Laser excited fluorescence and ionization for flame diagnostics

MISSING IMAGE

Material Information

Title:
Laser excited fluorescence and ionization for flame diagnostics
Physical Description:
x, 114 leaves : ill. ; 28 cm.
Language:
English
Creator:
Rutledge, Michael James, 1960-
Publication Date:

Subjects

Subjects / Keywords:
Laser spectroscopy   ( lcsh )
Fluorescence -- Data processing   ( lcsh )
Flame spectroscopy   ( lcsh )
Chemistry thesis Ph. D
Dissertations, Academic -- Chemistry -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1987.
Bibliography:
Bibliography: leaves 110-113.
Statement of Responsibility:
by Michael James Rutledge.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000947016
oclc - 16904871
notis - AEQ8996
sobekcm - AA00004847_00001
System ID:
AA00004847:00001

Full Text


















LASER EXCITED FLUORESCEtJCE AND IONIZATION
FOR FLAME DIAGNOSTICS







BY







MICHAEL JAMES RUTLEDGE


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


UtI'JVERSITY OF FLORIDA


1987




























To my parents and family,

whose unending support made all this possible.

















ACK J jOWLE DME NTS


I gratefully acknowledge the support of Dr. Benjamin Smith,

Dr. Edward Voigtman, and Dr. Nicolo Omenetto whose knowledge,

enthusiasm, and helpful suggestions were a continuing source of

inspiration.

I thank Dr. James D. Winefordner for the privilege of working

with him for the past four years and with the best spectroscopy group

in the world. His concern and encouragement during my four years

have made my work more pleasurable and an invaluable learning

experience.

My sincerest thanks go to Jeanne Karably and the secretaries of

J.D. Winefordner. They have been my constant friends for the past 4

years.

I also thank the members of other research groups in the

analytical department especially Ken Matuszak, Paul McCaslin, and

David Berberich whose help with my golf game may prove invaluable.
















TABLE OF CONTENTS


Page

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

LIST OF TABLES...................................................... vi

LIST OF FIGURES..................................................vii

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

CHAPTERS

1 INTRODUCTION.............................................. 1

2 MEASUREMENT APPROACHES FOR ATOMIC FLUORESCENCE AND
IONIZATION SPECTROSCOPIES................................ 4

Introduction to Laser Atomic Fluorescence Spectroscopy
and Laser Enhanced Ionization Spectroscopy Systems......4
Measurement Approaches and Instrumentation................10
Results and Discussion...................................17
Conclusions.............................................. 25

3 ESTIMATION OF ABSOLUTE NUMBER DENSITIES..................27

General Curve of Growth Introduction and Evaluation......27
Calculated Curves of Growth.............................33
Experimental Verification of Curves of Growth.............43
Saturation and Collisional Effects........................46
Conclusions...............................................62

4 SPATIAL DISTRIBUTIONS OF ATOMS IN INHOMOGENEOUS
FLAMES....................................................63

Concentration Modulated Absorption Spectroscopy..........63
Two-Wavelength Laser Enhanced Ionization and
Fluorescence: Spatial Distributions ...................76
Experimental Setup and Discussion........................78
Conclusions...............................................89

5 FINAL COMMENTS AND FURTHER STUDIES....................... 91











APPENDICES

A GLOSSARY OF TERMS AND SYMBOLS............................ 94

B COMPUTER PROGRAM LISTING FOR CALCULATION OF
FLUORESCENCE CURVES OF GROWTH INCLUDING SATURATION,
COLLISIONAL AND PREFILTER AND POSTFILTER EFFECTS.........96

REFERENCES........................................................ 109

BIOGRAPHICAL SKETCH...............................................113

















LIST OF TABLES


Table Page

2-1 Limits of Detection (ng/mL)............................... 18

3-1 Broadening Effects on the a-Parameter for Self-
Broadening............................................. .50

4-1 Ionization Processes, Optical Arrangements, Signals
and Noises for Two-Wavelength Laser Enhanced
Ionization Spectroscopy..................................85
















LIST OF FIGURES


Figure Page

2-1 Lasing Diagram for Copper Vapor Laser......................5

2-2 Experimental Setup........................................8

2-3 Boxcar Plus Lock-In Amplifier Signal Processing
Layout.................................................... 11

2-4 Boxcar-Active Baseline Subtraction Signal Processing
Layout................................................. 14

2-5 Bandwidth Limitation and Conventional Signal
Processing Layout........................................16

2-6 Bandwidth Limitation Effects..............................21

2-7 Noise Power Spectrum at 588.9 nm..........................24

3-1 Diagram for Right Angle Fluorescence.......................28

3-2 Expression for Right Angle Fluorescence...................29

3-3 Curves of Growth for Line Source Excitation................34

3-4 Curves of Growth for Continuum Source Excitation..........35

3-5 Curves of Growth for Two Pseudocontinuum Sources..........37

3-6 Curves of Growth for Mild, Medium, and Severe
Prefiltering--Line Source, Dark = prefilter region........38

3-7 Curves of Growth for Mild, Medium, and Severe
Prefiltering--Continuum Source, Dark = prefilter region...39

3-8 Curves of Growth for Mild, Medium, and Severe
Postfiltering--Line Source, Dark = prefilter region........41

3-9 Curves of Growth for Mild, Medium, and Severe
Postfiltering--Continuum Source, Dark = prefilter
region................................................. 42











3-10 Calculated and Experimental Curves of Growth for Na.......44

3-11 Calculated and Experimental Curves of Growth for Pb.......46

3-12 Expression for Collisional Broadening.....................48

3-13 Curves of Growth with Added Collisional Broadening.........52

3-14 Curves of Growth for Several Intensities for a
Line Source............................................. 55

3-15 Curves of Growth for Several Intensities for a
Continuum Source........................................ 56

3-16 Curves of Growth for Pb Direct-Line (----) and
Resonance (---) Fluorescence for Several Source
Intensities............................................. 57

3-17 Equations for Curves of Growth Equalities for
Saturating Irradiances...................................58

3-18 Prefilter Removal by a Saturating Source, Resonance
(- ) and Direct-Line (----) .............................60

3-19 Postfilter Effects with Saturation Effects Added,
Resonance (- ) and Direct-Line (----)...................61

4-1 Concentration Modulated Absorption Experimental Setup:
Co-linear Beams......................................... ..67

4-2 Experimental Setup for Spatial Diagnostics for COMAS......69

4-3 Results for Perpendicular and Parallel Burners
for COMAS.............................................. .71

4-4 Design of Surface Burner Used.............................72

4-5 Simple Absorption Results for Perpendicular and
Parallel Burners........................................75

4-6 Experimental Setup for Two-Wavelength LEIS and LAFS.......79

4-7 Spatial Profile Result for LEIS...........................82

4-8 Spatial Profile Result for LAFS...........................88

4-9 Three-Dimensional Spatial Profile for LEIS................90


viii
















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

LASER EXCITED FLUORESCENCE AND IONIZATION
FOR FLAME DIAGNOSTICS

BY

MICHAEL JAMES RUTLEDGE

May, 1987

Chairman: James D. Winefordner
Major Department: Chemistry


Measurement approaches involving electrical signal manipulation

and signal processing are investigated for a copper vapor laser

pumped dye laser system. Correction methods used involve modulation

of a pulsed laser output and subtraction of noise. Bandwidth

limitation of signals to reduce radio-frequency noise is also

discussed. A general purpose computer program for calculation of the

absolute number density of an atomic species present in an analytical

volume is presented. The program is written in FORTRAN with a high-

speed, high-precision approximation of the Voigt profile to

incorporate atomic broadening effects. Comparison to experimentally

measured fluorescence results is included.

A discussion of saturation and collisional broadening effects is

included. Calculated results are found to agree with experimental

results within an order of magnitude in all experimental cases.

Spatial profiling results are presented for two-wavelength excitation











of atomic species with detection via fluorescence and ionization of

the species of interest. A new form of absorption spectroscopy,

termed concentration modulated absorption spectroscopy, is

investigated for spatial profiling applications. These methods are

investigated for detection power, as well as spatial information

obtainable. Comparison to other methods of determining absolute

number densities and spatial information is included.
















CHAPTER 1
INTRODUCTION



Laser excited atomic fluorescence and laser enhanced ionization

spectroscopies have been used for many different types of

applications. The extreme sensitivity of these techniques has been

shown many times.- Some of the best limits of detection are

obtained using these methods. These sensitive techniques are applied

here to the investigation of absolute number densities and spatial

profiles in inhomogeneous flames.

A copper vapor laser is used for all investigations throughout

these studies. This laser has found limited application in

chemistry''5 but has found many applications outside this field.6

This laser system consists of a pulsed laser plus the dye laser(s)

for tunability to the wavelength(s) of interest. This is a pulsed

laser which operates at a 6 kHz repetition rate with an average power

of 25 W. Measurement approaches for atomic fluorescence and

ionization are presented as preliminary investigations using this

laser system in Chapter 2. These measurement approaches rely on

bandwidth limitation and/or removal of correlated background

noises. A more complete discussion of the use of correlation

functions to increase signal-to-noise ratios is given by Doerffel et

al.7 The extension of these methods to measurement of atomic











fluorescence for determination of absolute number densities is

discussed in Chapter 3.

Determination of absolute number densities is accomplished using

a computer program to model the expected relationship between

concentration of a species of interest and fluorescence signal

detected. The relationship between fluorescence intensity and

concentration, expressed in a log-log graph, is called a fluorescence

curve of growth (COG). The point of departure from linearity allows

determination of absolute number densities. Many methods have been

presented for determination of absolute number densities in flames,

plasmas and vapor cells. These methods range from the classical

approaches of the absolute intensity method8 and the integral

absorption method9,10 to some more recently introduced methods such

as laser induced fluorescence saturation spectroscopy'l and anomalous

dispersion.12 Several methods have also been presented which allow

absolute number density evaluation from the absolute magnitude of the

signal detected and the experimental detection efficiency.13 Several

COG methods have been presented which allow determination of absolute

number densities with a minimal knowledge of the atomic and geometric

parameters. Other methods include determination of number

densities from vapor pressure measurements of Na and Pb in laser

excited fluorescence experiments.15,16 Some of the simplest methods

for evaluation of number density rely on the supply of and

atomization efficiency of the analyte of interest. Experimentally,

many of the above mentioned methods are quite difficult or time-

consuming to implement. Many rely on an absolute calibration of the










detection optics and photodetector, while some require an additional

calibrated source. Some methods require a detailed knowledge of the

source characteristics including the source intensity and the

spectral profile (Gaussian or otherwise). A general overview of many

of these methods is given in two excellent works by Alkemade and

coworkers.13,17

Atomic fluorescence with two-wavelength excitation and a

multichannel image detector has been used for spatial imaging of OH

and for flow visualization.1820 No research has been presented

which utilizes two-wavelength excitation of atomic species for

spatial resolution with direct detection using laser atomic

fluorescence (LAFS) or laser enhanced ionization (LEIS)

spectrometry. This topic and some related areas are the focus of

Chapter 4. In a recent study by Turk et al.,21 two-wavelength atomic

spectra were measured with detection via ionization. In that study,

only the wavelengths are scanned and the three-dimensional spectra

are presented with axes of one wavelength ( 1) and the other

wavelength (X2) versus intensity. Here, the spatial distributions of

atoms using two-wavelength LEIS and LAFS and single-wavelength

absorption are investigated. An initial investigation into

absorption spectroscopy with a modulated pulsed laser is included.

