THE THEORY AND APPLICATION OF
A CONTINUOUS SOURCE IN ATOMIC
ABSORPTION FLAME SPECTROMETRY
WILLIAM WALTER McGEE III
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FU L FILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
I wish to thank the members of my committee;
Dr. L. A. Arnold, Dr. A. P. Black, Dr. W. S. Brey, and
Dr. R. C. Stoufer, for their help and advice.
I want to especially thank my research director and
committee chairman, Dr. J. D. Winefordner for the help and
encouragement he has given me not only in the writing of
this dissertation, but throughout my entire stay at the
University of Florida.
Also, a very grateful thank you goes to Dr. L.
de Galan. Dr. de Galan was the sounding board for many of
my ideas. His many suggestions were of invaluable assist-
ance to me.
And finally, my deepest thanks go to my wife, Mary
Ann, to whom I dedicate this dissertation. Without her
help this study would not have been possible.
TABLE OF CONTENTS
ACINOWLEDGMENTS ... . . . . . . . . ii
LIST OF TABLES. . . . . . . . . . v
LIST OF FIGURES . . . . . . . . . vi
INTRODUCTION. . . . . . . . .. 1
EXPERIMENTAL MEASUREMENTS DESIGNED TO ESTABLISH THE
RELATIONSHIP BETWEEN THE MEASURED SIGNAL AND THE
ATOMIC CONCENTRATION WHEN USING A CONTINUOUS
SOURCE OF RADIATION. . . . .. . . .. 5
Introduction . . . . . . . . .. 5
Theory . . . . . . . . . 6
Atomic absorption measurement with a line
source . . . . . . . 7
Atomic absorption measurement with a
continuous source. . . . . . . 10
A discussion of the curves of growth, total
absorption (AT), and the damping constant
and present means of calculation . . . 14
Experimental Conditions. . . . . . 22
Description of the experimental conditions. . 24
Experimental Results . . . . . . .. 36
Discussion . . . . . . . . . 47
Discussion of errors. . . . . . . 47
Comparison of results with literature data. . 49
EXPERIMENTAL MEASUREMENTS DESIGNED TO EXTEND
ANALYTICAL APPLICATIONS OF THE CONTINUOUS SOURCE 54
Introduction . . ... . . . . 54
Discussion . .
FUTURE WORK. . .
CONCLUSIONS. . .
APPENDIX I . . .
LITERATURE CITED .
a. . e * e e e * e e e
* * 0 ;
* C C C C
C C C *
. . .
. . .
. . .
. . .
. . .
S .* C C C C
S a .
* d 0 C
* C C * .
. . . .
. . ...
. . . .
LIST OF TABLES
1. Specific Components Used in Experimental
System for Measurement of the a Parameter. . 27
2. Values of Parameters Dependent Upon Flame Type 29
3. Values of Parameters Dependent Upon Element. .. 34
4. Values of Calculated Spectral and Flame
Compositional Parameters ... . ... 44
5. Values of Spectral and Flame Compositional
Parameters Taken from Literature . . . . 50
6. Specific Components Used in Experimental System.
for Measuring Limits of Detection. . . . 58
7. Experimental Conditions and Limits of
Detection Obtained for 21 Elements Using the
Continuous Source. . . . . . . . 61
LIST OF FIGURES
1. Theoretical curves of growth . . . ... 13
2. Instrumental set-up for the measurement of the
a parameter. . . . . . . . . 26
3. Experimental curves of growth for Zn, Cd, and
Mg . . . . . . . 38
4. Experimental curves of growth for Ag, Cu, and
Na . . . . . . . . . 40
5. Illustration of test for location of high
density asymptote and a parameter. . . . 43
6. Experimental set-up for determining limits of
detection with a continuous source . . . 57
7. Analytical absorbance curves for Ag, Cu, and
Na obtained using a continuous source. . .. 68
Atomic absorption flame spectrometry occupies a
unique position among analytical absorption (optical)
techniques in that it is the only one which does not use
a continuous source of radiation (e.g., tungsten or xenon
lamp) for making measurements. Instead, a line source (e.g.,
hollow-cathode discharge tube or electrodeless discharge
tube) emitting very intense radiation over a very narrow
range of frequencies is used.
