SATURATED ATOMIC FLUORESCENCE AS A DIAGNOSTIC TOOL
FOR FLAMES AND PLASMAS
JAMES N. BOWER
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY CF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF :-LORIDA
This thesis is dedicated
to my wife, Esther,
whose support and love have made the
difference between night and day.
Help and real human understanding have been available at all
times from my chairman, Dr. Winefordner, for these four years. He
has been a guiding light and a friend.
I would like to thank John Bradshaw for much of my basic under-
standing of analytical atomic spectroscopy. He has been a tireless
Dr. Nicolo Omenetto has been a great help in understanding the
physical processes of laser excitation.
Dr. Winefordner's research group is certainly one of the most
stimulating in the world. Thanks to all members for many discussions
and helpful suggestions.
Thanks are in order for two unnamed souls whose joyless scien-
tific interactions with me persuaded me to return to school and
TABLE OF CONTENTS
ACKNOWLEDGEMENTS. . . . . . . . . . . . iii
ABSTRACT . . . . . . . . ... . . . .. vi
1 INTRODUCTION . . . . . . . . ... . . 1
Quantum Efficiency . . . . . . . . . 1
Total Number Density . . . . . . . . 3
Noise Power Density. . . . . . . . . . 4
Reaction Rates . . . . . . . .... . 5
Tunable Dye Lasers . . . . . . . . 6
2 THEORETICAL CONSIDERATIONS . . . . . . . 8
Quantum Efficiency for a Two-Level Atom. . . . . 8
Quantum Efficiency for a Three-Level Atom. . . .. 17
Saturation . . . . . . . . ... 21
Quantum Efficiency Via the Slope Method. . ...... 22
Total Number Density . . . . . . . . 27
Possibility of Absolute Calibration. . . . . ... 27
Noise Power Density. .. . . . . . . 29
3 EXPERIMENTAL . . . . . . . .... .. .. 32
General Comments . . . . . . . .... . 32
Radio Frequency Shielding. . . . . . . .. 36
Photomultiplier. . . . . . . . . 37
Fluorescence Flux Collection. . . . . . 37
Detection Electronics. . . . . . . . . 38
Nitrogen Laser . . . . . . . . . 38
Dye Laser Operation. . . . . . . . . . 40
Flame System . . . . . . . . . . . 42
Solutions . . . . . . . . . . . 43
Measurement Procedure for Y and nT . . . . .. 43
Noise Power Density . . . . . . . . . 50
4 DATA . . . . . . . . . . . . . 54
Notation . . . . . . . .... ..... . 54
Strontium. . . . . . . . ... . .. 55
Sodium . . . . . . . . .. . . . 74
Calcium. . . . . . . . .. . . . 77
Indium . . . . . . . . . . .. 80
Noise Power Density. . . . . . . . . .. 87
5 RESULTS AND DISCUSSION . . . . . . ... 100
Quantum Efficiency and Total Number Density. . . ... 100
Noise Power Density. . . . . . . . . .. 103
6 CONCLUSIONS AND FUTURE WORK. . . . . . . ... 105
System Temporal Response . . . . . . .... 105
Complete Saturation. . . . . . . . . .. 105
Measurement of E . . . . . . . . .. 106
Photometer Calibration . . . . . . . . 106
General Considerations . . . . . . .. 106
Noise Power Density. . . . . . . . . .. 107
General Comments . . . . . . . .... 108
1 SUBSIDIARY MEASUREMENTS. . . . . . . . ... 10
2 COMPUTER PROGRAMMING AND PLOTTING. . . . . . 140
REFERENCES. . . . . . . . . ... ....... 166
BIOGRAPHICAL SKETCH . . . . . . .. .. ... 170
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy
SATURATED ATOMIC FLUORESCENCE AS A DIAGNOSTIC TOOL
FOR FLAMES AND PLASMAS
James N. Bower
Chairman: Dr. J.D. Winefordner
Major Department: Chemistry
A new experimental approach to the determination of three flame
diagnostic parameters is developed. The experimental application of
saturated atomic fluorescence to the measurement of quantum efficiency,
total number density, and noise power is discussed. Data for quantum
efficiency and total number density are compared to the scant litera-
ture values. Four elements and three different flame compositions
are investigated. Preliminary noise power spectra are discussed as
background for the use of saturated atomic fluorescence to measure
noise in the atomization and nebulization processes.
Characterization of physical and chemical parameters of flames
and plasmas has been a goal of chemists (1, 2, 3), physicists (4, 5, 6),
and combustion engineers (7, 8, 9) for many years. The goal of
physicists, physical chemists, and combustion engineers has been to
understand on a quite fundamental basis the processes which go on in
flames and plasmas, i.e., oxidations, reductions, free-radical reac-
tions, ionizations, etc. In recent years, an even larger impetus has
been seen for this research as sources of economical energy have
dwindled and better combustion understanding and design have become
paramount in importance. Analytical chemists, on the other hand, have
become interested in flames and plasmas because of the extremely
important role of these devices in quantitation of elemental species
in various samples. The coincidental research interests have brought
all together to try and provide a better understanding of basic
processes common to both.
A quantity of great interest to the analytical chemist and other
scientists is the quantum efficiency, the measure of the relative
yield of fluorescence processes to absorption processes. The quantum
efficiency has a strong influence on the performance of normally (low
intensity) excited atomic fluorescence spectrometry, which has become
an important flame diagnostic (3, 10, 11) and analytical tool (12, 13,
14) in recent years. In addition, quantum efficiency is an indication
of (rates of) deexcitation processes in flames and plasmas which can
be invaluable for unraveling many basic spectroscopic and fundamental
The traditional method of measurement for this quantity (Y) has
been to measure atomic absorption from a continuum source, then to
measure atomic fluorescence with the same source positioned at 90
degrees to the absorption axis. Great pains had to be taken to ensure
that solid angle effects, flame edge effects, etc. were eliminated
(15, 16). At best, this yielded a spatially unresolved picture of
deexcitation processes (i.e., absorption is a line of sight tech-
nique). In addition, no hope could be held cut for temporal resolu-
tion. Also, the measurement involved two highly different electronic
gains (those typical of absorption and fluorescence measurements),
which must be calibrated against each other.
The measurement of quantum efficiency via saturated atomic fluor-
escence, however, allows both spatial and temporal resolution under
the proper conditions. It is inherently easier to measure, as it
involves only one optical train for which the sclid angle must be
found. However, the photomultiplier tube and optical system must
be calibrated via a standard source. Along with the system used for
this investigation (a nitrogen-pumped dye laser) comes the need for
quite sophisticated electronics. Signal processing and cross-checks
for problems become much more involved.
Total Number Density
Total number density has been another sought after quantity in
flame and plasma spectroscopy. An important figure of merit in atomic
spectroscopy has been the efficiency by which a particular atomizer is
able to convert solution or solids introduced into it into the form of
neutral atoms. In most cases, this efficiency has been separated into
1. efficiency of nebulization-yield of fog droplets from a
volume of solution introduced,
2. efficiency of atomization-yield of neutral atoms from sub-
microscopic species, i.e., atoms, molecules, ions, etc.
The former quantity is a very tedious one to measure, involving long
periods of aspiration of a relatively high concentration solution of
a suitable element, collection of waste solution and washdown from
the nebulization chamber, and dilution to volume. Then, the concen-
tration of the analyte in the diluent must be determined. Typically
(for most nebulizers in atomic spectroscopy), this is a low efficiency
process. Therefore, the accuracy is subject to determinate errors in
The problem of determination of efficiency of atomization follows
directly on these results, unless a separate procedure is developed.
This is virtually impossible unless a vapor generation technique is
available. The efficiency measurement must be made via determination
of the total number density in the flame. The classical method for
this measurement has been to use absorption from a continuum source
(17, 18), which provides the desired quantity only after a tedious
calibration procedure, including correction for reflectivity of
multiple surfaces in the optical collection train. An expensive,
high-resolution monochromator is also required, as the atomic absorp-
tion line is quite narrow. In addition, this method does not allow
spatial or temporal resolution.