The absorption technique, termed concentration modulated

absorption spectroscopy (COMAS), is investigated for increased

sensitivity and spatial profiling applications. The highly sensitive

techniques of LEIS and LAFS are applied to probing small volumes in

inhomogeneous flames as a final study.
















CHAPTER 2
MEASUREMENT APPROACHES FOR
ATOMIC FLUORESCENCE AND IONIZATION SPECTROSCOPIES


Introduction to Laser Atomic Fluorescence Spectroscopy and
Laser Enhanced Ionization Spectroscopy Systems

In these experiments, a 20 W copper vapor laser (Cooper

Lasersonics model 251) and a flowing dye cell were used. This is a

pulsed laser system operating at -6 kHz with moderate peak powers of

-160 kW. Laser action in a cell containing copper vapor was first

reported by Walter et al. in 1966.22 Lasing action is accomplished

via the thermal production of ground state copper atoms and

electrical pumping of the ground state (2S) to the lowest resonance

levels (2P) resulting in spectral output at 510 and 578 nm (see

Figure 2-1). Repetition-rates of between 800 and 10,000 pulses per

second have been obtained in our laboratory. This high repetition

rate is accomplished using a thyratron-switched power supply using
-4000 W of electrical power. This high-power, high-frequency

switching produces a large amount of radio-frequency (RF)

interference which was found to be a major noise source in all

analytical experiments. From oscilloscope measurements and gate-

scanned boxcar average outputs, it was determined that the radio-

frequency has a high degree of pulse-to-pulse correlation. Shielding

of the detection electronics was attempted but was inefficient in the

reduction of RF interference.




















LOWEST RESONANCE
LEVELS
-. A 3d'0 4p





510nnm


ELECTRON
IMPACT
EXCITATION


3/2
5/2


3d94s2 2D
METASTABLE
LEVELS


3d10 4s 2S


Figure 2-1. Lasing Diagram for Copper Vapor Laser


GROUND STATE










Three models were used to correct for and/or reduce background

radio-frequency and other noise sources. Two methods rely on

synchronization of the laser with a high-speed chopper to block

alternate laser pulses. One of these methods involved the use of a

gated integrator and boxcar average equipped with an active baseline

subtraction circuit. Another used a gated integrator and boxcar

average plus lock-in amplifier to accomplish the same background

subtraction. The third method used a bandwidth limited amplifier to

reduce high frequency noise components present in the signal. For

comparison purposes, the conventional method of measuring pulsed

laser signals is also included. Analysis for several elements is

included to show consistency of the results.

Chemicals

Stock solutions of all elements were prepared from analytical

grade LiCI, NaC1, Fe wire, and In203 to give 1000 Ug/mL solutions.

Standard solutions were prepared by serial dilution of the stock

solutions. Laser dyes (Exciton Corporation) used included oxazine

720 (Li), and mixtures of Rhodamine 6G and Kiton Red 620 (In, Na, and

Fe).

Instrumentation

The frequency-doubled dye laser output (Molectron model DLII

pumped by a Cooper Lasersonics model 251 copper vapor laser with an

Inrad Autotracker II frequency doubler) was used to illuminate a

1 cm3 region of the flame. A flat mirror was mounted 5 cm from the

flame to allow for a second pass of the laser. For frequency-doubled

experiments, the laser had a pulse width of -20 ns, a pulse energy of











5-35 UJ, and a spectral bandwidth of 0.02 nm. For experiments

involving the fundamental wavelength of the dye laser, the frequency

doubler was removed from the optical path. The fundamental

wavelength experiments are characterized by a pulse width of 30 ns, a

pulse energy of -100-500 yJ, and a spectral line width of 0.03 nm.

Pulse temporal widths were measured using a gate-scanned boxcar

average with a 2 ns risetime photomultiplier tube and a 2 ns boxcar

gate. Additionally, line widths were measured by slowly scanning the

dye laser across an atomic transition and measuring the fluorescence

of a 1 Ug/mL solution of the element aspirated into the flame. The

flame was produced on a laboratory-constructed, brass capillary

burner23 (1 cm2) mounted on a commercial atomic absorption spray

chamber (Perkin Elmer model 303-0110). An approximately

stoichiometric air-acetylene flame was used for all studies.

Commercial-grade gases were pressure-regulated and flow controlled

using rotometers with needle valves. The experimental setup is shown

in Figure 2-2. Symbols are defined in Appendix A.

The fluorescence was produced in a flame volume of approximately

1 cm3 and detected using a small monochromator (SPEX 1670 Minimate,

f/4.0, 220 mm focal length) and RCA model 1414 photomultiplier

tube. The monochromator optical axis was 900 to the laser beam. No

additional optics were used since the acceptance angle of the

monochromator was filled. With the 1.25 mm slits used, this

monochromator had a 10 nm bandpass. The photocurrent pulse was

stretched slightly by a 1000 2 load resistor and connected directly

to the input of the gated integrator and boxcar average

























DYE LASER


-- REF CHOPPER
------------------------------







CIRCUIT
MND
ULSE
4APER H.V.-------
P I


SMONOCHRO- i
MATOR
I i



DETECTION
ct crTOMClreo I


_- I____Tfi,
I APRE-


Figure 2-2. Experimental Setup


COPPER
VAPOR
LASER


x2

P
S-







I


M I


LLJ-WI nwI'Iv-wj










(Stanford Research Systems model 250). The boxcar was triggered

using a photodiode positioned to receive a portion of the copper

vapor 510 and 578 nm output.

The laser enhanced ionization experiments were performed

similarly with the following exceptions. A 5 cm stainless steel slot

burner (Perkin-Elmer model 0040-0277) was used and served as the

detecting electrode. A -1500 V bias potential was applied to a

water-cooled stainless steel electrode placed -1 cm above the burner

surface. The laser beam was positioned to pass parallel to and

-0.3 cm below the bias electrode. The entire assembly was placed

inside a shielded and grounded housing. This experimental

arrangement is a modification of that presented by Travis et al.24

Ionization currents were capacitively coupled to reduce flame

background leakage currents and then converted to voltages using a

current-to-voltage converter (Princeton Applied Research model

181). Gains of 10 V/A were typically used. This signal was

connected directly into the input of the gated integrator. For the

boxcar plus lock-in amplifier portion of the experiments, the signals

from the boxcar were connected directly into the input of the lock-in

amplifier (Keithley model 840). Signals obtained in all experiments

were integrated using a voltage-to-frequency converter (Analog

Devices model 650) and counted for a 10 s integration period.25 This

method of signal integration was found to give an unambiguous

result. Limits of detection were determined using the IUPAC

convention,26 namely, a signal-to-noise ratio equal to three. Noise

levels were determined as the standard deviation of 16 blank










measurements. A more complete discussion of this and other methods

for calculating limits of detection is covered by Long and

Winefordner. 27



Measurement Approaches and Instrumentation

Gated Integrator and Boxcar Averager Plus
Lock-In Amplifier (BLIA)

This method of background correction is similar to background

correction methods in flame absorption or fluorescence using a

continuous wave (cw) laser or conventional source employing

modulation of the source intensity. For a complete discussion of

background correction methods, the reader is referred to the

excellent works of Alkemade et al.28 and Kirkbright and Sargent.29

In this experiment, a high speed chopper (Photon Technology) was

used free-running at 3 kHz, and the reference output of the chopper

was used to control the laser repetition rate. The reference from

the chopper was connected to a multiply-by-two circuit constructed

from a monostable multivibrator (74LS123) and an OR gate (74LS32).

This circuit was set up to give a single pulse output from both the

rising and falling edges of the chopper reference waveform. This

signal was used to trigger the laser externally at a 6 kHz repetition

rate. Initial attempts to control the high speed chopper from the

laser reference output failed due to the momentum and drift of the

chopper blade. Synchronization of the chopper had to be absolute

since any drift resulted in a complete reversal of the signal

polarity for at least a portion of the integration period. The

electrical signal processing system is shown in Figure 2-3.
























w
i-
0 Z 4-
o Z
o 00 0





00




z 0 0
W.-
2 i.

4 z -l-

0 o=
0 S
0. 0





-J-I







0I

to =


1 O

(0
o 2

UC- (
-,





-" =
Cl












Signals obtained from the fluorescence or the photoionization

were fed directly to the input of the gated integrator. The Stanford

boxcar average has two output signals, a last sample and an average

sample. The average sample is useful for reduction of signal

fluctuation and employs a gated resistor-capacitor low-pass (RC)

filter to average a variable number of samples. The last-sample

output is the most recent signal detected by the boxcar average.

The last-sample output was used in this instance because the signal

varies between signal plus noise on one pulse and noise only on the

following pulse, and the demodulation of this signal was accomplished

in the lock-in amplifier. The last-sample output was connected

directly to the input of the lock-in amplifier. The reference output

of the high speed chopper was connected to the reference channel of

the lock-in amplifier which was triggered at the 3 kHz repetition

rate. Synchronization of the lock-in reference waveform and the

boxcar signal is obtained using the phase adjustment on the reference

channel of the lock-in amplifier. The output from the boxcar, which

consisted of signal plus noise during one half-period and noise only

during the following half-period, was demodulated and background was

subtracted by the phase-sensitive detector in the lock-in

amplifier. Output signal ripple of the phase-sensitive detector was

smoothed using a 1 s time constant output filter. Signals output

from the lock-in were connected directly to the voltage-to-frequency

converter and then counted for a 10 s integration period.