The method of atomic absorption spectrometry was
used in astrophysics for investigation of the composition
of stellar bodies. Walsh (41), in 1955, indicated that
atomic absorption spectrometry should be a useful technique
if a flame cell were used to atomize the sample and a line
source was used to excite the atomic vapor. However, in
his classical paper, Walsh not only surveyed the technique
and equipment needed to perform analyses using a line
source, but also indicated that a continuous source might
have analytical use. He decided that because a mono-
chromator of very high resolution would be needed to resolve
the atomic lines under investigation when using a continuous
source, a line source of radiation would be easier and much
less expensive to use. In addition, in this paper he pre-
sented equations relating the decrease in the peak intensity
of the radiation emitted from the line source to the
concentration of solution aspirated into the flame. He
proved that the measured absorbance when using line sources
in atomic absorption analysis was linear with atomic
concentration of the sample vapor in the flame gases, and
that the technique should be extremely sensitivic, chat is,
limits of detectabilities in the ppm range. Because of
these interesting and useful results little work on sources
other than line sources was carried out during the next few
years. If Walsh had not obtained such excellent results
and had not deemphasized the use of a continuous source,
atomic absorption analysis might have easily developed
along such lines. Astrophysicists such as Ladenburg and
Reiche, and van der Held and Ornstein (28,37) in classical
experiments measured the amount of energy removed from a
continuous source of radiation to determine many classical
atom.parameters. Expressions for this energy parameter
(called total absorption) are well established (28), and in
addition to being directly proportional to the absolute
atom concentration for dilute atomic gases present in the
flame, are independent of the resolving power of the spectro-
Developments in atomic absorption analysis using a
line source have been numerous and varied. Advances con-
cerning the optical set-up, burner design, flame gases used,
and,most important, in analytical applications have tended
to make atomic absorption flame spectrometry a more useful
analytical technique. Relatively few advances in the line
sources have been made, and those that have been made, are
aimed at incorporating more than one metal into a line
source to enable multi-element analyses to be performed and
to increase the intensity of these sources. A few but
significant number of developments have occurred in atomic
absorption analysis using a continuous source. Gibson,
Grossman, and Cooke (16), in 1962, investigated the possi-
bility of using a continuous source for analytical measure-
ments. They concluded that a medium resolution monochromator,
when combined with a scale expansion technique, would provide
sensitivity comparable with the use of a line source.
Recently Ivanov and Kozireva (22), and especially Fassel and
co-workers (11,12) have demonstrated that excellent sensi-
tivities are possible when using a continuous source.
With this progress in mind, the aims of this dis-
sertation are then:
(1) To establish the mathematical relationship be-
tween atom concentration and measured signal when using
a continuous source; verification in the form of working
curves for twelve elements will be given. This working
curve is unique, not only in its shape which allows the
analyst to detect any deviations from linearity quickly,
but in the information concerning fundamental atom para-
meters which can be obtained from it; such parameters as
the a parameter or damping constant (ratio of Doppler and
collision half-intensity widths), absolute atom concentra-
tion, atom formation efficiency factor, and the total half-
intensity width of the line for the atom of concern can be
obtained; a discussion of the equipment and procedures used
to measure data for preparing the working curves as well as
for associated parameters will be given.
(2) To continue the development of the quantitative
aspects of atomic absorption using a continuous source;
limits of detection for twenty-one elements using an experi-
mental set-up which provides maximum versatility for per-
forming analyses will be given.
EXPERIMENTAL MEASUREMENTS DESIGNED TO ESTABLISH THE
RELATIONSHIP BETWEEN THE MEASURED SIGNAL AND THE
ATOMIC CONCENTRATION WHEN USING A CONTINUOUS
SOURCE OF RADIATION
When performing quantitative atomic absorption
measurements using a line source of radiation, the working
curve prepared is usually a plot of absorbance versus solu-
tion concentration. If a line source with a half-intensity
width less than the absorption line half-width is used, the
signal measured corresponds to variation in the maximum
absorption coefficient at the central wavelength of interest.