Saturation of the pertinent atomic transition provides a fluor-
escence signal (therefore a spatially resolvable signal in the usual
configuration), that is not a function of flame conditions (except as
a function of atom production capability), and is a function only of
basic atomic properties.
Noise Power Density
Noise power density measurements have been used typically by
information transfer and electrical engineers as a means of discovering
the sources of systematic and random noises in hopes of eliminating
them and being able to transmit data with fewer errors at ever in-
creasing rates. The techniques of noise measurement and theoretical
treatment have been growing in importance for analytical chemists
(19-28), as the science grows in depth and sophistication. The
attempts at measurement to date, however, have been troubled by lack
of definite calibration and lack of willingness to track down and
eliminate sources of instrumental noise in analytical techniques.
The difficulty of this task, while formidable, is not insurmountable
(29). If calibration can become complete, and the sophistication of
analytical chemists can be improved in areas which are now subsidiary
(flow engineering, electrical engineering, etc.),substantial improve-
ments should be forthcoming.
Saturated atomic fluorescence stands to provide a substantial
contribution to noise analysis in atomic spectroscopy. Saturation
provides several features:
1. elimination of dependence on quantum efficiency (flame
2. spatial resolution,
3. possible loss of dependence upon the atomization step
(if correct analyte is chosen),
4. possible relative freedom from emission background noise
(if conditions are chosen carefully), and
5. freedom from source variation.
Therefore, saturation can further the understanding of noise
sources in atomic analytical spectroscopy. It can provide separation
of noise of nebulization from noise of atomization (if such a noise
exists), allowing better understanding, and design, of nebulization
Since analytical flames and plasmas are used at atmospheric
pressure and are quite hot (2000 K to 6000 K), reactions are myriad
and fast. In most flames and plasmas, some sort of local thermodynamic
equilibrium (LTE) is assumed. In several cases, however, the mechanism
of excitation of atomic species is unclear (e.g., the induction coupled
plasma, which is not in LTE). Engineers and physicists share with
chemists the need to understand these and other phenomena kinetically.
Saturated fluorescence (atomic or molecular) can be a very useful
tool in this pursuit, as a means of simplifying kinetic schemes by
swamping. Reactions such as photoionization, quenching, and inter-
system crossing can all be studied by means of saturated fluorescence.
Tunable Dye Lasers
The advent of the tunable dye laser (30, 31) and the power it
gives to observe atomic and molecular populations have given flame and
plasma spectroscopists an invaluable tool for diagnostic procedures.
The general properties of these and other lasers include:
1. directionality (low divergence),
3. coherence (spatial and temporal), and
4. high irradiance.
All of these laser properties have been useful for analytical
atomic fluorescence spectroscopy since Fraser and Winefordner (32) first
used a dye laser to excite atomic fluorescence. Their work covered
nine elements in hydrogen-air and acetylene-air flames. Limits of
detection obtained were within 10 to 100-fold of those obtained with
conventional sources, and linear dynamic ranges were about four decades.
Progress in both theory and experimental achievements has been rapid
in the ensuing years (12, 33-38). With more and more powerful lasers
available (37, 39-45), the possibility and theory of non-linear
phenomena has been investigated (40-44). Several authors have ex-
perimentally achieved near or complete saturation of atomic (46-50)
and molecular systems (51, 52).
Laser saturation of atomic transitions in this work has made
possible diagnostic procedures with spatial, temporal, and flame com-
position independence. This will allow analytical chemists to measure
the quantum efficiency, total number density, and noise power density
spectrum in an atomizer without the above mentioned problems of linear
spectroscopy. However, the benefits of this freedom must be weighed
against the need for sophisticated measurement systems (fast time
scale) and laser temporal and spatial inhomogeneity effects.
Quantum Efficiency for a Two-Level Atom
A two-level atom is one which, although it may have many quantum
states, has only one excited state available for thermal population in
addition to the ground state. After it has been delivered to the
excited state by absorption of a photon, it can be returned to the
ground state by either of two paths, emission of a photon or colli-
sional deactivation (quenching). Quantum efficiency (Y) is a measure
of the efficiency of the radiative process with respect to the absorp-
tion process. A schematic diagram of a three-level atom and its
processes is depicted in Figure 1. A two-level atom can be pictured
by ignoring all processes connecting the third level and the other
two levels. A real example of a two-level atom is shown in Figure 2.
Only the 4s level is available for thermal population in analytical
flames (temperature range from about 2000 K to 40CO K, kT is therefore
0.2 to 0.37 eV).
The basic fluorescence radiance expression (53) is given by:
BF 47 Y21E 12 jk dv (1)
1 = path length in direction of detection system, mr
4r = number of steradians in a sphere (fluorescence is
Figure 1. Three level atom model photon and collisional
processes are illustrated. Symbols are explained
below. The lower state is symbolized by "1" and
the upper by "u."
B Einstein coefficient of induced absorption
lu (m3 j-1 s- Hz)
Bu Einstein coefficient of induced emission
ul (m3 j-1 s-1 Hz)
A Einstein coefficient of spontaneous emission
E,1 Spectral irradiance of exciting radiation
u(W m-2 Hz-1)
k Rate constant for collisional process con-
niecting state m wi h state n
Figure 2. Calcium term diagram levels 1 and 2 are indicated.
Y21 = fluorescence power (quantum) efficiency, W fluor-
Ev12 = spectral irradiance 9f exciting radiation at absorption
line, v12, W m"- Hz-1 (1 W = 1 Js-1)
f k dv = integrated absorption coefficient over absorption
0 line, m-1 Hz
The product Ev,2 f k dv is the power absorbed from the source by the
analyte atoms per meter cubed of atomic species. The integrated
absorption coefficient is given by:
hv, gn, -1
fk dv = n1 ) 812[ ], 1 Hz (2)
V 1 12'
Nh12 = energy of the exciting photon, j
c = speed of light, m s-1
B12 = Einstein coefficient of induced absorption, m3 J-
B s-1 Hz
g1' 92 = statistical weights of states 1 and 2, respectively,
nI, n2 = concentration of states 1 and 2, respectively, m-3
(note that n1 + n2 nT, the total concentration of
atoms in all states)
The bracketed quantity corrects for the effective decrease in absorp-
tion caused by stimulated emission from the upper state.
Using the rate equation approach, and invoking steady state, we
B 2Ev12 B 2112
(k12 + c = (k21 + 21 c A n2 (3)
k12' k21 = excitation and deexcitation non-radiational
(collision) rate constants, s"
A21 = Einstein coefficient of spontaneous emission, s-1
21 = Einstein coefficient of induced emission, m3 J-1
1 s-1 Hz
B12 = Einstein coefficient of induced absorption, m3 J'1
nI, n2 = concentrations of electronic states 1 and 2, m-3
c = speed of light, m s-1
The quantum efficiency is defined as
Y1 A 21 k (4)
Y21 A21 + k21
and A21 is related to B21 and 812 by
8rhv312 8hv 12 1
A21= (c3 ) B21= ) (2 12 (5)
Combining these expressions, one arrives at
6h (L-) Y E*1 (6)
BF= () 21E12n1 12) 12 (E*vl2 + E'12
where E*j12 is a modified saturation spectral irradiance (W m-2 Hz-1)
evaluated at frequency v12 and is defined as
The term E*v12 can be expressed in terms of the saturation spectral
irradiance Es 12, which is the spectral irradiance required to bring
about 50% of the maximum fluorescence radiance possible. If Es 12 is
expressed as a function of E .12, then
Es 12 = E* 12 9+ 12) (8)
Substituting for nI in terms of nT (nT = n1 + n, and using Eqs. 3,
6, and 8)
hK 12 F 12
BF )Y21E 12nT( c E 12
The maximum fluorescence is then
BFm (a )hA21 nT (g ) (10)
When E 12 = Es12 then BF = BFmax/2.
cA21 8Thv3 ()
12 1Y21 c2Y21
g1 7.6 x 1023 (12)
Es12 glg2) ( 2Y) (12)
A theoretical plot of BF vs. Ev12 appears in Figure 3. If a plateau in
fluorescence radiance can be observed as laser power is varied, and the
laser power at 50% of this maximum fluorescence radiance can be found,
then the quantum efficiency can be found from the quantity Es 12. For
calculations in this work the quantity Esx12 will be used (see Eq.