Gated Integrator and Boxcar Averager: Active Baseline
Subtraction (ABS)

This technique was a special method applicable only to the

Stanford Research Systems boxcar used here or similar systems with an

active baseline subtraction method. In this method, the same signals

were connected to the boxcar, but the boxcar was placed in the active

baseline subtraction mode (ABS). This mode caused the electronics in

the boxcar to switch the polarity of alternating samples before

adding each one to a moving average. The moving average circuit was

implemented using a gated RC filter30 with varying resistors and

capacitors depending on the number of samples to be averaged. This

method of correction is significantly different from the true phase-

sensitive detection methods employed in lock-in amplifiers. The

moving average switch was set at 3000 samples to keep an effective

1 s time constant, consistent with the lock-in amplifier case

(BLIA). Blocking alternating laser pulses resulted in a subtraction

of noise produced during each laser pulse. Signal output from the

boxcar average was connected directly to a frequency-to-voltage

converter and counted for a 10 s integration period. This is shown

diagrammatically in Figure 2-4.

Bandwidth Limited Signal Processing (BLSP)

Another method of diminishing the effect of high frequency noise

was investigated using a form of low-pass filtering. This "low-pass

filtering" was implemented using a bandwidth-limited amplifier.

Fluorescence photocurrents (-25 ns) were stretched and converted to

voltages using a 1 kn resistor giving a signal of approximately



























ACTIVE BASELINE SUBTRACTION METHOD


3kHz FL. SIGNAL.
3 kHz BKG 180 BOXCAR --- V/F + IO s
6 kHzREF------ COUNTER
6 kHz %
REF.
GATED INTEGRATOR POLARITY RC MOVING
+ A SWITCHING AVERAGE
BOXCAR AVERAGE AMPLIFIER j FILTER


Figure 2-4. Boxcar-Active Baseline Subtraction Signal Processing
Layout











100 ns full width at half maximum (FWHM), i.e., a bandwidth of

-10 MHz. Stretching occurred due to the RC time constant of the

photomultipler tube and connecting cables, due to the load resistor

and the stray capacitance in the system. The fluorescence signals

obtained were input to the bandwidth-limited amplifier and then

connected directly to the gated integrator and boxcar average. The

amplifier used, an Evans model 4131, has a 3 dB bandwidth of 10 MHz

at a gain of 10 times. Signals were stretched to approximately

120 ns (FWHM) by the Evans amplifier. The decreased response of the

amplifier in the high frequency region resulted in a significant

reduction of high frequency noise components. Actual low-pass

filtering would be difficult to implement since a time constant on

the order of 10 ns would be necessary to avoid significant signal

distortion. Stray capacitance and resistance make this low time

constant difficult to obtain. Signals from the Evans amplifier were

connected directly to the gated integrator and boxcar average.

Signals from the boxcar were connected to a voltage-to-frequency

converter and counted for a 10 s integration period as in all other

cases. The electrical system is shown in Figure 2-5 along with the

conventional method for measuring signals.

Conventional Method (CONV)

Signals obtained from fluorescence and photoionization were

input directly into the gated integrator with no modifications. This

method is included for comparison purposes as a reference.31 The

signal output from the boxcar average was connected to the voltage-

to-frequency converter and the signal was counted for a 10 s
































(0
oLz
0


LL









+<


N.






-.
x
0





Z cr



7-
pj











period. In all four measurement methods, the boxcar average was

operated with a 10 kHz first-order cutoff filter on the input to help

remove unwanted low frequency components.



Results and Discussion

All correction methods used show a significant improvement over

the conventional method (Table 2-1). While no attempt was made to

optimize fully system components or operating conditions, the limits

of detection obtained are very good and demonstrate the improvements

expected. Laser power was found to present a severe problem when

frequency doubling the copper vapor laser (CVL) dye laser output.

The relatively low peak power and squared dependence of frequency

doubling on input power32 resulted in low power outputs for In and

Fe. The low laser power is partially compensated for (in limits of

detection) by the lack of laser scatter and the associated shot noise

in the direct-line fluorescence of Fe and In. Compromises concerning

excitation wavelengths had to be made since the CVL fundamental

output is in the visible region at 510 and 578 nm. Additionally, RF

interference from this laser was and has been found to limit severely

its analytical applicability for laser enhanced ionization and

fluorescence.33

The simplicity and nearly equivalent results obtained using the

bandwidth limitation scheme (BW) for signal processing make this the

processing method of choice for most applications. As seen in Table

2-1, results obtained using the bandwidth limited approach for

fluorescence are nearly equivalent to those obtained using










Table 2-1. Limits of Detection (ng/mL).


CONVa ABSb BLIAC


Copper Vapor Dye Laser

Laser Enhanced Ionization Detection

Li 2.0 0.3 0.1
Fe 30 5.0 2.0
In 4.0 0.6 0.2

Laser Excited Fluorescence Detection

Li 14 0.8 0.4
Fe 98 18 5.0
In 43 8.0 5.0
Na 14 3.0 1.0


Excimer Dye Laser

Laser Enhanced Ionization Detection

Li 0.7 0.3 0.5
Na 4.0 1.0 2.0
In 0.2 0.2 0.4



BWd BW+ABSe BW+BLIAf


Copper Vapor Dye Laser

Bandwidth Limited Signal Processing--Fluorescence Detection

Li 2.0 0.6 0.3
Na 2.0 1.1 1.0
In 3.0
Fe 15


a CONV = Conventional method of measuring pulsed laser signals using
a boxcar average and gated integrator.

bABS = Active baseline subtraction method using a modulated pulsed
laser. Subtraction of background noise is done by the boxcar
special electronics.













Table 2-1--continued.



c BLIA = Boxcar average plus lock-in amplifier using a modulated
pulsed laser. Subtraction of background noise is accomplished in
the lock-in amplifier.

d BW = Bandwidth limited signal processing in which high frequency
laser noise is not amplified while the signal is amplified to
increase signal-to-noise ratio of fluorescence signals.

e BW+ABS = Bandwidth limitation as discussed above followed by active
baseline subtraction.

BW+BLIA = Bandwidth limitation followed by the boxcar average plus
lock-in amplifier.










more complicated schemes. No correction for the total number of

laser pulses per second has been made. In the BLSP scheme and the

CONV method, 6000 pulses per second pass through the flame and are

detected. In the ABS and BLIA experiments, there are only 3000

pulses passing through and being detected in the flame each second.

The remaining 3000 pulses were blocked by the chopper and the noises

detected in laser enhanced fluorescence and ionization were

subtracted out.

The background subtraction methods (ABS and BLIA) provide an

additional advantage over the BLSP scheme and the CONV method, namely

the removal of low frequency changes in the laser conditions. The

CVL operates in a continuous flow mode with a constant addition of

neon and a constant removal of neon and a small amount of copper

vapor. Conditions within the laser cavity are changing at a very

slow rate (relative to the 6 kHz repetition rate) due to changes in

the neon pressure, partial pressure of the copper vapor, and position

of the electrical arc within the laser tube. These changing

conditions result in small changes in the dye laser power output but,

more importantly, changes in the magnitude of the radio-frequency

signal detected within the boxcar gate. No correction for dye laser

power fluctuations could be obtained, since the fluctuations were in

the source prior to modulation. Correction for flame and analyte

background flicker was accomplished in both the ABS and BLIA methods.

The effect of bandwidth limited signal processing (BLSP) on

fluorescence signals is demonstrated in Figure 2-6. The fluorescence

signals in Figure 2-6 prior to bandwidth limitation demonstrate the



























BOXCAR


BW LIMITED
AMPURER


450


250
TIME ( s)


Figure 2-6. Bandwidth Limitation Effects










high frequency noise which is the limiting noise source for this type

of laser. The temporal effect of bandwidth limitation on

fluorescence signals is also shown in Figure 2-6.

The BLIA results show a marked improvement over CONV results in

all cases. The BLIA results are felt to be slightly better than the

ABS results because a true phase-sensitive detection method was

employed in this scheme. The ABS method used a gated RC filter to

add capacitively signals of alternating polarity to remove background

and was susceptible to capacitor leakage, noise added by the polarity

switching amplifier, and signal fluctuation when a low number of

samples is averaged.

The experiments with active baseline subtraction (ABS and BLIA)

provide some unique opportunities for signal manipulation due to the

extremely high repetition rate of CVL. At the 6 kHz repetition rate

used, signals obtained are well above the low frequency flicker noise

region of the flame (or a plasma) and analyte background flicker.

Additionally based on laser repetition rate, beam diameter, and flame

gas flow rate, it is possible to irradiate each atom present within

the flame with 3 to 10 laser pulses of moderate intensity as the

atoms pass through the flame volume. Indeed, the high repetition

rate of the CVL should make it possible to detect virtually every

atom passing through a flame (or any plasma).

For comparison purposes, results obtained using our excimer

laser (Lumonics model TE-861) are included for laser enhanced

ionization. This laser is capable of providing peak energies of up

to 3 mJ per pulse of fundamental power from the dye laser much higher










than the CVL system. Typical energies obtained at 25 Hz repetition

rate were in the range of -80-250 iJ per pulse of frequency doubled

energy in a bandwidth of 0.0015 nm. The inferior performance of the

two methods employing background correction (BLIA and ABS) should be

noted for results obtained using the excimer laser. These results

may be explained in one of two ways: diminished performance of our

lock-in amplifier at the 25 Hz repetition rate (where a larger amount

of flame background flicker and analyte flicker is present); and/or

lower correlation of the interfering noise from the excimer laser

compared to the copper vapor laser. Both effects are felt to

contribute to some extent. At 25 Hz, a much larger flicker component

is present than at the 6 kHz CVL repetition rate. The low frequency

analyte and background flicker component at the sodium wavelength

588.9 nm is shown in Figure 2-7. These signals were measured with

only the flame on and no laser present. Signals were measured using

a cross channel spectrum analyzer (Wavetek, model 5890B) and

represent the total signal present across the frequency range

shown. Several trends are apparent from this figure. Flame

background flicker noises are small and become very small above

approximately 1.0 kHz. Analyte flicker is much higher at low

frequency and decreases at higher frequency. The much higher flicker

noise of the analyte can be attributed to nebulizer noise while the

higher overall signal magnitude is due to analyte emission.

The other contribution to flicker noise is the pulse-to-pulse

fluctuation of the laser. The lower correlation of the pulse-to-

pulse noise from the excimer laser is felt to be a limiting factor in




































0
0









0




o


o
0





w
C.1

0,



o0




o
0










0
I


Ht/ASP 83MOd 3SION










the improvements obtained with the excimer laser. The excimer

discharge voltage, -32,000 V, is less easily regulated than the

7,000 V used in the copper vapor laser. Oscilloscope comparisons of

the noise from the two lasers confirmed the much higher noise

fluctuations for the excimer laser. No correction for laser noise

could be obtained unless the noise component was present on

successive laser pulses.