Because absorbance is proportional to the atomic absorption
coefficient, the working curve is linear only as long as the
absorption coefficient is proportional to solution concen-
When a continuous source of radiation is used, the
wavelength interval of continuous radiation isolated by the
monochromator depends upon the resolving power of the mono-
chromator. For the case of the medium resolution mono-
chromator used in this study, the spectral band width will
be much larger than the line width of the absorbing atoms
in the flame. Therefore, the measured absorption signal
will correspond to a function of the absorption coefficient
integrated over the spectral band width of the monochromator
and will always be proportional to concentration.
As seen from this brief comparison, certain basic
differences exist between the absorption signal measured
using a line source and a continuous source. Because of
these differences, this investigation was carried out to
determine if a more accurate and useful relationship (than
absorbance) existed between solution concentration and
measured signal when using a continuous source of radiation.
In order to indicate the similarities and the dif-
ferences in the use of a continuous source and a line source
in atomic absorption spectrometry, it is assumed that the
basic equipment consisting of a flame cell, a monochromator,
and an electrometer-read-out system are used in both cases.
Consequently, the instrumental proportionality factor Z,
relating read-out voltage to intensity of radiation reach-
ing the entrance slit of the monochromator, will be the
same in both cases.
All comparisons are made with respect to the type
of monochromator used in this study: a medium resolution
monochromator capable of isolating a single line, but not
capable of resolving the spectral line profile. This means
that the spectral band width of the monochromator s, is
considerably larger than the absorption line half-width
A T. In turn, the absorption line half-width is assumed to
be larger than the half-width of the line emitted by the
line source, AN.
Atomic absorption measurement with a line source
The photodetector signal due to radiation emitted by
the line source passing through the flame gases with only
blank solution being aspirated is given by
0I = / 7 (1)
where J, is the intensity per unit of wavelength. The
signal from the radiation passing through the flame with
sample solution being aspirated is given by
S Z-L (2)
where L is the path length in the flame.
In general, the atomic absorption coefficient k ,
is a complex function of wavelength and is described by the
following set of formulas (30):
\ = 4 Sc1,v ) (3)
ko ^7 2,.e1 )" 0 N/ (Li.)
o ~2- Al Ao NYC
Ak- = c T.,2- (5)
D c M
a = ay1 J
S(c~,v) r 2 V-. (6)
where e and m are the charge and mass of the electron,
c is the velocity of light,
Xo is the wavelength at the line center,
f is the oscillator strength,
N is the concentration of absorbing atoms in the flame,
T is the flame temperature,
R is the gas constant,
M is the atomic mass,
AAo is the Doppler half-intensity width of the absorption
a is the damping constant, given approximately by L
AkL is the collision (Lorentz) half-width of the
and v is equal to 2. (-
-'-jio o (*a v) describes the variation of i. over th
..:.. line., oweve, the integration i
u -...;i. ,::-o;ncOs over the emission line width,- k., nly c
;:ich is generally ,much smaller than the absorption line
.ith. If woe replace 1 with
.huo 1;: 'r<:ksents the average absorption coefficient over
-" -o;rce line iddth A, and v is an average over the
j from zero to AXNT, then the absorbance is given b;"
;,, n. = (8)
a for .al. concentrations, when AL ( / ,
.. - < (>v)L -A0
S llo fo uaions 8 and 9 that if a line sour-c
o .-. i'n ato.-ic absorption spectrometry, such a plot
.ho.i pro-duce a straight line as long as the atom concen-
ui ir th ..e .fla.de is proportional to its concentration
in so..utic W WLon a very narrow line. source is usco. (i.c
.\ K K v t' 0), this straight line should extonl
c.vr an e ;-trely large concentration range; naeiiely, up to
tle poi;t whore rsonanct- broadening causes the damping
c;.stant to change (0) (for molar solution concentration).
Normally, linearity over an extended range of concentration
is not found. Deviations can be attributed to:
(1) Ionization and compound formation of the atom
(2) A nonlinear relationship between atom concentra-
tion in the flame (N), and solution concentration (C).
(3) Variations in aspiration efficiency (43).
(4) The source line width is no longer negligible
when compared to the absorption line width.
Rubeska and Svoboda (34), and Vickers, Remington and
Winefordner (40) have discussed other sources of deviations
found in atomic absorption flame spectrometry.