Figure 3. Theoretical saturation curve.
log EA (relative)
-I---------- _,.-- ----_
The necessary conditions for the measurement of quantum efficiency
in this way are
1. Saturation must be completely achieved, so that BFmax can
2. One must measure E12 by measuring the following quanti-
a. laser peak power (W),
b. laser cross-sectional area (m2,
c. laser optical bandwidth (Hz).
3. Self-absorption must be negligible.
4. The source must be a spectral continuum with respect to
the absorption line width.
5. A steady-state condition must be achieved.
6. No coherence effects can be manifested. The rate equation
approach must be valid.
The achievement or measurement of these conditions is described in
Quantum Efficiency for a Three-Level Atom
The same approach (as used for the two-level atom) yields for a
three-level atom like Na (see Figure 4)
Es 13 g 31 k32 E*13 )
V13 93 91 k32 r!1
21/2, 1 1/2
21D 1 /2, 2 1/2
3 ... 3o
Figure 4. Sodium term diagram levels 1, 2, and 3 are indicated.
E*13 B131 c-h (14)
kmn = collisional rate constant connecting atoms in the
m m state with those in the n state
The E5,13 is again related to Y31, but no simple relationship
exists. The collisional constants, which are unknown in the litera-
ture, interfere with the interpretation.
For some limiting cases, where level 2 is close to level 3 (e.g.,
sodium, see Figure 4) or is close to level 1 (e.g., indium, see
Figure 5), certain assumptions can be made.
With sodium, for instance, if the assumption is made that
k23 >> k21 + A21
k32 k32 2
A21 + k21 + k23 = k23 93
E13 (1)( 2) E*13 (15)
93 1 gl g2)
1 + -- + --
E 13 g + g2 g3) E*13 (16)
This is equivalent to a two-level system, as the top 2 levels have
merged into one, effectively.
21 1/2, 2 1/2
Figure 5. Indium term diagram levels 1, 2, and 3 are indicated.
41/2, 2 1/2
For a three-level atom like In
s 13A= 32 + k32 + k21 E13
= t-~ 13 L k3 2 + k2 1 v13
93 k21+ k12 -
The same type of assumption should hold for levels 1 and 2 as
previously for level 2 and 3. Therefore, if
k21 >> k32 + A32
k21 1 1_ 91
21 + 12 1 2 + 2 1 + 2
1 + 1 + -
s g1 + 92 (17)
E13 =g1 + 92 + 93) E1317)
This is again equivalent to a two-level system.
Complete saturation means that the population of the excited
state and the ground state are equalized (with due regard to the re-
spective degeneracies). Therefore, the kinetic drains on the excited
state are negligible with respect to the optical pumping rate. A
further increase in irradiance cannot effect an increase in fluorescence
A measurement of this (saturated) state usually means that when
the irradiance is reduced, via some suitable filter, the fluorescence
signal does not decrease. By placing calibrated filters in the beam,
one can determine the point at which only half of the signal is left;
this point is the saturation spectral irradiance.
Quantum Efficiency Via the Slope Method
From the two-level theory, it can be shown that
1 1 1 (g4)
BF 12 g2 3 5 5
F '12 92 6.6 x 103 Y211hv211A21nT
+ ((1 2) 4h ) (18)
g2 hv12!A21n T
A plot of the theoretical shape of 1/BF vs. 1/Es is shown in Figure 6.
The pertinent features are the slope and the intercept. The slope
of this plot contains the quantum efficiency (and n-), and the inter-
cept contains nT.
For this procedure, all of the above mentioned criteria must be
met with the exception of completeness of saturation. The closer
saturation can be achieved, however, the less extrapolation is
necessary. In addition, the fluorescence collection depth, 1, must
be measured, and the collection photometer calibrated.
If both sides of Eq. 18 are multiplied by the quantity
then the 1/BF axis can be scaled absolutely, as the intercept is
Figure 6. Theoretical 1/BF vs. 1/E curves. Curves shown are
for two different quantum efficiencies (Y) and two
different total number densities (nT).
/-- -slope proportional
intercept proportional to nT
equal to the quantity
which is well known.
A rearrangement of the three-level fluorescence expres-
4w 1, 1
+ ( k32 +__
A21 + k21 + k23 g3
+ gl( )I
g3 Y31 3EXj
When this is rearranged, we have
1 g18rhc2 1 1 + 93 k32
1 1 1 i __e_ /_
B .5 E g A k + k (20)
Fl+3 g3Y31X13 13 3 21 21 23
A plot of these quantities cannot be scaled absolutely because the
intercept includes unknown collisional rate constants. The same
speculation as before can still be made, however, regarding the rate
constants, allowing calibration to be made in certain limiting cases.
That is, k23 can be expected to be much larger than either A21 or k21.
A21 + k21 + k23
then reduces to
By the law of mass action at equilibrium
k32 n3 2
T n- -g
23 2 93
The fluorescence slope method equation reduces to
1 4r 1 g2+g1 1
1 (_43 )T 92 91 +
BF13 A31hvl3)1 nT 93 93
Y( X l
Rearranged this equation resembles that for two levels
4 B F 1+3
gy 5 +
93 31 13E
g2 + 91
Indium. For an atom of the indium type, where the second ex-
cited state is very close to the ground state,the slope method equa-
1 g2 )(8hc2
)(+ -1 exp[-E2/kT])( 31E
93 gl V
+ g9 + g3
+ 2 exp[-E21/kT] + A32k 32
In this case k21 can be assumed to be much larger than A32 + k32'
and this term then is negligible.
Total Number Density
Two-Level Absolute Plateau Method
A measurement of BFx
B -(1 )[h 92
BF = (4, )[hvl2A21nT (g
produces nT easily if the measurement of BF is absolute (see Appendix
1) and if 1, A21, and the g's are known.
The calibration of the 1/BF axis is direct for total number density
if gl1 9g, h, v12, BF, and A21 are known, as the intercept of the
1/BF axis is
1 4 g 1 + 32
BF Ihv12A21n g2
Three-Level Case--Absolute Plateau Method
Unless assumptions or experiments are made for the relative size
of the collisional constants, calibration is impossible. Relative
measurements are still valid, and saturation still assures immunity
to quantum efficiency effects as in the two-level case.
Three-Level Case--Slope Method
Once again assumptions must be made for collisional constants in
order to permit calibration of BF in terms of n .
Possibility of Absolute Calibration
If the ratio of resonance to anti-stokes fluorescence is taken,
in saturation, we have
BF 92 g A31 + k31
13 1 + + -- exp[E21/kT] + k
1+3max 93 93 112 k
BF g92 A32 k32 (22)
f3max 1 + + exp[-E21/kT] + k21
2--3ax 93 93
A31 + k31
A32 + k32 A +k g
Ak32 k k32 312 3 exp[E2 /kT]
k1 2 k 21 gg2 12
are probably not much greater than one, as k21 is expected to be at
least comparable to either numerator. Then, if E21 >> kT, two con-
ditions are satisfied
1 g1 A31 + k31
93 92 k21
g2 A32 +k32
2. !3 exp[-E21/kT- << 32 k32
Equation 22 then reduces to
BF +2 1
1+3max 1 +--+--exp[E 21/kT]
1+3max g3 93 21
F 1 + A32 + k32(23)
1+3 i + g- +
23max g k21
If a flame of known temperature can be produced, then the ratio
(A32+k32)/k21 can be evaluated. This allows Y (and nT) to be obtained
in absolute units. For a flame with a temperature of 2300 K and
thallium as a probe (E12 = 0.966 eV),the exponential term is as
exp[+0.966/2300 x 0.861 x 10-4] = 131
BF 2 + A32 k32
The difference between the ratio and 2/135 = 0.0148 is the contribu-
tion of A32 + k32/k21. This value, when inserted into Eq. 21 (for
measurement of thallium, in this case), would allow the intercept of
axis to be completely calibrated, as the intercept then contains only
known quantities. If BF is known absolutely (calibrated),then nT can
also be arrived at unambiguously.