The two noise correction methods (BLIA and ABS) were combined

with the noise reduction method, bandwidth limitation (BW) to

determine the individual contributions to noise reduction, and

signal-to-noise enhancement. It was found that the combination of

these techniques results in minimal improvement beyond that obtained

with bandwidth limitation alone. This indicates that the major

improvement is due to bandwidth limitation and thus the major noise

source is RF interference from the laser system.



Conclusions

The methods presented are shown to provide a significant

increase in the detection power for LAFS and LEIS. For active

baseline subtraction, improvements of up to 6 times were obtained

while with the combined bandwidth limitation and boxcar plus lock-in

amplifier (BW+BLIA) improvements of up to 40 times were obtained.

While these improvements may be possible for other laser systems, it

is to be emphasized that many of these methods rely on a certain

stability in the noise levels associated with a given laser system.

Application of these correction methods to a poorly correlated noise






26



source may yield little or no improvement in the detection power (see

excimer laser results). A more highly correlated noise source should

yield correspondingly better results. Additionally, modulation of a

pulsed laser and subsequent loss of one-half of the laser pulses may

be more harmful than the gain in detection power allowed by

subtraction of any noises.
















CHAPTER 3
ESTIMATION OF ABSOLUTE NUMBER DENSITIES



General Curve of Growth Introduction and Evaluation

A general overview of methods for determination of the absolute

number density of a species is presented in Chapter 1. In this

chapter, general expressions and a computer program are presented for

evaluation of the fluorescence curve of growth and the absolute

number density determination. The computer program for these

calculations was written in FORTRAN-77 and was approximately 500

lines long. FORTRAN-77 was chosen for the programming language as a

compromise of speed, flexibility, and scientific compatability.

The computer program for evaluation of the absolute number

density calculates the fluorescence intensity vs the concentration of

the species of interest. A log-log plot of these terms is referred

to as a curve of growth (COG). The expression for right angle

fluorescence (diagrammed in Figure 3-1) is shown in Figure 3-2. The

expression for fluorescence may be divided into several parts: a

primary absorption term, a self-absorption term, and a term for the

optical collection efficiency. (The prefilter and postfilter terms

are considered separately below.) The primary absorption term is

derived from the Beer-Lambert law and can be used for an absorption

curve of growth in conjunction with an optical collection efficiency

term. The primary absorption term is responsible for some of the




















L pre-


J-0

EXCITATION
BEAM


POSTFILTER
REGION


FLUORESCENCE


Figure 3-1. Diagram for Right Angle Fluorescence























,-4




-J

(-4






-4






Jt
















r0

















I
'-
^

aL








-o



fe-<
lu


-4


-UJ










LUI
+





LLJ

















CL
x
-4



A
(a
+







~"-4








(L




Ii


u A
uJ




a.
x

L1L


k


-4
X


I;



C-4


x
LI




o-
I--.-


-4






0.
x


L ,"


-4



0.
LLI










curvature in a curve of growth at higher number density. The source

irradiance becomes increasingly absorbed at high number densities

until the entire source irradiance is absorbed at extremely high

number densities. The self-absorption term is a dimensionless factor

accounting for reabsorption of fluorescence photons, within the

excitation volume, as they traverse the atom reservoir. The self-

absorption term also contributes to the curvature of the curve of

growth and continues to increase with increasing number density.

This term may or may not be important depending on whether resonance

or direct-line fluorescence is measured. The remaining term to

account for the optical collection efficiency includes TE, the

transmittance of the collection optics; mL and mH, magnifications of

the length and height dimensions of the source on the atom cell; and

n, the refractive index of the medium (to account for the change of

solid angle of collection due to medium to air transfer of

radiation). The term AhAZAL is used to give the observed radiant

fluorescence flux, A0.

Shown in Figure 3-1 (with expressions in Figure 3-2) are regions

for prefiltering of the excitation source and postfiltering of the

fluorescence radiation. A prefilter region is one in which the

analyte of interest is present and is illuminated by the excitation

beam but the resulting fluorescence is not viewed by the detector. A

postfilter region is one in which the analyte of interest is present

but the region is not illuminated by the excitation source. These

terms become important in some analytical situations where it may not

be possible to avoid prefiltering and/or postfiltering.










Some simplifying considerations used in the development and

evaluation of the expressions are as follows:

(i) a single atomic transition is considered for the atomic

emission and absorption methods while results are

included for both direct-line and resonance fluorescence;

(ii) for the absorption and fluorescence cases, the excitation

beam is of rectangular cross section (Z x H) and is

spatially, as well as temporally, homogeneous;

(iii) the atom reservoir consists of a uniform ground state

number density distribution at a uniform temperature in

the absence of the excitation beam;

(iv) the absorption and fluorescence spectral line profiles

may be described by a Voigt function;

(v) no restrictions are made as to source intensity, but it

should be noted that saturation of a transition by an

apparent line source may result in saturation broadening

of the spectral excitation profile until the source no

longer may be considered effectively a line source. The

validity and applicability of many of these

considerations are discussed by Zeegers et al.

Definitions of all terms used in the fluorescence expression

(and other equations used in this text) are included in Appendix A.

The computer program used for evaluation of the fluorescence

expressions is included in Appendix B.

In this research, no simplifying assumptions were made other

than those noted above. Assumptions of line or continuum source










excitation would greatly simplify the equations and calculations

involved, but the much more difficult case of an intermediate source

is considered here. The shapes of curves for line or continuum

excitation have been studied thoroughly.35-37 A more complete

development of the curve of growth equation is given in these

sources.

The program for curve of growth calculation uses some

approximations to enable the calculation to be performed. The first

of these approximations is the substitution of a large polynomial

approximation for the Voigt profile. The Voigt profile is the

spectral convolution of a Doppler profile with added collisional

broadening and is used in calculating the atomic absorption

profile k(v) where

k(v) = koV(a,v)

and
+, 2
a + exp(-y )dy
V(a,v) = a
= 2 2
-_ a +(v-y)


The Voigt integral cannot be solved in closed form and a substitution

proposed by Hui et al.38 is used. This approximation is extremely

accurate (one part in 106) and evaluation is fast with no loss in

overall accuracy of the COG. The limiting accuracy of the COG is

determined by the accuracy of the integration limits used in the

evaluation of the other integrals of the fluorescence expression and

the accuracy of the experimental parameters used. The other major

approximation involved in the COG calculation is in the evaluation of

the primary, self-absorption, prefilter, and postfilter integrals.










The integrals should be evaluated from minus infinity to infinity.

Obviously, this integration is not possible and a tolerance for the

convergence of the integral must be chosen. In most cases, a

convergence tolerance of a maximum of 0.001 percent of the value of

interest was used. In some cases where a tighter convergence was

desired, an even smaller tolerance was used, although a greater

accuracy was not generally available.



Calculated Curves of Growth

The general COG in atomic spectroscopy is described by a linear

low concentration (or low number density) region and a noticeable

curvature at higher number densities. The degree of curvature and

the final slope of the COG at high number densities depends on the

source spectral width and the atomic parameters used in evaluation of

the fluorescence intensity expression. The curves of growth for line

and continuum sources are presented in Figures 3-3 and 3-4,

respectively. These curves are shown for several a values. The a

value or a-parameter is a measure of the ratios of the collisional

width to the Doppler width:



a = In 2((-) .
D


The collisional width, AAC, is largely dependent upon the mass and

number of collisional species present and the temperature of the

atoms. The Doppler width, AAD, depends upon the mass and temperature






















- o

0O
o -


o
0 /0



-4o
wx .





-~ o
/z
0.0




















I I I
ZO

L0

4-)












0 Q COQ)





32N3OS30on-d -138 00-1























0



0



C)
1 .c



co
So0








0
I O
z



cn
L4
0


Sm z








00
Z 0
z/z 0






So o o-


0 04

N "0







^ON33S38onl14 fl3d 90-1










of the atom. The expression for AD is presented below with

expressions for collisional broadening widths presented later.



AXD = 4wX/C /2 RT in 2/M .



The curves presented for line and continuum sources agree well

with the other literature and experimental results presented

previously. The abscissa in all curves is shown in relative units

since we are not concerned with the absolute signal magnitude but

rather the shape of the COG. Here, we are more concerned with the

intermediate case of a pseudocontinuum source, i.e., a laser.

The spatial line widths of dye lasers (measured as the full

width at one-half maximum (FWHM)) typically are on the order of -0.01

to 0.05 nm when only a grating is used as a tuning element. Atomic

spectroscopy line widths in analytical flames vary from about

0.0005 nm to 0.02 nm.39 Thus, the laser will vary from approximately

equal to, to approximately 100 times the line width of the atom of

interest. This can be considered a pseudocontinuum source for most

applications. The curves of growth for pseudocontinuum sources of

equal laser and atom line atomic width and a laser which is 10 times

the atomic absorption profile (FWHM) are shown in Figure 3-5.

Curves of growth are shown for prefilter and postfilter effects

in the next series of figures. These curves are presented for an a-

parameter value of 1.0, which is used as an average value of

approximately 40 elements in an air-acetylene flame.40-41 Shown in

Figures 3-6 and 3-7 are prefilter effects for mild, medium, and

































03

o
L
C3
0





<)


0



0











0)



o
0
0














)00
b-









C-



t-

.1-
I cr


30OoS3abon'li 73b 907






38








)
c-




..














ui-
II





/ -4
/



I /r




\ \ *


/ ',o a

1 0 U i
U-
-4









So 6 6 \
0




..J. \ Q
00 0








0
W;j














a



C.
















a,



o


E3
0

/1 o
C



I CO

I ,

I 0
(X0 0
"------.--------------------








SI



a)
az L





o,-


Sa.o E










u0 L
I I I I I
D\ \N3:SU0r7 1= 901



La
If | ~ v \
333SoJm \\~ 90
B; 5 -\s, N\ Q -
B| \\ ^
u- & x< \
uj Z *. v \ -
C 0 N \ .










severe prefiltering for line and continuum sources, respectively.

The equivalent curves of growth for line and continuum sources are

shown for postfiltering in Figures 3-8 and 3-9, respectively.