Atomic absorption measurement with a continuous source
With a line source, the wavelength region of interest
is determined by the width of the source line; whereas, for
the case of a continuous source, this region is determined
by the spectral band width of the monochromator, s. Over
this small range of frequencies the intensity of the source
is essentially constant; therefore, the blank photo-
detector signal is given by
I = Z / J; = J s (10)
The photodetector signal due to radiation passing through
the flame with sample solution being aspirated is then
Ic = 2 /Bc-7A L E J^ALi (I)
therefore, the fraction of intensity absorbed in the flame
is given by
Odc -C C r= d (12)
I; s s
where the integration limits are 0 to o because -6c spectral
band width is assumed to be much larger than the width of the
absorption line and where AT is known as the total absorption
(28,50,36). Curves representing AT as a function of the
concentration N, depend on the value of the damping constant
a, and are referred to as curves of growth (21,28930,38).
AT, a, and curves of growth will be discussed in the next
section. An important characteristic of A which can be
seen from the curve of growth is, AT, is proportional to N
at low concentrations and proportional to N at very high
concentrations. Where k << I for the low concentration
region, when Loppler broadening is assumed predominant,
equation 12 becomes
ac. 4ALzd- o4 _ _) L (13)
c S 21~122 T / *
which is independent of the absorption line profile. This
can be seen from the similarity of all the curves of growth
(see Figure 1) for the low concentration region. The
0-0 \ \"
general expression for the absorbance A, is then
A, () -) (14)
As can be seen from this relationship, absorbance is a
complicated function of the concentration of the element in
the flame. According to equation 14, the absorbance is
proportional to the total absorption AT, only for small
values of the ratio On the other hand, the fraction
of intensity absorbed in the flame ac is always pro-
portional to the total absorption, but will be linear only
over the linear portion of the curve of growth. De Galan
(7) has shown that both types of working curves absorbancee
or relative absorption versus concentration),when prepared
from measurements made with a continuous source, begin to
slope off when more than 10 per cent of the radiation is
absorbed in the flame. At lower values of relative absorp-
tion, linear expansion of AT in equation 14 is valid.
Therefore, both types of working curves will show a similar
linear range over the low concentration range.
A discussion of the curves of growth, total absorption (AT),
and the damping constant and present means of calculation
To establish the validity and to provide a better
understanding of these parameters, a discussion of the
general nature and present means of calculation is presented
in this section. The discussion of the value and use of
these parameters in this study will be presented in a later
Total absorption (A ), as defined by Ladenburg and
Reiche (28), is the fraction of energy removed over the wave-
length interval of the spectral line from a continuous
spectrum by atoms in an absorbing column of gas. It is
given mathematically by the expression
AT f=I-(-A )--, (15)
Minkowski (29), in a series of experiments, showed this para-
meter to be independent of the slit width of the monochromator
within the limits of experimental error.
The curve of AT plotted against the absolute number
of absorbing atoms is called a theoretical curve of growth.
Classically, this plot was used in experiments to determine
many fundamental parameters concerning the atom; some ex-
amples are the number of dispersion electrons associated
with the emission and absorption of a particular line, and
the natural lifetime of an atom. As seen from the theoreti-
cal curves of growth in Figure 1, each curve is characterized
by a linear range where AT is proportional to N in the low
concentration region; the curve gradually slopes off in the
high concentration range to become proportional to N .
While each curve is characterized by slopes of 1 and 1/2,
exact position of each curve is dependent upon the damping
constant or a parameter. This constant was previously de-
fined in equation 6 as
a = (6a)
Until recently, theoretical corves of growth were available
for only a few scattered a parameters (33). Recently
van Trigt, Hollander, and Alkemade (38) provided useful
calculations for preparing theoretical curves of growth for
a parameter values ranging from 0 to 10, with the values
from 0 to 1 given in 0.1 units. The latter range will be
of great use to flame spectroscopists because of conditions
found in the flame.
