content for
measurements produced using a high resolution spectrometer (error bars are + 1
standard deviation).









were certified for carbon; therefore, all soils were submitted for dry combustion CHN analysis.

Several soil samples were also collected independently from different arbitrary locations in

Marion County, Florida (referred to as locations A, B and C). At each location, samples were

taken from depths of 1', 4' and 7' using an auger. Soil samples from around the University of

Florida campus were also collected (termed S, SE and NE, referring to the collection location

relative to the Chemistry Department) and one sample was provided by Aerodyne Research, Inc

from Massachusetts. These samples were all air dried for one week and ground with steel balls

in a ball mill for half an hour to produce a fine powder much like the NIST standards. CHN

analysis was performed several times for all nine soils and the results are reported in Table 5-1.

It can be seen that for all three sampling locations, A, B and C, the soil samples taken from 4'

and 7' contained much less carbon than samples taken from 1'. This was expected, since the soil

organic matter (SOM) decreases significantly with soil depth. As such, these samples were not

included in the LIB S analysis.

Each sample was analyzed with 100 single laser shots, each on a fresh sample location.

The average C247 signal intensity was calculated and plotted against the %C determined by

CHN analysis (Figure 5-6). From this plot, one which the calibration curves produced from

graphite spiked silica and aluminum silicate, are also imposed, two distinct groupings of the soils

emerged. The grouping first is composed of the soil samples collected from Marion County,

plus the Aerodyne soil. These soils formed a steep slope slightly greater than that of graphite in

silica. The second group of soils exhibited a slope slightly less than that of the graphite in

aluminum silicate.

The equations of the line for both silica and aluminum silicate calibration curves were used

to predict the carbon content in the soil samples based on their average C247 intensity (Table 5-









furnace is used, the gases may then be passed through a catalyst furnace to ensure complete

oxidation. The dry combustion technique can be used to measure organic carbon (Corg),

inorganic carbon (Cinorg) and total carbon by controlling the combustion temperature; Corg

combusts between 650-720oC and the remaining Cinorg is converted to CO2 between 950-1100oC.

In some cases, Corg meaSurements may be inaccurate if carbonate decomposition or incomplete

organic carbon oxidation occurs at the lower temperature. While the actual dry combustion

method is relatively standard, the method of CO2 detection and quantification varies. A

summary of these different detection procedures is presented in Table 2-1. Dry combustion

carbon recoveries for dry, organic samples exceed 95%.72

In the wet oxidation technique oxidizing agents such as KMnO4 Or K2Cr207 in acidic

solutions are used to oxidize the carbonaceous materials in the sample. The chromic acid

method, referred to as the Schollenberger method, is the most commonly used wet oxidation

technique and uses solid K2Cr207 in the presence of H2SO4 Or a mixture of H2SO4 and H3PO4

The evolved CO2 TepreSents the total Corg COntent.73'74 Another wet oxidation technique that is

widely used is the Walkley-Black method, in which a known excess of K2Cr207 is used as the

oxidant and back titration is performed to determine the amount of oxidizing agent consumed,

which is proportional to the Corg-74,75 There are many other variations on the wet oxidation

technique, including the type of oxidizing solution, the digestion temperature and duration, and

the CO2 detection methods (the same as for dry combustion, listed in Table 2-1).

It is apparent that there are many differences between the two most common methods of

carbon quantification. While it is generally accepted that the dry combustion is more accurate

and standardized than the wet oxidation method, it requires more complex instrumentation at a

greater cost, and has a longer analysis time.73 A variety of automated dry combustion















Graphite In Aluminum Silicate


40- Graphite In Sllca


A B.
S350 / 350-

300 -1 300-

250 250-

S200 -200-

C~150 -~ 150-

100- 100-

50-1 50


02610 0 24 6 810
% Carbon % Carbon




Figure 5-5. Calibration curves produced from graphite in A) silica and B) aluminum silicate

(error bars are + 1 standard deviation).









CHAPTER 1
INTRODUCTION TO LASER-INDUCED BREAKDOWN SPECTROSCOPY

Introduction

Laser-induced breakdown spectroscopy (LIBS), sometimes referred to as laser-induced

plasma spectroscopy (LIPS), is a relatively young atomic emission technique that has found great

utility in the elemental analyses of a variety of materials. In brief, LIBS is achieved by focusing

a high-powered, short-pulse laser onto a sample target to produce plasma that is rich in electrons,

atoms and ions. The atoms and ions in the plasma emit radiation that is characteristic of the

elemental composition of the sample. The development of the LIBS technique has mirrored that

of the laser, beginning with its inception during the early 1960s. While first reported as a

potential spectroscopic technique by Brech at the Xth Colloquium Spectroscopicum

Internationale,l initially the LIBS plasma found greatest value as a micro-sampling source for an

electrode-generated spark. Regardless, the first LIBS instrument was introduced in 1967,

followed by the development of several commercial LIBS instruments by Jarrell-Ash

Coorporation and VEB Carl Zeiss.2 COmmercial success was elusive as the technique was

largely overshadowed by the development of high-performance elemental analysis techniques

such as inductively-coupled plasma spectroscopy (ICP) and conventional spark spectroscopy,

which were superior to LBES in both precision and accuracy. As a result, interest in LBES was

relegated to the physics of basic plasma formation and fundamental studies of breakdown in

gases.

During the 1980s, as the laser and other spectroscopic components decreased in both size

and cost, and the drive for more portable and versatile instruments increased, LIBS experienced a

rebirth. The potential of LIBS as a more convenient atomic emission technique became apparent

to both industrial and academic laboratories. Especially in the past decade, research in every












A.
"70-


60-



m 50-


S40-


30-


20-


S2 4 6 8 10 12 14 16 18 20 22

B. % Carbon









S70-


a 60-


0 50-



40-


S2 4 6 8 10 12 14 16 18 20 22
% Carbon



Figure 6-2. LBES C247 measurements on natural soils as a function of % carbon for A) first shot
measurements and B) second shot measurements. The error bars for % carbon
represent 1 standard deviation of the average CHN measurements.









LASER-INDUCED BREAKDOWN SPECTROSCOPY FOR THE DETERMINATION OF
CARBON IN SOIL


















By

LYDIA EDWARDS


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


2007





























Titrimetric


Non-aqueous titration


Table 2-1. Techniques for measuring CO2 in COmbination with dry combustion or wet oxidation.
Information from ref. 72


Type of Analysis
Photometric


Common means of analysis
Infrared analyzer


Type of measurement
Measures differential
absorption of infrared
energy between CO2 and
reference gas
Measures mass gain in trap
after combustion
Measures volume of CO2
released after combustion
or acid digestion
Measures differential
resistance between CO2 and
reference gas
Measures amount of CO2
consumed in solution by
back-titration with standard
solution


Gravimetric

Volumetric


Sorption trap

Gas burette


Conductimetric


Thermal conductivity detector









standard deviation could be reduced if multiple samples of each soil were measured and more

laser shots were averaged.

More recent research concerning carbon in soil analysis was reported at the fifth Annual

Conference on Carbon Capture & Sequestration in 2006. Harris et al. analyzed over 200 soil

samples from several Midwest states representing a wide variety of textures and sand-silt-cly

combinations. These soil samples were analyzed by dry combustion for carbon content and

pressed into pellets for LIBS analysis. The results indicated very different calibration slopes for

sand, loam or clay soils which correlated well with synthetic soils produced to mimic these

compositions. In agreement with previous research, it was determined that the smaller grain-size

soils (clay) exhibited a smaller slope than that of large-grain size soils (sand). The conclusion

was that the LIBS signal was very sensitive to soil texture and therefore, each field would have a

different response or sensitivity and require 'in-field' calibration.

Scope of Research

While the analysis of soil samples has received a great deal of attention over the last few

years, research concerning the determination of carbon in soil has been lacking, limited to a few

proof-of-principle demonstrations by research groups at Los Alamos and Oak Ridge National

Laboratories.54-57 Since the sequestration of carbon in soil may offer a means of mitigating the

effects of global warming (to be described in Chapter 2), it is becoming increasingly urgent to

develop a practical device that can accurately measure terrestrial carbon inventories in order to

understand the status of the global carbon cycle. LIBS has been proposed as an in-field carbon

monitor due to the robust, compact and low-power properties of its instrumentation.59 Before a

Hield-based LIBS instrument can be designed there must be laboratory-based demonstration of

the capability to determine carbon in soil. The purpose of our research was to investigate the

potential of LIBS for carbon in soil analysis with an emphasis on a minimal sample preparation









the LIBS soil spectra produced during this research. In some cases, quantification of these

elements may be more easily achieved over carbon. As such, this research has provided a firm

foundation for further studies on the characteristics of LIBS on soils and may be extended to a

variety of other analytes of interest









in the mounting tape). To ensure that this increase in signal intensity after the second sampling

shot was reproducible, a suite of 38 soil samples were analyzed with 100 laser shots on the same

spot. For each soil, the maximum carbon signal was achieved after varying numbers of laser

shots (between 3-6 shots) but never after the first shot. The same pattern was seen for iron,

though the shot number for the maximum iron signal did not correspond to that of carbon (this

was achieved between shots 2 and 4).

Microscope images of the tape-mounted soil after one laser shot show that there was no

discernable crater and the soil sample had not been ablated through to the tape surface (Figure 4-

10). The tape had a very distinctive surface composed of small bubbles and was easy to

recognize. The soil sample remaining after the first shot was a very Eine powder with no

individual particles visible at 4X magnifieation. The cause of this signal enhancement after the

second shot was not fully understood, however it was hypothesized that the first laser shot

prepares the sample surface to provide for maximum ablation efficiency during the second shot.

That is to say, the energy from the laser pulse interacts with more sample in the second shot than

it does in the first shot. The particles of the original soil samples were often larger than the

diameter of the focused laser beam and as such it is plausible that the plasma was formed on the

surface of one particle, shielding any sample adhered to the surface of the tape from ablation and

possibly wasting energy from the laser pulse if only a portion of the pulse was used on the

particle during ablation (Figure 4-11). However, the combination of the initial laser pulse, the

resulting plasma shockwave and the continuous air stream sweeps the area of any large particles,

leaving a thin, uniform monolayer of soil powder. As such, when the second laser pulse interacts

with the sample, the entire laser pulse interacts with sample without any large particles

obstructing part of it. Regardless of the mechanism for the second shot signal enhancement, it








emission line that possibly interferes with the C247 line and may be contributing to the signal.

This will be discussed further in the next section. Another potential explanation for such a large

'blank' signal is the same mounting tape contribution seen in measurements taken on the quartz

samples. While microscope images indicate that even after two shots on the soil sample the tape

is not visible, it is possible that it is not necessary to actually reach the tape in order for it to

cause some interference. As was previously discussed, it seems that as the particle size of the

sample decreases there is an amplification of the carbon signal observed from the tape relative to

the carbon signal produced from the tape alone. This phenomenon was prevalent for the quartz

particle size fractions on the order of tens of microns, and was not seen in the larger fractions

which resemble most of these soils. If this is the case, then is has to be assumed that this tape

contribution is relatively constant and the contribution is the same for each shot.

A number of normalization factors were applied to the LIBS soil measurements in an

attempt to minimize the matrix effects and to observe any increase in correlation. Some success

has been previously reported using silicon as an internal standard.54-5 However, no

improvement was found when the C247 signal was normalized with the Si288 line. This was

hardly surprising considering a simple visual inspection of the soils revealed how variable the

silicon content was (the large, translucent particles which can be seen in some soil samples in

Figure 6-1 are silica particles). In order for any other soil component to be used as an internal

standard, it must be shown that it exhibits a relatively constant concentration, or that the

concentration is proportional to carbon content. It is likely that any success that has been

achieved by normalizing the carbon signal with silicon was performed on soils with very similar

chemical compositions (i.e., they were collected from similar locations, had been subj ected to the

same environmental factors).










(17) Song, K.; Lee, Y.; Sneddon, J. Appl. Spectrosc. Rev. 2002, 37, 89-117.

(18) Yamamoto, K.Y.; Cremers, D.A.; Foster, L.E.; Ferris, M.J.; Appl. Spectrosc. 1996, 50,
222-233.

(19) Carranza, J.E.; Gibb, E.; Smith, B.W.; Hahn, D.W, Winefordner, J.D. Appl. Opt., 2003,
42, 6016-6021.

(20) Burgio, L.; Melessanaka, K.; Doulgendis, M.; Clark, R.J.H.; Anglos, D.; Spectrochim.
Acta 2001, 905, B56.

(21) Angelos, D. Appl. Spectrosc. 2001, 55, 1864.

(22) Angelos, D.; Balas, C.; Fotakis, C. Am. Lab. 1999, 31, 60-67.

(23) Giakoumaki, A.; Melessanaki, K.; Anglos, D. Anal. Bioanal. Chem. 2007, 387, 749-
760.

(24) Lopez-Moreno, C.; Palanco, S.; Laserna, J.J.; Delucia, F.; Mikiolek, A.W.; Rose, J.;
Walters, R.A.; Whitehouse, A.L.;J. Anal. At. Spectrosc. 2006, 21, 55-60.

(25) Castle, B.C.; Talabardon, K.; Smith, B.W.; Winefordner, J.D. Appl. Spectrosc. 1998, 52,
649-657.

(26) Chaleard, C.; Mauhien, P.; Andre, N.; Uebbing, J.; Lacour, J.L.; Geertsen, C.; J. Anal. At.
Spectrom. 1997, 12, 183-188.

(27) Panne, U.; Haisch, C.; Clara, M.; Niessner, R.; Spectrochim. Acta, Part B. 1998, 53B,
1969-1981.

(28) Gondal, M. A.; Hussain, T. Talanta 2007, 72, 642-649.

(29) Rai, V.N.; Yueh, F.Y.; Singh, J.P. Appl. Opt 2003, 42, 2094-2101.

(30) Knopp, R.; Scherbaum, F.J.; Kim, J.I. Fres. J. Anal. Chem. 1996, 355, 16-20.

(31) Pichahchy, A.E.; Cremers, D.A. Spectrochim. Acta 1997, 25-39.

(32) Caceres, J.O.; Tornero Lopez, J.; Telle, H.H.; Gonzalez Urena, A. Spectrochim. Acta B
2001, 56, 831-838.

(33) Neuhauser, R.E.; Panne, U.; Niessner, R. Anal. Chim. Acta 1999, 392, 47-54.

(34) Carranza, J.E.; Fisher, B.T.; Yoder, G.D.; Hahn, D.A. Spectrochim. Acta B 2001, 56,
851-864.

(35) Panne, U.; Neuhauser, R.E.; Theisen, M.; Fink, H.; Niessner, R.; Spectrochim. Acta B.
2001, 56, 839-850.









thank my step-father, Keith, my grandparents, Roy and Jean, and my brothers and sisters, Elliot,

Yvonne and Daryll who are my motivation.

Finally, I would like to acknowledge my husband, Daniel, without whom none of this

would be possible. Because of his love and support, seven years ago I took a chance and stayed

in America while my family moved to England, and I' d like to think that none of this would have

been possible if not for that.









the C247 signal intensity from the first and second shots were proportional, with an average

increase of 1.7 observed for the second shot. However, due to the weak nature of the C247 line,

the increase in sensitivity was exploited and second shot measurements were employed for the

majority of this research.


















2000 -

1800-
Fe
1600-1 247 836

1400 -Fe
248 30439
1200- I 25127
Fe
1000-24261 2499

800 -

600-

400 ,
24/ 0 247 5 248 0 248 5 2

Wavelength, nm


1600-

1400-

1200-

1000-

800-

600-

400-

200-

O
24 0


247 83695


249 29001


249 855


49 0 2495 2500


247 5 248 0

Wavelength, nm


248 5 249 0


249 29001


200-



2470 2475 2480 2485 2490 2495 2500
Wavelength, nm


Figure 6-5. Extent of the iron spectral interference. A) Pure iron foil. B) Pure graphite and C)

An iron and graphite mixture.










































126


1400-
247 83695

1200-


u,1000-

so -00 Fe
1 I \248 30


d6000 -247 26515












1100-
CN bands
1050-

1000-

950-

c 900-

850-

c 800-

750-

700-

650-

600
385 386 387 388 389 390

Wavelength, nm




Figure 5-3. The CN molecule, with a band head at 388 nm, is barely detected in a soil sample.









can be incorporated into a robust, portable system. The spectral range of the spectrometer is

another important factor, especially when multi-element analysis is to be performed. In general,

the larger the spectral window, the more information can be retrieved from the sample.

Detectors

The type of detector that is used for LIBS is generally dependent upon the spectrometer.

Photodiodes (PD) and photomultiplier tubes (PMT) are highly sensitive devices that produce an

amplified electrical signal proportional to incident photons. While these types of detectors offer

valuable advantages, such as sub-nanosecond resolution (useful for temporal-based plasma

studies), each device can only monitor one wavelength at a time, precluding broad spectral

bandwidths. A more common type of detector is one that involves placing many photosensitive

elements in an array at the exit slit of the spectrometer. Such an assembly will respond to a

much wider spectral range and therefore is more convenient for multi-element analysis.

Examples include photodiode arrays (PDA), charge-inj section devices (CID) and charge-coupled

devices (CCD). One of the major limitations of these devices is that they must accumulate

incident radiation for a period of time before the readout is processed, and when it is processed,

each pixel is read sequentially, further increasing the response time. Such time-integration

precludes detailed temporal studies of the plasma emission. In order to achieve time-resolved

LBES, a micro-channel plate (MCP) must be coupled to the detection system. The MCP

regulates when the incident radiation is allowed to reach the detection device, and in doing so it

also amplifies the light by converting it to electrons which are then reconverted back to light

before detection. Combining an MCP with a time-integrating detector such as a CCD is referred

to as an intensified charged-coupled device or (ICCD).19









was clear that sensitivity could be increased across the spectrum by utilizing repetitive shot

sampling; one initial preparatory shot and a second shot on the same spot from which data is

collected.

Conclusion

As a result of the preceding observations, a sampling regime optimized for possible soil

analysis was evaluated. The advantages of mounting the soil samples on a microscope with

adhesive tape, namely simplicity and maintaining particle size integrity far outweighed pressing

the samples into pellets. Furthermore, it was determined that for at least one soil sample, the

precision of C247 LBES analysis on the slide exceeded that of the same soil in pellet form.

While this may not be the case for all soils, it was deemed an acceptable trade-off considering

the ease with which slides could be prepared in-Hield. While evidence indicates that there may

be some physical or chemical process occurring between the small sample particles and the tape

adhesive that resulted in enhanced carbon signal from the tape, this was assumed to be a constant

contribution for each laser shot. A study of matrix grain size on LIBS signal revealed that there

was a significant effect on the C247 signal intensity and RSD. Ideally, when measuring different

soils they should be all be ground and homogenized to approximately the same size in order to

minimize grain size effects between samples and to increase precision. However, since this

would not be a feasible option for in-field analysis, care should be taken that only soils

containing similar particle size distributions are compared.

With respect to the sampling population, it was established that a minimum of 20 laser

shots should be taken on a soil sample in order to account for the heterogeneity of soils. Taking

many more than 20 laser shots did not seem to improve the RSD. Finally, it was determined that

using a preparatory laser shot on the soil surface before the sampling shot significantly increased

the signal intensity of the entire spectrum. It should be noted however that for 38 soil samples,









LIST OF ABBREVIATIONS

AAS Atomic absorption spectrometry

AES Atomic emission spectrometry

CCD Charge coupled device

COG Curve of growth

F TSD Fiber-to-sample distance

FWHM Full-width at half-maximum

ICCD Intensified charge coupled device

ICP Inductively coupled plasma

LBES Laser-induced breakdown spectroscopy

LIF Laser-induced fluorescence

LTE Local thermal equilibrium

LTSD Lens-to-sample distance

Nd:YAG Neodymium doped yttrium-aluminum-garnet

OES Optical emission spectroscopy

PSI Pounds-per-square inch

RSD Relative standard deviation

S/B Signal-to-background ratio

SOM Soil organic matter










spectra almost instantaneously after each laser shot. The laser can be run in external trigger

mode, which means that the software specifies the laser repetition rate and the acquisition delay

time. The software can also control the number of laser shots fired, the number of cleaning shots

used, any necessary background correction, and average or individual data collection. There is a

spectral line database available to identify specific lines; however at the time of this research, it

had not been developed enough to provide accurate results and therefore was not used. In

addition to this there is the option to monitor specific lines during the analysis, including

calculating the mean intensity and area, standard deviation and RSD after each subsequent laser

shot. This is especially helpful for rapid analysis (the mean intensity of a chosen line after 100

shots is immediately available after the last shot); however for this research a more advanced

data analysis program separate from the OOILIBS software was used.

Optimizing Instrumental Parameters

Laser Energy

Before any measurement can be undertaken, it was necessary to determine the stability of

the laser pulse. While the manual specifications indicated that the stability was 2%, there are

many factors that may change this, including the age and use of the laser and the chosen energy.

Instability in the laser pulse energy will result in variable plasmas and poor precision; therefore it

is essential to ensure that the laser exhibits a tolerable energy instability.

To measure this, the focusing lens was removed from the LIBS system and a power meter

was positioned on the sample stage. The laser was run in continuous mode at a repetition rate of

1 Hz and an oscilloscope displayed the measured energy. Thirty-two measurements were taken

at expected laser energies of 30, 35, 40, 45 and 50 mJ. The mean energy, standard deviation and

relative standard deviation (RSD) for all laser shots, the first ten laser shots and the remaining 22

laser shots are reported in Table 3-2. Surprisingly, it was determined that the maximum laser









individuals are responsible for reducing emissions and/or creating carbon sinks and trading

carbon credits to meet these targets.67 Potential carbon stocks have been valued at

approximately $8 billion in the agricultural industry alone.68,69

While terrestrial soil is the second largest carbon sink, it is currently not recognized as a

sink that can be traded as an off-set of carbon emissions. The reason for this is simple: reliable

measurements of soil carbon stocks on a national scale are scarce, and high-precision estimates

of annual carbon inputs to the soil are even scarcer.70 Izaurralde et al. concluded that although

there is a "compelling scientific basis for believing that a substantial Canadian C sequestering

activity is occurring, the science and supporting measurement protocols have not been developed

to the point that this can be adequately quantified at the farm or regional scale and accepted for

marketing purposes."n The Soil Science Society of America maintains that in order for carbon

sequestration in soil to effectively reduce atmospheric CO2 mOre research is needed to better

quantify carbon sequestration and improved monitoring and verification protocols must be

established.66 As such, while sequestering carbon in soil is one of the most promising short-term

methods to mitigate increasing atmospheric CO2, it is severely limited by the lack of analytical

techniques to provide quantitatively defensible carbon levels in soil.

Current Methods for the Determination of Carbon in Soil

Typical analytical methods for the determination of carbon in soils and sediments rely on

converting carbonaceous material to CO2 followed by direct or indirect measurement.72 The

two most common methods for carbon analysis are dry combustion and wet oxidation coupled

with a variety of different methods to measure evolved CO2.

Dry combustion involves combusting the soil sample in a stream of CO2-free Oz gaS at

temperatures above 900oC. These resulting combustion gases are passed through a series of

traps that are designed to remove particulates and other interference. If a medium temperature











LIST OF FIGURES


Figure page

1-1 Number of LIB S publications since 1960 ................. ...............35........... .

1-2 Significant events during plasma formation.. ............ ...............36.....

1-3 Temporal evolution of the LIB S plasma ................. ...............37..............

1-4 Typical schematic of a LBES set-up ................. ...............38..............