The COGs for all prefilter and postfilter calculations are shown

with log (koL) as the abscissa. This is used, instead of the log

number density, to partially compensate for the reduced absorption

pathlength involved. The relationship between the peak absorption

coefficient (for pure Doppler broadening), ko, and the number

density, n, is presented below.



22
/4a1n2 e X nf
k -
o 2
me AAD



This relationship is developed in Mitchell and Zemansky10 and is

based on the absorption oscillator strength, f, and the central

wavelength, Xo, and Doppler width, AAD, (FWHM) of the atom of

interest (other terms are defined in Appendix A).

General trends which may be noted for these curves are

relatively simple. Prefiltering of the source results in an increase

in the rate of curvature for line and continuum sources. This

increase in curvature is accompanied by a general decrease in the

signal strength. This is due primarily to the decreased fluorescence

volume observed. The postfilter effect results in a general decrease

in the overall fluorescence intensity. A more complete discussion of

the prefilter and postfilter effects is included in a later section

on collision broadening and saturation effects. From the figures and




















SII


L

0

// / /

/ / 3






















bO





0 a)
/ F -O
,-4


\\ \ \







Sj \
5 --











obo
-.-J
0 \. -
w E









O 0


3 0
XXa




-a






*-4
















CI
Qa




c
co
00

o




1-



c,


i C i/
uI




\ \
0)






-Io
S\, \ -J

\\ \ 8-
\01d 0)



I-N_ O a








0











oo
3 LOO U 10 cc'-










discussion presented above, it can be seen that it is important to

minimize prefilter and postfilter effects (to achieve higher signal-

to-noise ratios and maximum linear dynamic range). It is generally

possible to do so by simple modification of the optical design.



Experimental Verification of Curves of Growth

Figure 3-10 shows the calculated curve of growth for sodium

resonance fluorescence in an air-acetylene flame. The experimental

data points are included on this curve also. The curve of growth was

calculated for a flame of 2500 K using a Gaussian source of 0.03 nm

(FWHM) and a damping parameter of 1.0. These represent the line

width of the copper vapor pumped dye laser and reasonable parameters

for Na in an air-acetylene flame.40 The fit of the curve of growth

to the experimental points is obtained by transposition of the COG

until this curve overlaps the experimental points. The overlap of

these curves gives a direct relationship between the concentration of

the species introduced and the absolute number density present in the

flame.

The experimental curve of growth was obtained using the same

experimental arrangement as described and diagrammed earlier in

Chapter 2 (Figure 2-2). Aqueous solutions of sodium were prepared by

dissolving sodium chloride in deionized water. Solutions below

1000 mg/L were prepared by dilution, while solutions above 1000 mg/L

were prepared directly by weighing.

The fluorescence results are in good agreement with a calculated

estimate of the absolute number density. This estimate is based on







44



















9 z

SL





0 -





z a. M



S ,-4
r r a.







S o
1 / 2 1









0- z a)
a: w-






0 co







1-4
w 8. 4 Q,
- 0, >- 0)

n o I rd











-et t
S nO z 0 )


4 U \ i


















,-4


30N3oS3 Onij 13" 0-7










the solution concentration, the solution uptake rate, the

nebulization and atomization efficiencies, and the flow rate and

thermal expansion of flame gases. These terms are grouped to form a

single term, 8, which should be within the range shown in Figure

3-10.

To confirm these experimental results and to extend the curve of

growth calculations to direct-line fluorescence, the same experiments

were performed using another laser system. Poor doubling

efficiencies with the copper vapor laser system (see Chapter 2) make

the wavelength used for Pb fluorescence difficult to obtain, at least

at a laser intensity level which will provide enough sensitivity for

these studies. The COG for lead with 283.3 nm excitation and

resonance (283.3 nm) and direct-line (405.7 nm) fluorescence were

calculated and measured experimentally. The curves were calculated

for a 0.019 nm source spectral line width (FWHM) with a-parameters of

1.0 (283.3 nm) and 1.5 (405.7 nm). These are estimates based on

39
calculations given by Parsons et al. since no known measured values

are available. The 0.019 nm source line width is obtained from our

Nd:YAG laser pumped frequency doubled dye laser (Quantel, model

YG580). The Pb experimental results were measured by Dr. Benjamin

Smith. His measurements were performed using an experimental setup

similar to that previously described (Figure 2-2). In this

experiment, two curves must be transposed to match with two sets of

experimental curves. The two experimental curves, shown in Figure

3-11, were found to give reasonable agreement between themselves and

are within one order of magnitude agreement with the estimated number





























0
rt



o

o






0 Z

-O


rv


3N3S3 90

33N33S38onii 138 001










density (from the nebulization and atomization efficiencies, etc.).

The agreement of these curves with the experimental values

demonstrates the accuracy and applicability of these methods.



Saturation and Collisional Effects

A further extension of the curve of growth calculations involves

the inclusion of some of the lesser noted effects in atomic

fluorescence spectroscopy. The expression for fluorescence presented

previously (Figure 3-2) and merely be extended here. The expressions

for saturation effects are included in Figure 3-2 but were not

previously discussed. The collisional broadening terms have not been

discussed yet and will be presented here.

Collisional broadening effects may be considered from two

separate aspects: collisional broadening due to analyte collisions

with analyte termed self-broadening or resonance broadening and

collisional broadening due to collisions with other species. The

first will be referred to as self-broadening and the second will be

termed collisional broadening. The general expression for

collisional broadening is presented in Figure 3-12.42 The theory of

collisional broadening was originally developed by Lorentz and later

extended by Lenz and Weisskopf. A complete development is presented

by Breene42 and Mitchell and Zemansky.10 The general expression for

collisional broadening simplifies when only resonance broadening is

considered. The summation of all atomic and molecular concentrations

simplifies to involve only the analyte concentration. The mass





















COLLISIONAL BROADENING


c 2N X
^^-rT C-


[2TkT


RESONANCE


BROADENING


2N ra"?2
0= a [4-rkTNA- F2)


Figure 3-12. Expression for Collisional Broadening


NAM


I 4













term, 1/mA + 1/mQ, simplifies to involve only the mass of the analyte

(see Figure 3-12). Worst case collisional broadening effects will be

seen for situations of small atomic mass, long wavelength

transitions, and large collisional cross sections.

The self-broadening effects for two elemental cases have been

included in Table 3-1. The elements evaluated represent the two

extremes encountered in atomic spectroscopy: a high mass, low

wavelength case and a low mass, high wavelength case. The

collisional cross sections used here represent the range of values

40
measured.4 From curve of growth calculations and the results in

Table 3-1, it can be seen that no significant differences due to

self-broadening are expected for experimentally realizable

situations. No appreciable changes are seen in the a-parameter until

number densities of greater than 1017 atoms/cm3 are approached.

Based on previous measurements in Chapter 2, this corresponds to an

initial atomic concentration of greater than 100 M and an unrealistic

aspirated concentration of 108 parts-per-million. Additionally, at

high number densities where self-broadening effects should become

apparent, the primary absorption term will typically have reached

absorption saturation and any change in the a-parameter and line

shape will not affect the absorption of radiation. Very weak

transitions (ko/n < 10-15) may show some small self-broadening

effects at high number densities since the absorption terms are very

small and allow considerable transmittance at high number densities.
















Table 3-1. Broadening Effects on the a-Parameter for Self-Broadening


Best Case Worst Case
(Low Collisional Broadening) (High Collisional Broadening)

Pb Li
A = 283.3 nm X = 670.8 nm
T = 2500 K T = 2500 K
mA = 207.2 g/Mol mA = 6.94 g/Mol


a-Parameters

a2 (A2)
30 100 n


1.0007 1.0023

1.0068 1.0227

1.0680 1.2270


(cm-3)


1016

101

1018


a2 (A2)
30 100


1.0035 1.0097

1.0350 1.0970

1.350 1.9700


n(cm-3)


1016

10
1017

1018
1 o1


~


~











Collisional broadening effects due to an added matrix species,

termed simply collisional broadening effects, are included also. The

collisional broadening effects are somewhat different than the self-

broadening effect in that these effects are seen over the entire

range of analyte concentrations and are due to some constant

concentration of an added species. The effect of collisional

broadening may or may not be noticeable depending on the spectral

width of the source used and the spectral width of the atomic

profile. For line sources, the wings of the absorption profile

(described by the a-parameter) will not matter since absorption

occurs only at the line center. As the source width increases, the

wings of the absorption profile may affect the curve of growth. At

low atomic concentrations, however, the a-parameter does not affect

the curve of growth (see Figures 3-3 to 3-9).

One of the few situations where collisional broadening may need

to be considered is the technique of graphite furnace atomic

spectroscopy. In many instances, a high concentration of species is

added to serve as a matrix modifier. The added species will be

vaporized in a small volume of the graphite furnace and result in a

high atomic concentration. 3

In a recent publication, Schlemmer and Welz44 used an aliquot of

Pd and Mg for matrix modification, resulting in an atomic

concentration of 1016 to 10 in a typical tube furnace atomizer.

Collisional broadening effects become barely noticeable at this

level. The curves of growth for Li with added Mg matrix modifier are

shown in Figure 3-13. These curves are calculated in the same manner

























e-

a
o0


>-.
!'- .




a o
O .










J E
0











.. o
o <
0 0-


L
I C





9)










as before (Figures 3-3 to 3-5) but with the added term for

collisional broadening. Each of the relationships discussed

previously, a-parameter dependence and source width dependence, are

shown in this figure. The expected source dependence is seen in the

curves of this figure. The differences calculated for the

pseudocontinuum (AA = 0.19 A) and the line source
s

(AS = 0.005 A) are too small to be seen here but occur in a region

somewhat smaller than that seen in the other pseudocontinuum

sources. These results represent an extreme case, not likely to be

seen in conventional flame spectroscopy and only infrequently seen in

nonflame spectroscopies. Generally, collisional and self-broadening

effects will have a very small influence, and in fact may not be

apparent at all.

Saturation (optical) effects are dependent on the source

spectral irradiance and do not depend upon the analyte of interest,

at least not in the same manner as the collisional broadening does.

As seen in the expression for right-angle fluorescence (Figure 3-2),

an increase of the source spectral irradiance E(X), above the

saturation spectral irradiance E (A), will result in an effective

decrease in the absorption coefficient. This decrease will occur at

all wavelengths at which the source irradiance is above the

saturation spectral irradiance. Prefilter terms are affected in the

same manner as the primary absorption term. The self-absorption term

may or may not be affected by saturation, depending upon whether the

fluorescence wavelength is within the saturating source spectral

profile. A postfilter region will not be affected by a saturating










irradiance since the source does not pass through the postfilter

region. A complete discussion of saturation effects and practical

measurement of saturation curves is given by Alkemade.45

Curves of growth for several source intensities are presented in

Figures 3-14 and 3-15 for line and continuum sources, respectively.