Interest in the a parameter stemmed from the funda-
mental data which can be calculated from it. Information
concerning the interaction of perturbing and emitting
particles (i.e., the optical cross-section), the resultant
wavelength distribution of emitted intensity (i.e., the
line profile), and in combination with a curve of growth,
the absolute atom concentrations, and efficiency of atom
formation in (e.g.) flames, are just a few of the more
From now on, all formulas will be given in terms of
frequency units. Those formulas previously calculated in
terms of wavelength units can be converted by the following
expressions: d = C/% and A) = Cc/X)AA.
important parameters which can be calculated. Experimental-
ly, a is a difficult parameter to determine; therefore, most
measurements are made on the emission intensity from flames
of low burning velocity (e.g., acetylene-air). In general,
few a parameters have been measured (even in these flames)
because of the restrictions placed on the types of flames
and flame conditions which can be used to prepare curves of
growth in emission from which they are determined. The re-
quirements for the preparation of the curves of growth in
emission are (21):
(1) A homogeneous distribution of flame gas and metal
(2) A homogeneous distribution of temperature over
the region of observation.
(3) Complete elimination of self-reversal.
(4) Complete elimination of ionization.
(5) Linearity of aspiration and photodetector-ampli-
For these reasons, no measurements have been made in flames
of high burning velocity (e.g., acetylene-oxygen and
hydrogen-oxygen) which are of use in analytical flame
Two methods based on measurement of the variation of
intensity of radiation emitted by atoms in the flame for
preparing curves of growth are available for accurate de-
termination of the a parameter. The newest method developed
by van Trigt, Hollander, and Alkemade (38) utilizes experi-
mental and theoretical curves of growth in combination with
an experimental duplication curve to determine the a para-
meter and absolute atom concentration directly. The
readers attention is called to this article for a more de-
tailed discussion of the method used.
The older and more widely known method is the one
developed by Hinnov and Kohn (18,19). In their classical
papers, the authors presented equations for the high and
low density asymptotes of the curve of growth which take
into account their dependence on the a parameter. They
showed that the absolute ordinate of the point of inter-
section of the asymptotes (see Figures 3 and 4) is
log AT 2 = (16)
Thus from a knowledge of the temperature of the flame and a
series of experimentally measured AT values, the a parameter
could immediately be determined. The AT values necessary
to prepare the experimental curve of growth were determined
from the measured variation of intensity from atoms in a
series of metal solutions aspirated into a flame of low
In more recent papers by Behmenburg and Kohn (1,2),
the effects of hyperfine structure due to nuclear spin and
isotopic shift on the value of the ordinate were studied.
They found the ordinate value equal to log na, where n is a
constant, the.value of which is dependent upon the atom under
burning velocity. The measured emission intensity was con-
verted to total intensity (absorption) by comparison with
the radiation from a standard lamp.
The curve of growth prepared in this manner is an
experimental curve of growth because the abscissa of this
plot is solution concentration. The intersection point of
the high and low density asymptotes is also necessary for
converting experimental solution concentration to absolute
atom concentration in the flame gases.
K = 4- (17)
allowed calculation of the proportionality constant K which
related the theoretical and experimental curves of growth to
CI, which is the solution concentration at the intersection
point. Therefore, at any solution concentration C, the
concentration of atoms in the flame capable of absorbing the
line under consideration was given by
N = eC AdJ KC (18)
j ZxL >fL -22T-
This value of Nj represents the population in energy level j,
and can be converted to the total number of free atoms of
The proportionality constant K is labeled Q in
Hinnov and Kohn's (18,19) work.
the element by using the Boltzmann distribution relation-
Once the a parameter and absolute atom concentration
have been determined, many other parameters of interest can
The efficiency of atom formation PI, at the inter-
section point is calculated from the following relation-
S -N I (19)
I Total x lO2 (c
where NI, is the absolute atom concentration at the inter-
Total, is the total number of atoms present regard-
less of form;
S is the solution flow rate in cc./min.;
n298 and nT, are the total number of moles of flame
gas products present at 2980 K and at the flame temperature
Q is the flow rate of unburned gases in cc./min.;
S, is the aspiration efficiency as defined by
Winefordner, Mansfield, and Vickers (43).
As defined, PI, is a factor used to account for
atomic losses due to ionization and dissociation of the salt
introduced and compounds formed between the atomic species
of interest and various gas products.