Noise Power Density
The theory and practical measurement of power spectra (power vs.
frequency data) have been adequately covered by Blackman and Tukey (54).
Two approaches are possible to measure noise power spectra. The first
is to compute the autocovariance function for the data collected. The
Fourier transform of this is the power spectrum. The second procedure
is to compute the Fourier transform of the data, then to square and
add the frequency and phase components to find the power spectrum.
Although both methods were available, the second one was chosen for
its conceptual familiarity. The use of a discrete data collection
system (analog to digital converter) required a discrete Fourier
transform, with attendant aliasing and bandwidth problems.
The application of saturated atomic fluorescence will result in
an ability to separate the noises resulting from the nebulization pro-
cess from those stemming from the atomization of the analyte. Since
these two sources of noise are independent, their noise power spectra
add linearly. Since the measure being computed in our case is the
noise current spectrum, these would add quadratically. If an analyte
such as copper, which is completely atomized in a standard flame, is
measured, the noise power under known conditions can be found.
This noise power should be reproducible in nature with the ex-
ception of certain non-stationary noise sources. If this proves to be
the case then the total noise and the 1/f noise pertinent to most
spectroscopic measurements can be investigated with respect to design
elements and operating characteristics common to atomic spectroscopy.
These include burner, chamber, torch, and nebulizer design, as well as
gas flows and flame stoichiometry. Noises involved in the flame or
plasma processes up to the point of atomization could then be sub-
tracted from the noises seen for an incompletely atomized analyte,
producing the noise inherent in the atomization process.
The unfortunate truth, however, is that many of the underlying
noise sources in analytical spectroscopy are non-stationary.
Examples of non-stationary noises are nebulizer clogging, RF noise,
60 Hz mains noise, some source drift noises, stray light noise (room
and daylight), and some electronic drift noise. These are not really
approachable via the technique or theory of noise power spectra, though
often a reasonable idea of the noises can be found (55).
A variety of experimental systems were used for collection of data
pertinent to these measurements. The basic system shown in Figure 7
is the laser excited atomic fluorescence system described by Weeks
et al. (13). The system has been modified slightly, as reported by
Bower et al. (56) and Bradshaw et al. (10). Several of the features
changed since the measurements of Weeks et al. are of high importance.
The spatial effects of saturation in the wings of focused Gaussian
laser beams, as observed by Blackburn et al. (50), and previously
predicted by Rodrigo (57) and theoretically approached by Daily (58)
have been eliminated by moving the observation point approximately 10
feet further from the laser. At this point, the spatially homogeneous
section of the laser beam was enveloping the whole seeded flame (see
In order to facilitate alignment, it was necessary to add folding
mirrors (M) with adjustments in the horizontal and vertical planes.
A problem involving dye changes leading to a different angle of
emergence has been circumvented by the use of the folding mirrors
and the combined aperture-scatter shields (ASS). The two apertures
initially were set to define the beam line of interaction of the dye
laser and the flame. When subsequent dye changes produced non-colinear
Figure 7. Laser excited atomic fluorescence system. System
components are as follows: LPS = laser power supply,
NPL = nitrogen pump laser, DL = dye laser, M = folding
mirrors, TG = trigger generator, B = boxcar, SCR =
strip chart recorder, ASS = aperture-scatter shield,
LT = light trap, BU = burner, L = lenses, MO = mono-
chromator, PMT = photomultiplier tube, PS = high
volLage power supply. For source of major equipment
see Table 1.
Laser Excited Atomic Fluorescence Equipment List
Model UV-14 Nitrogen Laser
Model DL-300 Tunable Dye Laser
Model FL-2000 Tunable Dye Laser
Model H-10 monochromator
Model 412B High Voltage
Model R106 Photomultiplier
Model 160 boxcar
Model 162 boxcar mainframe
Model 163 boxcar plug-in
Model 164 boxcar plug-in
Type S-2 sampling heat
Servoriter strip chart recorder
Apertures (iris diaphragms)
Molectron Corp., Sunnyvale, Cal.
Molectron Corp., Sunnyvale, Cal.
Lambda Physik, Gottingen, W. Ger.
Perkin-Elmer Corp., Norwalk, Conn.
Jobin-Yvon, % ISA, Metuchen, N.J.
John Fluke, Seattle, Wash.
Hamamatsu Corp., Middlesex, N.J.
Princeton Applied Research,
Princeton Applied Research,
Princeton Applied Research,
Princeton Applied Research,
Tektronix, Portland, Oregon
Texas Instruments, Dallas, Texas
Edmund Scientific, Barrington, N.J.
beams, these could be adjusted with the mirrors to pass through the
The collection system was aligned at a 90 degree angle to the in-
coming beam by the use of helium-neon alignment laser crossed with the
dye laser beam.
Radio Frequency Shielding
Radio Frequency (RF) shielding of the entire Molectron nitrogen
pump laser enabled the output to be freed of the large drift-type noise
associated with people and objects moving in proximity to the experi-
mental system. This involved enclosing the whole pump laser and power
supply in a brass screen Faraday case. All electronics were connected
via grounding straps to a salt-bed earth ground. All cables leading
from the photomultiplier to the detection electronics were double-
shielded by a coaxial woven shield augmenting the internal shield.
These modifications enabled RF noise-free measurements into the milli-
volt level, which corresponded (with a 50 n 2-fold attenuator) to
40 uA of peak current. This was possible even with a background
current of 100 pA. For most measurements, the limiting noise was the
electronic noise of the boxcar average.
The photomultiplier was wired for fast pulse, high current
output, as reported by Fraser and Winefordner (32) and operated at
-1000 V. Cable of 50 n impedance (RG-58U) was used for all
Fluorescence Flux Collection
A Jobin-Yvon H-10 low dispersion monochromator was used
for collection of fluorescence. Dispersion is low and collection
efficiency high, as suggested by Weeks et al. (13). A two lense
collection system with an intervening aperture was used to help
cut down on scatter for resonance collection cases. Scatter in
these cases was reduced by liberal use of light traps (LT), aper-
ture scatter-shields (ASS), and black felt cloth. In many cases,
this scatter signal was Rayleigh scatter (indicated by an increase
when the flame was turned off), and could not be completely elimi-
nated. In such cases, the noise limit for measurements was due to
laser carried (peak to peak variation) scatter noise. The image
formed by the two lenses at the slit of the monochromator was of
A PAR 162 boxcar average was used for detection and measurement
of fast fluorescence pulses. Two different inserts (plug-ins) were
used in measurement and alignment. The 164 plug-in offered fixed
sensitivity (100 mV full scale) and several integration times. It was
quite hardy, and was used for preliminary alignment, when a very large
scatter signal was used to set the gate delay. The 50 a input im-
pedance was always used to match the cable and minimize ringing.
Some ringing was still apparent, however. The 163 plug-in with an .S-2
sampling head was used for the main measurement scheme. The gate was
75 ps (S-2 sampling head); jitter in the boxcar gating circuits led to
some distortion of this aperture. This head was used to find plateau
(saturation) values on top of the fluorescence temporal pulses, which
corresponded to saturation of the atomic fluorescence.