2-1 Magnitudes of the reservoirs of actively cycling carbon in gigatons (Gt). ................... ....47

2-2 Carbon cycle. ............. ...............48.....

2-3 Constituents of humus, the major contribution to soil organic carbon. ................... ..........51

3-1 Configuration of LBE S instrumentation. ........... .....__ .. ...............62

3-2 Ablation area produced on copper foil by a 50 mJ laser pulse and a 2.5 inch focal
length lens.. ............ ...............63.....

3-3 A soil spectrum that clearly shows the discontinuities between different spectrometer
channels in the LIBS2000+ unit. ............. ...............64.....

3-4 Arrangement of the individual fiber apertures in the fiber optic probe ........._..._................65

3-5 Controlling the CCD emission acquisition delay time by manipulating the laser Q-
sw itch. .............. ...............67....

3-6 Results of changing the spectrometer delay times on soil spectra ................. ................68

3-7 Graphite pellet spectra .............. ...............69....

3-8 C247 signal intensity (the average of 10 shots) as a function of focal point offset...........70

3-9 Calculation of intensity (I), background (B) and area (above the red line). ................... ...71

4-1 An image of a pressed pellet and soil mounted on a microscope slide..............................86

4-2 C247 signal as a function of shot number on the same pellet spot for five different
locations. ............. ...............87.....

4-3 Analysis of bare mounting tape.. ............ ...............88.....

4-4 Microscope images (4X magnification) of the craters produced after the indicated
shot number ................. ...............89.................

































Figure 3-4. Arrangement of the individual fiber apertures in the fiber optic probe









carbon determination; however, it did not take into account many other variables, including

varying soil grain size, applicability of the calibration curve to a broad range of different soil

samples, differing Si content, and Fe interference with the C247.8nm line.

Subsequent research by the same group investigated the use of the C193 nm line, which is

not susceptible to interference from Fe as is the C247.8 nm line In addition to this,

standardization of the C signal was achieved by taking the ratio of said line to the sum of the

All99 nm and Si212 nm lines, C/(Al+Si). The soils chosen were noted to have similar physical

and chemical properties. The resulting C signal was well correlated with the C concentration

determined by dry combustion (R2=0.95) and exhibited great reproducibility over a 30 day

period. It was acknowledged that the Al and Si content of different soil samples will vary as a

function of the soil's mineralogy and texture, which would significantly limit the use of both Si

and Al as internal standards. However, it was assumed the those properties were constant for

soils as similar as the ones employed in this study.56

Martin et al. extended on the results of this previous research by determining carbon and

nitrogen in a variety of different soils. A Q-switched Nd:YAG laser was used at 266 nm with

typical pulse energies of 23 mJ/pulse. Emission collection was via a fiber optic bundle and

detection was achieved with an ICCD. Because this study was only concerned with the soil

organic matter (SOM), inorganic soil carbon was first removed by acid washing. Soil was then

dried, ground and pressed into pellets. The C247.8 nm emission was averaged from 10 laser

shots and the resulting calibration curves showed good correlation up to 4.3% (the highest

concentration of carbon). The RSD's for the soil samples were higher for the low C

concentration soils due to the increased interference of the Fe peak. It was found that the










(54) Cremers, D.; Ebinger, M.; Breshears, D.; Unkefer, P.; Kammerdiener, S.; Ferris, M.;
Catlett, K.; Brown, J. J. Environ. Equal. 2001, 30, 2202 2206.

(55) Ebinger, M.H.; Breshears, D.D.; Cremers, D.A.; Unkefer, P.J. Earth and Environmental
Sciences, Progress Report, 1998 2000

(56) Ebinger, M.; Lee Norfleet, M.; Breshears, D.; Cremers, D.; Unkefer, P.; Kammerdiener,
S.; Ferris, M.; Unkefer, P.; Lamb, M.; Goddard, K.; Meyer, C.; Am. J. Soil Sc. 2003, 67,
1616 -1619.

(57) Martin, M.; Wullschleger, S.; Garten, C.; Palumbo, A. Appl. Opt. 2003, 42, 2072 2077.

(58) Harris, R.; Martin, M.; Labbe, N.; Ebinger, M.; Clegg, S.; Wullschleger, S. Advances in
Calibrating Laser-Induced Breakdown Spectroscopy (LIBS) for Measuring Total Soil
Carbon. In Steps Toward Deployment, Proceedings of the 5th Carbon Capture &
Sequestration Conference, Alexandria, Virginia, May 8-11, 2006.

(59) Freedman, A.; Wormhoudt, J.C. Portable Soil Carbon Monitor, Aerodyne Research, Inc.,
Billerica, MA, 2006.

(60) Bruce, J.; Frome, M.; Haites, E.; Janzen, H.; Lal, R.; Paustian, K. Carbon sequestration in
soils. In Carbon Sequestration in Soils Workshop, Proceedings of the Soil and Water
Conservation Society Meeting, Calgary, Alberta, May 21-22, 1998.

(61) Sulzman, E.W. The Carbon Cycle Global Change Instruction Program, 2000, University
Corporation for Atmospheric Research: Boulder, CO.

(70) Assessment M~ethods for Soil Carbon; Lal, R.; Kimble, J.M.; Follett, R.F.; Stewart, B.A.,
Eds.; Advances in Soil Sciences; Lewis Publishers: Boca Raton, FL, 1998.

(71) Izzaurralde, R.C.; McGill, W.B.; Bryden, A.; Graham, S.; Ward, M.; Dickey, P. In
Management of Carbon Sequestration in Soil. Advances in Soil Science; Lal, R Kimble,
J.; Follett, J.; Stewart, R. Eds.; Lewis Publishers: Boca Raton, FL, 1998; pp 433-446.

(72) Charles, J.M.; Simmons, M. S. Analyst 1986, 111, 385-390.

(73) Schollenberger, C.J. Soil Sci. 1927, 24, 65-68.

(74) Tan, K.H. Soil Salmpling, Preparation, and Analysis; Marcel Dekker, Inc.: New York,
1996.

(75) Walkley, A.L.; Black, A. Soil Sci. 1947, 63, 251-263.

(76) Viscarra Rosel, R.A.; Walvoort, D.J.J.; McBratney,A.B.; Janik,L.J.; Skj emstad, J.O.
Geod'ernza 2006, 131, 59-75

(77) Reeves, J.B.; Follett,R.F.; McCarty,G.W.; Kimble, J.M. Conan. Soil Sci. Plant Anal.
2006, 37, 2307-2325.










111


21~


Figure 6-1. Microscope images of natural soil samples (3.5X magnified).


C,


















~CL


Laser Pulse


I I


Plasma


Figure 4-11. Proposed mechanism for the second shot signal enhancement on soil.




Full Text

PAGE 1

1 LASER-INDUCED BREAKDOWN SPECTROSC OPY FOR THE DETERMINATION OF CARBON IN SOIL By LYDIA EDWARDS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 2007 Lydia Edwards

PAGE 3

3 Dedicated to my Mum and my husband, Daniel

PAGE 4

4 ACKNOWLEDGMENTS I extend my deepest gratitude to my a dvisor, Dr. James D. Winefordner for his unwavering support and guidance during my four years in his research group. His knowledge and patience are legendary and have contribut ed to a research environment that fosters cooperation and mutual respect. I am incredibly proud to have been a member of his group, and even more honored to have been one of his last students before his reti rement. I have great appreciation for Dr. Nicolo Omenetto and Dr. Ben Smith, who have nurtured my interest in applications of spectroscopy and provided countle ss insightful observations about this research. I also extend a special thank you to Dr. Igor Gornushkin, who was so patient with me and wrote data analysis programs that saved me from weeks, if not months, of incredibly tedious and repetitive data processing. Additionally I th ank Dr. Andy Freedman and Dr. Joda Wormhoudt of Aerodyne Research, Inc, who initially proposed and provided funding for this research. I thank all the past and present members of the Winefordner/Omenetto/Smith research group, who have become like a second family to me over the years. I have learned so much from each of them beyond chemistry and know that lif e-long friendships have been developed. I directly thank Mariela Rodriguez for understa nding the frustrations involved in LIBS applications and for her many illuminating ideas, and also Benoit Lauly for his elucidation of plasma fundamentals. Past group members whom I acknowledge for significantly influencing my research include Dr. Galan Moore and Dr. Kirby Amponsah-Manager. No list of acknowledgements would be complete without thanking Jeanne Karably and Lori Clark for keeping the often forgotten paper-work in line. I cannot begin to express that gratitude I owe to my family for their love and support during my time in both undergraduate and graduate school. I owe so much to my Mum, Kay, who is my inspiration and never lets me forget that anything can be accomplished. I especially

PAGE 5

5 thank my step-father, Keith, my grandparents, Roy and Jean, and my brothers and sisters, Elliot, Yvonne and Daryll who are my motivation. Finally, I would like to acknowledge my husband, Daniel, w ithout whom none of this would be possible. Because of his love and supp ort, seven years ago I took a chance and stayed in America while my family moved to England, and Id like to think that none of this would have been possible if not for that.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES.........................................................................................................................9 ABSTRACT...................................................................................................................................13 CHAP TER 1. INTRODUCTION TO LASER-INDUCED BREAKDOWN SPECTROSCOPY ................ 15 Introduction................................................................................................................... ..........15 Characteristics of LIBS...........................................................................................................16 Instrumentation................................................................................................................ .......18 Laser................................................................................................................................18 Spectral Resolution.......................................................................................................... 19 Detectors..........................................................................................................................20 Advantages of LIBS............................................................................................................. ..21 Sample Preparation..........................................................................................................21 Micro-destructive Nature.................................................................................................22 Portable and Stand-off LIBS........................................................................................... 22 Rapid Analysis.................................................................................................................23 Challenges to LIBS............................................................................................................. ....23 Precision..........................................................................................................................24 Matrix Effects..................................................................................................................24 Applications of LIBS for Soil Analysis .................................................................................. 26 Scope of Research...................................................................................................................33 2. CARBON IN SOIL................................................................................................................. 39 Sequestration of Carbon in Soil..............................................................................................39 Current Methods for the Determination of Carbon in Soil..................................................... 42 The Challenges of Soil Analysis............................................................................................. 44 3. INSTRUMENTATION AND DATA ANALYSIS................................................................ 52 Instrumental Components.......................................................................................................52 Laser................................................................................................................................52 Optics...............................................................................................................................53 Fiber Optics.....................................................................................................................53 Spectrometer................................................................................................................... .53 Software...........................................................................................................................54 Optimizing Instrumental Parameters...................................................................................... 55

PAGE 7

7 Laser Energy....................................................................................................................55 Optimization of the Q-Switch Delay............................................................................... 56 Optimization of Fiber Probe Position.............................................................................. 57 Optimization of Lens-to-Sample Distance...................................................................... 59 Data Analysis..........................................................................................................................60 4. SAMPLE PRESENTATION AND SA MPLING CHARACTERISTICS ............................. 72 Sample Presentation............................................................................................................ ....72 Pressed Pellets vs. Mounting Tape..................................................................................72 Analysis of Bare Mounting Tape....................................................................................75 Grain Size..................................................................................................................... ...76 Sampling Characteristics........................................................................................................80 Number of Laser Shots.................................................................................................... 80 Repetitive Laser Shots..................................................................................................... 81 Conclusion..............................................................................................................................84 5. THE SPECTROSCOPY OF CARBON AN D CALIBR ATION CURVES WITH SIMULATED SOILS............................................................................................................. 99 The Spectroscopy of Carbon..................................................................................................99 Choice of Analytical Line...............................................................................................99 Curves of Growth..........................................................................................................101 Experimental Calibration Curves......................................................................................... 102 Silica and Aluminum Silicate Artificial Soils............................................................... 102 Analysis of Powdered Soil Samples.............................................................................. 103 6. LIBS MEASUREMENTS OF NATURAL SOILS: IS QUANTIFICATION POSSIBLE? ...................................................................................................................... ....114 Introduction................................................................................................................... ........114 CHN Analysis................................................................................................................... ....114 LIBS Measurements of Natural Soils................................................................................... 115 High-Resolution LIBS Measurements oF Soils................................................................... 118 7. CONCLUSION AND FUTURE WORK............................................................................. 130 LIST OF REFERENCES.135 BIOGRAPHICAL SKETCH.......................................................................................................140

PAGE 8

8 LIST OF TABLES Table page 2-1 Techniques for measuring CO2 in combination with dry combustion or wet oxidation. Reproduced from ref. 72.................................................................................. 49 2-2 The three major groups of soil size fracti ons are sand, silt and clay. These can be further subdivided into finer size fractions. .......................................................................50 3-1 Specifications for the Big Sky Ultra Nd:YAG pulsed laser.............................................. 61 3-2 Mean pulse energies of the Big S ky laser m easured with a power meter.......................... 66 4-1 Statistics for the analysis of quart z size fractions spiked with graphite ............................ 92 4-2 C247 statistics for intra-sample and inter-sample measurements...................................... 95 5-1 Carbon content of soil samples determined by CHN analysis......................................... 111 5-2 Predicted carbon content of soils using the applicable silica or alum inum silicate calibration curve...............................................................................................................113 6-1 Results of dry combustion CHN analysis performed on a suite of natural soil sam ples........................................................................................................................ .....122

PAGE 9

9 LIST OF FIGURES Figure page 1-1 Number of LIBS publications since 1960. ......................................................................... 35 1-2 Significant events during plasma formation...................................................................... 36 1-3 Temporal evolution of the LIBS plasma............................................................................ 37 1-4 Typical schematic of a LIBS set-up...................................................................................38 2-1 Magnitudes of the reserv oirs of actively cycling carbon in gigatons (Gt). ......................47 2-2 Carbon cycle............................................................................................................... .......48 2-3 Constituents of humus, the majo r contribu tion to soil organic carbon.............................. 51 3-1 Configuration of LIBS instrumentation............................................................................. 62 3-2 Ablation area produced on copper foil by a 50 m J laser pulse and a 2.5 inch focal length lens.................................................................................................................... ......63 3-3 A soil spectrum that clearly shows the di scontinuities between di fferent spectrom eter channels in the LIBS2000+ unit........................................................................................ 64 3-4 Arrangement of the individual fibe r apertures in the fiber optic probe ............................. 65 3-5 Controlling the CCD emission acquisiti on delay tim e by manipulating the laser Qswitch.................................................................................................................................67 3-6 Results of changing the spectro m eter delay times on soil spectra..................................... 68 3-7 Graphite pellet spectra.................................................................................................... ...69 3-8 C247 signal intensity (the av erage of 10 shots) as a function of focal point offset. .......... 70 3-9 Calculation of intensity (I), back ground (B) and area (above the red line). ...................... 71 4-1 An image of a pressed pellet and soil m ounted on a microscope slide.............................. 86 4-2 C247 signal as a function of shot number on the sam e pellet spot for five different locations.............................................................................................................................87 4-3 Analysis of bare mounting tape......................................................................................... 88 4-4 Microscope images (4X magnification) of the craters produced after the indicated shot num ber.................................................................................................................... ....89

PAGE 10

10 4-5 Images of different quartz size fractions spiked with graphite (4X m agnification)..........90 4-6 C247 peak area as a function of carbon c oncentration (wt %) for different particle sizes of quartz m atrix......................................................................................................... 91 4-7 Microscope images of three so ils to illustrate heterogeneity. ............................................ 93 4-8 RSD of the C247 signal intensity as a f unction of the num ber of shots on separate spots...................................................................................................................................94 4-9 C247 and Ca422 signal intensities as a f unction of shot num ber on the same spot..........96 4-10 Microscope images of soil af ter the indi cated number of laser shots (4X magnification)................................................................................................................. ...97 4-11 Proposed mechanism for the sec ond shot signal enhancem ent on soil.............................. 98 5-1 Full spectrum of carbon taken on a graphite pellet with prom inent lines and bands identified..........................................................................................................................106 5-2 C247 peak area as a function of spectrome ter delay tim e. The integration time is fixed at 200 ns................................................................................................................ ..107 5-3 The CN molecule, with a band head at 388 nm is barely detected in a soil sample....... 108 5-4 Theoretical curves of growth for carbon in a silicon dioxide m atrix at three different plasma temperatures......................................................................................................... 109 5-5 Calibration curves produced from graphite in................................................................. 110 5-6 Results of LIBS measurements on finely ground soil samples........................................ 112 6-1 Microscope images of natura l soil samples (3.5X m agnified)........................................ 121 6-2 A histogram of the RSDs from LIBS C247 m easurements on natural soils using the first shot and the second shot........................................................................................... 124 6-3 LIBS C247 peak area as a function of % carbon............................................................. 123 6-4 Normalizing the C247 peak area by soil surface density increa ses the correlation between LIBS signal an d carbon concentration............................................................... 125 6-5 Extent of the iron sp ectral interference. ..........................................................................126 6-6 A soil spectrum produced using the Ac ton SpectraPro showing the resolved carbon and iron lines. ................................................................................................................ ...127 6-7 LIBS C247 peak area as a function of carbon concentration for m easurements produced using a high reso lution spectrometer............................................................... 128

PAGE 11

11 6-8 A plot of the C247 peak area normali zed by surface density vs. carbon content for m easurements produced using a high resolution spectrometer........................................ 129

PAGE 12

12 LIST OF ABBREVIATIONS AAS Atomic absorption spectrometry AES Atomic emission spectrometry CCD Charge coupled device COG Curve of growth FTSD Fiber-to-sample distance FWHM Full-width at half-maximum ICCD Intensified charge coupled device ICP Inductively coupled plasma LIBS Laser-induced breakdown spectroscopy LIF Laser-induced fluorescence LTE Local thermal equilibrium LTSD Lens-to-sample distance Nd:YAG Neodymium doped yttrium-aluminum-garnet OES Optical emission spectroscopy PSI Pounds-per-square inch RSD Relative standard deviation S/B Signal-to-background ratio SOM Soil organic matter

PAGE 13

13 Abstract of Dissertation Pres ented 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 SPECTROSC OPY FOR THE DETERMINATION OF CARBON IN SOIL By Lydia Edwards December 2007 Chair: James D. Winefordner Major: Chemistry Laser-induced breakdown spectroscopy (LIB S) is a relatively young atomic emission technique that has found great utility in the elemental analyses of a variety of materials. In brief, LIBS is achieved by focusing a high-powered, shor t-pulse laser onto a samp le surface to produce plasma that is rich in electrons, atoms and ions. The atoms and ions in the plasma emit radiation that is characteristic of the elemental com position of the sample. While LIBS has been extensively applied to soil anal ysis, its use as a means of determining carbon in soil has been lacking, limited to a few proof-ofprinciple demonstrations by re search groups at Los Alamos and Oak Ridge National Laboratories. Since th e sequestration of car bon in soil may offer a means of mitigating the effects of global warming, it is beco ming increasingly urgent to develop a practical device that can accurately measure terrestrial carbon inve ntories in order to understand the status of the global carbon cycle. Due to the unique advantages of LIBS, including the potential for in-situ analysis, little-to-no sample preparation, simultaneous multielement detection and relatively simple instrume ntation, the technique ha s the potential to fulfill the requirements of an in-field carbon monitor. Before a field-based LIBS instrument can be designed there must be a laboratory-based demons tration of the capabilit y to determine carbon in soil.

PAGE 14

14 The purpose of our research was to investig ate the potential of LI BS for carbon in soil determination with an emphasis on minimal sample preparation and the potential for in-situ analysis. While the quantification of carbon in soil by LIBS was not achieved in this study, an optimal soil sampling regime was established that increased sensitivity and measurement precision. The LIBS instrumentation was optimi zed specifically for so il samples, including the laser energy, optical configura tion and spectrometer delay time. The sample presentation was extensively studied, including the effect of soil grain size on LIBS signal and the advantages of mounting the sample on tape as a thin monol ayer rather than as a pressed pellet.

PAGE 15

15 CHAPTER 1 INTRODUCTION TO LASER-INDUCED BREAKDOWN SPECTROSCOPY Introduction Laser-induced breakdown spectroscopy (LIBS) som etimes referred to as laser-induced plasma spectroscopy (LIPS), is a relatively young atomic emission technique that has found great utility in the elemental analyses of a variety of ma terials. In brief, LIBS is achieved by focusing a high-powered, short-pulse laser onto a sample target to produce plasma that is rich in electrons, atoms and ions. The atoms and ions in the plasma emit radiation that is characteristic of the elemental composition of the sample. The development of the LIBS technique has mirrored that of the laser, beginning with its inception during the early 1960s. While first reported as a potential spectroscopic te chnique by Brech at the Xt h Colloquium Spectroscopicum Internationale,1 initially the LIBS plasma found greatest value as a micro-sampling source for an electrode-generated spark. Regardless, the first LIBS in strument was introduced in 1967, followed by the development of several comm ercial LIBS instruments by Jarrell-Ash Coorporation and VEB Carl Zeiss.2 Commercial success was elusive as the technique was largely overshadowed by the development of high-performance elemental analysis techniques such as inductively-coupled plasma spectroscopy (ICP) and conventional spark spectroscopy, which were superior to LIBS in both precision and accuracy. As a result, interest in LIBS was relegated to the physics of basic plasma forma tion and fundamental studies of breakdown in gases. During the 1980s, as the laser and other spec troscopic components decreased in both size and cost, and the drive for more portable and vers atile instruments increased, LIBS experienced a rebirth. The potential of LIBS as a more convenient atomic emission technique became apparent to both industrial and academic la boratories. Especially in the past decade, research in every

PAGE 16

16 area of LIBS has seen remarkable progress, in cluding plasma characterization, instrumentation, and applicability. Indeed, it is the broad range of the latter that is fueling the increased interest in LIBS, which has been successfully applied to an extraordinary variety of samples, from priceless works of art to sewage; from environmental aerosols to forensic analysis. The revival of LIBS is reflected by the numbers of publishe d journal articles on the subject, which began to increase in the late 1980s and continues to do so each year thereafter, as illustrated in Figure 11.3 Today the potential of LIBS as a qualitativ e and potentially quantitative elemental analysis technique is well documen ted in virtually every scientific field. The unique advantages of LIBS that have contributed to widespread inte rest include remote sens ing capabilities, in-situ analysis, little-to-no sample prep aration, micro-destructive nature applicability to all media, simultaneous multi-element detection capability a nd relatively simple instrumentation. While LIBS may currently be limited by its poor precis ion (typically 5-10%), hi gher detection limits (relative to conventional techniqu es), and difficulty in preparing matrix-matched standards, the preponderance of applications reported in the literature may indicate that the advantages overcome the limitations. Characteristics of LIBS The principles of laser ablation and plasm a formation on a solid target have been extensively covered in previous literature.4-11 As such, only a brief description of the mechanisms that contribute to atomic emissi on during the LIBS technique will be covered. LIBS is an atomic emission spectroscopic (AES ) technique that shares similar fundamental principles to ICP-OES, microwave induced plas ma (MIP-OES), direct current plasma (DCPOES), flame-OES, arc-OES and spark-OES. All atomic emission techniques exhibit the following basic steps:

PAGE 17

17 atomization/vaporization of the sample to pr oduce free atomic species consisting of atoms and ions, excitation of these atoms, detection of the radiation emitted by these atomic species, calibration of the intensity to c oncentration or mass relationship, determination of concentration, masses, or other information.4 In LIBS, the first step, atomization and vaporiz ation of the sample, as well as the second step, atomic excitation are both achieved by focu sing a short-pulsed laser of adequate power density onto the surface of a sample. Laser pulses of 5-10 ns are most commonly employed. Increasing research is being performed on LI BS achieved with picosecond or femtosecond lasers.11-15 Once the laser pulse strikes the target, si ngleor multi-photon absorption, dielectric breakdown and other, often undefined mechanis ms contribute to the surface temperature instantly increasing beyond the vaporization temperature of th e material. The subsequent dissipation of energy through vaporization is slow relative to the rate of energy deposition from the laser pulse. This results in vaporization of the surface layer and causes the underlying material to reach critical temperature and pressure, finally causing the surface to explode. Particles, free electrons and highly ionized atoms are released into the surrounding atmosphere and begin to expand at a velocity much faster than the speed of sound, forming a shockwave in the process. If the laser pulse is of long enough duration, as is the case for nanosecond lasers, the expanding plume and the material within will co ntinue to absorb energy from the laser and intense plasma up to a few millimeters in height is visually observed. After several microseconds, collisions with ambient gas specie s cause the plasma plum e expansion to slow down while the shockwave detaches from the plasma front and continues pr opagating at a rate of approximately 105 m/s. As the plasma begins to decay through radiative, quenching, and electron-ion recombination proce sses, high-density neutral specie s and clusters of dimers or trimers are formed via condensation and thre e-body collisions and with the thermal and

PAGE 18

18 concentration diffusion of species into the ambient gas. This usually occurs within hundreds of microseconds after the plasma has been ignited.2 The temperature of the plasma and the electron number densities are time dependant and range from 104 to 105 K and on the order of 1015 to 1019 cm-3, respectively (Figure 1-2).16 The temporal properties of the LIBS plasma are reflected in the emission radiation. During the first 1 s of the plasma, emission is domi nated by broad background continuum emission that is caused by bremsstrahlung a nd recombination radiation resulting from the interaction between free electrons and ions as the plasma cools. While the atoms and ions in the plasma do emit their characteristic radiation dur ing this interval, weak er lines can not be differentiated from the continuum. However, after approximately 1 s, the continuum begins to subside while the atomic and ionic species in the plasma continue to emit. Consequently, collection of the plasma emission at time inte rvals after the decay of the background continuum results in superior signal-to-background (S/B) and narrower peaks (Figure 1-3). Instrumentation One of the greatest app eals of the LIBS techni que is the simplicity of the instrumentation involved in the analysis. The most basic LIBS analysis requires a pulsed laser, focusing and collection optics, a spectral sel ection device and a detector. The nature of these components can vary greatly depending on the purpose of the analysis (Figure 1-4). Laser The choice of laser for LIBS is often dependa nt on the purpose of the analysis. Several considerations must be made regarding the lase r parameters, including: (1) size and portability potential, (2) maximum pulse energy and re petition rate (3) power source and cooling components, (4) beam quality and stability and, (5) safety. Pulsed CO2 lasers (with a wavelength of 10.6 m) and excimer lasers (typical wavele ngths of 193, 248, and 308 nm) have all been

PAGE 19

19 successfully applied to LIBS.17 However, these types of lasers are limited by size, the need for specialized optics and safety con cerns, and are therefore used al most exclusively in laboratory applications with no eye towards portability and in-field analysis The primary advantages of LIBS dictate that the optimum laser should be sma ller, robust and accessible to researchers in all scientific fields. The most common type of laser that fits these requirements is the solid-state, flash-lamp pumped, Q-switched Nd:YAG laser with a fundamental wavelength of 1064 nm and a pulse width range of 6-15 ns. The Nd:YAG laser is a compact, reliable laser that is produced in a variety of sizes, from a bench-top model for la boratory use to a hand-held instrument that has been successfully incorporated into a portable LIBS system.18 Output peak pulse energies can vary from a few J to 1 J and can be delivered at re petition rates exceeding 50 Hz. Nd:YAG lasers can also be easily fre quency doubled, tripled and quadrupled to produce wavelengths of 532, 355 and 266 nm, further increasing th e versatility of the laser. Spectral Resolution One of the most im portant components of th e LIBS instrumental set-up is the spectral dispersing system. There are many different type s of spectrometers available, ranging from inexpensive hand-help models with small spectra l bandwidths and low resolution, to larger, more expensive echelle spectrographs which span 190 to 800nm and exhibit superi or resolution. The type of spectrometer used is dictated by the nature of the analysis, however some basic considerations include: (1) ad equate spectral resolution, (2) si ze and maintenance requirements, and (3) the spectral range. While higher reso lution spectrometers offer obvious advantages, including fewer spectral interferen ces, they are overall more expe nsive, exhibit smaller spectral ranges and are less robust and portable. Size and maintenance requirements are obvious considerations when assessing the portability of the LIBS system, and in some cases it may be more beneficial to sacrifice some resolution in favor of a smaller, less expensive system which

PAGE 20

20 can be incorporated into a robus t, portable system. The spectra l range of the spectrometer is another important factor, especially when multi-elem ent analysis is to be performed. In general, the larger the spectral window, the more information can be retrieved from the sample. Detectors The type of detector that is used for LIBS is generally dependent upon the spectrom eter. Photodiodes (PD) and photomultiplier tubes (PMT) are highly sensitive devices that produce an amplified electrical signal proportio nal to incident photons. While these types of detectors offer valuable advantages, such as sub-nanosecond resolution (useful for temporal-based plasma studies), each device can only monitor one wave length at a time, precluding broad spectral bandwidths. A more common type of detector is one that involves placing many photosensitive elements in an array at the exit slit of the sp ectrometer. Such an assembly will respond to a much wider spectral range and therefore is more convenient for multi-element analysis. Examples include photodiode arrays (PDA), char ge-injection devices (CID) and charge-coupled devices (CCD). One of the ma jor limitations of these devices is that they must accumulate incident radiation for a period of time before th e readout is processed, and when it is processed, each pixel is read sequentially, further increasi ng the response time. Such time-integration precludes detailed temporal studies of the plas ma emission. In order to achieve time-resolved LIBS, a micro-channel plate (MCP) must be coupled to the detection system. The MCP regulates when the incident radiation is allowed to reach the detection device, and in doing so it also amplifies the light by convert ing it to electrons wh ich are then reconverted back to light before detection. Combining an MCP with a timeintegrating detector such as a CCD is referred to as an intensified charged-coupled device or (ICCD).19

PAGE 21

21 Advantages of LIBS No discussion about the characteristics of LIBS would be complete without a closer look at the advan tages that have fueled increased interest in the technique. In terms of figures of merit such as precision and limits of detection (LOD), LIBS is mostly surpassed by more traditional techniques such as ICP-MS or ICP-OES. Howeve r, the unique advantages that LIBS offers are significant enough to render it a serious cons ideration in many analytical applications. Sample Preparation One of the most attractive features of LIBS is that the analysis can be perfor med without any sample preparation, in virtually any location. Unlike more conventional techniques such as ICP-OES, the sample does not need to be digest ed or converted to a solution. As such, the analysis time is drastically decreased and ther e is less potential for contamination. This is especially appealing for industrial applications of LIBS, whic h may require continuous sample analysis of raw materials. It also allows for th e analysis of samples the can not easily be digested for ICP analysis, such as extremely hard material s like ceramics and steels. The lack of sample preparation required for LIBS is also applicable to samples of all media. Indeed, one of the greatest benefits of the technique is that it can be performed on solids, liquids, gases and aerosols with minimal instrumental variation. It is important to note that while LIBS can be performed on virtually any sample without preparation, in some cases this may not provide the best analytical resu lts. For example, the element of interest may not be homogeneous thr oughout the raw sample. If the ultimate goal is to quantify this element then the best results (and increased precision) will be achieved once the sample has been homogenized, which may require more extensive preparation. In most cases, qualitative analysis can be performed on virtually any samp le without any preparation.

PAGE 22

22 However, quantitative analysis may require some minor preparation, dictated by the type of sample and the required degr ee of precision and accuracy. Micro-destructive Nature While LIBS can not truly be considered a nondestructive technique since som e amount of sample is always permanently removed from the surface, it can be considered micro-destructive due to the microscopic craters that are achievabl e under the correct laser conditions. While this may not be so important for applications where th ere is an abundance of sample, it is imperative in applications where the sample is scarce or prec ious, or has aesthetic qua lity and can not appear damaged. Because of this advantage, LIBS ha s been extended to applications involving the pigment analysis or authentication of priceless works of art20-22 and ancient artifacts.23 Portable and Stand-off LIBS Arguably one of the greatest advantages of LIBS, and possibly th e reason why lim itations such as poor precision and sensitivit y are tolerated, is the ability to take the instrument out into the field with minimal instrument al modification. There is increased momentum in the world today to achieve analytical information about a variety of samples in a shorter time frame. In some cases, the time it takes to retrieve a sample transport it back to the laboratory and analyze it may be too long in terms of preventing the threat the sample poses. In other cases, it may be necessary to analyze hundreds or thousands of samples to get accurate results, a monumental task if each sample is to be individually colle cted and sent back to the laboratory. The underlying factor is that it is simply more convenient to take th e instrument to the sample rather transport the sample to the instrument. The relatively simple and versatile instrument ation that LIBS employs makes it an ideal portable system for field-work. Small, solid-state lasers are often used in portable LIBS systems with the laser head eith er mounted in a small probe or laser pulses delivered to the sample via

PAGE 23

23 fiber-optics. Emission is colle cted using the same fiber-optic and delivered to a miniature spectrometer; data are collected and pro cessed on a laptop computer. One common configuration of these components is to house th e spectrometer, laser a nd battery power supplies in a back-pack and the fiber-optics delivery and co llection system in a long probe, a system that is referred to as man-portable LIBS. LIBS can also be used for stand-off analysis, often referred to as remote sensing, whereby no component of the system is remotely close to the sample. This is especially important when interrogating samples that may be hazardous or are inaccessible (though they must be optically accessible). In order to achieve stand-off LIBS the laser pulse mu st be focused onto the desired sample surface using a long focal length optical system. Plasma emission is usually collected using the same optical system. Recent research has demonstrated successful detection and identification of organic materi als from a distance of 30 m.24 Rapid Analysis The speed with which a LIBS analys is can be pe rformed is attributed to both the lack of sample preparation that the technique requires and the very short lifetime of the plasma. Assuming a rapidly responding spectrometer and ad equate spectral proces sing software, results can be obtained from a one shot analysis in less than one second. In practice however, especially in quantitative work, it is best to take data from many shots in order to increase pr ecision. Still, depending on the laser repetition rate this can still be accomplished in a fraction of the time necessary to achieve the same results from more conventiona l analytical techniques. Challenges to LIBS As previously m entioned, regardless of the multiple benefits of LIBS, there are several limitations which may arise during specific applications. Often the effects of these limitations

PAGE 24

24 can be reduced to acceptable levels. In other cas es, the standards of analytical performance are lowered in favor of the versatil ity and applicability of the tec hnique to a specific analysis. Precision It is g enerally accepted that the precision of LIBS is one of its greatest disadvantages. Common relative standard deviat ions (RSD) reported fall between 5-10%, though values can be significantly higher depending on the sample. Th ere are many different f actors that contribute to such poor precision, the most important of which is more a property of the sample than the technique itself. The lack of sample prepara tion, which makes LIBS so attractive, also means that sample remains in its raw state, which most often is completely non-uniform. This, combined with the very small mass of sample th at is interrogated, ensures that the sampled mass may not be representative of the bulk sample, a nd subsequent laser shots on different areas may produce very different spectra. While increasing the number of laser shots may increase the precision, this is still limited by th e inherent inhomogeneity of th e sample. There are many other factors that will significantly affect the precision, including: choice of analytical emission line, emission signal temporal development, sample translation velocity, detector gate delay, background correction methods, perturbations of the plasma, changing lens-to-sample distance (LTSD), changing fiber-to-sample distance (FTSD) and laser-pulse shot -to-shot instability.25 In many cases computing the ratio of the element of interest to another fixe d-concentration element in the sample will alleviate the effect these parameters have on the precision. Matrix Effects Matrix effects are on e of the greatest chal lenges to the LIBS analysis and represent a significant obstacle to quantitative analysis. The term matrix effect refers to the physical properties and chemical composition of the sample that affect the element signal such that changes in concentration of one or more of the elements forming the matrix alter an elements

PAGE 25

25 emission signal even though the element concen tration remains constant. For example, the signal intensity for carbon in si lica is very different from that of carbon in alumina, despite having the same carbon concentration. This is an example of a chemical matrix effect and is especially problematic for samples that appear the same (such as soils) but may exhibit small unknown variations in composition, such as iron or aluminum content. This makes it very difficult to produce a universal calibration curve th at will encompass a broad range of soils. Some success in correcting matrix effects has been found by normalizing signal intensities to excitation temperatures and electron densities using the Saha-Boltzmann equilibrium relationships.26 Physical matrix effects are also of concern in a LIBS analysis. Examples of physical matrix effects include differences in specifi c heat, latent heat of vaporization, thermal conductivity, and absorption between matrices. Each of these factors can affect the mass of sample ablated and therefore influence quantitati ve results. Compensation for these physical matrix effects can be achieved by producing a ratio between the element of interest and an internal standard, if it can be assumed that the relative ablated masses of both remain constant. Another method of compensation that has been reported involves acous tic monitoring of the spark sound.27 An additional example of an important physical matrix effect is grain size, or particle size. It has been well documented that the LIBS signal is dependant on the particle size and arrangement (pressed pellet, fr ee powder, etc.) of the sample. Matrix effects are either the primary focus or at least an inte gral part of much of the current research concerning LIBS. While there is still much work to be done, great progress has been made in identifying the source of different matrix interferences and conceiving different ways to prevent or correct for them.

PAGE 26

26 Applications of LIBS for Soil Analysis A great deal of research has concen trated on applying LIBS to environmental analysis, with emphasis on detecting hazardous contaminants in a variety of media. LIBS has been applied to liquids (such as waste water, river and ocean samples), 28-31 ice,33 aerosols (such as factory effluents),33-36 plants,37-39 bactera,40 paint,41-42 and slags (the residue formed by the smelting of metallic ores).43 Much research has also been focused on LIBS of soils and sediments, especially for the det ection of metals and qua ntification of toxic metals. This section will provide a review of some of the more pertinent research concerning LIBS of soils. Early work with LIBS of soils focused prim arily on determining toxic metals and studies evaluating the numerous matrix effect s that soils pres ent. Eppler et al.44 were able to determine Ba and Pb in spiked soil samples with detecti on limits of 42 and 57 ppm respectively, well below the Los Alamos screening levels of 5400 and 400 pp m. The soil samples were significantly ground and homogenized before bei ng pressed into pellets and ther efore can not be considered natural soil sample. Precision was improved to 4% by use of a cylindrical lens to focus the laser beam and produce larger ablation areas whic h decreased the effects of sample homogeneity. The effect of analyte speciation on the emissi on signal was also studi ed, and while it was determined that the origin of the analyte (for ex ample, nitrate or oxide) a ffected the signal, this could not be correlated with a ny physical property of the orig inating compound. However, the absorptivity of the bulk matrix was also found to affect the emission signal and this was well correlated with increasing plasma electron de nsity which caused signifi cant changes in the concentrations of ionized species. Barbini et al .45 found that correcting integrated line in tensities of Fe in soil by the plasma temperature substantially improved correlation between LIBS Fe concentration and ICP Fe concentration. Such a correction was achieved by constructing a Saha-Boltzmann plot to obtain

PAGE 27

27 the plasma temperature, and applyi ng this factor to the raw LIBS signal. The poor correlation of the uncorrected LIBS with ICP measurements is a ssumed to be a function of the matrix effects, which cause significant changes in the plasma te mperature and thus the emission signal. The efficacy of using such a technique alone to corr ect for matrix effects is limited by the assumption of local thermal equilibrium (LTE). Another matrix effect of great significance to LIBS analysis of natural soils is water content, which was studied by Bublitz et al .46 They proposed the use of LIBS to monitor Ba ions that are added to soils as tracer s in percolation studies. The influence of wate r content on the intensity of LIBS spectra was investig ated using soil samples with 2.5% BaCl2 added. It was found that with increasing water content to abou t 6% the relative inte nsity of the 455.4 nm Ba line increased rapidly, and decrease d thereafter exponentially as wa ter content increased. The reason for the initial increase in intensity with increased water content was explained by three different processes. First, the shockwave on th e wet sample creates less dust, reducing scattering and absorption of the laser radiat ion and increasing laser energy that is coupled into the sample. Second, the Ba ions form complexes with aquatic colloids such as fulvic acid and clay minerals, which causes a decrease in breakdown threshold and increased emission.30 Finally, the cohesion introduced by the water content may lead to a larger surf ace concentration in cont rast to dry soil. However, for soil samples with water content grea ter than 6%, the emission intensity decreases due to plasma cooling by the water as a result of increasing collisions that dissipate the energy from the plasma plume. In addition to this, th e exciting laser radiation is absorbed by the water decreasing the efficiency of plasma ignition. Capitelli et al.47 compared LIBS to ICP for the detec tion of heavy metals in a variety of standard reference soils which had been homoge nized, dried and presse d into pellets. The

PAGE 28

28 detection limits achieved with LIBS from relativ ely linear calibration cu rves were between 30-50 ppm for elements such as Cr, Cu, Ni, Pb, and Zn. These values fall well below the European Union (EU) limits in both sewage sludge and so ils established by the European Commission in 1986. In comparison to ICP results, the RSDs fo r the LIBS measurements were all significantly higher, with the exception of Cr In the extreme case of Ni the RSD for LIBS was 14.18% compared to a value of 0.79% for ICP. It was hypothesized variations in plasma temperature as well as other matrix effects contributed to this poorer precision rela tive to ICP results. One method to circumvent trouble some matrix effects and decr ease detection limits in soil is to couple LIBS with laser-induced fl uorescence (LIF), first reported by Gornushkin et al.48 Hilbk-Kortenbruck et al. extended this technique to atmospheri c conditions for the determination of As, Cr, Cu, Ni, Pb, Zn, Cd and Tl in presse d soil pellets. The latter two elements were determined by LIF, which was accomplished by directing a laser pulse tuned to the wavelength of interest into the LIBS plasma plume to excite fluorescence. Because this laser must be tuned to a specific wavelength, LIF can only be used fo r the detection of one element, typically one that requires detection lim its significantly below what LIBS can achieve.49 Some of the more practical aspects of LIBS analysis of soils were reviewed by Wisburn et al.50 A variety of different soil s were dried and pressed into pellets for analysis with a Qswitched Nd:YAG laser. Emission collection was ach ieved with a fiber optic. One of their most significant findings was the production of a persis tent aerosol above the sample surface at high laser repetition rates focused on the same sample s pot. This aerosol was characterized as ablated residual material from prior laser pulses that ha d not relaxed back to the sample surface and was available in the plasma spot location for the next pulse. At lower aerosol concentrations, this will provide increased emission signal, however at higher concentrations, it will actually absorb

PAGE 29

29 more of the incoming laser light, thus reducing the portion of the laser pulse that reaches the solid sample, effectively decrea sing the signal. It was found th at the optimal repetition rate varied as a function of the sample mean part icle size and for a soil sample was 1 Hz. The crater formed during laser interrogation of the sample spot with multiple pulses was also studied. It was found that there was a slight increase in the emission signal from the second laser pulse that could not be attributed to persistent aerosol due to the much lower repetition rate. Instead it was theorized that th e sampling site was enriched with larger particles after the shockwave from the first laser pulse swept away the smaller particles. After the second laser pulse, a sharp decrease in signal was observed which was explained as a crat er effect: the deeper the crater became the greater the plasma-to-fiber distance and enhanced shielding of the plasma by the crater walls from the fiber. Another very important aspect of LIBS soil an alysis that was investigated by Wisburn was the effect of soil grain size. The importance of this parameter on any analys is is rather dependent on the analyte of interest and how it is distribut ed in the sample, for instance if it is a surface contaminant or if it is bound w ithin the particles. To inve stigate this, so il samples were separated into four different grain sizes by si eving (0.38-1 mm range). Each was contaminated with Cd, which adheres to the surface of the pa rticles. Good linear correlation between signal intensity and Cd concentration for each grain size was found, as well as a decrease in slope as the particle size decreased. While th ese results are significant in that they show a great dependence of the LIBS signal on the particle size, two important factors must be considered for practical soil analysis. First, most raw soil samples are co mposed of a broad distribution of particle sizes that exceeds the range studied in this research. In addition to this, these results were based on the assumption of only surface contamination. For many di fferent analytes of interest this is not the

PAGE 30

30 case; rather they may be found tightly bound within the particle, a property that is not often known before the analysis.50 The issue of soil grain size was also raised by Theriault et al.,51 who performed LIBS of soil with a cone penetrometer system which pr obes deep into the soil. They found that the detection limits of Pb, Cd, and Cr increased as the particle size was decreased to less than 106 m. This variation was explained by considering the soil surface area over which the contaminant was spread. Because a fine-grained ma terial has a larger tota l surface area per unit mass than a coarse-grained material, the to tal surface area over which equal contaminant concentrations by mass are dispersed is larger fo r the fine-grained material and smaller for the coarse grain material, resulting in a thicker su rface layer of contamin ation on coarse-grain material. If the laser spot size is fixed and ab lates the contamination on th e surface of the grains, there is an overall larger volum e of contaminant in the plasma for coarse-grained material, resulting in lower detection limits. One of the most interesting ap plications of LIBS on soils ha s been for the purpose of space exploration. It has been propos ed that stand-off LIBS can be used for the exploration of planetary surfaces that are not accessible for analysis by other techniques such as X-ray fluorescence and alpha-proton X-ray spectrometry (APXS), which both requi re close proximity to the sample, and in the case of APXS, exhib it very slow data acquisition. Because of the similarity between the geological properties of the Martian surface and some Earth soil samples, preliminary research regarding the feasibility of incorporating a LIBS system into a space mission have been conducted using soil samples. Knight et al.52 studied the effect of decreasing pressure on both the plasma formed on the soil an d the emission intensities. It was determined that down to pressures of 1 torr th ere is an increase in plasma dimensions and emission signal.