The collision (Lorentz) half-intensity width AL
(30,36) is given by
and the total half-intensity width of the line Av) (30,56)
is given by
A = (+ C& ) ) (21)
Normally, natural broadening and resonance broadening will
not be significant when compared to Doppler and collisional
broadening. Natural broadening is on the order of 10 A,
resonance broadening does not become appreciable until
molar solutions are used, Doppler and collisional broadening
are on the order of 101 102 A. For this reason natural
and resonance broadening will not be considered in this
The effective cross-section for collision (Lorentz)
broadening (30,36) -L' is given by
2 A JL
o- L ,`- ,5 10 P+ > (22)
where P,, is the pressure on the system during measurement
(normally taken to be 760 mm);
M, is the atomic weight of the absorbing species; and
Ma, is an effective molecular weight of the flame gas
species including water which causes collision
This parameter (p-L) is of interest because it can be used
to indicate the nature and type of interaction between per-
turbing and emitting species.
As can be seen from the theory developed for the
signal measured using a continuous source, the absorption
method can be used to determine total absorption (Am) di-
rectly.* A valid test of the relationship would be to
prepare working curves (experimental curves of growth), to
check for correct shape (i.e., the correct slopes), and to
calculate reasonable a parameters. The method could then
be used to calculate other parameters such as atom formation
efficiency, total line width, and effective collisional
cross section; however, these values could not be used as
conclusive proof of the method because they represent the
first to be calculated in this type of flame.
The preparation of the curves of growth from absorp-
tion measurements also offered other advantages which
A,/s is obtained from the experimental curve and is
converted to AT by multiplying by the spectral band width s.
overcome some of the restrictions inherent in the emission
method.' Some of these advantages are:
(1) Self-reversal will not influence the value of a
measured absorption signal.
(2) For those atoms which do not form compounds or
ionize readily, small variations in temperature will have
little effect on absolute atom concentration (8).
(3) Atoms which emit radiation too weak to be detec-
ted reliably can still be measured using an absorption
(4) The absolute number of atoms producing the ab-
sorption is obtained directly. Generally the total number
of atoms producing absorption is approximately the same as
the total number of atoms in all states. However, if this
situation is not valid, then the atomic concentration calcu-
lated from equation 18 can be converted to the total number
of atoms in all states by use of the Boltzmann equation. An
example of this case is Ni (3414 A) where the absorption
transition originates from a low lying energy level.
In this study, curves of growth will be prepared for
twelve elements from measurements of the variation of the
fraction of intensity absorbed with concentration of
aspirated solution using a total-consumption aspirator burner
for five flames of interest in analytical flame spectrometry.
The choice of hydrogen and acetylene as a fuel and
oxygen as the oxidant, and their respective flow rates to
produce stoichiometric or fuel-rich flames was determined
by their relative usefulness for analytical flame spectrom-
etry. The argon-hydrogen-entrained air (Ar/H2-E.A.) flame
was added to this list because of recent successes with
this flame in this laboratory (39,46).
With respect to the requirements for preparing curves
of growth in emission (referred to earlier), those not
covered by the inherent advantages of the absorption method
will be met by:
(1) Reducing ionization by addition of an ionization
buffer to solutions of those metals with ionization poten-
tials below 5 e.v.
(2) Reducing the variation of temperature and compo-
sition in the flame region viewed by judicious choice of
the location and size of the flame region viewed.
The a parameters and absolute atom concentrations
will be calculated from the curve of growth by the method
developed by Hinnov and Kohn (18,19).
Description of the experimental conditions
(a) Conditions for measurement of the a parameter.
The instrumental set-up used for making the flame
absorption measurements is shown in schematic form in Figure
2. The specific components used in the experimental set-up
are given in Table 1, and the operating conditions for the
measurements are given in Table 2.
SPECIFIC COMPONENTS USED IN EXPERIMENTAL SYSTEM FOR
MEASUREMENT OF THE a PARAMETER
1. Spectral Continuum
4. Detector-Power Supply
Xenon arc, 150-watt (Englehard,
Hanovia, Newark, N. J.) powered
by a regulated a.c. supply
(Sola Electric Co., Chicago,
Single pass, chopped at 320
cps. Chopper consists of a
disc with eleven equally
spaced holes rotated by an
1800 rpm synchronous motor
(Bodine Electric Co., Chicago,
Ill.). L1 and L2 are quartz
Jarrell-Ash, Model 82000, 0.5
meter Ebert Mount, grating
spectrometer (Jarrell-Ash Co.,
Waltham, Mass.). Grating is
ruled for 1250 lines/mm. and
is blazed at 5000 A (for Zn,
Cd, and Mg a 3000 A blazed
grating was used).