The synchronous trigger of the laser could not be used, as RF
feedback occurred at the time of each trigger and interfered with data
collection. A separate trigger generator was used, which had a delayed
pulse feature. This enabled correct timing between laser firing and
The nitrogen pump laser was run according to manufacturer'sin-
structions. The high-voltage power supply was kept at 22 kV to give
better longevity of the critical components (thyratron and capacitor),
as these had been problematical in the past. The nitrogen flow rate
was 15 liters per minute for all experiments. Operating pressure was
maintained at 55 torr. The pressure transducer had to be replaced
and calibrated twice, however. This involved approximately half a
day to a day's work, since the entire pump head had to be removed from
the Faraday cage. All connections had to be made again outside the
cage in order to test fire and calibrate the unit.
The repetition rate was kept at about 15-20 Hz for all experi-
ments. Greater repetition rates led to lower peak powers. High peak
power was necessary for saturation. However, a rate below ca. 20 Hz
would lead to signal leakage problems on the boxcar mainframe.
An interesting phenomenon was observed at the 20 Hz repetition
rate. A beat frequency appeared on the dye laser output. This was
first attributed to the power supply of the laser, as its voltage
dipped when the laser was fired. This idea was discarded when it was
noticed that the beat frequency was about 2 Hz. The observation was
made that as the dye fluoresced in the cuvette on each nitrogen laser
pulse, the magnetic stirrer appeared to be processing at about 2 Hz.
The motor provided was an AC motor running at a multiple of 60 Hz.
The stirring of the dye cell seemed to be incomplete enough that when
the nitrogen pulse came at a sub-multiple of the stirring rate, a
beat frequency due to excitation of spent dye appeared. A need for
improvement in stirring was quite apparent. This beat frequency
effect had never appeared in the literature.
Voltage drifts during the day were so bad (1 V on regulated
supplies, 10 V on unregulated lines) that severe trigger drift became
apparent. The nitrogen pump laser was very susceptible to trigger
drifts through its high voltage threshold performance. A voltage
change of 1 V on the 120 VAC line can induce a 20 ns drift in the
trigger to firing delay. Using a 164 plug-in on the boxcar, this was
annoying. When a 163 plug-in is used the signal can entirely dis-
appear inside of 5 min time. Therefore, all of the experiments had
to be done at night when stability was much better, but still trouble-
some, however. Generally during each series of measurements the
trigger had to be reset at least once.
An optical trigger was used for subsequent experiments. The
163 plug-in, however, was not used. Indications were that this
arrangement was much more stable with respect to jitter, especially
if triggered from the pump laser. Use of the optical trigger necessi-
tated placing a delay line (V155Z050, Allen Avionics, Mineola, NY)
into the signal line from the photomultiplier to produce a 75 ns delay
necessary to compensate for the internal gate delay of the boxcar.
A 50 0 attenuator was placed before the delay line to produce a
voltage pulse from the current pulse. Indications were that distor-
tion and noise introduction by the delay line were minimal.
Dye Laser Operation
The dye laser was also operated according to manufacturers in-
structions. Dyes were prepared according to Table 2. Adjustments of
the dye cell carriage were made after each new dye cell was introduced
into the beam line. This should not have been necessary if the laser
was functioning correctly. There was some indication that this was
an artifact due to the state of our dye cells. Possibly the anti-
reflective coating had worn off in several years of use (see Appen-
Laser Dye Preparation
Dye Concentration Solvent Range (nm)
M) (10% points)*
DPS Saturated p-dioxane 396-416
1.2 x 10-3
Bis-MSB 1.2 x 10-3 p-dioxane 411-430
7D4MC 10-2 ethanol 440-478
R6G 5 x 10-3 ethanol 568-605
*The 10% points are the wavelengths at which laser output is 10%
of the peak output.
Dye was changed frequently to avoid power loss due to dye deteri-
oration. This was done in one of two ways. The method of exchange
recommended by the manufacturer was to remove the dye via a pipet,
rinse the cell with fresh dye, and refill with fresh dye. This was
moderately successful. Some dyes (7D4MC in particular) seemed to
function better when the dye cell was removed entirely from the dye
head, rinsed thoroughly with alcohol, then fresh dye, and then re-
filled with fresh dye. Peak power with 7D4MC exchanged in this manner
was enhanced at least 2-fold over the more cursory method of exchange.
Bubbling of pure nitrogen for two minutes through the dyes (as sug-
gested by Exciton) did not seem to make a substantial change in dye
longevity. Perhaps nitrogen saturation during preparation and storage
would have helped more.
A gas-flow stabilization system consisting of individual pressure
gauges and rotameters was used for each gas flow. The entire system
was calibrated by means of a linear mass flow meter (ALK-50K, Hastings,
Hampton, Virginia). Each time a flow was measured, the pressure was
readjusted, and the flow and rotameter readings recorded. The com-
bination of a fixed pressure and the rotameter readings led to a very
reproducible and precise system. Premixing of flame gases was done
to avoid background noise due to poor mixing in the chamber of the
burner. The flows passed into the appropriate ports of a nebulizer-
flow chamber (Perkin-Elmer Corporation, Norwalk, Conn.). The same
composition (premixed) passed into a separated shielding flame for
hydrogen-based flames. The fog and premixed combustion gases left
the flow chamber and passed into a capillary, flame-shielded burner
for hydrogen based flames (previously described by Snelleman (59)).
There they were combusted. The flame-shielded flame was surrounded
by an inert gas sheath to avoid background DC emission and noise as
much as possible.
Acetylene based flames were combusted above a capillary burner (Haragu-
chi and Winefordner (60)) supported on the same nebulizer chamber.
Flows for the different flames used are shown in Table 3.
Stock solutions of 1000 Ig/ml for all elements were made from
reagent grade chemicals in deionized water per Parsons et al. (61)
(see Table 4).
Measurement Procedure for Y and n
The necessary criteria for evaluation of quantum efficiency and
total number density via the two different schemes are
1. Saturation must be complete (for plateau method) or nearly
complete (for the slope method); that is, an increase in laser
power must not result in an increase in fluorescence
2. One must measure laser spectral irradiance via measurement
of the following quantities:
a. laser optical bandwidth--via some type of wavelength
Flame Gas Flow Rates
Air 9.88 1/min
Acetylene 1.65 1/min
Hydrogen-Oxygen-Inert Gas Flame
Hydrogen 1.62 1/min
Oxygen 0.81 1/min
Nitrogen or Argon 4.51 1/min
Source of Reagents
Mallinckrodt Chemical Works,
St. Louis, Mo.
Mallinckrodt Chemical Works,
St. Louis, Mo.
Fair Lawn, N.J.
Apache Chemicals, Seward, Ill.
b. laser peak power--via calibrated photosensitive device,
c. laser cross-sectional area--via geometrical considera-
tions; b and c measurements may be combined.
3. Self-absorption must be negligible. (This is automatically
achieved if saturation is complete, as the absorption coef-
ficient then goes to zero.)
4. The source must be a (pseudo-)continuum across the line-
width of absorption; i.e., the wavelength spread of the laser
must be measured and determined not to vary substantially
across the absorption linewidth.
5. Steady state fluorescence must be achieved. The temporal
behavior of the atomic fluorescence pulse must be observed.
Measurement of the saturated value must be during a suitable
6. The rate equation approach must be valid. No coherence
effects may be seen.
7. The fluorescence depth 1 must be found (Figure 8).
In addition, if the possibility of absolute values of Y and nT
is. to be evaluated for a three-level atom, the ratio of resonance to
stokes fluorescence must be found in a flame with a known temperature.
Since the proof of validity of these criteria is somewhat
tedious, it will be assigned to Appendix 1, with the exception of the
first point, completeness of saturation, because this is the basis
of the whole measurement scheme.
For each dye change, alignment of the laser beam with the burner
head was undertaken. Side to side position of the laser was checked
with the flame off, by the use of the aperture-scatter shields and a
Laser beam illumination
Inner (seeded) flan
Flame volume investigated
Figure 8. Fluorescence illumination and collection schematic.
card over the burner. Vertical positioning was accomplished by using
the card to project an image of the laser beam through the first col-
lection lens onto the slit-shaped aperture. After spatial alignment
was completed, the gate delay was peaked up on a scatter signal.