PAGE 31

31 At lower pressures, the plasma readily expands into the lower pressure regions increasing in volume relative to those formed in atmospheri c pressures. The signal enhancement at lower pressures is the result of increased mass ablati on due to decreased shielding of the sample surface from the laser pulse by the more diffuse plas ma. It was also determined that pressures of less than 10 torr were required to detect O(II) in the soil due to the increased electron density which causes the ionic oxygen atom to c onvert to the neutral oxygen atom. Arp et al.53 reported on LIBS of ice/soil mixtures which were synthe tically produced to mimic the polar surfaces of Mars. They determined detection limits for elements in these mixtures (at a pressure of 7 torr), which were typically 10% soil, we re not significantly changed from those obtained by analyzing dry soil. The determination of carbon in soil is arguably one of the most pressing issues concerning soil analysis due to the current attention to global warming. Ho wever, very few papers have been published regarding the potential of LIBS to achieve quantitative data on carbon in soil. Ebinger et al. first reported on the analys is of carbon in dried soil samples from farms in Colorado and fields surrounding Los Alamos, New Mexico. Each sample was sieved to <2mm and placed in a quartz tube. A 50-mJ, 10 ns, Nd: YAG laser was focused into the quartz tube and the resulting emission collected via fiber optics and delivered to a 0.5-m spectrograph with an intensified photodiode array det ector. Shot-to-shot variations were normalized by taking the ratio of the 247.8 nm C line to a Si line. Veri fication of the C content was achieved by dry combustion. The results showed good correlati on between LIBS signal and the carbon concentration, with an R2 value of 0.966. In turn, the resu lting calibration curve predicted C concentration in other soil samples with an accu racy of 3-14%. The detection limit was reported as 300 ppm.54,55 This research certainly demonstrated the feasibility of LIBS as a method for

PAGE 32

32 carbon determination; however, it did not take into account many other variables, including varying soil grain size, applicability of the calib ration curve to a broad range of different soil samples, differing Si content, and Fe interference with the C247.8nm line. Subsequent research by the same group investig ated the use of the C193 nm line, which is not susceptible to inte rference from Fe as is the C247.8 nm line In addition to this, standardization of the C signal was achieved by ta king the ratio of said line to the sum of the Al199 nm and Si212 nm lines, C/(Al+Si). The soils chosen were noted to have similar physical and chemical properties. The resulting C signa l was well correlated with the C concentration determined by dry combustion (R2=0.95) and exhibited great reproducibility over a 30 day period. It was acknowledged that th e Al and Si content of different soil samples will vary as a function of the soils mineral ogy and texture, which would significantly limit the use of both Si and Al as internal standards. However, it wa s assumed the those properties were constant for soils as similar as the on es employed in this study.56 Martin et al.57 extended on the results of this prev ious research by determining carbon and nitrogen in a variety of different soils. A Q-switched Nd:YAG laser wa s used at 266 nm with typical pulse energies of 23 mJ/pulse. Emi ssion collection was via a fiber optic bundle and detection was achieved with an ICCD. Because this study was only concerned with the soil organic matter (SOM), inorganic soil carbon wa s first removed by acid washing. Soil was then dried, ground and pressed into pellets. The C 247.8 nm emission was averaged from 10 laser shots and the resulting calibra tion curves showed good correlati on up to 4.3% (the highest concentration of carbon). The RSDs for the soil samples were higher for the low C concentration soils due to the increased interference of the Fe peak. It was found that the

PAGE 33

33 standard deviation could be reduced if multiple samples of each soil were measured and more laser shots were averaged. More recent research con cerning carbon in soil analysis was reported at the fifth Annual Conference on Carbon Capture & Sequestration in 2006. Harris et al.58 analyzed over 200 soil samples from several Midwest states representing a wide variety of textures and sand-silt-cly combinations. These soil samples were analyzed by dry combustion for carbon content and pressed into pellets for LIBS analysis. The resu lts indicated very differe nt calibration slopes for sand, loam or clay soils which correlated well with synthetic soils produced to mimic these compositions. In agreement with previous researc h, it was determined that the smaller grain-size soils (clay) exhibited a smaller slope than that of large-grain size soils (sand). The conclusion was that the LIBS signal was very sensitive to soil texture and therefore, each field would have a different response or sensitivity and require in-field calibration. Scope of Research While the analysis of soil samples has received a great deal of attention over the last few years, research concerning the determination of carbon in soil has been lacking, limited to a few proof-of-principle demonstrations by research groups at Los Alamos and Oak Ridge National Laboratories.54-57 Since the sequestration of carbon in soil may offer a means of mitigating the effects of global warming (to be described in Ch apter 2), it is becoming increasingly urgent to develop a practical device that can accurately measure terrestria l carbon inventories in order to understand the status of the global carbon cycle. LIBS has been proposed as an in-field carbon monitor due to the robust, compact and low-power properties of its instrumentation.59 Before a field-based LIBS instrument can be designed there must be la boratory-based demonstration of the capability to determin e carbon in soil. The purpose of our research was to investigate the potential of LIBS for carbon in soil analysis wi th an emphasis on a minimal sample preparation

PAGE 34

34 and the potential for in-situ analysis. The instrumentation used for the majority of the research was compact, light-weight and similar to the co mponents that would be incorporated into a portable device.

PAGE 35

35 0 20 40 60 80 100 120 140 160 180 20019 6 0 1980 198 1 198 2 198 3 198 4 198 5 198 6 198 7 198 8 198 9 199 0 199 1 199 2 199 3 1 9 94 1 9 95 1 9 96 1 9 97 1 9 98 1 9 99 2 0 00 2 0 01 2 0 02 2 0 03 2 0 04 2005 2006 2007YearNumber of Publications Figure 1-1. Number of LIBS pub lications per year since 1960.

PAGE 36

36 Figure 1-2. Significant events durin g plasma formation. A) The laser pulse strikes the sample surface. B) Energy is dissipated through the sample. C) Vaporization of the underlying material causes the sample surf ace to explode. D) Plasma expands while still absorbing energy from the residual laser pulse and a shockwave is formed. E) Radiation is emitted. F) Plasma cools and slows. G) A crater remains after the ablation event. A B C F E D G

PAGE 37

37 Figure 1-3. Temporal evoluti on of the LIBS plasma. Td is the detector delay time and tg is the gate pulse width, also referred to as the in tegration time. Constructed with data from ref. 4. 1 ns 10 ns 100 ns 1 s 10 s 100 s Strong continuum emission Ions Neutrals Molecules Laser pulse Continuum td tg

PAGE 38

38 Figure 1-4. Typical schematic of a LIBS set-up Pulsed Laser Fiber Optic Sample Focusing lens 400410420430440450460 -50 0 50 100 150 200 250 300 350 400 407.747 421.607 460.829 Y Axis Title Emission Collection Spectrometer Laser-Induced Plasma

PAGE 39

39 CHAPTER 2 CARBON IN SOIL The im portance of accurately measuring car bon content in soils has become readily apparent with the increased awareness of the causes of global warming. In December 1997, the Kyoto Protocol was established to set target CO2 emission reductions for ratifying countries. For the United States, the Protocol calls for the reduction in atmospheric carbon emissions by 7% from the 1990 level by the year 2012.60 In order to meet these challenging targets, and mitigate the harmful effects that global warming presents, an approach must be formulated that not only reduces current CO2 emissions but also captures CO2 from the atmosphere and stores it in long term sinks. A full understanding of the carbon cycl e and the rate of carbon fluxes, as well as accurate inventories of carbon in the terrestrial ecosphere, ar e needed to optimize potential emission reduction techniques. This chapter w ill introduce the sequestrati on of carbon in soil as well as current methods that are employed to de termine soil carbon levels for the purpose of developing an accurate terrestri al carbon inventory. Finally, th e challenges of soil sampling, preparation and analysis will be briefly discussed. Sequestration of Carbon in Soil Carbon is the funda mental building block of life and as such exis ts in appreciable quantities in every ecosp here: the atmosphere, hydrosphere (oceans and seas), biosphere (living matter) and the geosphere (the earth). The global amount of carbon is fixed, however the carbon is readily exchanged between each of these spheres during a process that is referred to as the carbon cycle.61 Figure 2-1 shows the amount of carbon, in gigatons (Gt), that is actively cycling through each sphere. Figure 2-2 illustrates mech anisms that contribute to the carbon cycle as well as approximate carbon fluxes in Gt/year.

PAGE 40

40 Data from CO2 trapped in ice co res and long term CO2 monitoring has indicated that the current CO2 levels are the highest they have been in 650,000 years, at 390 ppm in 2007.62 Such elevated levels can be attributed to increased land clearing (destruction of vast rain forests, depleting woodlands) and burning of fossil fuels (o il, coal, and natural ga s), which contributes over 5 Gt of CO2 into the atmosphere each year. Indeed, the atmospheric concentration of CO2 was 270 ppm prior to the indus trial revolution of the 18th century, after which it exponentially increased to the current value of 390 ppm. After the complete expenditure of all fossil fuels in approximately 300-400 years, the at mospheric concentrations of CO2 are predicted to be 1200ppm.63 Increased CO2 levels in the atmosphere contribute to a phenomenon referred to as global warming. Carbon dioxide, along with other gases such as water vapor, methane (CH4), and nitrogen oxide (N2O), are referred to as greenhouse gases because they trap heat from the sun in the earths atmosphere, which results in increa sed global temperatures, a process referred to as global warming.64 The ramifications of global warmi ng are ominous, and include decreased vegetation production and accelerat ed polar ice-cap melting. There is little argument that atmospheric CO2 levels need to be reduced; however there is much debate on how to achieve this goal. Th e obvious recourse is to reduce fossil fuel consumption however current economics depend on fossil fuels and the stunted development of alternate energy sources is reflective of this.63 An alternative to stemming the CO2 source is to enhance natural carbon sinks (where carbon is stored for long periods of time). As illustrated in Figure 2-1, the largest carbon sink by far is the oc ean; however the ability to manipulate such a vast sink is limited and may take hundreds of y ears to produce appreciable changes. The next largest carbon sink is terrestrial soil, the carbon content of which is double that of carbon in

PAGE 41

41 vegetation, such as plants, trees, crops and grasses. Due to poor land management and agricultural practices during the first half of the 20th century, over half of the maximum soil organic carbon (SOC) has been lost. However th is has resulted in a greater capacity of many soils to sequester carbon. The majority of carbon is introduced into soil as decaying biomass. Plants and other vegetative material withdraw CO2 from the atmosphere during photosynthesis and convert it to sugars, starch and cellulose. Upon death, soil microorganisms cause decomposition and respire CO2 in the process. However, an organic residu e remains after decay and is referred to as humus. In the absence of oxidation, this organic matter will remain in the soil for hundreds, possibly thousands of years.65 Carbon sequestration in soil ca n be significantly increased by up to 0.1% per year by converting marginal arable land to forest and gras sland, planting high yield crops, and not tilling crop lands, especially in warm, moist climates. Even such a seemingly small increase in carbon sequestration can ha ve a huge impact on the global (atmospheric) carbon balance. Current estimates indicate that successful sequestration of carbon in soil can potentially offset 30,000-60,000 million metric tons of carbon that is released by fossil fuel combustion over the next 50 years.66 A direct result of the Kyoto Protocol is th e establishment of a so-called carbon stock market. In brief, it allows count ries that have large carbon sinks, such as forests or vegetation to deduct certain amounts from their CO2 emission targets, thus making it easier for them to achieve the desired net emission level. Countries that have already met their emission targets, and have additional CO2 sinks may wish to sell the benefit th at these sinks provide (an offset of the target emission levels) to countries that are not on target to meet required emission reductions. Such a cap-and-trade system has also been proposed for the national level, where

PAGE 42

42 individuals are responsible for reducing emissi ons and/or creating ca rbon sinks and trading carbon credits to meet these targets.67 Potential carbon stocks have been valued at approximately $8 billion in the agricultural industry alone.68,69 While terrestrial soil is the second largest ca rbon sink, it is currently not recognized as a sink that can be traded as an off-set of carbon em issions. The reason for this is simple: reliable measurements of soil carbon stocks on a national scale are scarce, and high-precision estimates of annual carbon inputs to the soil are even scarcer.70 Izaurralde et al concluded that although there is a compelling scientific basis for believ ing that a substantial Canadian C sequestering activity is occurring, the science and supporting measurement protoc ols have not been developed to the point that this can be adequately quantifie d at the farm or regional scale and accepted for marketing purposes.71 The Soil Science Society of America maintains that in order for carbon sequestration in soil to eff ectively reduce atmospheric CO2 more research is needed to better quantify carbon sequestration and improved monito ring and verification protocols must be established.66 As such, while sequesteri ng carbon in soil is one of th e most promising short-term methods to mitigate increasing atmospheric CO2, it is severely limited by the lack of analytical techniques to provide quantitatively defensible carbon levels in soil. Current Methods for the Determination of Carbon in Soil Typical analytical m ethods for the determina tion of carbon in soils and sediments rely on converting carbonaceous material to CO2 followed by direct or indirect measurement.72 The two most common methods for carbon analysis are dry combusti on and wet oxidation coupled with a variety of different methods to measure evolved CO2. Dry combustion involves combusting the soil sample in a stream of CO2-free O2 gas at temperatures above 900oC. These resulting combustion gase s are passed through a series of traps that are designed to remove particulates and other interferences. If a medium temperature

PAGE 43

43 furnace is used, the gases may then be passed through a catalyst furnace to ensure complete oxidation. The dry combustion technique can be used to measure organic carbon (Corg), inorganic carbon (Cinorg) and total carbon by controlling the combustion temperature; Corg combusts between 650-720oC and the remaining Cinorg is converted to CO2 between 950-1100oC. In some cases, Corg measurements may be inaccurate if carbonate decomposition or incomplete organic carbon oxidation occurs at the lower temperature. While the actual dry combustion method is relatively standard, the method of CO2 detection and quantification varies. A summary of these different detection procedures is presented in Table 2-1. Dry combustion carbon recoveries for dry, organic samples exceed 95%.72 In the wet oxidation technique oxidizing agents such as KMnO4 or K2Cr2O7 in acidic solutions are used to oxidize the carbonaceous materials in the sample. The chromic acid method, referred to as the Schollenberger method, is the most commonly used wet oxidation technique and uses solid K2Cr2O7 in the presence of H2SO4 or a mixture of H2SO4 and H3PO4. The evolved CO2 represents the total Corg content.73,74 Another wet oxidati on technique that is widely used is the Walkley-Black method, in which a known excess of K2Cr2O7 is used as the oxidant and back titration is performed to determine the amo unt of oxidizing agent consumed, which is proportional to the Corg.74,75 There are many other vari ations on the wet oxidation technique, including the type of oxidizing solution, the digesti on temperature and duration, and the CO2 detection methods (the same as for dry combustion, listed in Table 2-1). It is apparent that there are many differences between th e two most common methods of carbon quantification. While it is generally accept ed that the dry combustion is more accurate and standardized than the wet oxidation method, it requires more complex instrumentation at a greater cost, and has a longer analysis time.73 A variety of automated dry combustion

PAGE 44

44 instruments are commercially ava ilable for the determination of total carbon content. Most analyze hydrogen and nitrogen as well as carbon and are termed CHN analyzers. More recently, the potential of visible, near infrared, mid in frared and diffuse reflectance spectroscopy have been investigated as a mean s of analyzing soil, specifically for carbon content.76-78 The Challenges of Soil Analysis The greatest challenge to the analysis of soil, regardless of analytical technique or analyte of interest, is the difficulty in obtaining a truly re presentative sample of soil. There are several factors that contribute to this difficulty, incl uding sampling errors, soil variability and soil homogeneity. Soil variability refers to different types of soils located within a relatively small sampling region. Many environmental and anth ropogenic factors can affect the spatial variability of soil, including localized decompos ition events and farming practices. Variability can occur in both the horizontal and vertical directions, and must be compensated for with a careful sampling regime. The simplest way to account for it is to take multiple samples; however the hundreds, or possibly thousands of samples that may be needed to obtain representative data can be cumbersome. Due to the importance of determining carbon pools on a regional scale, much attention has been directed toward the spatial dist ribution of carbon in a variety of different lands.79-81 Soils are also highly heterogeneous even within the same soil type and therefore create a so-called resolution problem. Soils are compos ed of many different c onstituents that are unevenly distributed throughout a given sample. Inorganic constituents such as quartz are present in three dis tinct particle size distribut ions referred to as sand, si lt and clay, each of which exhibits different physical and chem ical properties related to the different surface area. Table 22 shows the physical size ranges of these particle fractions, as well as the sizes of sub-groups

PAGE 45

45 within the fractions. Soil organic matter is composed of humus and biological organisms. Humus is classified as either non-humic substa nces (carbohydrates, lipid s and other metabolic products of organisms) or humic substances (humic and fulvic acids synthesized by microorganisms). Figure 2-3 illustrates the break down of soil organic matter. Each of these organic compounds is heterogeneous ly distributed throughout the so il, adsorbing to the surface of inorganic particles in different degrees and often binding di fferent particles together as aggregates.74 Most analytical methods are considered microanalysis techniques, and as a result sampling will not be representative of the bulk sample. Grinding and sieving is often used to produce a homogeneous sample of uniform partic le size; however this often disturbs other properties of the soil and is a time consuming process.82 While soil variability and heterogeneity are the two major challenges in soil analysis, there are several other factors that must also be consid ered. Soils of higher water content have greater vertical variation due to the downward leaching of water solubl e components. Water content also affects the pH and porosity of the soil, wh ich may impact specific analytical methods. The amount of water in the soil is of ten unknown and must first be dete rmined if the technique has a tolerable limit. If the soil must be dried prio r to analysis, the drying method must be carefully considered especially if it induces changes in the soil properties. For instance, oven drying soil may result in the oxidation and subsequent lo ss of carbon, therefore ai r drying is recommended for soil carbon analysis. Soil density is also a challenging fact or, especially when it is necessary to report the concentration of an analyte per unit volume of soil. Soil density is a function of both the particle density and the bulk density. Th e determination of bulk density is especially important when assessing carbon pool sizes and carbon fluxes, which are usually reported as kg/

PAGE 46

46 m2 to a depth of 50cm.70 It is apparent that in order for quantitative results to be meaningful, both bulk soil density and particle density must be assessed at the time of analysis. Clearly, for accurate and precise soil analysis a carefully considered analytical scheme must be developed. Consideration must be given to the inherent variable s in soil and the effect these will have on the analysis method. LIBS may be a suitab le technique for practical soil analysis due to its portable nature (which precludes costly sample transport) and high-throughput (which allows for the orders of magnitudes of sa mples necessary to gain representative data). However, as also discussed, LIBS itself may su ffer from poor precision due to a variety of instrumental and physical variables. As suc h, any analysis of soils by LIBS requires an experimental plan that reduces th e effects of these variables, a nd balances the required accuracy and precision with the limits of the technique and sample. The research described in the following chapters describes an experimental scheme that optimizes these parameters and assesses the feasibility of using LIBS for in-situ carbon determination

PAGE 47

47 39120 754 2190 Oceans Atmosphere Terrestrial Biosphere Figure 2-1. Magnitudes of the re servoirs of actively cycling carbon in gigatons (Gt). Produced with data from ref. 61.

PAGE 48

48 Figure 2-2. Carbon cycle. The flux of carbon be tween the sources and sinks are represented by the purple arrows. (Reproduced with permission from Teachers Domain http://www.teachersdomain.org/resources/t dc02/sci/life/eco/ ccycle/index.html)

PAGE 49

49 Table 2-1. Techniques for measuring CO2 in combination with dry combustion or wet oxidation. Information from ref. 72 Type of Analysis Common means of analysis Type of measurement Photometric Infrared analyzer Measures differential absorption of infrared energy between CO2 and reference gas Gravimetric Sorption trap Measures mass gain in trap after combustion Volumetric Gas burette Measures volume of CO2 released after combustion or acid digestion Conductimetric Thermal conductivity de tector Measures differential resistance between CO2 and reference gas Titrimetric Non-aqueous titra tion Measures amount of CO2 consumed in solution by back-titration with standard solution

PAGE 50

50 Table 2-2. The three major groups of soil size fr actions are sand, silt and clay. These can be further subdivided into finer size fractions. Diameter (mm, unless otherwise stated) SAND Very coarse sand 2.00-1.00 Coarse sand 1.00-0.50 Medium sand 0.50-0.25 Fine sand 0.25-0.10 Very fine sand 0.10-0.05 SILT Coarse silt 0.050-0.020 Medium silt 0.020-0.005 Fine Silt 0.005-0.002 CLAY Coarse clay 2.0-0.2 m Medium clay 0.2-0.08 m Fine clay <0.08 m

PAGE 51

51 HUMUSDecomposed Organic Matter Comprises 50-85% of total soil organic matter Non-humicSubstances -35% of humicmatter Composed of metabolic products of organisms Humicsubstances -75% of humic matter Synthesized by microorganisms Black/brown color High molecular weight Carbohydrates Amino acids Lipids Lignins HumicAcids FulvicAcids Figure 2-3. Constituents of humus, the major contribution to soil organic carbon.

PAGE 52

52 CHAPTER 3 INSTRUMENTATION AND DATA ANALYSIS The com ponents of a LIBS system are very si mple, as explained in chapter 1. The basic requirements are a pulsed laser, focusing optics, some method of emission collection, a spectrometer coupled with a detector and a computer. Naturally, there are many variations for each of these components, and the choice of whic h to incorporate into the LIBS system is dependent on the type of analysis and availabl e resources. The instrumental components that were employed in the majority of this research were selected because they were comparable to the type of instruments that may be incorporated in to a portable LIBS instru ment. That is to say, they are compact and robust, have low power requi rements and require little maintenance. Each of these components will be discussed in depth in the following section. The second half of this chapter will describes the optimization of these instrumental components. Instrumental Components Laser The laser us ed for these studies is manufactur ed by Big Sky Laser Tech nologies. It was a flashlamp-pumped Nd:YAG laser operated at the fundamental wavelength of 1064 nm and with a specified maximum pulse energy of 50 mJ. The pulse energy can be changed in increments of 5 mJ. The pulse duration is 6-8 nm and the lase r can be operated at repetition rates of up to 20 Hz. Table 3-1 provides a comp lete summary of the laser specifications. While these specifications state that the laser energy stabil ity is 2%, it was assumed that this was an underestimation of the value and that it had to be measured independently pr ior to use. The laser consists of two units, the laser head where the pu lse is generated, and the integrated cooler and electronics (ICE) system from which the energy and repetition rate are controlled. For this research, the laser head was m ounted vertically with the lase r pulse directed down onto the

PAGE 53

53 sample stage. This type of laser is classified as a Class 4 laser which means that the output beam, by definition, is a safety and fire hazard. As a result, protective ey ewear was worn at all times to prevent accidental exposure to both direct and reflected beams. Optics A quartz plano-convex focusing lens with a diameter of 1 and a focal length of 2.5 was used to focus the laser. The lens was housed in a 1 lens tube which was m ounted in a steel cage system (Figure 3-1). This allowed for the le ns to be easily moved up or down while still remaining on axis with the laser. The lens was located approximately 11 from the laser aperture. The diameter of the ablation area prod uced on a copper surface, with a pulse energy of 50 mJ was approximately 0.35 mm (Figure 3-2). Fiber Optics The fiber optic used to collect the plasm a emission was produced by Ocean Optics, Inc. especially for the LIBS2000+ sp ectrometer unit. The fiber bundl e actually consists of seven individual fibers, each of which collects a cert ain portion of the plasma radiation and transmits it to an individual spectrometer. The fiber bundle has a nylon jacket and PVC sheathing. The fiber cores are pure fused silica and the cladding is doped fused silica. The core diameter of each of the individual fibers is 600 m. The numerical aperture of th e optical fibers and the probe is 0.22, which yields an acceptance an gle of 24.8 in air. The probe houses a collimating lens which has a focal length of approximately 1. Spectrometer The spectrom eter used for this research wa s the LIBS2000+ and was also manufactured by Ocean Optics (Figure 3-1). As the name implies, it is a commercial spectrometer that is uniquely designed for LIBS applications. It is in fact not one spectromete r but seven individual miniature spectrometers which each cover a specific spectral range (approximately 100 nm). Each of the

PAGE 54

54 fibers in the fiber optic bundl e connects with one spectromete r in the LIBS2000+ unit. The resolution at full-width half-maximum (FWHM) is 0.1 nm, though this value varies depending on the spectrometer. Radiation was dispersed by a grating, a nd detection achieved by a CCD with 2048 pixels for a combined total of 14,336 pixels. The integration time was fixed in spectrometer mode at 2.1 ms. Det ection could be delayed up to 135 s in 500 ns intervals. The LIBS2000+ system is very convenient for LI BS applications due to its broad spectral range, portability and ease of use. However, it does suffer from several disadvantages, the most significant of which is the poor sensitivity. This is due in part to the fact that each channel receives only 1/7th of the total radiation collected. In some cases it may not be necessary to acquire the whole spectrum, but ra ther a specific spectral range which may correspond to one of the seven spectrometer channels. If this is the ca se, a single fiber can be coupled to the channel of interest. As a result instead of receiving 1/7th of the collected radiation, that channel will receive all of it, which will improve the sens itivity. Another disadvantage to the seven spectrometer system is the discontinuity between each channel in the total spectrum. Often, very significant differences in background and overa ll sensitivity can occur between adjacent channels causing the spectrum to exhibit sharp jumps (Figure 3-3). One of the causes of these differences in background and sensitivity between channels is the orientation of the individual apertures in the fiber probe. As illustrated in Figure 3-4, the arrangement is a circle of six apertures and the seventh in the center. Depending on the positioning of the fiber probe, each of these fibers will view a slightly different portion of the plasma. It is often necessary to optimize the emission collection of one or two channels at the expense of ot her less important channels. Software The software that is used to process the sp ectral data is called OOILIBS, also produced by Ocean Optics, Inc. OOILIBS is a simple, user friendly p rogram that displays the individual