The general trend seen is an increase in the linear regions of these

curves. No change of shape in the COG is seen. Curves of growth

were also calculated for the same Pb transitions previously measured

(283.3 nm resonance fluorescence and 405.7 nm direct-line

fluorescence). These curves are shown in Figure 3-16 and are

calculated for the same 0.019 nm source spectral width (FWHM) used

previously. As seen in this figure, it is possible to extend the

resonance fluorescence COG using a saturating source to equal linear

dynamic range and even beyond that of the direct-line fluorescence

case.

The source intensity at which equality occurs for the direct-

line and resonance fluorescence curves can be shown to be related to

the fractional populations of the two lower levels used. The

equations relating the source intensity necessary for equality of the

direct-line and resonance curves are shown in Figure 3-17. These are

derived from the initial curve of growth equation. As seen in Figure

3-16, the equality for the resonance and direct-line curves occurs at

a source irradiance approximately 100 times greater than the

saturation irradiance for the lead transitions. This agrees with the

ratios of the populations of the two levels 98.6% to 1.05%, a ratio

of approximately 100 to 1. This is calculated using the continuum















































En
o z
-0

LLI



0
O
.J



















































33N3OS3?On7j "138 901













































































30N30S3Om1zi -i73a 901


r.
O









oo
U )
- 0)












z
0 c )
.,.

C)
0

- U) 0



z C
r-f






C3




O




0 00
o 00

C-





.0
o i
C
02




D a





--4
3 0
o 002













GENERAL CASE


e Le E+E dv
Jfl- e-k(V')L dv


LINE SOURCE CASE


S
koV(a,v) E

k"oV(ao vI


CONTINUUM


SOURCE CASE


E
S^E^ES

5 (y k(V/)
/ov k ?7, E


F1
F


Figure 3-17. Equations for Curves of Growth Equalities for
Saturating Irradiances


ko
ko


F2
" ^










source approximation since the source spectral width (FWHM) is

roughly 25 times larger than atom profile and as such closely

approximates a continuum.

The prefilter and postfilter effects with a saturating laser

source (0.019 nm) are shown in Figures 3-18 and 3-19. A 0.5 cm

prefilter length and a 0.5 cm postfilter length were used in the

corresponding curves. A 0.5 cm absorption path and a 0.5 cm

fluorescence path length are used in all curves here. In both cases,

it is possible to extend the calculated curve of growth to some

extent. Curves with prefiltering show the greatest extension. This

is due to removal of prefilter effects by the saturating laser and an

extension of the primary absorption linear region. The constant

difference between the curves of growth with and without prefilter is

due to the decreased source irradiance passing through the detected

volume. This corresponds to an additional primary absorption term

with no additional fluorescence. For the postfilter curves of growth

in Figure 3-19, the difference between the curves with and without

postfilter is seen to increase with increasing source irradiance.

The increasing difference between curves is due to the increase in

linearity of the curve which does not have postfiltering while the

curve with postfilter does not have the same increase in linearity.

If the source intensity is increased drastically, the overall curve

of growth will become limited by the curvature due to the postfilter

region. In no manner is it possible to remove postfilter effects by

using a more intense source.







60












I c




a)

oc








C1

I S=
Q .0
------ -7 -------------- C ,--






















0
V
ca

CLI^ / ^








Cl)
c3
S4-
I I
1 oM














0)

o
0 2-













C..
^x 3



























,,.






61










0

0
/ / /

I / / C



S I / a
,' I ,
I / 0C
I /I /
\' // //O --
/, /



I / /

'' 1 / / )
n / 0s: i




I 1 -
Z




11
CD
00
cJ oc

ra

C-

CI)

0
0 0 0
I 1


30N33S38onTii 173 001





-4










Conclusions

A general purpose computer program has been presented and is

investigated for applications to determine the absolute number

density of an atomic species. Various aspects of this curve of

growth program have been investigated including prefilter and

postfilter effects, saturation using various sources and collisional

broadening due to resonance broadening and broadening due to

quenchers.

It was found that collisional broadening, in all forms, should

present very minimal effects if they are seen at all. Prefilter and

postfilter regions are shown to cause a premature curvature of the

curve of growth. Prefilter and postfilter effects may be accounted

for and an absolute number density obtained if the exact geometry is

known. The prefilter effect may be removed to a large degree by a

saturating excitation source. The use of a saturating source may not

be desirable, as it adds another complication to the curve of growth

measurement and calculation of the absolute number density. Much of

the work of this chapter has been theoretical in nature and some

would be difficult to verify experimentally. Nevertheless, the

experimental results obtained for two of the simpler cases discussed

in this chapter show extremely good agreement between the

calculations and the experimentally measured results.
















CHAPTER 4
SPATIAL DISTRIBUTIONS OF ATOMS
IN INHOMOGENEOUS FLAMES


Concentration Modulated Absorption Spectroscopy

A technique for increased sensitivity in atomic absorption

spectroscopy with pulsed lasers was recently introduced by Langley et

al.6 The technique was employed by the authors to yield absolute

number densities for atoms in flames and for molecules in

solutions. This experiment is based on a pump-probe arrangement of

optical beams and relies on a linear relationship between source

intensity and fraction absorbed. This technique, along with two

other related techniques using pulsed lasers in atomic spectroscopy,

are applied for the determination of spatial distributions of atoms

in inhomogeneous flames and are described in this chapter.

Many of the techniques presented in Chapter 1 are not applicable

to spatial profiling. This is due to the moderate to low sensitivity

of those techniques and the small region, and thus the small atomic

concentration, probed by these methods. The combination of these two

limitations and the limitations presented by pulse-to-pulse

instabilities and the spectral bandwidth of some pulsed dye lasers

make many techniques unsuitable for spatial diagnostics.

Concentration modulated absorption spectroscopy (COMAS) was

initially applied by Langley et al.6 for absolute number density

determination. The development of the COMAS expression follows from










the interaction of two focused Gaussian beams and is based to a large

degree on the Beer-Lambert absorption expressions. The beams are

assumed to be derived from the same pulsed laser source with a

spatially Gaussian electric field amplitude. From the Beer-Lambert

absorption expression, the change in the number of laser photons in

an incremental length dz will be



di = anAi dz
P p


where ip is the number of laser photons, A is the equilibrium

fractional population between levels 1 and 2, and a is the absorption

cross section. This fractional population is equal to the difference

of levels 1 and 2 divided by the total analyte concentration

A = (N1 N2)/n. Based upon expressions for focused Gaussian beams

and the interaction volume for these beams, an expression for the

modulation of the probe signal is obtained:



dipr = (2 nA/X)i ip
pr p pr


Obtaining the fractional change in the number of probe photons or

gain gives



di /ip = G = (2~ 2nA/X)i .

prSubstituting the original Beer-Lambert law n(/T) = n and

Substituting the original Beer-Lambert law ln(1/T) = anA and










substituting for a gives G = (27i /(nAA))(ln(1/T))2. A plot of gain
p
times concentration vs In 1/T2 will give straight-line plot, the

slope of which is the proportionality factor in the relationship

between concentration and analyte number density C = xNA where C is

the analyte concentration. Thus, this experiment was viewed as a way

of obtaining an absolute number density which required only measuring

readily available analytical signals in a relatively simple optical

arrangement. Some optical restrictions were applied in the

development of the expressions which limit the applicability of these

equations.

The development of the COMAS expressions is not questioned here,

and in fact extremely good results are obtained by Langley et al.46

for molecular solution-phase analyses, but the assumption of a true

Beer-Lambert relationship for excitation of atoms in flames with a

relatively broad spectral width dye laser is not valid. The Beer-

Lambert relationship works well for molecular analyses since the

molecular absorption band is typically much wider spectrally than the

line width of the dye laser excitation source used. The results for

molecular analyses by Langley et al.46 were easily checked assuming

no solution phase decomposition and were found to be 7.4x1015

molecules/cm3 experimentally for a solution concentration of 6.0x1015

molecules/cm3. When considering the atomic experiments, the dye

laser spectral bandwidth is much wider than the atomic line width.

Using the expression in Figure 3-2 for the primary absorption term of

the fluorescence curve of growth, a difference in signals of 5 times

(for the low-density linear region) is calculated for the signals











expected with a narrow line source and the signal obtained with a

source width 10 times greater than the Doppler width of the atom.

This will result in a direct error of the absolute number density

determined by this factor. A correction for this error may be

obtained by evaluating the overlap of the two profiles expected and a

correction factor obtained.

The concentration modulated method involves a pump-probe

arrangement of optical beams as shown in Figure 4-1. In this

arrangement the probe is approximately 5% of total dye laser

output. This method was originally investigated for Na in an air-

acetylene flame. The same copper vapor laser and dye laser system

presented in Chapter 2 were used for this application. The bandwidth

limitation approach was used for the absorption measurements.

An initial investigation of the method presented by Langley et

al. was carried out. This arrangement of counterpropagating co-

linear beams involves modulation of one of the two beams. The COMAS

signal is obtained by subtraction of the absorption signal with and

without the pump beam present. Subtraction of the two signals is

obtained using the active baseline subtraction method employed in

Chapter 2. The absorption signals were detected by a fast photodiode

(United Detector Technology, model PIN-1ODP) and connected to a

current-to-voltage converter (Thorn EMI Gencom Inc., model Al).

These signals were connected directly to the gated integrator and

boxcar average. The COMAS experiment conducted by Langley et al.46

was confirmed and a limit of detection of 1 part-per-million (ppm)













































DETECTION
ELECTRONICS


67













FPD
5% R

I







FLAME








M


BDU

SPRE-
AMP


Figure 4-1.