Next, the flame was lit, and analyte was introduced. The dye laser
was wavelength scanned, using the scan control until a substantial
signal was found. Since saturated atoms show a broadened response
to wavelength (see Appendix 1), a calibrated neutral density filter
was set in an accessory optics holder in the laser beam. In this way,
a linear response to laser illumination was ensured. Neutral density
filters of about 2.0 to 3.0 were generally required. When illumina-
tion was in the linear region, the dye laser output could be easily
centered on the wavelength of the atomic transition. When the
neutral density filter was removed, the atoms were saturated.
Generally analyte concentration was chosen to yield a high enough
signal so that a linear portion of the laser fluorescence vs. power
plot could be reached as the laser power was decreased. In some
cases, this involved using a higher concentration at low laser power
to ensure sufficient signal to noise ratio. Linearity of fluorescence
with respect to concentration was always checked when this was done.
The upper limit of concentration was set to avoid problems of vapor-
ization (especially prevalent in the low temperature hydrogen-based
flames used here) and self-absorption. No absorption was noticed,
however, with solutions up to 1000 Pg/ml in concentration, although
the measurement of absorption was complicated by severe shot to shot
peak power variations on the dye laser. In most cases, concentra-
tions were in the 1 to 20 vg/ml range.
The procedure for measurements started with a check for PMT
linearity. Even with the use of a low resistance (high current)
dynode chain and capacitor charge storage for each dynode, the PMT
output current could become non-linear with respect to illumination
when the light levels became high enough. Each time data were taken,
a 0.3 neutral density filter was inserted in front of the mono-
chromator. If the signal did not decrease by 50%, a lower concen-
tration was aspirated, and checked for linearity. Approximately half
of the full scale (1 V) was usable for the 163-S-2 plug-in combina-
tion. This corresponded (with a 50 0 2-fold attenuator) to a 10 mA
peak current. This compared with a limit of 0.3 mA peak current limit
of Olivares (46), for a 220 kn resistor chain with no added capaci-
tance, for a 1P28 PMT. A linear dynamic range of 30-fold higher
(based on current) was obtained in our case.
After the linearity check, the actual saturation curve measure-
ment was performed. This consisted of measuring full scale fluorescence,
blank scatter, and several reduced fluorescence and scatter values
when calibrated neutral density filters were placed in the laser beam
singly or in combination. Scatter corrections for resonance measure-
ments were generally significant until the laser intensity was re-
duced 1 to 2 orders of magnitude by neutral density filters.
These fluorescence measurements were converted to absolute
radiance values following calibration of the optical collection
system, the boxcar, and the recorder used to collect data (see
From the absolute radiance measurements, values were computed
for quantum efficiency and total number density. In some cases rela-
tive data only were computed for different flames.
Noise Power Density
The measurement of noise power density spectra in academic analy-
tical chemistry is generally accomplished via the discrete fast
Fourier transform (DFFT). This approach is necessitated by the lack
of continuous measurement power spectrum analysis instrumentation in
most laboratories. The dedicated instrumentation required by this
method is generally too inflexible to be affordable under tight
budgeting restrictions of this time. On the other hand, the general
availability of digital computers and analog to digital data collec-
tion systems makes discrete (or sampled) data collection fit reason-
ably within the price and flexibility range of most laboratories.
This leaves one with only the DFFT approach to noise power spectra
evaluation possible. This enforced choice leads to a loss of flexi-
bility in one way, and to a gain in another direction. The disadvan-
tages of discretely sampled records for approximating a power spectrum
are very well covered in Blackman and Tukey (54) and include aliasing
(necessitating very well designed filters for spectra with wide dynamic
ranges and wide frequency ranges), intermodulation distortion, and
The flexibility advantages are those which stem from the com-
1. ease of collection,
2. ease of calibration,
3. fast data reduction,
4. ease of presentation,
5. flexibility of handling of data and power spectra.
The primary purpose of noise power spectra collection in this work
was to establish a background for later possible saturated noise power
spectra, which should show the advantages enumerated in the Introduc-
The primary sources of noise for atomic spectroscopy consist of
1. Source related noise
b. whistle or proportional,
c. flicker or 1/f.
2. Background noise-non-source related
b. whistle or proportional,
c. flicker or 1/f.
3. Sample related noise
b. whistle or proportional,
c. flicker or 1/f.
The preliminary experiments presented
types of noises. The first type will
here measured the latter two
be the subject of later experi-
Background noise can be somewhat artificially separated into two
types. The first type is related to the bulk properties of the flame
gases; this type could be exemplified by the emission from the C2
molecule, a product of partially completed combustion. This molecule
has a major emission at 516.5 nm called the Swan band. While it does
not interfere with most atomic transitions, other bands can interfere
with DC and fluorescence measurements with pulsed sources. The
Swan band is sensitive and can be used as an indicator of formation of C2,
and therefore as an indicator of combustion completeness. This
molecule has been used as a measure of success for premixing com-
bustion gases (29). Noise power spectra collected under fixed con-
ditions also provide a useful figure of merit for the evaluation of
success of mixing.
A second type of background emission noise source is one which is
more indicative of external conditions. An example is the OH molecule,
which is strongly indicative of the success of sheathing a flame.
The pertinent (and sensitive) band is centered on 306 nm, and extends
to the red far enough to interfere severely with copper and silver
measurements while a companion band at 285 nm interferes with
lead and magnesium measurements. Noise power spectra at this band
can hopefully lead to better design and construction for flame
Sample related noise can be evaluated as a preliminary to satura-
tion measurements by using atomic emission. In order to collect data
on both nebulization and atomization noises without source noises
interfering, it is necessary to measure the noise involved in atomic
emission, unless saturated measurements lead to an improvement in
signal to noise ratio. An atom must be chosen which has strong emis-
sion, and is not near a background spectral region which would
interfere. Strontium is a likely candidate. Its emission is in a
relatively background-free region, and should therefore yield infor-
mation on nebulization and atomization efficiency noise. Strontium
is quite sensitive to flame temperature, and therefore composition.
It should be a good candidate for the proposed separation of information
from the two sources of noise. In addition, it is among the most
strongly saturated atoms found in this work. The general scheme of
investigation is as follows. The nebulization chambers and nebulizers
of three companies are to be investigated within the guidelines
established above. Conditions to be varied are use of an inert gas
sheath, premix chamber, and removal of the nebulizer. In addition,
data collection will be made in three frequency ranges (0-50 Hz,
0-500 Hz, 0-10,000 Hz).
A general notation system is developed here for the reporting of
the experimental results in this chapter. Ratios will be listed as "R,"
with addition of one or more of the following subscripts for clarifica-
tion of the meaning:
S Slope method of calculating a quantum efficiency or ratio
of quantum efficiencies
E Es method of calculating a quantum efficiency or ratio of
B BF method of calculating a total number density or ratio
ofmotal number densities
I Intercept method of calculating a total number density or
ratio of total number densities
A Measurements made in the argon diluted hydrogen based flame
N Measurements made in the nitrogen diluted hydrogen based
Y Quantum efficiency
nT Total number density
References to the use of these methods can be found in Chapter 2
(Theoretical Considerations). Ratios "R" will always refer to argon
diluted flame values over nitrogen diluted flame values.
The need to use the most versatile and least complicated equipment
possible was recognized at the commencement of this work. One of the
goals of the work was to find the optimal instrumentation for flame
diagnostic measurement via saturated atomic fluorescence, choosing from
the available choices in this laboratory. A series of measurements were
made under set conditions for the comparison of three boxcar average
measurement systems. The first pair of measurements (Figure 9) shows
saturation curves under identical illumination conditions for the
Princeton Applied Research (PAR) 160 boxcar and the PAR 162 mainframe
with the 164 plug-in. The two curves are very nearly identical, though
the scatter on this data was fairly great. (A substandard burner system
was used for these two preliminary curves. It was replaced after these
two experiments with the system described in the text.) The gates on
both boxcars were the same (15 ns). The 10 us stretch feature of the
160 boxcar did not seem to affect the extent of saturation. The second
pair of measurements (Figure 10) shows a comparison of the 160 boxcar
with the 162 mainframe using a 163 plug-in and a S-2 sampling head (75 ps
aperture). In addition to the boxcar change, a change of dye was per-
formed. As expected, the extent of saturation improved. The shorter
gate "viewed" mainly the temporal plateau section of the fluorescence
pulse (see Appendix 1), while the longer gate averaged substantially
unsaturated portions of the pulse with the saturated portion. The higher
power output of the new dye probably accounted for some portion of the
Figure 9. Strontium saturation curves using the PAR 162-164 boxcar
(x) and the PAR 160 boxcar (0). Actual BF values were
different. Plotted values were ratioed to highest signal
observed for each boxcar. A = e = 460.7 nm.