PAGE 55

55 spectra almost instantaneously after each laser sh ot. The laser can be run in external trigger mode, which means that the software specifies the laser repetition rate and the acquisition delay time. The software can also control the number of laser shots fired, the number of cleaning shots used, any necessary background corr ection, and average or individual data collection. There is a spectral line database available to identify specific lines; however at the time of this research, it had not been developed enough to provide accurate results and therefore was not used. In addition to this there is the option to monitor specific lines during the analysis, including calculating the mean intensity and area, standard deviation and RSD after each subsequent laser shot. This is especially helpful for rapid analys is (the mean intensity of a chosen line after 100 shots is immediately available af ter the last shot); however for this research a more advanced data analysis program separate fr om the OOILIBS software was used. Optimizing Instrumental Parameters Laser Energy Before any m easurement can be undertaken, it wa s necessary to determine the stability of the laser pulse. While the manual specifications indicated that the stability was 2%, there are many factors that may change this, including the age and use of the laser and the chosen energy. Instability in the laser pulse energy will result in variable plasmas and poor precision; therefore it is essential to ensure that the laser exhibits a tolerable energy instability. To measure this, the focusing lens was removed from the LIBS system and a power meter was positioned on the sample stage. The laser was run in continuous mode at a repetition rate of 1 Hz and an oscilloscope displayed the measured energy. Thirty-two measurements were taken at expected laser energies of 30, 35, 40, 45 and 50 mJ. The mean energy, standard deviation and relative standard deviati on (RSD) for all laser shots, the first ten laser shots and the remaining 22 laser shots are reported in Ta ble 3-2. Surprisingly, it was determined that the maximum laser

PAGE 56

56 pulse energy of the laser was act ually 57 mJ rather than 50 mJ as stated by the manufacturer. With respect to the energy stability, it was f ound that when all laser shots were included in calculating the mean, the average RSD (for all en ergies) was 6.1%. However, when the first ten shots were abandoned, the average RSD d ecreased to 3.6% (when omitting the 50 mJ measurements, which exhibited an unusually high RSD). The stability for the highest pulse energy of 57 mJ, which was the one that will most likely be used for soil analysis, was 2.3% when the first ten shots were abandoned. Due to the significantly increased instability of the laser during the first ten shots, the data from these shots were discarded in subsequent measurements. Optimization of the Spectrometer The only way that de layed acquisition of emission can be performed is to control the timing of the laser pulse rela tive to the fixed timing of the CCD integration. Figure 3-5 illustrates how the Q-switch can be controlled by the software to fire the laser, and create a plasma at different times relative to the commen cement of CCD acquisition. Positive values of the Q-switch delay correspond to plasma forma tion after the CCD begins acquisition, thus radiation from the entire lifetime of the plasma is integrated, and there is no true delay. If the Qswitch delay is set to zero, acquisition occurs at the same time as plasma formation. Negative Qswitch values correspond to emission acquisition at 500 ns intervals after plasma formation. For instance, if the Q-switch delay time is set to -0.5, radiation from the first 500 ns of the plasma will not be collected. Both positive and nega tive Q-switch delays have advantages. When emission from the entire plasma is collected w ith no delay signal inte nsities are greater and species that have prohibitively hi gh ionization potentials may be detected. However, without a delay, the initial con tinuum is collected, which increas es the background and can obscure sensitive emission lines.

PAGE 57

57 In order to determine the best Q-swit ch delay time for the C247.8 nm line, 15 measurements were taken on a soil sample and averaged for delays of +4 s to -4 s. From this data, the minimum detectable concentration of carbon, CL was determined for each delay time (Equation 3-1). B S RSDkCCB L1 ) ( (3-1) S is the intensity of the 247.8 nm signal after the backgr ound, B, is subtracted, RSDB is the relative standard deviation of the background, C is the concen tration of carbon and k is a statistical factor equal to 3. For positive Q-switch delay times, S and B remain relatively constant (Figure 3-6). This is expected because there is no delay (emission acquisition begins at different time periods before the plasma if formed ) and as such radiation from the entire plasma is collected each time. However, once the Qswitch delay time becomes negative and delays occur at intervals of 500 ns, both S and B decrease and S/B increases. A Q-switch delay time of -1 s was found to yield the lowest minimum detect able concentration of C and employed for subsequent studies. It should be noted that this corresponds to a concentration of 1.4% C; however for practical purposes C was still detect ed at concentrations lower than this. Optimization of Fiber Probe Position The placem ent of the fiber optic probe is a crucial parameter in LIBS analysis. The fiber must be placed in a position that renders it as insensitive as possible to the spatial inhomogeneity and shot-to-shot instability of the plasma. Co mmonly the fiber is either mounted close to the plasma at a 30-45o angle, so called side-collection, or pl asma emission is focused on to the fiber by the same lens that focuses the laser and a dich roic mirror, termed back-side collection. Both orientations provide unique advantages. The back-side viewing is less sensitive to spatial

PAGE 58

58 inhomogeneities of the plasma because emission co llection is from the top rather from a single side, which may not be representative of the whole plasma. In addition to this, back-side collection is also insensitive to the vertical position of the pl asma, which may change if the sample height is not constant. One of the draw backs of back-side emissi on collection is the use of a dichroic mirror. It is very difficult to find a dichroic mirror that will transmit the laser at 1064 nm while reflecting all UV a nd visible radiation in the sp ectrometer spectral range of 200900 nm. In addition to this, the longer path length that emission must travel to the fiber, coupled with the incomplete reflection at the mirror will result in radia tion losses that decrease signal intensity. Side-collection does not suffer from these limitations, and also requires less optical alignment. To determine the best fiber position, spectra from a graphite pellet were collected using both collection orientations. For the back-sid e emission collection, a 1 dichroic mirror was mounted in a cube at a 45o angle along the laser axis. The mirror transmitted the 1064 nm laser light and reflected UV and visible light. Em ission from the plasma was focused onto the dichroic mirror and reflected to a plano-convex le ns that focused the light onto the fiber probe. The emission was slightly defocused to ensu re that each channel on the 2 mm probe was illuminated. Side-collection was achieved by mounting the fiber on a movable arm and carefully aligning it until the position that provides maximum signal intensity was reached. No spectrometer delay was used in order to collect the maximum emission possible. The results of this study prove d that the side-collection conf iguration provided superior sensitivity in all channels of the spectrometer (Figure 3-7). The C247 signal and background obtained by placing the fiber so close to the plasma were almost 12 and 2 times greater respectively than those obtained using back-s ide collection. Even w ith the higher background,

PAGE 59

59 the S/B was nearly 7 times greater for the side -collection configuration. Clearly, placing the fiber closer to the plasma provided significan tly increased sensitivity and was consequently chosen as the best position of the fibe r probe for the remainder of the study. Optimizing Lens-to-Sample Distance The lens-to-sam ple distance (LTSD) is an important parameter that must be held as constant as possible in order to obtain reproducible spectra. This can be a challenge when analyzing samples of different heights, or when a single sample does not have a uniform surface. If the sample is moved farther from the focal point of the laser, the irradiance (W/cm2) of the sample decreases. In cases where the breakdown th reshold of the sample is low, or the laser energy is high, this can be beneficial because the laser beam samples a greater area. However, if the spot size is too great, a breakdown will not occur. Depending on the physical and chemical properties of the sample, there is an optimal LTS D that must be kept constant for a set of samples. In order to determine the best LTSD a graphi te pellet was placed on the sample stage and spectra were collected at 1 mm increments above and below the focal point of the lens (measured as 2.5). Ten shots at maximum laser energy were recorded and averaged. The results indicated that the greatest C247 signal was found when the fo cal point was 2 mm above the sample surface (Figure 3-8). Incidentally, this position also yielded the lowest standard deviation in measurements. It is important to note that the measured focal point of the plano-convex lens used in this research was measured using room light and may be slightly different for the 1064 nm laser. Regardless, the results of this experiment provide the optimum LTSD and fiber-tosample-to distance (FTSD) based on maximum C247 signal intensity. In order to ensure a constant LTSD and FTSD, regardless of different sample heights, a laser positioning system was constructed. With the graphite pellet at the he ight that provides the best signal, a small helium-

PAGE 60

60 neon laser was directed over the sample surface toward a photodiode connected to a voltmeter. Half of the laser beam was blocked by the sample and the voltmeter measurement for the blocked beam was recorded. When samples of different heights were placed on the sample stage, they were moved either up or down until they blocked the same portion of light as the graphite pellet in its optimal position. While using this positioning system allowed for each sample to be analyzed under the same condi tions, little could be done to account for the variations in height within the sample itself, such as the case for particulate soil matter. Data Analysis Once the req uired number of spectra had been collected for a sample, it was necessary to garner the required information by processing the data. As prev iously mentioned, the OOILIBS software had a mode which allowe d the user to collect statistics during the analysis; however the information provided was limited to simple statistic s, with no regard to the trends or patterns which may appear within a sample. Instead, the data collected during this research was processed using Microcal Origin and a Matlab program written specifically for the analysis of LIBS data. Both the peak area and the backgrou nd subtracted peak intensity were calculated (Figure 3-9). A comparison between the statisti cs (standard deviation, standard error) produced from calculating either the intensity or integral revealed that they were both very similar, however in general, computing th e integral yielded slightly lower RSDs than the intensity, most likely due to the dependence on more data points than the one pixel used to calculate intensity. The Matlab program also calculated the intensity and integral of the background (B), the noise (N), and RSDB.

PAGE 61

61 Table 3-1. Specifications for the Bi g Sky Ultra Nd:YAG pulsed laser. Repetition Frequency (Hz) Up to 20 Pulse nnergy (mJ)a Up to 50 Beam divergence (mrad)b <7 Energy stability (%)c 2 Beam diameter (mm)d 2.8 Pointing stability ( rad)e 50 Jitter ( ns Q-Switch)f <2 Energy drift over 8 hours <10% a) Nominal FWHM. b) Angle containing 86.5% ener gy. c) Variation from me an for 99% of shots d) Nominal value. e) Full a ngle, 99% of shots. f) Measur ed from Q-Switch sync output.

PAGE 62

62 Figure 3-1. Configuration of LIBS instrumentation. Laser Head Spectrometer Air stream Focusing Lens Fiber Sample Fiber optic probe with collimating lens

PAGE 63

63 Figure 3-2. Ablation area produ ced on copper foil by a 50 mJ laser pulse and a 2.5 inch focal length lens. (Note, only th e black area is considered the ablation area).

PAGE 64

64 Figure 3-3. A soil spectrum that cl early shows the discontinuities be tween different spectrometer channels in the LIBS2000+ unit. 200300400500600700800900 0 1000 2000 3000 4000 Intensity (Counts)Wavelength (nm)

PAGE 65

65 Figure 3-4. Arrangement of the individual fiber apertures in the fiber optic probe

PAGE 66

66 Table 3-2. Mean pulse energies of the Big Sky laser measured with a power meter. Expected pulse energy (mJ) Mean pulse energy of 32 laser pulses (mJ) Mean pulse energy without first ten laser pulses (mJ) Mean pulse energy of first 10 laser pulse (mJ) 25 29 1.7 (5.9%) 30 2.2 (7.2%) 29 1.4 (4.8%) 30 35 1.9 (5.3%) 36 2.4 (6.7%) 34 .3 (3.8%) 35 40 1.5 (3.8%) 40 1.7 (4.3%) 40 1.4 (3.6%) 40 46 2.3 (5.1%) 47 2.9 (6.2%) 45 1.7 (3.7%) 45 51 3.7 (7.2%) 54 2.7 (5.0%) 50 3.7 (7.4%) 50 57 1.8 (3.2%) 59 2.2 (2.3%) 57 1.3 (2.3%)

PAGE 67

67 Dead time of the CCD ~180 s Acquisition 2.1 ms 170 s 190 s Plasma Flash lamp fired SOFTWARE CONTROLLED Emission collected Plasma Q-switch is fired before CCD aperture is opened PositiveQ-Switch values Q-switch is fired after CCD aperture is opened NegativeQ-Switch Values Figure 3-5. Controlling the CCD emission acquisition delay tim e by manipulating the laser Qswitch.

PAGE 68

68 Figure 3-6. Results of changing the Q-switch de lay times on soil spectra. A) The C247 line drastically changes with spectrometer dela y. B) C247 signal intensity vs. delay. C) C247 S/B vs. delay. D) CL vs. delay. 247.6247.8248.0248.2248.4248.6 0 300 600 Signal IntensityWavelength, nm +4 +3 +2 +1 0 -1 -2 -3 -4 Soil Spectrum Varying Q-Switch (2-2 L : 20.38% C) Positive No delay Negative Delayed (1000ns increments) 247.681 247.837 248.098543210-1-2-3-4-5 0 50 100 150 200 250 300 350 Signal Intensity, CountsQ Switch Delay (us increments) C247 Signal Intensity vs. Q-Switch Delay543210-1-2-3-4-5 0.01 0.02 0.03 0.04 0.05 0.06 Minimum Detectable wt% C Q-Switch Delay (us increments)A. B. C. D.543210-1-2-3-4-5 -5 0 5 10 15 20 25 30 35 40 45 50 S/BQ Switch Delay (us increments) S/B vs. Q-Switch Delay

PAGE 69

69 Figure 3-7. Graphite pellet spectra with A) back -side emission collecti on and B) side emission collection (Note: Only the C247 line is identified because it was used to determine increase in signal intensity) 200300400500600700800900 200 400 600 800 1000 1200 1400 1600 1800 Intensity (Counts)Wavelength (nm) C247 200300400500600700800900 200 400 600 800 1000 1200 1400 1600 1800 Intensity, CountsWavelength, nm C247 A B.

PAGE 70

70 -6-4-20246 400 450 500 550 600 650 700 750 800 850 C (247.8nm) net intensityFocal Distance Offset (mm) Figure 3-8. C247 signal intensity (t he average of 10 shots) as a function of focal point offset (error bars are 1 standard deviation).

PAGE 71

71 Figure 3-9. Calculation of intensity (I), b ackground (B) and area (above the red line). 247.2 247.6 248.0 248.4 0 100 200 300 400 Signal Intensity, CountsWavelength, nm I B

PAGE 72

72 CHAPTER 4 SAMPLE PRESENTATION AND SAMPLING CHARACTERISTICS One of the most important advantages of th e LIBS technique is the ability to analyze a variety of materials without the need for extensive sample preparation. Indeed, one of the motivating factors behind this research is the potential to analyze soil samples in their natural, unperturbed state with a portable LIBS system. However, before this can be achieved the technique must be validated in the laboratory, and this presents the question of how to prepare the sample for measurements with as little modifi cation as possible. In addition to this, an optimized sampling regime must be establishe d that accounts for the chemical and physical heterogeneity of soil samples and maximizes th e information obtained during the analysis. Sample Presentation Pressed Pellets vs. Mounting Tape The m ajority of research that has been conducted on soils has involved grinding soil samples in a ball mill and using a pellet press to produce tight, compact pellets.44,47-50,54-58 There are many advantages to using pres sed pellets for LIBS; most significantly, the sample surface is completely flat and there is uniform mass ablation. However, pressing pellets is time consuming, can introduce contamination and ofte n requires the use of binders, many of which contain carbon and are therefore pr ecluded in carbon analysis. In addition to this, the pressure used during the pellet production will have a direct effect on the mass ablated and the LIBS signal. As such, it must be ensured that the exact same pressure is applied for each sample, no easy feat with a manual pellet press. Finally, and possibly the most compelling disadvantage, pressing soils into pellets completely destroys the integrity of the soil sample such that subsequent LIBS analysis will in no way simulate the analysis of the same soil in its natural state. Simply, if laboratory anal ysis is conducted on pressed pellets then in order for the results

PAGE 73

73 to be valid for in-field analysis, samples must also be pressed into pellets. Obviously, this additional sample preparation defeats the high throughput capability desi red for the portable instrument. An alternate method of sample presentation is to mount the sample on a microscope slide using double-sided foam mounting tape. This is achieved by attaching the desired area of tape onto a standard microscope slide and removing the backing to expose the adhesive. The tape is then firmly pressed onto a thin layer of the sample and the ex cess, unstuck sample is gently tapped away. The tape is then fully covered with a monolayer of pract ically undisturbed soil particles. The so-called m ounting-tape sample preparation method is very quick, cheap, introduces few contaminants, and minimizes sample modification. Due to these very attractive qualities, it was the method of choice for sa mple preparation during this research. In order to study the effects of the using the mounting tape sample preparation, LIBS analysis was performed on a pressed soil pellet a nd the same soil mounted on a microscope slide (Figure 4-1). The soil pellet was prepared by applying in excess of 16,000 psi for 30 seconds. One drop of water was added to th e pellet before pressing to make the sample more cohesive. In general, a binder would be used for this task ; however common binders such as cellulose and wax contain carbon and would therefore interfer e with the LIBS carbon signal. The resulting pellet was 13 mm in diameter with a height of approximately 4 mm. Both samples were analyzed using the same laser and spectrometer parameters (pulse energies of 57 mJ, a delay time of 1 s and repetition rate of 1 Hz). For the slide sample, two la ser shots were used for each of 50 sample spots and data from each shot were collected (termed first-shot and second-shot). The purpose for this double-shot analysis will be di scussed later in the chapter. The resulting 50 spectra were collected, and the average integrat ed C247 peak was determined for both the first

PAGE 74

74 and second shot measurements. For the pellet sample, measurements from 5 separate sample spots were collected. Each spot was subjecte d to 5 cleaning shots (to remove contaminants imparted by the pellet press) and 10 subseque nt measurement shots. The C247 signal was averaged for each of the five spots, and then the average C247 integral was determined for the whole sample. The soil sample chosen for this compar ison contained 9.15% carbon. All soil carbon concentrations in our research were measur ed by dry combustion at the Department of Chemistrys CHN laboratory using an Elemental Analyzer from Carlo Erba Instruments operated at a temperature of 1020oC and using a sample mass of approximately 2 mg. The average C247 peak area produced from 50 second shots on th e microscope slide was 35.4 4.6, an RSD of 11.7%. The average for 5 spots of 10 shots on th e pellet was 18.1 2.9, an RSD of 16.3%. One possible reason for significant decr ease in C247 signal for the pell et was that much less mass was ablated than for the microscope slide mounted sa mple. As the pellet was produced under great pressures, the sample was highly compacted and the laser energy penetrates less into sample, ablating less mass. Moreove r, inspection of the C247 signal as a function of shot number on the pellet indicated that crater effects were contributing to the poor precision. As the number of laser shots increased on a spot, the crater became deeper which resulted in a greater LTSD and FTSD as well as shielding of the plasma by the crater walls. This is evident by the decrease in C247 signal as the shot number increased (Figure 4-2). Based on these results, the use of microscope slides to mount the soil samples was judged superior to pressed pellets for LIBS meas urements due to increased mass ablation and measurement precision. In addition to this, moun ting the soils on microsc ope slides maintained

PAGE 75

75 the original physical properties of the soil, and as such allowed fo r further studies of the effects of grain size on LIBS signal. Analysis of Bare Mounting Tape The tape that was used to m ount the sample to the microscope slide was produced by Scotch 3M and consisted of a foam center coat ed on both sides with an adhesive. The foam backing and the adhesive coating are composed of acrylic and vinyl pol ymers which contain high concentrations of carbon. For th is reason, the tape was extensiv ely analyzed to determine the extent of carbon interference. Fi rst, nine separate measurements on different spots of the bare tape were collected. The C247 lin e was not detected in any of them (Figure 4-3a). Following this, 12 shots were collected on the same spot of the tape. Each of th e resulting spectra were compared, and it was determined that carbon was not detected after the fi rst shot; however, it was detected after subsequent s hots (Figure 4-3b). A plot of the shot number directed on the same spot and the resulting C247 intensity showed that the maximum amount of carbon (approximately 55 counts under these specific inst rumental conditions) was detected for the second laser shot, after which time the signal decr eased, until a plateau was reached after 6 shots (figure 4-3c). To investigate the reproducibility of this phenomenon, 10 pairs of shots were taken on the tape. In each instance, the first sh ot yielded no carbon peak and the second shot produced a carbon peak of with a net intensity of approximately 60 count s (Figure 4-3d). Images taken of the ablation spots after a number of different laser shot s illustrated that the first laser shot did not produce a discernible crater and the surface of the tape merely exhibited signs of localized melting. It is possible that the breakdown threshold of the acrylic adhesive is not reached under these sampling parameters; howeve r the first shot modified the surface of the tape, allowing for the second shot to produce plasma and for carbon detection to occur. Regardless of the mechanism, this occurrence suite d the purposes of this research well. During

PAGE 76

76 the soil measurements using the mounting tape substrate, no more that two consecutive measurements were taken on the same spot. If th e soil acted as the first barrier and the acrylic adhesive of the tape acted as the second barrier, it would have required at least three shots on the same spot to produce carbon in terference from the tape. Grain size All natural soils are classified according to % sand, % silt, % clay and as such exhibit a wide range of particle sizes both between and within a soil sa mple. Previous research by Wisburn et al revealed that the slope of the calibration curve decreases with decreasing grain size for analytes that are primarily present as surface contaminants.50 In order to study the effect of matrix grain size on the C247 si gnal, five different particle size fractions of quartz (SiO2, Alfa Aesar) were spiked with graphite (pure carbon) and measured. Qu artz was chosen as the matrix due to its physical and chemical similarities to the main compone nts of soil. Separation of the quartz sample into different size fractions was achieved at the University of Floridas Soil Science Department by sieving and water separation. The resulting fractions had particle sizes in the following ranges: >500 m, 250-500 m, >30 m, 18-30 m, <18 m and 0.5-1 m. However there was not enough of the last sample for LIBS measurements. Samples of each size fraction were then spiked with 0-19 % gra phite by weight. The method of sample homogenization was critical in maintaining the un iform particle size of the sample. The usual method involved adding small, methacrylate balls to the sample and blending in a ball mill for a specific period of time. The methacrylate ball s produced a homogeneous matrix by breaking up and mixing the particles; however, this significantly decreased the grain sizes. In addition to this, microscopic inspections of the samples after blending with plastic mixing balls indicated that the quartz particles were not uniformly cove red with the graphite. This uneven coating is possibly due to the mixing balls continuously knocking the graphite off the particle. In order to

PAGE 77

77 maintain the same grain sizes and ensure uniform coating, the samples were milled for 15 minutes without mixing balls. Microscope images were taken of each sample after it had been affixed to a microscope slide using a Nikon Labophot microscope with a 4X magnification (Figure 4-5). It is clear that graphite was much more homogeneously distributed in the quartz samples of smaller grain size than the larger ones. In addition to this, as the concentration of graphite in large grain size quart z matrix increased, the number of quartz particles decreased and the graphite coated larger areas of the tape surface. This resulted in some of the laser shots sampling 100% graphite and significantly a ffected the precision of the sampling. The average integrated C247 peak of 25 separate measurements on each sample was determined and a plot of signal vs. carbon concen tration was constructed for each of the particle fractions (Figure 4-6). The statistics for each size fraction (Table 4-1) illustrate how the precision of the measurements significantly increased as the particle size of the matrix decreased. This trend was expected considering the higher degree of heterogeneity associated with the larger grain sizes. The homogeneity of the smaller grain sizes resulted in sampling that was more consistent and representative of the bulk sa mple. Table 4-1 also illustrates how the slopes of C247 signal vs. carbon concentration decreased w ith the particle size. This was in agreement with the results of Wisburn et al .50 The decreasing slope was due to the decreased mass of carbon that was sampled per shot as the particle size of the qu artz matrix decreased. As illustrated in the microscope images of the sample s (Figure 4-5), the larger quartz particles were completely coated in graphite which concentrated more graphite into the ablation area of the laser beam. However, as the particles became sma ller and closer in size to the graphite particles almost no coating occurred, and the graphite a nd quartz were more inde pendently distributed resulting in fewer concentrated areas of carbon and a more homogenous mixture.