Concentration Modulated Absorption Experimental Setup:
Co-linear Beams


DYE LASER


COPPER
VAPOR
LASER











was found for an interaction volume of approximately 1 cm3. This

agrees closely with previous results found.46 The COMAS technique

relies on the linear interaction of the two beams and allows direct

determination of absolute number densities. Better limits of

detection are obtainable with simple experiments using hollow cathode

lamps.48

An extension of COMAS was attempted to determine spatial

concentration profiles. The arrangement of optical beams was

modified to form a pair of crossing beams (Figure 4-2). A scanning

motor (General Scanning Inc., model G3) with a mirror attached and a

pair of lenses was added to scan a focused probe beam through the

flame. The scanner in conjunction with the first lens serves to

position the probe beam within the flame. With the first lens

positioned at a distance of one focal length from the scanning

mirror, beams from the scanning mirror pass through the flame

perpendicular to the pump beam. The second lens served to turn the

probe beam and redirect the beams to the photodiode detector. The

scanning is controlled by the computer interface (Stanford Research

Systems, model 24S) and allowed rapid collection of the entire

spatial profile of the flame. For these experiments the computer

controlled the scanning of optical beams and collection of all

data. For increased spatial resolution several optical elements were

added. A small pinhole, diameter 0.5 mm, was used in the optical

path of the probe beam to limit the beam size. Additionally, a lens

was used in the optical path of the pump beam to produce a smaller

beam. Relatively long focal length lenses (200 mm, 2" dia.) were



























5% R


ROTATING
MIRROR


- IBM-PC

INTERFACE


Figure 4-2. Experimental Setup for Spatial Diagnostics for COMAS


69





































o I1
I











used in all cases. These lenses provide a relatively uniform beam

waist over the width of the flame.49 The volume probed by the

interaction of these two beams is approximately 0.1 mm3 (0.4 mm

diameter probe by 0.5 mm diameter pump). This interaction volume is

calculated using the method of Jackson et al.50 and relies on the

overlap of two spatially Gaussian beams. The beams were measured and

closely approximate a Gaussian spatial profile. Additionally, the

spectral line profile of a dye laser has been measured here by

scanning the dye laser across an atomic transition and elsewhere by

this and other51,52 methods and closely approximates a Gaussian

spectral profile. The optical arrangement used allows scanning while

maintaining the same optical axes and beam size. Horizontal beam

placement was found to be extremely important for a uniform

interaction region. The results obtained for scanning the flame

produced by an inhomogeneous surface burner23 are shown in Figure

4-3. The surface burner used in these experiments was provided

courtesy of Dr. R.J. Krupa and the construction is shown in Figure

4-4. The COMAS results are shown for an initial solution

concentration of approximately 5000 ppm Na. The COMAS results are

shown for a single region located above one row of analyte

capillaries in both the parallel and diagonal burner orientations.

The measurements were taken in a region approximately 6 mm above the

surface of the burner head. This corresponded to the top of the

flame cones53 for gas flow rates of 1.6 1/m acetylene, 2.2 1/m

oxygen, and 8.3 1/m nitrogen with this type of surface burner. Based

on a measured nebulization efficiency of 0.14 and an estimated





















5000 ug/mL Na


PERPENDICULAR
BURNER


5000 ug/mL Na


DIAGONAL
BURNER


CONCENTRATION MODULATED
ABSORPTION
Figure 4-3. Results for Perpendicular and Parallel Burners for COMAS


SI I I I t I


SI I a I I a I





















OXIDANT







I cm.


00000
000000
00000
00000
00000
000-o000


0 OXIDANT HOLES


(0.031" dia.)


0 FUEL CAPILLARIES (0.042" i.d., 0.058"o.d.)

Figure 4-4. Design of Surface Burner Used










atomization efficiency54 of 0.6, the analytical region of 0.1 mm3

calculated for the COMAS crossed beam experiments should have -1013

atoms present during each laser pulse. The relative insensitivity of

this technique is seen in the high concentration of analyte necessary

to obtain any analytical signal. This low sensitivity is due to

several experimental problems. The relatively large spectral

bandwidth of the dye laser (0.3 A) is approximately 6 times greater

than the Doppler-width (FWHM) of Na in an air-acetylene flame with an

estimated temperature of 25000C. This large bandwidth results in

unabsorbed source irradiation reaching the detector and increased

background shot noise. The peak-to-peak fluctuation of the dye laser

intensity also limits the minimum absorption measurable.55 Radio-

frequency noise is also thought to contribute to some degree (see

Chapter 2). At any rate, COMAS has been found to be too insensitive

for applications in spatial profiling.

The minimum detectable signal may also be calculated based upon

the spectral bandwidth of the laser, the parameters for Na in an air-

acetylene flame, the region probed in the interaction volume and the

minimum detectable absorbance which may be typically measured.56 The

same curve of growth program used in Chapter 3 is used for this

calculation with only the primary absorption term calculated (see

Figure 3-2). The minimal concentration detectable based on this

calculation and a minimum detectable absorbance was found to be

approximately 1000 ppm Na. This does not consider the effect of

pulse-to-pulse fluctuations of the laser or any shot noise or radio-

frequency (RF) noise added by the laser.











Results were also obtained for a simple atomic absorption

experiment with only the relatively weak probe beam and optical

scanning system (no pump beam). Results for the parallel and

diagonal burner configuration are shown in Figure 4-5. Again a

solution concentration of 10 ppm Na was used for these experiments.

The increased sensitivity of this technique is seen immediately since

these results are for an initial atomic concentration 50 times lower

than the COMAS results. This increase in sensitivity is accomplished

at a loss in spatial information since the analytical signal is

obtained for the entire absorption path length (1 cm or greater for

the burner used) and are not obtained for a much smaller interaction

region. The interaction volume is approximately 1.3 mm3 (0.4 mm

diameter by 10 mm (or greater in the diagonal burner case)),

approximately 13 times greater than in the COMAS case. This accounts

for a significant portion of the increased sensitivity. The spatial

information obtainable with this system is still somewhat

remarkable. Since the laser beam is coherent and has an extremely

low degree of divergence, good spatial resolution may be obtained.

An attempt was made to perform the same series of experiments with a

hollow cathode lamp. The hollow cathode lamp, with its much narrower

spectral bandwidth (compared to the dye laser), should have had a

higher sensitivity. No suitable optical arrangement could be found

to give a tightly focused beam within the flame volume. With the

high repetition rate of this laser and the fast optical beam scanning

method employed, entire spatial profiles of the atom cell may be

obtained in a very short period of time.

























10 ug/mL Na

PERPENDICULAR
BURNER


A ~10 ug/mL Na


DIAGONAL
BURNER


*OO

*000000
00!0!0!09
*coo*



00



SIMPLE ABSORPTION
Figure -5. Simple Absorption results for Perpendicular and Parallel





Burners
SIMPLE ABSORPTION

Figure 4-5. Simple Absorption Results for Perpendicular and Parallel
Burners











In an attempt to decrease the spectral bandwidth of the dye

laser, an etalon was added within the oscillator cavity of the dye

laser.57 The decreased spectral bandwidth would have resulted in an

increased sensitivity in all absorption results. The etalon was

found to reduce the oscillator cavity gains to such an extent that

lasing would not occur. The dye laser was then switched to an

oscillator-only configuration with 100% of the copper vapor laser

output used to pump a single dye laser cell. The etalon was found to

increase cavity losses5 and decrease the stability of the dye laser

(in terms of pulse-to-pulse fluctuations) and gave only a slight

reduction in the spectral bandwidth (0.3 A to 0.18 A). The only

slight reduction in the spectral bandwidth of the dye laser is due to

the relatively poor quality of the etalon used.

Concentration modulated absorption spectroscopy was initially

investigated by us as a means of performing spatial profiling and

determining absolute number densities. The low sensitivity of the

technique prevented this application; in addition, complications far

beyond those expected were present. As a result of these

complications, other methods were investigated for increased

sensitivity in spatial profiling applications.



Two-Wavelength Laser Enhanced Ionization
and Fluorescence: Spatial Distributions

In several recent articles,18-20 the use of laser excited

fluorescence for spatial distribution measurement has been studied











using a planar laser beam and a 2-dimensional image detector. The

spatial resolution element for this technique was found to be =1 mm3

with concentrations between 1 and 20 ppm being used for spatial

profiling results. In other studies, spatial profiles have been

measured in inductively coupled plasmas59 and in flames6 with a lens

and image detector or lenses and monochromators used to obtain the

spatial profiles. One advantage of image detection has been pointed

out in a recent publication where spatial information was obtained in

an entire plane within the flame volume on a single laser pulse.61

Other studies have included probing of local electrical fields in

flames using LEIS which involved measurement of the atomic line

widths.62 The electrically-broadened atomic line width is a measure

of the strength of the local electrical field. Several of these

methods are applied later in this chapter for similar applications

with some modifications.

Two of the most sensitive methods for atomic species

determinations in flames are laser excited atomic fluorescence

spectrometry (LEAFS) and laser enhanced ionization spectrometry

(LEIS). The sensitivity of these techniques has been discussed

previously in Chapters 2 and 3 and the discussion is extended here.

Recent results presented by Magnusson et al. 636 and Axner et al.6

in graphite furnaces and flames, respectively, show limits of

detection which approach 1 part-per-trillion (ppt) for single

wavelength excitation. For two-wavelength excitation in a graphite

furnace, results approaching 1 pg absolute limits of detection have

been obtained. Two-wavelength excitation was found to give an










ionization signal enhancement of up to 6000 times over single

wavelength excitation.64 With the high temperatures present in an

inductively coupled plasma (ICP) and the hazards of arcing of the

high power radio-frequency field, an optical method of detection of

laser induced ionization was presented by Turk et al.66'7 This is

instead of an electrode placed within the atom cell. As yet, no one

has applied two-wavelength LEIS and LEAFS for spatial measurements.