Figure 10. Strontium saturation curves using the PAR 162-163 boxcar
(x) and the PAR 160 boxcar (0). Actual BF values were
different for each boxcar. Each plotted value was ratioed
to the largest value observed for that boxcar.
m = ex = 460.7 nm.
Based on the results of these measurements the rest of the measure-
ments were made with the 162 mainframe 163 plug-in combination to achieve
an accurate representation of the saturation condition.
Flame vs. Inductively Coupled Plasma (ICP)
An acetylene-air flame and an ICP were compared to determine whether
saturation resulted in information about quantum efficiency. Two
parameters were evaluated under equal illumination in the flame and ICP,
namely the saturation curve and the saturation broadening (Appendix 1)
for strontium. The half-power points Ds for the saturation curve (re-
lated to E, see Chapter 2) is inversely proportional to the quantum
efficiency of a flame or plasma. For a given element and excitation
line, here strontium at 460.7 nm, the ICP was expected to have a higher
quantum efficiency than the flame. This was due to the prevailing argon
atmosphere of the ICP, as opposed to the predominantly molecular atmo-
sphere of the flame. The saturation curves (Figure 11) indicated that
the quantum efficiency of strontium in the ICP was slightlyhigher than in the
flame. A second measure of the quantum efficiency of strontium in the
flame or plasma was the saturation broadened excitation profile (Appen-
dix 1 and Reference (62)). A broader profile indicates a higher quantum
efficiency, as shown by:
p(X ) 1/2
6, = b6L (1 + )-- )
exc L S
6e = full width at half maximum (FWHM) of the excitation profile
exc of the atom (nm)
SAL = FWHM of the Lorentzian (undisturbed) profile of the atom
Figure 11. Strontium saturation curves in the air-acetylene flame (D)
and the induction coupled plasma (ICP) (x). The ICP had a
slightly lower E than the flame, and therefore, a slightly
higher quantum efficiency. The log-log scale of this
figure reduces the visualization of this small change.
Se= X e = 460.7 nm.
-I i I I I
-3 -2 -1 0 1 2
log E (relative)
P(A;) = power density of the laser (W/m2
ps (X) = saturation power density (W/m2)
The term p (AX) is inversely proportional to the quantum efficiency (see
Chapter 2). Therefore, the exc depends upon quantum efficiency.
Saturation broadened exc are given below for strontium in the flame
and in the ICP under identical power density conditions.
Atom Cell 6xexc (full power)
The increase in 6Xexc in the ICP also bears out the expected increase in
quantum efficiency in the argon atmosphere.
Hydrogen Based Flames
Hydrogen based, rather than hydrocarbon based, flames were chosen
for the major investigation of saturation because of the expected im-
provement in saturation due to a much less complex and reactive small
molecule (quenching) mixture in the flame gas. Two inert gas diluents
were investigated to find the role they play in quantum efficiency in
flames. Nitrogen was chosen for its obvious place in any air oxidant
flame. Nitrogen is a good diatomic quencher. It is often replaced for
atomic fluorescence purposes by a monatomic species, argon, which is a
very poor excited state quencher. Therefore, it allows a higher quantum
efficiency. This yields a higher sensitivity in application of atomic
fluorescence to analysis.
Saturation curves for the two diluent gases are shown in Figure 12.
The two curves are very close in plateau values and in shape. This was
an indication that the change in flame gas had very little effect in
Figure 12. Strontium saturation curves in two hydrogen based flames.
The flames were diluted with argon (x) and nitrogen (0).
While the plotted values were ratioed to the individual maxi-
mum values, the relative magnitude between the curves was
retained. The value of Ex for the data plotted here was
measured using a round (1 cm diameter) aperture.
x = X = 460.7 nnm.
either the atomization of strontium or the quenching of excited state
strontium atoms. Relative information was derived for Y and nT from
this plot. Relative saturation powers (proportional to Es) were found
from the points at which the fluorescence was 50% of the maximum. The
saturation powers are proportional to 1/Y (see Chapter 2). In addition,
the slopes of the 1/BF vs. 1/E, curves were taken as measures of 1/Y also.
The values for RYS and RYE were computed, and are shown below.
The plateaus of fluorescence were used as relative measures of nT
(see Chapter 2). The intercept values of the 1/BF vs. 1/E, curves were
used also as relative measures of nT. The values for RNI and RNB are
The agreement of the two methods for finding relative Y's and nT's
for the two flames was quite good. The percent relative error for the
Y measurement was 9%. The percent relative error for nT was less than
The errors in the ratios are probably random in nature. The percent
relative standard deviation for these measurements was about 5-10%.
Absolute Y's were measured by means of the E5 method using equation
12 from Chapter 2 under the conditions of Figure 12, and additionally
under the conditions of Figure 13 using the flame-shielded flame and
the small slit mentioned in Appendix 1 placed in front of the photodiode.
The results are shown below. Both flame stoichiometries were the
Figure 13. Strontium saturation curve for the argon diluted flame.
The values plotted were ratioed to the highest value
observed. The Ex values were measured using the slit
shaped aperture mentioned in Appendix 1.
e = Xem = 460.7 nm.
X X XXX
I I I
-2 -1 0 1 2
log Ex (relative)
Figure 14. Composite boxcar scan of fluorescence pulse of
strontium. The gate was halted at several delay
times (DT) to record the saturation curves seen
in Figure 15. x = X = 460.7 nm.
Strontium Y21,in capillary burner 0.50
Strontium Y21,in flame-shielded flame 0.24
As the results indicate, careful attention must be placed on select-
ing a correct and representative aperture. The capillary burner, with
no flame shield, certainly had an edge effect, producing a substantially
poorer quantum efficiency. This was entirely masked, however, by the
inadequacy of the measurement of the E5 caused by using an aperture
larger than the laser beam. As can be seen in equation 12, the quantum
efficiency is inversely proportional to EX, and therefore directly pro-
portional to the measured laser beam cross-sectional area. The approxi-
mately 2-fold increase of quantum efficiency corresponded in this case
to the same decrease in the estimate of Es
Total number density was figured from the slope of the 1/BF vs.
1/Ex curve after calibration of the intercept, and then the slope, under
the conditions of Figure 13. This procedure resulted in nT = 1.3 x 1012
cm-3, a very reasonable nT for 100 ppm Sr in a hydrogen based flame. A
second value, derived from equation 10, yielded a value of nT = 3.1 x 1011
cm-3. This value agreed very well with the prior value. The small
discrepancy could have been due to calibration of the boxcar plug-in
module or the photometer. The boxcar module had been under repair prior
to these experiments, but had not been calibrated. Unfortunately the
equipment required for a check of this value was not available.