PAGE 78

78 One troubling detail evident in Figure 4-6 is the large C247 si gnal that was produced with the blank quartz samples, especially for the three smaller particle size fractions. The initial assumption was that the bulk quartz sample cont ained some carbon contamination. To measure this, each of the five size frac tions was analyzed by dry combus tion. The results indicated that trace amounts of carbon were pres ent, with the concentration increasing as the particle size decreased (Table 4-1). Indeed, the 0.5-1 m quartz fraction, which wa s not used for the LIBS analysis due to limited sample, had a carbon conc entration of 0.118%, 10 times greater than the carbon concentration of the 500 m fraction. In spite of this, such a contamination is still very small, less than 0.1% for the quartz samples used in this study, and could not account for such a large carbon signal. To further study this, the smallest grain si ze fraction was pressed into a pellet and measured by LIBS. Several cleaning shots were used to remove surface contaminants introduced by the pellet press, and a total of 20 measuremen ts were taken. Interestingly, the carbon peak was not detected. Based on these results, it can be concluded that th e quartz sample was not contributing to the large carbon sign al in the blank particle size fractions. The only available source of carbon was the mounting tape, yet prev ious studies had exha ustively shown that no carbon was detected on the tape after the first shot, and only a small peak was detected for subsequent shots, the magnitude of which did not account for the large carbon si gnal. In spite of this, LIBS measurements were taken on bare ta pe and then immediatel y on fresh quartz covered tape to monitor the difference in carbon measur ements. As expected, no carbon was detected on the fresh tape, however on the quartz covered tape a large carbon peak appeared. The mechanism behind this amplified carbon det ection on the tape is unknown; however, it seems that small particles adhere to the viscous adhesive resulting in a physical or chemical

PAGE 79

79 change to the LIBS plasma causi ng the carbon ablation from the tape to increase. To evaluate this supposition, extensive studies of the plasma diagnostics as a func tion of particle-tape composition would be required. Such an investig ation would be the content of a full PhD study. However, it should be noted that the magnitudes of the carbon signal for each of the three blank quartz size fractions were very similar, indicating th at this may be a fairly constant interference that could be accounted for by merely subtrac ting the blank signal from the spiked sample signals. Another assumption that could be made is that this tape-carbon enhancement effect seems limited to samples of very fine powders, unlik e the rather coarse soil samples used in this research. The results of this quartz particle size study illustrated the great de pendence of LIBS signal on matrix particle size and analyt e distribution. It is clear that LIBS measurements performed on soils that have a high % sand comp osition can not be directly comp ared with soils that have a high % clay composition. It is also clear that soils with high % sand compositions will likely yield intolerable precision whic h will preclude accurate quantita tive results. However, there were many variables that were not considered in this study which limited the direct application of these results to natural soils For instance, the form of car bon in this study, graphite, in no way simulated the actual sources of carbon in soil mostly organic acids, biological materials and inorganic carbonates. This difference will undoubtedly significantly affect the plasma properties and the resulting carbon signal. In addition to this, the distribution of graphite in quartz is not truly representative of how carbon is distributed in natural soils, where it may either be trapped inside particles, heterogeneously coating them or as free organic matter. One final consideration is the actual matrix itself, which is certainly more complex than simple quartz and may contain many elements that may chemically or physically interfere with carbon. Natural soil matrices

PAGE 80

80 also exhibit a wide particle si ze distribution unlike the almost uniform quartz particle size fractions which were used in this experiment. Sampling Characteristics Number of Laser Shots The num ber of measurements on a soil sample was a very important parameter that was directly related to the precision of the analysis As previously discu ssed, soils are inherently heterogeneous and any given soil sample will be composed of a variety of particles exhibiting an array of different physical and chemical compositions To illustrate this fact more clearly, Figure 4-7 shows microscope images of three soil samples which exhibit different organic content and particle sizes. The yellow ci rcles superimposed on the images represent the approximate area that would be ablated by the focused laser. This was determined by the diameter of the crater produced on copper foils under the same conditions and was about 0.35 mm2. It is plain to see that each laser shot will be sampling very different sized particles of visibly different composition (for reference, the dark particles are likely organi c matter whereas the translucent particles are likely silica). This heterogeneity can partially be comp ensated for during LIBS measurements by increasing the mass analyzed (d efocusing the laser to increase the ablation area) and averaging out the differences by ta king a many more measurements. However, defocusing the laser risks decreasing the en ergy density of the laser and not forming a breakdown. Instead, increasing the number of laser shots, and th erefore the tota l mass analyzed was preferred. To investigate how the number of laser shot s (on separate sample spots) affected the precision, LIBS measurements on two dried soil sa mples were collected. One of the soils had low organic content and high % sand and the other had high organi c content and was less coarse. The samples were fixed to a microscope slide a nd subjected to 50 laser shots; the sample was

PAGE 81

81 manually moved between each shot to ensure fresh sampling. The RSD of the background subtracted C247 intensity was noted after each subs equent shot (Figure 4-8). It was found that for both soils, regardless of their physical or chemical differences, the RSD was relatively constant after 20 shots on different spots. To fu rther investigate the effect of the number of laser shots on the RSD, three microscope slides were prepared of the same soil and each slide was subjected to as many separate laser shots as sa mple availability would allow, approximately 290 shots per slide and 856 shots total. The aver age C247 signal was then calculated for each individual slide, and for all 856 shots (Table 4-2) Surprisingly, increasing the shot number up to 290, and even to 856, did not improve on the RSD ach ieved after only 20 shots. In addition to this, the differences in both average C247 peak area and the associated RSD between the slides were very small, indicating that the intra-sample precision for the technique, when using greater than 20 shots, is high. However, the inter-sampl e RSD of 19% was rather high and will certainly affect the accuracy of the results. However, these experiments revealed that the greatest contribution to this poor precisi on is not the LIBS technique, but rather the soil heterogeneity itself, something that can only be improved by excessive grinding and milling. Therefore, if little sample preparation is desired, the inherent heterogeneity of the soil will contribute to the poor precision. Repetitive Laser Shots W isburn et al also reported on the occurrence of a pe rsistent aerosol that was formed over the sample surface after the first shot on the soil and contributed to enhanced signal intensity after the second shot in the same spot. Howeve r, it was reported that this phenomenon was only achieved at repetition rates in excess of 1 Hz and all results were produced on soil pellets.50 However, measurements obtained from the same s pot on tape-mounted soil revealed that even at repetition rates of 1 Hz, greater signal intensity was achieved for the second shot on the same

PAGE 82

82 spot. This was surprising since to the naked eye, it looked as though there was no soil left to sample after the first shot, least of all the s econd shot. Initially it was thought that the carbon signal increased because the s econd shot was measuring only the carbon from the mounting tape, and indeed, based on the adhesive -particle carbon enhancement re ported previously, this may be a contributing factor. However, this significant increase in signal occurred across the entire spectrum and not just for carbon. A plot of si gnal intensity vs. shot number for carbon and calcium (which was certainly not present in the ta pe) revealed the same trend (Figure 4-9 a,b). The carbon signal was prevalent until around the 175th shot on the same spot and the calcium signal was prevalent until the 80th shot. One explanation was that airborne particles that had been liberated from the sample by the force of the plasma shockwave were either entering into the plasma or interacting with the laser beam along its axis an d forming mini-plasmas. When watching the sampling in the dark, it was easy to ascertain these small plasma events occurring up to an inch above the sample surface. The em ission from these small independent plasmas was likely contributing to the long lifetime of the calcium and carbon signals. To reduce this effect, which decreased the am ount of laser energy av ailable for the next sampling spot, a stream of air was directed over the target surface to sw eep any loose particles from the sampling areas. This drastically decrea sed the number of laser s hots that still produced carbon or calcium emission, however, even with the air stream carbon and calcium could be detected until approximately the 20th shot (Figure 4-9 c,d). The greatest difference was that without the air, the emission intensity for both elements was very random, increasing or decreasing with no discernable pattern. Yet with the air stream, the emission intensity reached a single maximum, usually after the first or second shot, and exponentially decreased to zero after 1020 shots (it takes longer for the carbon signal to disappear due to the carbon being measured

PAGE 83

83 in the mounting tape). To ensure that this increase in signal intensity after the second sampling shot was reproducible, a suite of 38 soil samples were analyzed with 100 laser shots on the same spot. For each soil, the maximum carbon signal was achieved after varying numbers of laser shots (between 3-6 shots) but never after the fi rst shot. The same pattern was seen for iron, though the shot number for the maximum iron signa l did not correspond to that of carbon (this was achieved between shots 2 and 4). Microscope images of the tape-mounted soil af ter one laser shot show that there was no discernable crater and the soil sa mple had not been ablated through to the tape surface (Figure 410). The tape had a very distinctive surface composed of small bubbles and was easy to recognize. The soil sample remaining after the first shot was a very fine powder with no individual particles visible at 4X magnification. The cause of this signal enhancement after the second shot was not fully understood, however it was hypothesized that the first laser shot prepares the sample surface to provide for maxi mum ablation efficiency during the second shot. That is to say, the energy from the laser pulse in teracts with more sample in the second shot than it does in the first shot. The particles of the orig inal soil samples were often larger than the diameter of the focused laser beam and as such it is plausible that the plasma was formed on the surface of one particle, shielding any sample adhe red to the surface of the tape from ablation and possibly wasting energy from the laser pulse if only a portion of the pulse was used on the particle during ablation (Figure 4-11). However, the combination of the initial laser pulse, the resulting plasma shockwave and th e continuous air stream sweeps the area of any large particles, leaving a thin, uniform monolayer of soil powder. As such, when the sec ond laser pulse interacts with the sample, the entire la ser pulse interacts with sample without any large particles obstructing part of it. Regardless of the mech anism for the second shot signal enhancement, it

PAGE 84

84 was clear that sensitivity could be increased across the spectrum by utilizing repetitive shot sampling; one initial preparatory shot and a seco nd shot on the same spot from which data is collected. Conclusion As a result o f the preceding observations, a sampling regime optimized for possible soil analysis was evaluated. The advantages of m ounting the soil samples on a microscope with adhesive tape, namely simplicity and maintaining particle size integrity far outweighed pressing the samples into pellets. Furthermore, it was de termined that for at least one soil sample, the precision of C247 LIBS analysis on the slide exceeded that of the same soil in pellet form. While this may not be the case for all soils, it was deemed an acceptable trade-off considering the ease with which slides could be prepared in-field. While evidence indicates that there may be some physical or chemical process occurring between the small sample particles and the tape adhesive that resulted in enhanced carbon signal from the tape, this was assumed to be a constant contribution for each laser shot. A study of matrix grain size on LIBS signa l revealed that there was a significant effect on the C247 signal intensit y and RSD. Ideally, when measuring different soils they should be all be ground and homogenized to approximately the same size in order to minimize grain size effects between samples and to increase precision. However, since this would not be a feasible option for in-field an alysis, care should be taken that only soils containing similar particle si ze distributions are compared. With respect to the sampling population, it wa s established that a minimum of 20 laser shots should be taken on a soil sample in order to account for the heterogeneity of soils. Taking many more than 20 laser shots did not seem to im prove the RSD. Finally, it was determined that using a preparatory laser shot on the soil surface before the sampling shot significantly increased the signal intensity of the entire spectrum. It should be not ed however that for 38 soil samples,

PAGE 85

85 the C247 signal intensity from the first and s econd shots were proportional, with an average increase of 1.7 observed for the second shot. Howe ver, due to the weak nature of the C247 line, the increase in sensitivity was exploited and second shot measurements were employed for the majority of this research.

PAGE 86

86 Figure 4-1. An image of a pressed pellet and soil mounted on a microscope slide.

PAGE 87

87 01234567891011 12 14 16 18 20 22 24 26 28 30 Integrated C247 PeakShot number Figure 4-2. C247 signal as a functio n of shot number on the same pellet spot for five different locations.

PAGE 88

88 Figure 4-3. Analysis of bare mounting tape. A.) Spectra from 9 different locations of the tape surface. B.) Spectra from 12 consecutive shots on the same tape spot. C). The C247 signal intensity as a function of shot number on the same spot. D.) Spectra from the first and second shot on tape. 012345678910111213 160 170 180 190 200 210 220 230 Intensity of C 247.889 nm (Counts)Shot number 247.2247.6248.0248.4248.8 160 165 170 175 180 185 190 195 200 205 210 215 220 225 230 Intensity, CountsWavelength, nm Spot Location 1 2 3 4 5 6 7 8 9 246.0246.4246.8247.2247.6248.0248.4248.8249.2249.6250.0 160 170 180 190 200 210 220 230 Intensity (Counts)Wavelength (nm) shot 1 shot 2A C 246.0246.4246.8247.2247.6248.0248.4248.8249.2249.6250.0 160 165 170 175 180 185 190 195 200 205 210 215 220 225 230 Intensity, CountsWavelength, nm Laser shot number 1 2 3 4 5 6 7 8 9 10 11 12B D

PAGE 89

89 Figure 4-4. Microscope images (4X magnificati on) of the craters produced on mounting tape after the indicated shot number. 1mm 100 shots 3 shots 10 shots 50 shots 1 shot 2 shots 20 shots

PAGE 90

90 Figure 4-5. Images of differen t quartz fractions spiked with graphite (4X magnification). 0 2 4 6 9 14 19 >500 m 250-500 m >30 m18-30 m <18 m Particle Size 1 mm % Graphite

PAGE 91

91 Figure 4-6. C247 peak as a function of carbon concentrati on (wt %) for different particle sizes of quartz matrix (error bars ar e 1 standard deviation). 0246810121416182022 0 10 20 30 40 50 60 70 80 90 100 Integrated C247 Peak% Carbon >500 um Quartz 0246810121416182022 0 10 20 30 40 50 60 70 80 90 100 Integrated C247 Peak% Carbon 250-500 um Quartz 0246810121416182022 0 10 20 30 40 50 60 70 80 90 100 INtegrated C247 Peak% Carbon 18-30 um Quartz 0246810121416182022 0 10 20 30 40 50 60 70 80 90 100 Integrated C 247 Peak% Carbon >30 um Quartz 02468101214161820 0 10 20 30 40 50 60 70 80 90 100 INtegrated C247 Peak% Carbon <18 um Quartz A. B. C. D. E.

PAGE 92

92 Table 4-1. Statistics for the analysis of qua rtz size fractions spiked with graphite Quartz particle size Average RSD for all carbon concentrations (%) Correlation coefficient (R) Slope (C247 area/% C) Carbon concentration (wt %)a >500 m 40 20 0.9143 2.74 0.01 250-500 m 26 21 0.7554 1.83 0.001 >30 m 12 21 0.8691 1.52 0.016 18-30 m 10 2 0.9657 1.25 0.020 <18 m 7.2 1.2 0.9898 1.30 0.034 0.5-1 ma 0.118 a) Determined by dry combustion. b) No t used in study due to limited sample

PAGE 93

93 Figure 4-7. Microscope images of three soils to illustrate he terogeneity. The yellow circles represent the ablation ar ea of the focused laser beam, approximately 0.35 mm2. (Note: the 10X magnification are not the sa me areas as the 4X magnification areas) 4X Magnification 10X Magnification

PAGE 94

94 0102030405060 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 RSD (%)Shot Number Figure 4-8. The RSD of the C247 signal intens ity as a function of the number of shots on separate spots.

PAGE 95

95 Table 4-2. The C247 statistics for intrasample and inter-sample measurements. Slide Number of spots sampled C247 Area RSD (%) A 272 28 5 19 B 293 28 6 19 C 291 27 5 18 Average of the shots taken on all slides 856 28 5 19

PAGE 96

96 Figure 4-9. The C247 and Ca422 signal intensities as a function of shot number on the same spot A). without air stream B). with air stream. 0255075100125150175200225 0 50 100 150 200 250 300 350 C247 IntensityShot Number20406080100120140160 0 500 1000 1500 2000 2500 3000 3500 4000 Ca422 IntensityShot count 020406080100120140160180 0 50 100 150 200 250 300 C247 IntensityShot Number 020406080100120140160 0 500 1000 1500 2000 2500 3000 3500 4000 Ca422 IntensityShot Number B. A.

PAGE 97

97 Figure 4-10. Microscope images of soil after the indicated number of laser shots (4X magnification). 1 2 3 4 5 6 7 8 9 10 20 30 40 50 100 1 mm

PAGE 98

98 Figure 4-11. Proposed mechanism for the second shot signal enhancement on soil. Plasma Laser Pulse

PAGE 99

99 CHAPTER 5 THE SPECTROSCOPY OF CARBON AND CALI BRAT ION CURVES WITH SIMULATED SOILS The emission properties of carbon have been well studied, especially for the detection and potential quantifica tion of molecular fragments such as C2 and CN.84-90 Much of the previous work concerning atomic carbon emission has been achieved by monitoring the 247.8 nm line. This chapter will discuss the characteristics of this line and why it was chosen for this research, as well as alternate aspects of the carbon sp ectrum which were considered for quantitative analysis. The second half of this chapter descri bes the results of calibration curves that were produced from graphite spiked silica and alumin um silicate powders and subsequent validation with natural soil samples. The Spectroscopy of Carbon Choice of Analytical Line To determ ine the best emission line of car bon for quantitative purposes using the Ocean Optics LIBS2000+ spectrometer, emission from the ablation of a graphite pellet was studied. Initially, spectra were taken with no acquisition delay. This was because the ionization energies of the carbon atoms are relatively high, and are therefore expected to be more prominent early in the plasma lifetime, where the temperatures ar e high enough to sustain emission. When no delay was used, the graphite spectrum showed four em ission lines in the UV region: three ions that produce emission lines at 229.6 nm, 251.1 nm and 283.8 nm and have ionization energies of 53.5 eV, 27.5 eV and 27.5 eV respectively and one stro nger atomic line that emits at 247.8 nm and has an ionization potential of 7.7 eV (Figur e 5-1). When an acquisition delay of 1 s was applied, the three ionic lines were not detected and the emission intensity of the atomic line decreased. To further study the temporal proper ties of the C247 emission, spectra were taken on

PAGE 100

100 a graphite pellet using an Acton Spectra-Pro-500i spectrometer with an ICCD detector which allowed for the integration time to be changed. Spectra from the graphite pellet were collected for 200 ns at increasing delay times. The maximum C247 emission was achieved approximately 200 ns after plasma formation, and steadily decrea sed thereafter as the plasma cooled (Figure 52). However, these data did not reflect the high continuum emi ssion that was also occurring at such early intervals in the pl asma life time, and even though the C247 emission had substantially decreased by 1 s, earlier research, discussed in Chapte r 3, revealed that the S/B was at a maximum. It was initially proposed that the CN bands might be used for quantitative analysis of carbon in soil. However, previous research has indicated that the CN molecule in the LIBS plasma is principally formed through the react ion of a carbon dimer and atmospheric nitrogen: C2 + N2 2CN Evidence indicates that the majority of the C2 available for this reaction is from the photodissociation of compounds in the sample th at contain carbon double bonds or delocalized carbon bonds rather than the recombina tion of carbon atoms in the plasma.90-92 Because of this, the use of the CN band for qua ntitative analysis seems rele gated to unsaturated organic compounds and not for carbon in soil, which origin ates from a variety of organic and inorganic sources. Besides, inspection of the CN bands in a spectrum of soil showed that due to the high background in the 300-400 nm spectrometer channel and the multitude of interfering species, the CN band head at 388 nm was unresolved and barely detected, and the two lower wavelength bands at 387 and 386 were not re solved at all (Figure 5-3). The carbon atomic emission line at 247.8 nm was the most distinct line in the graphite spectrum and was the obvious choice to monito r. Information concerning the energy levels,

PAGE 101

101 relative intensities and exact wavelength for car bon were obtained from the NIST atomic line database.93 In order to assess the acc uracy of the LIBS2000+ spectrometer, the actual measured wavelength of 247.837 nm was compared with th e true wavelength of 247.856 nm. A difference of 0.019 nm corresponded to less than one pixel a nd indicated high accuracy in the UV channel. The precision of the spectrometer was also assessed by collecting 100 spectra on a pure, homogenous graphite pellet. The RSD of th ese measurements was an acceptable 2.4%. Curves of Growth To determine the linear concentration limits of calibration linearity, theoretical curves of growth (COG) were constructed using a softwa re program developed in our group. At high concentrations of analyte, a cal ibration curve is exp ected to exhibit non-l inear behavior, although the actual concentration at which this will occu r varies as a function of plasma temperature, electron number density, emission line monitored and matrix composition. The theoretical COGs for the C247 line were calculated assuming a matrix of SiO2 and plasma temperatures of 8000, 1200 and 16000 K. While the use of such a simplified matrix undoubtedly introduced some error in the results, it w ould be impossible to account for al l components of the soil matrix as well as their relative concentrations. From the calculated theoretical COGs shown in Figure 5-4, it is clear to see that the theoretical integrated C247 peak varies greatly with plasma temperature. For instance the relative C247 peak areas are 0.0053:0.26:1 for plasma temperatures of 8000 K, 12000 K and 16000 K respectively. Despite the signal disparities for different plasma temperatures, all three COGs indicated a departure fr om linearity for carbon concentrations in excess of 8 %. While these COGs provide useful informati on concerning the theore tical LIBS response, it is important to realize that th e inhomogeneity of soil samples pr ecludes the direct use of these calibration curves. As previously discussed, an d is evident from Figure 4-7, each plasma that is

PAGE 102

102 formed on a soil surface will vary greatly in chemical composition due to the heterogeneity of the soil composition. In addition to this, the slig ht differences in LTSDs that the uneven surface topography produces will cause diffe rent laser irradiances. Both these factors contribute to highly variable plasma temperatures and electr on number densities and because of this, a theoretical calibration curve based on such broad assumptions will not produce accurate results when applied to natural soil samples. Experimental Calibration Curves Silica and Aluminum Silicate Artificial Soils Before any LIBS m easurements on soil sample s were undertaken, an attempt was made to produce calibration curves for carbo n in matrices with chemical compositions similar to the main components of most soils. Due to the complex na ture of soils, it would be impossible to produce perfect matrix matched standards; instead, fine silica (SiO2) and aluminum silicate (Al2SiO5) powders were used. Each matrix was spiked with 0-10% graphite powder by weight. The powder mixtures were then homogenized with a ball mill for 15 minutes and mounted on microscope slides as previously described. A pproximately 20 spectra were taken on each slide and the average C247 intensity calculated. No preparatory laser shot was used for these analyses because unlike natural soils, these samples were very extremely fine powders and the entire layer was ablated with one shot. Two things were evident from the calibration curves shown in Figure 5-5. First, both matrices exhibited a definite linear correlati on between LIBS carbon si gnal and the % carbon. Second, the slope for the graphite in silica curve is greater than that of the graphite in aluminum silicate. This is a prime exampl e of how dependent LIBS is on matr ix effects; two matrices that are similar in chemical composition, save for th e presence of aluminum in one, produced very different LIBS responses. One im portant thing to note when cons idering these curves is that