Experimental Setup and Discussion

The experimental arrangement of optical beams for two-wavelength

LEIS and LEAFS for spatial diagnostics is shown in Figure 4-6. This

is an extension of the technique presented previously for COMAS with

an additional wavelength of excitation and fluorescence and

ionization detection instead of an absorption detection method. The

same copper vapor laser used earlier was also used here. The beam

from this laser was used to pump two dye lasers. The dye lasers used

(Hansch-design) were pumped in the oscillator-only configuration.58

That is, only one flowing dye cell in each dye laser was excited by

the copper vapor beam. This was found to be a more efficient method

for pumping the dye lasers since the power per pulse of the copper

vapor laser is relatively low. Approximately 40% of the copper vapor

beam was used to pump the first dye laser with the remainder used in

the second. A fast photodiode placed at the end of the second dye

laser received the small fraction of the copper vapor beam which was

not reflected by the 99% reflector in the second dye laser. The

signal from this photodiode was used to trigger the boxcar detection




























r------------ ---------"------------

DYE DYE PD

LASER LASER I




COPPER REF....
I --------t
VAPOR CHOPPER__

LASER I kT


Scut





IBM PC MONO-
CHROMATOR


ROTATING
--- MIRROR +



_-------* H
RS 232 -TBOXCARS [
GATE B.D. I ,

U r; IjJ--T'-r -.----~-~-r--~---~-- --C-----.
`- 7 -'--B1- : Pl. SIGNAL
- TRIGGER

OSCILLOSCOPE



**
L !: -::----~1-


Figure 4-6. Experimental Setup for Two-Wavelength LEIS and LAFS










electronics and the oscilloscope. The same signal, used to trigger

the boxcars, was divided by multiples of 10 to trigger the computer

interface. This division was necessary in many instances since the

computer interface was only capable of transferring data to the

computer at a rate of 960 samples per second. The necessity of

division was dependent on the number of data points taken since the

computer interface contained a buffer memory which accumulated

untransferred data points. Scans across the burner in one direction

were obtained using the rotating mirror. To obtain the total profile

across the burner, the burner itself was translated using a

micrometer stage, with profiles obtained at several positions using

the rotating mirror. An initial attempt to obtain profiles by

positioning the Al beam at different places within the flame volume

was investigated. Voltage field variances and voltage field

collection effects were found which complicated all spatial results

obtained. With a water-cooled bias electrode (-1500 V) placed

directly in the flame, spatial profiles were obtained. When the XI

laser beam was positioned directly under and parallel to the

electrode, signals were found to reach a maximum. Translation of the

laser beam, while maintaining the same burner and electrode position,

to a horizontal distance of approximately 0.5 cm from directly under

the electrode, but still at approximately the same atomic

concentration (see COMAS and absorption results), the analyte signal

was found to drop by more than one order of magnitude. This is due

to field collection effects and voltage field differences. Other

field collection effects have been investigated by other










authors.62,68 To minimize voltage field effects, the burner head was

translated and was found to give no appreciable voltage field effects

when measured by the method of Axner and Berglind.62 The placement

of a large electrode within the flame volume does disturb the

laminarity and shape of the flame and may result in somewhat

distorted profiles. For these reasons, fluorescence was investigated

as an alternate means of detection.

In the course of these investigations, several important factors

were discovered. To obtain a uniform interaction volume and an equal

sensitivity over the entire optical path probed, the two optical

beams had to be horizontal and intersect or remain a fixed distance

apart, over the entire distance probed. Additionally, the burner and

electrode also had to be horizontal to maintain a constant distance

between the laser beam and the electrode and to reduce voltage field

effects.

A typical profile obtained in a single scan is shown in Figure

4-7. This profile was obtained approximately 7 mm above the surface

of the burner at the tips of the flame cones. This profile was

obtained for an initial atomic concentration of 1 ppm Na with

excitation at 588.9 nm and 568.8 nm. From the signal-to-noise ratio

apparent in this figure, the atomic concentration produced by 1 ppm

Na is well above the limit of detection. These results were obtained

for a 1000 point scan, each point representing an average of 10 laser

shots (for reasons discussed earlier). With the high repetition rate

of the laser and the data acquisition system, these results required

only 1.67 s to obtain. Thus, the speed, ease, and high capacity for











information gathering is shown. The slightly irregular profile

obtained for this case is due to a slight flow constriction in one of

the analyte capillaries. This only served to illustrate the

analytical usefulness of this technique, as it was possible to

correct this flow restriction in later studies. Two arrangements of

optical beams were investigated for these studies. These

arrangements involved swapping the positions of the two laser

beams. Each arrangement had its own advantages. When the first

wavelength laser (AX) was positioned directly under the electrode, a

large pulsed ionization current was created by AI. The ionization

current created by A2 occurs only in the interaction region. Thus

the current produced by X1 had to be subtracted from that due to

X2+X1 to obtain the interaction volume. The current due to X1

produced a background shot noise which could not be corrected for.

An additional consideration was the thermal population of the lowest

excited state. Thermally excited atoms are excited by X2 and

collected.

In the first configuration with A1 directly under the electrode,

the collection of thermally excited and optically excited (X2) atoms

was limited by the electric field decrease with distance away from

the electrode. In the second configuration with X2 directly under

the electrode, there was a lower background current produced due to

X1 but a higher current which could not be corrected for due to

thermal excitation of atoms. The best analytical scheme would be

that one which gives the lowest background signals. It would seem





























I /
.j

44
0


U











0 2 4 6 8 10 12 14

(mm)
0

N
0








0 2 4 6 8 10 12 14

(mm)


Spatial Profile Result for LEIS


Figure 4-7.











that as long as A1 is of higher energy than kT, the first scheme

would be best but this is complicated by the electric field effects

and the intensities of the lasers employed. For these experiments,

the first configuration was found to give better signal-to-noise

ratio. The contributions of the thermal, electric field, and optical

arrangement effects to signals obtained, corrections obtainable and

noises for each ionization process are summarized in Table 4-1.

The best arrangement for subtraction of signals due to

ionization outside the interaction region would involve modulation of

both 1X and 12. A modulation scheme which would give equivalent

results for both optical arrangements would be to sequentially excite

the analyte with 1I, then 12, and then the combined beams, )1+A2.

The analytical signal would be obtained by subtracting the single

wavelength signals from the combined signals A1+A2. Such a

modulation scheme would involve custom design of the optical choppers

or the use of acousto-optic deflectors and electronic circuits to

form the modulation pulses. Demodulation of the signals obtained

would be complicated also and would require a custom design lock-in

amplifier. Considering the complications expected from the

discussion above and the minimal improvement expected, this method of

correction was not attempted. This method of modulation would also

use three pump laser pulses for each data point and shift the

effective data taking rate to 2 kHz for this laser system. For any

pulsed laser system other than copper vapor (or mode-locked cavity-

dumped lasers), this modulation scheme would result in a shift in the

data taking rate back into a flicker noise dominated region (i.e.,











Table 4-1.


PROCESSES


Ionization Processes, Optical Arrangements, Signals and
Noises for Two-Wavelength Laser Enhanced Ionization
Spectroscopy


SIGNALS


Nonspecific thermal ionization of atoms and molecules
Ionization due to A1 and thermal processes
Ionization due to thermal processes and X2
Ionization due to A1+12 in interaction region


OPTICAL ARRANGEMENTS


11 directly beneath electrode, 2 perpendicular
X2 directly beneath electrode, A1 perpendicular


PROCESSES

CORRECTION
METHODS


NOISES


D.C. signals
blocked by
boxcar input
capacitor.

Shot noise
due to D.C.
signal. Equi-
valent noises
for both
optical
arrangements.
Magnitude
will vary for
different
elements.
Analyte &
background
fluctuations
corrected for
by modulating
A1 or X2.


Modulation
of X2


Signal magni-
tude and
noise greater
for arrange-
ment 1. Shot
noise uncor-
rectable.


Modulation
of X1


Signal magni-
tude and
noise greater
for arrange-
ment 2 but
smaller than
that due to
process 2 in
arrangement
1 (due to
lower popu-
lation of
excited state).


Analytical
signal. No
correction
necessary.


Noises and
signal are
the same
for both
optical
arrangements.






86



<100 Hz). The only laser system which comes close to the high-

repetition rate needed is a 500 Hz excimer system, with an effective

data taking rate of approximately 160 Hz. This is still within the

flicker noise region for a typical analytical flame.

Similar results were obtained for two-wavelength laser excited

atomic fluorescence spectrometry (LEAFS). A lower sensitivity is

obtained for this technique compared to LEIS due to several

reasons. The detector in this instance is placed a significant

distance from the region excited by the two beams. Since the

fluorescence is isotropic, this detector will only receive a small

fraction of the fluorescence emitted. Using two-wavelength

excitation, atoms are excited to levels within several kT of the

ionization continuum. Losses of these atoms to the ionization

continuum are likely and make the fluorescence from the second

excited level to the first excited level less sensitive. Monitoring

the fluorescence from the first excited level to the ground state is

possible and the change in this fluorescence, when A2 is added, is

referred to as fluorescence dip spectroscopy.69 Monitoring

fluorescence from the first excited state to the ground state is

somewhat complicated by shot noise from analyte emission and noise

due to analyte and laser flicker. Fluorescence from the second

excited state to the first excited state is limited by shot noise of

the scatter of X2 and losses to the ionization continuum. For

monitoring fluorescence, the best method would primarily depend upon

the proximity of the second excited level to the ionization

continuum. For these results, the fluorescence from the second level










to the first was found to yield the best results. Omenetto et al.70

have investigated two-wavelength fluorescence of ions produced in an

inductively coupled plasma. The results approach low part-per-

trillion levels and benefit from the lower losses to the ionization

continuum (the doubly ionized species is especially difficult to form

in most instances) compared to the two-wavelength atomic fluorescence

case.

A typical result for two-wavelength excitation and fluorescence

detection is shown in Figure 4-8 for 10 ppm Na at a height of

approximately 4 mm in the same burner used previously. The increased

separation of the spatial distribution for each capillary within the

burner are evident in this figure. Approximately the same

representation was obtained at a height of 7 mm as shown for

ionization results in Figure 4-7. These two figures demonstrate that

this burner has very good laminarity low in the flame which

deteriorates rapidly. The fluorescence was monitored at X2. The

presence of fluorescence was confirmed (versus scatter) by detuning

X2. One major advantage of the fluorescence method is that it is not

necessary to place an electrode within the flame volume, which makes

it possible to detect signals without disturbing the flame. This

also makes the fluorescence method more applicable, especially in

situations where it is not possible to place an electrode within the

cell volume.

As a direct result of these techniques, it was possible to

spatially map the entire combustion zone. By translating the burner

and obtaining successive profiles across the burner, it was possible




















































(mm)


Figure 4-8.


Spatial Profile Result for LAFS











to map a horizontal plane within the flame volume. If multiple

horizontal planes are obtained a four-dimensional structure may be

developed, three spatial dimensions and a concentration dimension. A

single horizontal plane located at approximately the top of the flame

cones is shown in Figure 4-9 using ionization detection for a 1 ppm

solution of Na.



Conclusions

Single-wavelength and two-wavelength methods for obtaining

spatial information within inhomogeneous flames have been

presented. While the single wavelength concentration-modulated

absorption results were not particularly sensitive or useful, the

two-wavelength LEIS and LEAFS results demonstrated exceptional

spatial resolution and very good sensitivity. The spatial results

shown here represent the first application of two-wavelength LEIS and

LAFS to this area. The high-repetitive rate laser used and an

optical method of scanning the beams through the flame were shown to

allow rapid determination of spatial profiles of analyte

concentrations. While the applications here are limited in scope, it

is felt that these techniques will find many applications.

































































Figure 4-9.


Three-Dimensional Spatial Profile for LEIS