A temporal scan of the fluorescence pulse from the photomultiplier
produced an interesting observation for strontium. Several overlapping
scans were taken to define the composite temporal shape shown in
Figure 14. The delay time for the boxcar was set at 4 different points
on the peak, and 4 sets of saturation measurements were taken. These
data are shown in Figure 15. As expected, the leading edge of the pulse
was not saturated. Surprisingly, however, the trailing edge of the pulse
was saturated. Normally, assuming two levels available for laser and
thermal processes, a decrease in signal could not be observed while the
system was saturated. This meant that a third, presumably photon-
induced, process was necessary to explain the phenomenon. Two possi-
bilities existed for the explanation of these data, compound formation
and laser-induced ionization. Compound formation was rejected as a cause
of this effect, as a two body collision would be necessary. Presumably
the only suitable species available in this flame for compound formation
would be oxygen atoms or hydroxyl molecules. Even with favorable orien-
tation of the collisions, they probably could not happen fast enough to
account for the subnanosecond phenomena which were seen here. A more
probable explanation was the absorption of a second photon of wavelength
460.7 nm, which could bring the atom within 0.3 eV of its ionization
potential. A thermal event, not so selective as a compound forming
collision, could then easily ionize the atom. Unfortunately, an experi-
ment to measure electron current generated via laser-induced ionization
with the aid of the nitrogen pumped dye laser was unsuccessful (63).
In private communication with the investigator it was found that poor
results were attributed to laser pulse width considerations. (While
the instantaneous power produced ions, the time spread on the electron
bunch produced only a small signal. The best type of laser for this
experiment has been found to be one with a longer pulse width and
highpower, i.e., high pulse energy.)
Figure 15. Strontium saturation curves collected at several
delay times (DT). The plotted values do not contain
relative magnitude data, which can be inferred from
Figure 14. The plotted values are for DT1 (x),
DT2 (0), DT3 (A), and DT4 (+). Note the substan-
tially less saturated data from the leading edge
data (x). x = x = 460.7 nm.
2B m $A
-1 0 1 2
log E (relative)
Measurement of sodium parameters was hampered by an experimental
inconsistency. In all other measurements, the monochromator bandwidth
was narrow enough to exclude any other radiating wavelengths from the
saturated atoms. For sodium, however, both resonance lines fell within
the monochromator bandpass. Estimation of the nT values from BFmax and
continuum absorption required inclusion of both sodium resonance lines
in the theoretical treatment. The nT value for sodium was arrived at
via two different routes. A value was calculated from the saturation
plateau for the argon diluted flame by the use of equation 10 from
Chapter 2. In addition, a continuum absorption experiment was used to
give an independent measure of nT, as described by DeGalan and Winefordner
(17). The lack of a suitable bandwidth monochromator ( mono 5 x atom)
in a suitable position necessitated substitution of an argon ion pumped
dye laser as a continuum source for this technique. The dye laser out-
put bandwidth was measured by the use of a high-dispersion monochromator
(used in the second order). The optical bandwidth was 0.174 nm. As
the sodium atomic absorption linewidth was only ca. 0.005 nm (61), this
laser was considered a continuum source. The monochromator bandwidth
described in (17) was taken as the laser bandwidth. Measurements of a
(fraction absorbed) were made in both the argon and nitrogen diluted
flames. In addition, saturation curves were taken in both flames
(Figure 16). The nT's calculated from both methods are shown below.
Method Value (1 ppm)
continuum absorption 3.6 x 1010 cm-3
BF 2.9 x 1011 cm-3
Figure 16. Sodium saturation curves for hydrogen based flames.
Relative magnitude for the argon diluted (x) and the
nitrogen diluted (0) flames has been preserved. The
value of Ex used for calculations has been measured
using the slit shaped aperture mentioned in Appendix 1.
Aex = 589.0, Aem = includes both 589.0 nm and 589.6 nm.
x x D
Sx D 0
-2 -1 0 1 2
log E (relative)
The agreement between these values was considered quite good in
light of the fact that the methods are independent, and several calibra-
tions of different types are involved. Calibrations of absolute optical
radiances are considered to be quite difficult and error prone. An
agreement of 2-fold is considered state of the art for inexperienced
The value for RNI was determined, and augmented by an RN derived
from the continuum measurements. However, in the case of sodium in the
nitrogen diluted flame, the saturation was incomplete. Therefore, the
RNB value could not be calculated. Ratio values (RN) are given below.
The valueswere corrected for differing aspiration rates for the two
RN, continuum absorption 0.89
RN, intercepts 0.84
The agreement is quite good.
Since the ratios calculated by these two methods did agree so well,
the intercept ratio was used to calculate the saturation plateau value
for the nitrogen diluted flame. This enabled evaluation of the E.
parameter, which was otherwise unavailable. This allowed a second
means of determining Ry. The values are shown below.
In Figure 17, saturation curves for calcium in HOA and HON flames
are given. The Ry values for the two flames via two methods are given
Figure 17. Calcium saturation curves in hydrogen based flames.
Relative magnitude for the argon diluted (x) and the
nitrogen diluted (0) values has been preserved. The
value of Ex used for calculations was measured using
the round (1 cm diameter) aperture. x = x = 422.7 nm.
-2 -1 0 1 2
log E (relative)
A Y value was calculated from equation 12 for the HOA flame, and is
An nT value calculated from equation 10 is shown below.
nTHOAB 2.6 x 1011 cm-3
Values of RN derived from the two methods are shown below.
Indium was marginally saturated in the stoichiometric air-acetylene
flame (Figure 18), which is a highly quenching flame with respect to the
In contrast with this flame Figure 19 shows the saturation curve in
the HOA flame. The much lower power and much more complete saturation
shown in the resonance (A = A = 410 nm) measurement in this figure
indicated a much higher quantum efficiency. In addition, the observation
of A32 (451 nm) fluorescence followed the expected behavior in showing
the same saturation shape as A31 observation. The 2 3 transition
(451 nm) could also be marginally saturated, as seen in the same figure.
The measurement of power for these curves, however, was not reliable
enough to provide anything but the relative data given above.
A more reliable power measurement was made for the data shown in
Figure 20. The values for Y and nT in the HOA flame given below were
arrived at by the calibration of the 1/BF vs. 1/E. curve.
Figure 18. Indium saturation curve for an air-acetylene
flame. X = e = 410 nm.
0 1 2
log EX (relative)
Figure 19. Indium saturation curves in the argon diluted hydrogen
flame. ex = em = 410 nm (+), Xex = em = 451 nm (0),
Xex = 410 ni em = 451 nm (x). The plotted data do not
retain relative magnitudes of the three curves.
log E (relative)
Figure 20. Indium saturation curve in two hydrogen based flames.
Relative magnitude of values for the argon diluted (x)
and nitrogen diluted (0) flames has been retained.
nTHOA 2.0 x 1010 cm-3 (1 jg/ml)
An Ry value was available only from the slope method. The plateaus of
the two flames were not developed enough to find Es values.
An RN value was only available from the intercept method, and is shown
Noise Power Density
Noise power density (W/Hz1/2) measurements preliminary to satura-
tion noise power density were made under the conditions shown in Figures
21-26. Since the measurement of saturation noise power density will
only be possible in the lower frequency ranges (up to about 50 Hz, due
to repetition rate limitations on pulsed lasers), the low frequency
range (0-50 Hz) data collectedwere chosen for these figures. The
higher collection rate (higher frequency spread) data had no interesting
features except for harmonics of 60 Hz, and a band of discrete fre-
quencies (centered on 6 kHz) introduced through the electronics power
Figure 21. Noise power spectrum of the OH region. Slit
width = 500 pm; A = 306 nm, no premixing of com-
bustion gases, viewing height = 2 cm above capil-
laries, Perkin-Elmer nebulizer and chamber, acqui-
sition rate 100 Hz, low pass filter 3 dB point-30
Hz, DC current = 0.57 pA, no sheath, stoichiometric
U I --
Figure 22. Noise power spectrum of the OH region with sheath,
DC current = 0.8 x 10-9 A, all other conditions
same as Figure 21.
0 25 50
Figure 23. Noise power spectrum of the C2 region. = 516.5 nm,
slit width = 250 pm, DC current = 0.38 x 10- A, no
sheath, other conditions as Figure 21, the zeroeth
harmonic has been dropped (to expand the scale).
0 25 50
Figure 24. Noise power spectrum of the C2 region. With sheath,
DC current = 0.31 x 10-7 A, slit width = 500 um,
other conditions as Figure 23.