PAGE 103

103 after concentrations of carbon in excess of 6% both curves reach a plat eau. When points after 6% were removed from the curv e, a linear fit could be applied that produced a correlation coefficient, R, of 0.9880 for silica and 0.9932 for alum inum silicate. It should also be noted that both curves exhibit a non-zero intercept. Sample s of silica and aluminum silicate were submitted for dry combustion CHN analysis, which revealed that both contained car bon at concentrations of approximately 0.1 %. However, such a concen tration can not explain su ch a large signal. It was determined that the same particle-adhesive effects were contributing greater carbon from the tape. However, the linearity of these curves up to 6% provide further evidence that this contribution is likely constant for all sampling shots. When constructing these potenti al calibration curves, it was realized that their use for natural soil samples would be limited. The effect of the two different matrices on the LIBS signal is significant, and such a matrix effect was expected to be more pronounced for soil samples, which are much more varied in ch emical and physical composition. Based on this knowledge, the possibility of constructing a univer sal LIBS calibration curv e for all soil samples is not reasonable. Furthermore, it was consid ered highly unlikely that a calibration curve produced from graphite spiked silica or alum inum silicate powders would be highly accurate predictors of % carbon in natural soils. Howeve r, the level of accuracy required for carbon in soil measurements is still in question, and these results provided a good foundation for studying carbon in a soil-like environment. LIBS Measurements of Powdered Soil Samples Before any m easurements of natural soils we re undertaken, soil standards were obtained from NIST. These four soils, Buffalo Rive r Sediment (SRM 2704), Estuarine (SRM 1646), River Sediment (SRM 1645) and Montana Soil (SRM 2710) were very fine powders and considered completely homogenous. Only two of the soils, Buffalo River and Montana Soil

PAGE 104

104 were certified for carbon; therefore, all soils were submitted for dry combustion CHN analysis. Several soil samples were also collected indepe ndently from different arbitrary locations in Marion County, Florida (referred to as locations A, B and C). At each location, samples were taken from depths of 1, 4 and 7 using an auger. Soil samples from around the University of Florida campus were also collected (termed S, SE and NE, referring to the collection location relative to the Chemistry Department) and one sample was provided by Aerodyne Research, Inc from Massachusetts. These samples were all ai r dried for one week a nd ground with steel balls in a ball mill for half an hour to produce a fine powder much like the NIST standards. CHN analysis was performed several times for all nine soils and the results are reported in Table 5-1. It can be seen that for all three sampling locatio ns, A, B and C, the soil samples taken from 4 and 7 contained much less carbon than samples take n from 1. This was expected, since the soil organic matter (SOM) decreases significantly with soil depth. As such, these samples were not included in the LIBS analysis. Each sample was analyzed with 100 single lase r shots, each on a fresh sample location. The average C247 signal intensity was calculate d and plotted against the %C determined by CHN analysis (Figure 5-6). From this plot one which the calibration curves produced from graphite spiked silica and aluminum silicate, are also imposed, two distinct groupings of the soils emerged. The grouping first is composed of the soil samples collected from Marion County, plus the Aerodyne soil. These soils formed a steep slope slightly gr eater than that of graphite in silica. The second group of soils exhibited a slop e slightly less than that of the graphite in aluminum silicate. The equations of the line for both silica and al uminum silicate calibration curves were used to predict the carbon content in the soil samples based on their average C247 intensity (Table 5-

PAGE 105

105 2). The accuracy of each curve to predict the carbon content varied greatly, with a range of 5.2 155 %. If the soil that produced an accuracy of 155 % is omitted, the range was reduced to 5.2 64.5 % with averages of 34.9 % for the aluminum s ilicate curve and 23.3 % for the silica curve. Still, one problem with these re sults is that they were obtaine d with advance knowledge of which curve to apply to which soil (based on Figure 5-6) It is still unknown wh at properties of the soil matrix caused such separate and distinct slopes, though particle size can be dismissed as a cause because all soils were extensively ground to a fine powder. Without firm evidence as to why one soil aligned with the silica curve and another w ith the aluminum silicate curve, it would be impossible to know which to apply to an unknown soil. Furthermore, the results of this research indicate that even in the abse nce of particle size e ffects, it would be impossible to produce a universal carbon calibration curve that is applicable to all soil samples. This includes calibration curves produced from simulated soils or from a range of standard soils.

PAGE 106

106 Figure 5-1. Full spectrum of carbon taken on a gra phite pellet with prominent lines and bands identified. 200300400500600700800900 200 400 600 800 1000 1200 1400 1600 Intensity, CountsWavelength, nm200210220230240250260270280290300 200 400 600 800 1000 1200 1400 229.653 247.837 251.153 283.859 Intensity, CountsWavelength, nmC III C IC II C II383384385386387388389390 400 600 800 1000 1200 1400 1600 CN Intensity, CountsWavelength, nmCN 388.388 387.204 386.186 CN490500510520530540550560570580590600 200 400 600 800 1000 1200 1400 1600 1800 2000 Intensity, CountsWavelength, nmC2 Swan516.535

PAGE 107

107 02505007501000125015001750200022502500275030003250 0 2000 4000 6000 8000 10000 12000 C247 Peak AreaDelay time (ns) Figure 5-2. C247 peak area as a f unction of spectrometer delay time. The integration time is fixed at 200 ns.

PAGE 108

108 385386387388389390 600 650 700 750 800 850 900 950 1000 1050 1100 Intensity, CountsWavelength, nm CN bands Figure 5-3. The CN molecule, with a band head at 388 nm, is barely detected in a soil sample.

PAGE 109

109 Figure 5-4. Theoretical curves of growth for carbon in a silicon dioxi de matrix at three different plasma temperatures. 0102030405060708090 0 2000 4000 6000 8000 10000 12000 Integrated C247 Peak % C in Quartz2 Plasma Temperature 8000 K 1200 K 16000K0102030405060708090 0 5 10 15 20 25 30 35 INtegrated C247 Peak (x107) % C in Quartz 8000 K02 04 06 08 0 0 500 1000 1500 2000 2500 C247 Integral (x107) % C in Quartz 12000K0102030405060708090 0 2000 4000 6000 8000 10000 12000 C247 Integral (x106) % C in Quartz 16000K

PAGE 110

110 Figure 5-5. Calibration curves produced from graphite in A) silica a nd B) aluminum silicate (error bars are 1 standard deviation). 024681 0 0 50 100 150 200 250 300 350 400 C247 Intensity% Carbon Graphite in Silica024681 0 0 50 100 150 200 250 300 350 400 C247 Intensity% Carbon Graphite in Aluminum SilicateB. A.

PAGE 111

111 Table 5-1. Carbon content of soil samples determined by CHN analysis Sample Name % C (1st CHN Analysis) % C (2nd CHN Analysis) %C Average (% C) Buffalo 3.271 4.58 1.3 3.9 0.1 River 5.029 5.090 0.06 5.1 0.1 Estuarine 1.624 1.758 0.13 1.7 0.1 Montana 1.730 1.876 0.14 1.8 0.1 A1 1.043 1.157 0.11 1.1 0.1 A4 0.143 0.128 0.01 0.1 0.1 A7 0.093 0.074 0.02 0.1 0.1 B1 0.168 0.186 0.02 0.2 0.1 B4 0.049 0.068 0.02 0.1 0.1 B7 0.056 0.058 0.002 0.1 0.1 C1 1.576 1.524 0.05 1.6 0.1 C4 0.085 0.071 0.01 0.1 0.1 C7 0.034 0.033 0.001 0 0.1 Aerodyne 1.024 1.043 0.02 1.0 0.1 S 7.190 7.06 0.13 7.1 0.1 SE 2.901 2.741 0.16 2.8 0.1 NE 4.091 3.712 0.38 3.9 0.1 Average = 0.1

PAGE 112

112 01234567891011 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 Soil Samples Graphite in Silica Graphite in Aluminum SilicateC247.8 Intensity, Counts% Carbon Figure 5-6. Results of LIBS measurements on fi nely ground soil samples A) With graphite in silica and aluminum silicate calibration curves B) The two separate, distinct slopes formed by the soils (error bars are 1 standard deviation). 01234567 70 80 90 100 110 120 130 140 150 160 170 180 River S SE S+River NE Montana Estuarine Buffalo C247 Intensity, Counts% Carbon0.20.40.60.81.01.21.41.61.8 50 60 70 80 90 100 110 120 130 140 150 160 170 A1 B1 C1 Aero C247 Intensity, Counts% CarbonA. B.

PAGE 113

113 Table 5-2. Predicted carbon conten t of soils using the applicable silica of aluminum silicate calibration curve. Sample Name True Carbon Content (%) Predicted Carbon Content (%) Absolute Difference (%) Relative Difference (%)a From graphite in SiO2 curve A1 1.1 1.8 +0.7 +38 B1 0.2 0.5 +0.3 +64 C1 1.6 2.3 +0.7 +31 Aerodyne 1.0 1.1 +0.1 +5 From graphite in Al2SiO5 curve Average=34 Estuarine 1.7 1.9 +0.2 +10 Buffalo 3.9 3.4 -0.5 -14 Montana 1.8 1.9 +0.1 +7 NE 3.9 2.5 -1.4 -54 SE 2.8 2.2 -0.6 -30 River 5.5 2.2 -3.3 -155 Average=45 a.) Calculated from ((Predicted %C True %C)/ Predicted %C)x100

PAGE 114

114 CHAPTER 6 LIBS MEASUREMENTS OF NATURAL SOILS: IS QUANTIFICATION POSSIBLE? Introduction The com plex nature of soil samples presents numerous difficulties to any analytical technique, but more so to LIBS, which is especia lly sensitive to sample heterogeneity. In order to assess the potential of LIBS for in-situ soil measurements, the previously described sampling regime must be applied to natural soils that have not be extensively modified. To achieve this, 38 soil samples collected from various locations on the Gulf Coast of Florida were obtained from the University of Floridas Soil Science Department The soils had been previously analyzed for carbon content by dry combustion and represented a carbon concentration range of 0-20%. All samples had been completely dried and mildly ground to break up excessive ly large particles and to reduce heterogeneity. Figure 6-1 shows micros cope images of each soil at a magnification of 3.5X and it is clear that the soils are very differe nt with respect to particle size and chemical composition. The darker soils have higher carbon content due to the black/brown contribution of humus, the major contributor to soil organic matter. CHN Analysis The original dry com bustion CHN analysis that was conducted by the Soil Science department was performed on the coarse soil sa mples and as such it was unlikely that the reported carbon concentrations by weight were re presentative of the bulk sample. Accurately knowing the carbon content of these soils was a critical factor, and while dry combustion is relatively accurate (specified as within 4 % of th e true value), it is still a microanalysis technique, and as such is sensitive to sample heterogeneiti es. To study the accuracy and precision of dry combustion for carbon determination, all 38 soils were submitted for multiple CHN analyses over several months. Table 6.1 presents the average, RSD and range of the reported carbon concentrations for each soil sample. The average RSD for all soil samples is 37%. This poor

PAGE 115

115 precision is undoubtedly a combination of the grea t sample heterogeneity and the small mass of sample analyzed. As such, it is unlikely th at dry combustion can yi eld a highly accurate measurement of carbon concentration that is representative of the bul k sample. In the absence of an alternate method, such poor accuracy was tole rated and the average carbon concentration was used for correlation with LIBS signal. Howeve r, as a consequence, there will be some unknown error associated with the stated carbon concentration of each soil. LIBS Measurements of Natural Soils All soil samples were su bjected to LIBS us ing the previously established optimized sampling regime. Because the soils had not been characterized in terms of texture and particle size, the surface density of the so il on tape was determined and a ssumed to be proportional to the average grain size of the soil. Surface density was determined by weighing the microscope slide and the adhesive tape before and after the applic ation of the soil layer. The difference between these two measurements represented the weight of the applied soil sample, which was then divided by the surface area of the tape (approximately 1.9 cm2). The resulting surface densities ranged from 3.7 to 38 mg/cm2. The soils that exhibited the greatest surface densities corresponded to the soils that were composed of the largest particle size s. Since it had been previously established that th e LIBS signal exhibits some de pendence on particles size, these surface densities were used to correct the LIBS C247 signal for soil grain size. LIBS measurements of the soil samples were obtained using the maximum laser energy of approximately 57 mJ at a repetition rate of 1 Hz. This allowed ample time to manually maneuver the slide to fresh sample locations. Forty separate locations were sampled, for a total of 80 laser shots, with two shots used per spot. A gentle air stream was directed over the soil surface to remove liberated particles. Once the data were co llected, the average C247 peak area was separately calculated for the 40 first shots and the 40 second shots. Figure 6-2 shows the

PAGE 116

116 correlation plots of LIBS signal vs. % carbon for both the first and second shot data. Previous studies on soils indicated that the C247 signa l from the second shot on the same spot is proportional to that of the first shot (by a factor of between 1.6 and 2.0) and as such, besides an increase in signal, the correlation plots were not expected to be si gnificantly different. This is supported by Figure 6-2, which clearly shows a ve ry similar trend between first and second shot measurements. One important advantage of using the two shot sampli ng method is the increase in precision that it provided relative to first sh ot measurements. Figure 6-3 is a histogram that compares the range of both first and second shot RSDs associated with the average C247 peak areas for each of the natural soil samples. From this histogram, it can be seen that the majority of second shot soil measurements provide lower RSDs and thus higher precision than the same measurements collected from only the first s hot. This higher precisi on for the second shot measurements is most likely due the homoge nizing action of the fi rst preparatory shot. There are several important feat ures to note about the plots in Figure 6-2. First, there was no evident correlation between LIBS C247 signal and the average % carbon determined from CHN analysis across all different soil samples. Th is was expected based on previous results that established the dependence of LI BS on properties such as matrix particle size and chemical composition. Due to this, it would not be feasible to expect one correlation plot to produce a single calibration curve applicable to all soils, regardless of text ure or chemical composition. Second, the plots showed that there was an upper concentration limit of approximately 6%, after which the C247 signal plateaus. This would indicate that under these measurement conditions LIBS would not be applicable to high carbon soil sa mples. The third point of interest from these plots was the high carbon signal that the soils containing less than 1% carbon produced. There are two possible explanations for this. Unlike the quartz particle size fractions, which also produced a large signal for tr ace carbon, these soil samples contain iron, which has a weak

PAGE 117

117 emission line that possibly interferes with the C2 47 line and may be contributing to the signal. This will be discussed further in the next section. Another potential explanation for such a large blank signal is the same mounting tape contri bution seen in measurements taken on the quartz samples. While microscope images indicate that even after two shots on the soil sample the tape is not visible, it is possible that it is not necessary to actually reach the tape in order for it to cause some interference. As was previously disc ussed, it seems that as the particle size of the sample decreases there is an amplification of th e carbon signal observed from the tape relative to the carbon signal produced from the tape alone. This phenomenon was prevalent for the quartz particle size fractions on the order of tens of microns, and was not seen in the larger fractions which resemble most of these soils. If this is the case, then is has to be assumed that this tape contribution is relatively constant and the contribution is the same for each shot. A number of normalization factors were appl ied to the LIBS soil measurements in an attempt to minimize the matrix effects and to observe any increase in correlation. Some success has been previously reported using silicon as an internal standard.54-56 However, no improvement was found when the C247 signal was normalized with the Si288 line. This was hardly surprising considering a si mple visual inspection of the so ils revealed how variable the silicon content was (the large, translucent particles which can be seen in some soil samples in Figure 6-1 are silica particles). In order for any other soil component to be used as an internal standard, it must be shown that it exhibits a relatively consta nt concentration, or that the concentration is proportional to carbon content. It is likely that any success that has been achieved by normalizing the carbon signal with silicon was performed on soils with very similar chemical compositions (i.e., they were collected from similar locat ions, had been subjected to the same environmental factors).

PAGE 118

118 Normalizing the C247 signal to surface density proved to be the most promising method. Figure 6-4 shows such normalization transforms the raw measurements and produces a much more linear correlation plot. Based on these re sults, it would seem that correcting the LIBS signal for the physical rather than then chemical matrix effects is the most promising method to relate C247 signal to car bon content. This is advantageous because the physical properties of any given soil are easier to assess during in-field analysis than the chemical composition. In order to correct the LIBS signal for chemical ma trix effects, the chemical composition of the soils would have to be determined prior to LIBS measurements. In some cases, this may be as simple as determining high iron content based on color; however, even subtle differences in chemical compositions will contribute to significant chemical matrix effects and reliable analytical techniques would be necessary to clas sify soils into groups of similar composition for LIBS measurements. High-Resolution LIBS Measurements on Soils One potential problem with th e C247 line was the possibility of an interference from the weak iron ionic line at 247.8 nm and atomic line at 247.9 nm. To study the extent of this interference, a pure iron powder was analyzed, as well as a 1:1 mixture of iron and graphite powders (Figure 6-5). The spectru m of iron powder revealed that the iron atomic line is detected at 247.99 nm. A small shoulder on the left of the peak corresponded to the ionic iron line. The spectrum of the graphite/iron mixture revealed that the Fe247.98 line was not resolved from the C247.83 line. While the extent of this interference was expected to be minimal due to the low iron content of these soil samples, it was still necessary to ensure that the poor correlation between the LIBS carbon signal a nd the carbon content was not due to the unresolved iron peak contributing to the carbon signal.

PAGE 119

119 To investigate this, LIBS measurements were performed on the same soil samples from the previous study and emission radiation was spect rally resolved using an Acton SpectraPro-500i spectrometer with a 3600 grooves/mm grating wh ich provided a resolution of approximately 0.02 nm. This resolution was high enough to fully resolve the carbon and iron lines (Figure 6-6). Most other experimental condi tions were maintained, including a laser pulse energy of approximately 60 mJ, spectrometer delay time of 1 s and an integration time of 100 s. Samples were mounted vertically and shifted ma nually after every two sa mpling shots to ensure a fresh sample location. Emission was coll ected by a 2 lens and focused onto the 35 m slit of the spectrometer. The focusing lens produced 1:1 inverted image of the plasma onto the entrance slit of the spectrometer, and alignment was achieved usi ng a flat graphite pellet. However the soil samples used in these studies were far from flat, and the positioning of the laser induced plasma could be horizontally translat ed by up to 1 mm depending on the size of the particle where the plasma was formed. As suc h, this optical configuration was highly sensitive to the coarse textures of soils and emission collection efficien cy was not always optimal or reproducible. The resulting LIBS measurements produced with the high resolution spectrometer are shown in Figure 6-7. It is clear that there is no increase in co rrelation between the LIBS carbon signal and the carbon content; indeed, the trend is very similar to the results achieved using the lower resolution Ocean Optics spectrometer. Norma lizing these results by the surface density of the samples (Figure 6-8) did not significantly impr ove the correlation as it did for the previous data. The C247 peak produced using the high reso lution spectrometer were completely resolved, and therefore wholly due to ca rbon alone. If the poor correl ation observed with the Ocean Optics spectrometer were due to the iron spectral inte rferences, the fully resolved carbon peak would significantly improve the correlation. Th is does not occur, a nd therefore it can be

PAGE 120

120 concluded that the lack of co rrelation between LIBS signal and carbon observed with the Ocean Optics measurements must be attributed to ot her matrix factors and is not caused by an iron interference. One other important aspect to note concerning these results is the poor precision. As previously mentioned, this was likely due to the optical arrangement of the emission collection. For many soils, the RSD was double that achieved using fiber optic emission collection. It is likely that the precision can be increased by fu rther grinding the soils to fine powders or analyzing pressed pellets; the flatter surfaces will minimize the plasma translation with respect to the spectrometer entrance slit. However, su ch additional samples preparation steps are cumbersome, and fiber optics is likely the best method of emission collection when analyzing natural soils of varying textures.

PAGE 121

121 Figure 6-1. Microscope images of na tural soil samples (3.5X magnified). Mounting Tape 2-10 2-10 2-9 2-9 2-6 Layer 2-7 2-8 Layer 2-9 2-9 2-10 2-9 2-7 La y er 2-8 Layer 2-8 La y er 2-8 Layer 2-6 La y er 2-6 La y er 2-6 2-5 2-5 2-5 2-4 2-4 2-4 2-4 2-3 2-3 2-3 2-3 2-3 2-2 2-2 2-2 2-2 2-2 2-1 2-1 2-1 2-1

PAGE 122

122 Table 6-1. Results of dry combustion CHN analysis performed on a suite of natural soil samples. Soil Label Number of CHN analyses Average carbon Concentration (%) RSD of CHN analyses (%) 1 3 7.1 16 2 6 12.1 13 3 3 9.2 13 4 6 6.0 13 5 3 9.2 43 6 6 8.2 10 7 3 0.3 26 8 3 19.4 6 11 6 2.0 18 12 3 1.2 7 13 3 6.9 56 14 3 0.8 30 15 3 2.6 33 16 3 0.4 9 17 3 0.5 54 18 3 5.7 17 19 3 0.4 51 20 5 0.0 66 21 3 9.5 64 22 3 6.8 62 23 6 0.9 26 24 3 0.5 5.0 25 6 5.2 44 26 3 0.7 43 27 3 0.6 20 28 6 15.6 20 29 3 14.1 18 30 3 9.5 16 31 3 0.7 26 32 3 1.7 68 33 3 0.5 14 34 3 11.6 62 35 3 8.0 58 36 3 8.5 49 37 3 11.4 60 38 3 12.5 58

PAGE 123

123 Figure 6-2. LIBS C247 measurements on natural soils as a function of % carbon for A) first shot measurements and B) second shot measur ements. The error bars for % carbon represent 1 standard deviation of the average CHN measurements. A. B. 0246810121416182022 20 30 40 50 60 70 C247 Peak Area% Carbon0246810121416182022 40 50 60 70 80 C247 Peak Area% Carbon

PAGE 124

124 Figure 6-3. A histogram of the RS Ds from LIBS C247 measurements using the first shot and the second shot. 51 01 52 02 5 0 1 2 3 4 5 6 7 8 FrequencyRSD (%) Second shot First shot

PAGE 125

125 0246810121416182022 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 C247 Peak Area (Normali zed to Surface Density)% Carbon Figure 6-4. Normalizing the C247 peak area by so il surface density incr eases the correlation between LIBS signal and carbon concentration (error bars ar e 1 standard deviation).

PAGE 126

126 Figure 6-5. Extent of the iron spec tral interference. A) Pure iron foil. B) Pure graphite and C) An iron and graphite mixture. 247.0247.5248.0248.5249.0249.5250.0 400 600 800 1000 1200 1400 1600 1800 2000 247.26515 247.9928 248.30439 249.29001 249.85 9 Intensity, CountsWavelength, nm 247.836 Fe Fe Fe247.0 247.5 248.0 248.5 249.0 0 200 400 600 800 1000 1200 1400 1600 247.83695Intensity, CountsWavelength, nm C247.0247.5248.0248.5249.0249.5250.0 0 200 400 600 800 1000 1200 1400 247.26515 247.83695 249.29001 249.85 9 Intensity, CountsWavelength, nm 248.30 C FeA. B. C.