EMI 9558 QBophotomultiplier
(1650-9000 A). Regulated d.c.
power supply (No. 418A, Fluke
Mfg. Co., Seattle, Wash.).
A.C. Photomultiplier output
signal was fed into an O.R.N.L.
No. 7 a.c. amplifier turned to
320 cps. The plate and fila-
ment voltages were taken from
a dual power supply (No. R
100 B, Philbrick Researches,
Boston, Mass.). The a.c.
signal was rectified and the
d.c. signal was recorded on a
potentiometric recorder (Model
TR, E. H. Sargent and Co.,
Chicago, Ill.). A scale ex-
pansion technique was used
6.' Gas Pressure and Flow
The gas flow is regulated by
a Beckman High Precision
regulation unit (No. 9220,
Beckman Industries, Fullerton,
Calif.). The resultant flow
is monitored by rotammeters
(No. 4-15-2 Ace Glass Co.,
Inc., Vineland, N. J.).
Total-Consumption Type (Carl
Zeiss Inc., New York, N. Y.).
" 0 00 CT.
S CM4 CM 1 r-4 -4
,O * H
0 C- N4
CM4 CM4 CA ..
O0 0 4 M
NM C4 p- C4 CM(
a M < (u *
Stock solutions of 20,000 ppm for each of the twelve
elements were prepared by dissolving the appropriate salt
(chloride or nitrate) in aqueous or the metal in acidified-
aqueous solution. More dilute solutions were made by suc-
cessive dilution of the stock solution. The ionization
buffer salt (CsNO3 or KNO ) was added to each dilution of
the stock solution for the alkali metals in quantities as
prescribed by Hoffman and Kohn (20) (i.e., 2 x 10-)_ X CO3
for Li and Na, 2.5 x 10l i K2 CO3 for Rb, and 1 x 10-1
CsNO for K).
The flow rates for the gases used to produce the
flames used in this study are given in Table 2. The region
of the flame chosen for investigation in the total-consump-
tion aspirator burner was, in all cases, located a short
distance above the inner cone in the interconal region of
the flame (see Table 2 for respective heights). This region
was chosen to insure that equilibrium statistics controlled
the concentration of the species present in this region (17).
The entrance optics and slit height of the experi-
mental set-up were used to control the size of the region
viewed and to reduce it to a minimum according to the
restrictions discussed by Hollander (21).
The temperature of the interconal region of the flame
was measured by the line-reversal method for the Ar/H2-E.A.
and H2/02 flames. The line-reversal method of Fery (15) as
discussed by Gaydon and Wolfhard (15), is based on the
assumption that, on introducing metal atoms in the flame,
statistical equilibrium is established between the electronic
degrees of freedom of the metal atoms and flame gases. The
metal atoms thus emit and absorb their spectral lines as
thermal radiators. The line-reversal method is based on the
following principle: If a blackbody is placed behind a flame
containing sodium atoms which are emitting the yj,., sodium
D doublet, and a spectrometer is aligned to see both the
blackbody and flame simultaneously, then there will be some
temperature of the blackbody at which its brightness for the
specific wavelength region equals the brightness of the
blackbody transmitted through the flame, plus the brightness
of the D lines from the flame. At this temperature, only a
continuous spectrum of the blackbody will be recorded. At a
lower temperature of the blackbody, the sodium line will
appear in emission superimposed upon the continuous back-
ground; and at a higher temperature, the sodium line will
appear in absorption. The brightness temperature of the
standard tungsten lamp was calculated as a function of lamp
current from data for the radiance at 35 amperes supplied as
a calibration with the lamp and the emissivity of tungsten
at 5890 A calculated by de Vos (10).
The two-line method was used for the hotter C2H2/02
flames and also for the H2/02 flames in order to compare
the results of the two methods.