PAGE 127

127 247.6247.7247.8247.9248.0248.1248.2 10000 15000 20000 25000 30000 35000 247.68 247.811 247.978 248.24Intensity, CountsWavelength, nm Carbon Iron Figure 6-6. A soil spectrum produced using th e Acton SpectraPro spectrometer showing the resolved carbon and iron lines.

PAGE 128

128 0246810121416182022 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000 C247 Peak Area% Carbon Figure 6-7. The LIBS C247 peak area as a f unction of carbon concentration for measurements produced using a high resolution spectrometer (error bars are 1 standard deviation).

PAGE 129

129 0 5 10 15 20 0 100000 200000 300000 400000 500000 600000 700000 800000 C247 Peak Area (Norm. to surface density)% Carbon Figure 6-8. A plot of the C247 peak area norm alized by surface density vs. carbon content for measurements produced using a high reso lution spectrometer (error bars are 1 standard deviation).

PAGE 130

130 CHAPTER 7 CONCLUSION AND FUTURE WORK Natural so ils are arguably the most challenging of all samples for quantitative LIBS measurements. There are many variables that mu st be accounted for that each contributes to detrimental chemical and physical matrix effects. Chemical matrix effects can include spectral interferences from other soil com ponents, speciation of the analyte of interest, water content and the overall chemical composition, which may alter the samples breakdown threshold. Physical matrix effects include the broad range of different particle sizes, the distribution of the analyte throughout the sample and extensive differences in physical properties such as reflectivity and heat of vaporization on both the macro and micro s cale. The initial purpose of this research was to assess the potential of LIBS for quantitative carbon measurements in soil; however, it became readily apparent that a more fundamental ap proach was necessary before any conclusion concerning quantitative results co uld be achieved. As such, the majority of this work was devoted to establishing optimal instrumental and sampling parameters exclusively for LIBS measurements on soils. While the ultimate goal of this research was for carbon determination in soils, most of these studies can be extended to any other an alyte of interest. A number of modifications were made to the LIBS instrumental components for the purpose of soil analysis, including the use of a positioning laser to ensure constant LTSD and FTSD for soils of variable heights. A gentle air stream directed over the sample was found to eliminate the interaction of the laser beam with libera ted, air-borne particles along the laser axis, thus increasing measurement reproducibility. Careful alignment of the fiber-optic probe, positioned close to the plasma, was found to provid e superior emission coll ection relative to the popular back-side emission collection. A spectrometer delay time of 1 s was found to produce the best S/B for the C247 line, as well as the lowest limit of detection for carbon in soil.

PAGE 131

131 Establishing an optimized sampling regime for soils is much more difficult than for most solid samples encountered in LIBS. Due to the highly heterogeneous physical and chemical nature of all soils, combined with the mi cro-sampling property of LIBS, single shot measurements were not representa tive of the bulk sample. As such, it was found that a minimum of 20 sample spots must be measured to achieve the minimum, if not always acceptable, RSD. In reality, it might take thousands of sampling shots to produce a precise measurement that is representative of the bulk sample. The particulate nature of the soil samples also greatly affects the LIBS measurements. It was confirmed that the LIBS signal was greatly dependent on the matrix particle size and that the increased homogeneity imparted by samples composed of smaller particles yielded much higher measurement pr ecision. It was also determined that when a single, preparatory laser shot was used to mo dify the sample surface, the subsequent second laser shot produced increased emission intensity across the entire spectrum. Microscopic images of the soil samples before and after the sampli ng indicated that surface modification resulted from the removal or break-up of large particle s which most likely allowed more sample to interact with the second laser pulse, thus increasing the mass ablated. Under these sampling conditions, the precision of the C247 signal fo r all 38 soil sample did not exceed 18%. The optimized sampling regime was applied to natural, dried soil samples to assess the potential of LIBS as a quantit ative technique. However, the results revealed no correlation between the uncorrected carbon p eak area of all soil samples and their reported carbon concentrations. There are many factors that must be considered before quantification can be ruled out. For instance, all of these soils suffere d from very different matrix effects and therefore could not be all be compared together or us ed to produce one single calibration curve. In

PAGE 132

132 addition to the different matrix effects, other factors independen t of the LIBS measurements may have contributed to such poor correlation. For instance, the significant variability in the CHN analyses introduced significant er ror to the carbon concentrations, which in turn will influence the correlation. In order to minimize this e rror, for future work, soil samples should be extensively ground prior to dry combustion and anal yzed multiple times. Another consideration is whether the mounting tape contributes to the LIBS carbon signal of soil samples, and whether this contribution is relatively constant. Other sample mounti ng methods that do not involve a carbon based substrate should be explored. Fo r instance, future work could produce LIBS measurements on loose soil samples. Correlation is likely to incr ease when soils of very similar chemical and physical compositions are analyzed together rather than as suite of ve ry different soils. Additional research must focus on identifying the chemical or physical properties that contributed most to the LIBS matrix effects. For instance, this re search exploited the particle size dependence of LIBS signal and found an increase in correla tion by normalizing the carbon signal to surface density, which is related to the particle size. In addition to particle size other variables may influence the LIBS signal and could be used to ca tegorize soils or to correct the LIBS signal. Examples include the chemical form of carbon, chemical composition, and mineral content. The use of these soil properties would require extensive prior knowledge about each soil sample and collaboration with soil scient ists, who often work with fully characterized soils, would be a valuable resource. Future work should also consider the use of a multivariate analysis of the entire soil spectrum. In the absence of prior knowledge about a specific soil, the LIBS spectrum can provide valuable information conc erning the relative concentrati ons of elemental components.

PAGE 133

133 At the very least, inspection of the entire spectrum can be used to group soils based on major constituents. More advanced multivariate algorithms may be applied to entire spectral regions instead of simply one carbon line and will account for varying peak intensities of other elemental constituents which may be producing specific matrix effects. The quantification of carbon in natural soils by LIBS was not achieved during this research. While it is possible that this will be accomplished in future work, it is important to reflect on the original motivation of this research: to assess LIBS as a means of in-situ carbon determination in natural soils. During this research, every attempt was made to minimize sample preparation and reduce cumbersome in strumentation in order to mimic the LIBS measurements that would be produced using a po rtable LIBS system. Future work may prove that quantification is achievable on finely ground, pelletized samples and precision drastically increased by taking thousands of measurements. However, once the sample preparation negates in-situ analysis and requires extensive laboratory ba sed work, the advantages of LIBS become redundant and more conventional techniques such as dry com bustion or wet oxidation may provide more accurate results. The inability of LIBS to determine carbon in soils at this point does not diminish the importance of the sampling and instrumental stud ies that were performed using for the purpose of soil measurements. For instance, studies c oncerning the dependence of LIBS signal on matrix particle size are applicable to nearly all types of particulate samples. In addition to this, qualitative analysis of soils is successfully achieved using the soil sampling characteristics studied here. Indeed, the increa se in emission intensity that is observed after a single laser preparatory shot will increase LIBS sensitivity for trace toxic metals such as Hg, As or Pb. A variety of important elements in soil, including Mg, Mn, Na, Fe, Cu can be readily detected in

PAGE 134

134 the LIBS soil spectra produced during this research. In some cases, quantification of these elements may be more easily achieved over carbon. As such, this research has provided a firm foundation for further studies on the characteristic s of LIBS on soils and may be extended to a variety of other analytes of interest

PAGE 135

135 LIST OF REFERENCES (1) Brech, F.; Cross, L. Appl. Spectrosc. 1962, 16, 59. (2) Rusak, D.A.; Castle, B.C.; Sm ith, B.W.; Winefordner, J.D. Trends Anal. Chem. 1998, 17, 453-461. (3) SciFinder, version 2006; Chemical Abstra cts Service: Columbus, OH, 2006; Accessed June 3, 2006); calculated using ACD/Labs software, version 8.14; ACD/Labs 1994-2006. (4) Laser-Induced Breckdown Spectroscopy (L IBS): Fundamentals and Applications; Miziolek, A. W.; Palleschi, V.; Schechter, I., Ed.; Cambridge Un iversity Press: New York, 2006. (5) Weyl, G.M.; Root, R.G.; Baldis, H.A, Hauser, A.A. In Laser-Induced Plasmas and Applications; Radziemski, L.J.; Cremers, D.A., Eds. ; Marcel Dekker, Inc.: New York and Basel, 1989. (6) Lee, Yong-Ill, Song, Kyuseok, Sneddon, Joseph. Laser Induced Breakdown Spectroscopy ; Nova Science Publishing, Inc.: New York, 2000. (7) Sneddon, J.; Lee, Y.; Curr. Top. Anal. Chem. 2004, 4, 111-117. (8) Winefordner, J. D.; Gornushkin, I. B.; Correll, T.; Gibb, E.; Smith, B. W.; Omenetto, N. J. Anal. At. Spectrom. 2003, 19, 1061-1083. (9) Radziemski, L. J. Spectrochim Acta, Part B. 2002, 57B 1109-1113. (10) Lee, Y.; Sneddon, J Adv. At. Spectrosc. 2002, 5, 235-288. (11) Radziemski, L.J.; Loree, T. R.; Cremers, D. A. Springer Series in Optical Sciences 1983, 39, 303-307. (12) Gunratne, T.; Kangas, M.; Singh, S.; Gross, A.; Dantus, M.; Chem. Phys. Lett. 2006, 423, 197-201. (13) Le Drogoff, B.; Margot, J.; Vidal, F.; Lav ille, S.; Chaker, M.; Sabsabi, M.; Johnston, T. W.; Barthelemy, O. Plasma Sources Sci. Technol. 2004, 13, 223-230. (14) Le Drogoff, B.; Chaker, M.; Margot, J.; Sabsabi, M.; Barthelemy, O.; Johnston, T. W.; Laville, S.; Vidal, F. App. Spectrosc. 2004, 58, 122-129. (15) Angel, S. M.; Stratis, D. N.; Eland, K. L.; Lai, T.; Berg, M. A.; Gold, D. M. Fresenius' J. Anal. Chem. 2001, 369, 320-327. (16) Lenk, A.; Witke, T.; Granse, G. Appl. Surf. Sci. 1996, 195, 96-98.

PAGE 136

136 (17) Song, K.; Lee, Y.; Sneddon, J. Appl. Spectrosc. Rev. 2002, 37, 89-117. (18) Yamamoto, K.Y.; Cremers, D.A.; Foster, L.E.; Ferris, M.J.; Appl. Spectrosc. 1996, 50, 222-233. (19) Carranza, J.E.; Gibb, E.; Smith, B.W.; Hahn, D.W, Winefordner, J.D. Appl. Opt., 2003, 42, 6016-6021. (20) Burgio, L.; Melessanaka, K.; Doul gendis, M.; Clark, R.J.H.; Anglos, D.; Spectrochim. Acta 2001, 905 B56. (21) Angelos, D. Appl. Spectrosc. 2001, 55, 1864. (22) Angelos, D.; Balas, C.; Fotakis, C. Am. Lab. 1999 31, 60-67. (23) Giakoumaki, A.; Melessanaki, K.; Anglos, D. Anal. Bioanal. Chem 2007 387, 749760. (24) Lopez-Moreno, C.; Palanco, S.; Laserna, J.J.; Delucia, F.; Mikiolek, A.W.; Rose, J.; Walters, R.A.; Whitehouse, A.L.;J Anal. At. Spectrosc. 2006 21, 55-60. (25) Castle, B.C.; Talabardon, K.; Smith, B.W.; Winefordner, J.D. Appl. Spectrosc. 1998, 52, 649-657. (26) Chaleard, C.; Mauhien, P.; Andre, N. ; Uebbing, J.; Lacour, J.L.; Geertsen, C.; J. Anal. At. Spectrom. 1997, 12, 183-188. (27) Panne, U.; Haisch, C.; Clara, M.; Niessner, R.; Spectrochim. Acta, Part B. 1998, 53B 1969-1981. (28) Gondal, M. A.; Hussain, T. Talanta 2007, 72, 642-649. (29) Rai, V.N.; Yueh, F.Y.; Singh, J.P. Appl. Opt 2003, 42, 2094-2101. (30) Knopp, R.; Scherbaum, F.J.; Kim, J.I. Fres. J. Anal. Chem. 1996, 355, 16-20. (31) Pichahchy, A.E.; Cremers, D.A. Spectrochim. Acta 1997, 25-39. (32) Caceres, J.O.; Tornero Lopez, J.; Telle, H.H.; Gonzalez Urena, A. Spectrochim. Acta B 2001, 56, 831-838. (33) Neuhauser, R.E.; Panne, U.; Niessner, R. Anal. Chim. Acta 1999, 392, 47-54. (34) Carranza, J.E.; Fisher, B.T.; Yoder, G.D.; Hahn, D.A. Spectrochim. Acta B 2001, 56, 851-864. (35) Panne, U.; Neuhauser, R.E.; Th eisen, M.; Fink, H.; Niessner, R.; Spectrochim. Acta B. 2001, 56, 839-850.

PAGE 137

137 (36) Radzeimski, L.J.; Loree, T. R.; Cremers, D.A.; Hoffman, N.M. Anal. Chem. 1987, 997981. (37) Kaiser, J.; Samek, O.; Reale, L.; Liska, M. ; Malina, R.; Ritucci, A.; Poma, A.; Tucci, A.; Flora, F.; Lai, A.; Mancini, L.; Tromba, G. ; Zanini, F.; Faenov, A.; Pikuz, T.; Cinque, G. Microsc. Res.Tech 2007, 70, 147-153. (38) Samek, O.; Lambert, J.; Hergenroeder, R.; Liska, M.; Kaiser, J.; Novotny, K.; Kukhlevsky, S. Laser Phys. Lett. 2006, 3, 21-25. (39) Sun, Q.; Tran, M.; Smith, B. W.; Winefordner, J. D. Can. J. Anal. Sci. Spectros. 1999, 44, 164-170. (40) Morel, S.; Leone, N.; Adam, P.; Amouroux, J. Appl. Opt 2003, 42, 6184-6191. (41) Haisch, C.; Liermann, J.; Panne, U.; Niessner, R.; Anal. Chem. Acta 1997, 346, 23. (42) Marquardt, B.J.; Goode, S.R.; Angel, S.M. Anal. Chem. 1996 68, 977-981. (43) Kraushaar, M.; Noll, R.; Schmitz, H.-U. Appl. Spectrosc. 2003, 57, 1282-1287. (44) Eppler, A.S.; Cremers, D. A.; Hickmott, D.; Ferris, M.; Koskelo, A. Appl. Spectrosc. 1998, 50, 1175-1181. (45) Barbini, R.; Colao, F.; Fantoni, R.; Palucci, A.; Capitelli, F. Appl. Phys. A 1999, 69, S175 S178. (46) Bublitz, J.; Dolle, C.; Schade, W.; Hartmann, A.;Horn R. Euro. J. Soil Sci 2001, 52 505312. (47) Capitelli, F.; Colao, F.; Provenzano, M. R; Fantoni, R.; Brunetti, G.; Senesi, N. Gerderma 2002, 106, 42-65. (48) Gornushkin, I.B.; Kim, J.E.; Smith, B.W.; Baker, S.A.; Winefordner, J.D. Appl. Spectrosc. 1997, 51, 1055-1059. (49) Hilbk-Kortenbruck, F.; Noll, R.; Wintjens, P.; Falk, H.; Becker, C. Spectrochim. Acta: Part B 2001, 56, 933-945. (50) Wisbrun, R.; Schechter, I.; Nie ssner, N.; Schroder, H.; Kompa, K. Anal. Chem. 1994, 66, 2964-2975. (51) Therlault, G.; Bodensteiner, S.; Lieberman, S. Field Anal. Chem. Tech 1998, 2, 117-125. (52) Knight, A.; Scherbarth, N.; Cremers, D.; Ferris, M. Appl. Spectrosc. 2000 54, 331340. (53) Arp, Z.; Cremers, D.; Wiens, R. ;Wayne, D.; Salle, B.; Maurice, S. Appl. Spectrosc. 2004, 58, 897 909.

PAGE 138

138 (54) Cremers, D.; Ebinger, M. ; Breshears, D.; Unkefer, P.; Kammerdiener, S.; Ferris, M.; Catlett, K.; Brown, J. J. Environ. Qual. 2001, 30, 2202 2206. (55) Ebinger, M.H.; Breshears, D.D.; Cremers, D.A.; Unkefer, P.J. Earth and Environmental Sciences, Progress Report, 1998 2000 (56) Ebinger, M.; Lee Norfleet, M.; Breshears, D.; Cremers, D.; Unkefer, P.; Kammerdiener, S.; Ferris, M.; Unkefer, P.; Lamb, M.; Goddard, K.; Meyer, C.; Am. J. Soil Sc 2003, 67, 1616 1619. (57) Martin, M.; Wullschleger, S.; Garten, C.; Palumbo, A. Appl. Opt 2003, 42, 2072 2077. (58) Harris, R.; Martin, M.; Labbe, N.; Ebinger, M.; Clegg, S.; Wullschlege r, S. Advances in Calibrating Laser-Induced Br eakdown Spectroscopy (LIBS) for Measuring Total Soil Carbon. In Steps Toward Deployment Proceedings of the 5th Carbon Capture & Sequestration Conference, Alexa ndria, Virginia, May 8-11, 2006. (59) Freedman, A.; Wormhoudt, J.C. Portable Soil Carbon Monitor, Aerodyne Research, Inc., Billerica, MA, 2006. (60) Bruce, J.; Frome, M.; Haites, E.; Janzen, H. ; Lal, R.; Paustian, K. Carbon sequestration in soils. In Carbon Sequestration in Soils Workshop, Proceedings of the Soil and Water Conservation Society Meeting, Ca lgary, Alberta, May 21-22, 1998. (61) Sulzman, E.W. The Carbon Cycle Global Change Instruction Program, 2000, University Corporation for Atmospheri c Research: Boulder, CO. (70) Assessment Methods for Soil Carbon ; Lal, R.; Kimble, J.M.; Follett, R.F.; Stewart, B.A., Eds.; Advances in Soil Sciences; Lewis Publishers: Boca Raton, FL, 1998. (71) Izzaurralde, R.C.; McGill, W.B.; Bryden, A.; Graham, S.; Ward, M.; Dickey, P. In Management of Carbon Sequestration in Soil. Advances in Soil Science; Lal, R Kimble, J.; Follett, J.; Stewart, R. Eds.; Lewis Publishers: Boca Raton, FL, 1998; pp 433-446. (72) Charles, J.M.; Simmons, M.S. Analyst 1986, 111, 385-390. (73) Schollenberger, C.J. Soil Sci. 1927, 24, 65-68. (74) Tan, K.H. Soil Sampling, Preparation, and Analysis ; Marcel Dekker, Inc.: New York, 1996. (75) Walkley, A.L.; Black, A. Soil Sci. 1947, 63, 251-263. (76) Viscarra Rosel, R.A.; Walvoort, D.J.J.; McBratney,A.B.; Janik,L.J.; Skjemstad, J.O. Geoderma 2006, 131, 59-75 (77) Reeves, J.B.; Follett,R.F.; McCarty,G.W.; Kimble, J.M. Comm. Soil Sci. Plant Anal. 2006, 37, 2307-2325.

PAGE 139

139 (78) Wight, J.P.; Allen, F.L. ; Zanetti,M.Z.; Rials,T.G. The Science of Conservation Tillage: Continuing the Discoveries, Proceedings at the 2005 Southern Conservation Tillage Systems Conference, Clemson Univ ersity, Florence, SC, May 27-29, 2005. (79) Grieve, I.C.; Davidson, D.A.; Ostle, N.J.; Bruneau, P.M.C.; Fallik, A.E. Soil Biol. Biochem. 2006, 38, 229-234. (80) Jian-Bing, W.; Du-Ning, X.; XingYi, Z.; Xiu-Zhen, L.; Xiao-Yu, L. Environ. Monit. Assess. 2006 121, 597-613. (81) Pye, K.; Blott, S.J.; Croft, D.J.; Carter, J.F. For. Sci. Int. 2006, 163, 59-80. (82) Conklin, A.R. In Introduction to Soil Chemistry ; Winefordner, J.W.; Ed.; Chemical Analysis: A Series of Monographs on Analytical Chemistry and its Applications; WileyInterscience: Hoboken, NJ, 2005; Vol. 2. (83) 3M, Foam Mounting Tape MSDS, http: //multimedia.mmm.com/mws/mediawebserver. accessed July 8, 2007. (84) Portnov, A.; Rosenwaks, S.;Bar, I. Appl. Opt 2003, 42, 2835 2842. (85) Wormhoudt, J.; Iannarilli, F.J.; Johnes, S.; Annen, K.D.; Freedman, A. Appl. Spectrosc. 2005, 59, 1098-1102 (86) Tran, M.; Sun, Q.; Smith, B.; Winefordner, J.D.; J. Anal. At. Spectrom 2001, 16, 628632. (87) Acquaviva, S.; De Giorgi, M.L. J. Phys. B 2002, 35, 795-806. (88) Harilal, S.S.; Bindu, C.V.; Issac, R.; Nampoori, V.P.N.; Vallabhan, C.P.G. J. Appl. Phys. 1997, 82, 2140-2146. (89) Aguilera, J.A.; Aragon, C.; Campos, J. Appl. Spectrosc. 1992 46, 1382 1387. (90) St-Onge, L.; Sing, R.; Bechard, S.; Sabsasi, M. Appl. Phys. A 1999, 69, S913 S916. (91) Locke, R.J.; Morris, J.B.; Forch, B.E.; Miziolek, A.W.; Appl. Opt. 1990, 33, 4987-4992. (92) Vivien, C.; Hermann, J.; Perrone, A.; Boulmer-Leborgne, C.; Luches, A.; J. Phys D 1998, 31, 1263-1272. (93) Ralchenko, Yu., Jou, F.-C., Kelleher, D.E., Kramida, A.E., Musgrove, A., Reader, J., Wiese, W.L., and Olsen, K. (2007). NIST Atomic Spectra Database (version 3.1.2), [Online]. Available: http: //physics.nist.gov/asd3 [ 2007, July 12]. National Institute of Standards and Technology, Gaithersburg, MD. (94) Gornushkin, I.B.; Anzano, J.M.; King, L. A.; Smith, B.W.; Omenetto, N.; Winefordner, J.D. Spectrochim. Acta. B 1999, 54, 491-503.

PAGE 140

140 BIOGRAPHICAL SKETCH Lydia Edwards was born in Shrewsbury, England, in 1981 and is the oldest of four siblings. In 1994 she m oved to Woodbridge, Virgin ia, with her family and attended C.D. Hylton High School from which she graduated summa cum laude in 1999. After attending Northern Virginia Community College and working full -time for one year, she moved to Blacksburg, Virginia, to attend Virginia Tech, where she stud ied chemistry. Lydia graduated from Virginia Tech in May 2003 with a Bachelors of Science in Chemistry. In August 2003 she enrolled in the chemistry graduate program at the University of Florida in Gainesville, Florida and joined the research group of Dr. James D. Winefordner in the analytical ch emistry department. In addition to performing graduate research, she also took classes in Forensic Drug Chemistry and Forensic Toxicology. In August 2007 Lydia completed her doctoral research as well as a Master of Science in Forensic Drug Chemistry. In 2008 sh e will receive a Master of Science in Forensic Toxicology from the UF Department of Veterinary Sciences. Lydia has been married to Daniel Breckenridge since 2006. She has accepted a po sition in atomic spectroscopy with BristolMyers Squibb in New Brunswick, New Jersey.