The Ornstein two-line method (3) involved the
measurement of the intensities of two iron lines; Fe 3734.87
A, and Fe 3737.14 A, from which the temperatures can be
calculated by the following formula:
log log A + (22)
Ib ASb X-
where Ia and Ib are the relative measured intensities at
ha = 5734.87 A and Xb = 3737.14 A lines, respectively; Aga
and Ag, represent the product of the transition probability
and statistical weight for the two lines (the ratio of Aga
to Agb was calculated from data given by Crosswhite (6)
and found to be equal to 11.'6); Ea and b are the energy
of the transition corresponding to the line; k is the
Boltzmann constant; and T is the flame temperature of
Crosswhite (6), and Broida and Lalos (4), have dis-
cussed some of the requirements for use of this method. A
few more of the more important requirements are:
(1) Both lines chosen must have negligible self-
absorption and self-reversal or at least have them of equal
(2) Lines must be close enough for rapid but accurate
(3) Both lines must be close enough together to
insure that the photocathode has essentially the same
response to both lines or a suitable correction must be made,
The agreement of the temperatures obtained by the two metho&G
was excellent and within experimental error. The experi-
mental set-up used for the temperature measurements was the
same as shown in Figure 2, except that the continuous source
was replaced by a standard tungsten lamp. The temperatures
determined for these regions are given in Table 2.
The spectral band width o h,;Lu onochromavor (see
Table 3) was determined by scanning the 3650.15 3654.83 A,
and 5769.59 5790.65 A lines from a low pressure Hg dis-
charge lamp. The relative recorded distance between the
two lines was converted to a known distance, and this scale
was used to measure the half-intensity widths of the lines
involved. The relative standard deviation in measuring the
spectral band width was less than 10 per cent for all slit
widths used. This method is preferred to the scanning of a
single line (e.g., the iron lines used in the temperature
measurements) because of the difficulty encountered in
reproducing the exact scanning speed. The experimental set-
up used for this measurement was the same as that shown in
Figure 2, except the total-consumption aspirator-burner was
replaced by the Hg lamp.
Because metal concentrations ranged from 1 104 ppm,
the output of the detector-amplifier system was checked for
linearity of response by use of neutral density filters of
VALUES OF PARAMETERS DEPENDENT UPON ELEXNT
Band Width (A)
known transmission and found to be satisfactory. The rela-
tive error due to deviations from linearity of aspiration
was also checked and found to be less than 3 per cent for
all solutions with metal concentration less than 1.5 : 101
b. Conditions for measurement of the atom efficiency
The aspiration efficiency C is a meaasU u th.
ability of the flame to remove the solvent from the salt
solution aspirated into the flame (31,43). As such, it is
a difficult quantity to measure because of its strong de-
pendence upon height in the flame and flame gas composition.
Parsons and Winefordner (51) have suggested a method of
measuring ( based on a comparison of the measurements of
the light reflected from water droplets aspirated from the
burner, with and without the flame. Using their method,
the aspiration efficiencies for the five flames under
investigation for the specific region of interest were
measured (see Table 2).
The value of the path length of radiation through the
flame (L) for a total consumption burner (see Table 2) is
difficult to measure due to the irregular shape of the turbu-
lent flame. The value of L can be approximated by measuring
the length of the path over which continuum radiation is
scattered when solution is aspirated into the burner (flame
The solution flow rate 1 was determined from tho
time required for aspiration of a known volume of salt
solution (see Table 2). The standard deviation in all cases
was less than 1 per cent.
The value of Q, the flow rate of unburned gases, can
be calculated from a simple conversion of the flow rates of
gases used to produce the flame (see Table 2).
The value of the paramoor n2 /n which corr/ts
for the expansion of the total number of moles of flame
gases for the temperatures indicated was calculated and
found to be equal to the value determined by Winefordner and
Vickers (44,45). The value of this ratio is 0.83.
Curves of growth were prepared from absorption
measurements for each element in each of the five flames.
Each point on the curve represented the average value of
nine determinations with a relative standard deviation of
less than 1 per cent. Examples of several curves are given
in Figures 3 and 4, for Zn in C2H2/02 (stoichiometric),
cadmium in H2/02 (fuel-rich), and Mg, Na, Cu, and Ag in
Ar/H2-E.A. From the value of the total absorption A., taken
from the intersection point of the asymptotes of a specific
experimental curve of growth and the Doppler half-intensity
width calculated from flame temperatures, the a parameter
Fig. 3.-Experimental curves of growth for Zn, Cd,
One Zinc, 2138 A line, in C2H2/02 ST.
Two Cadmium, 2288 A line, in H2/02 F.R.
Three Magnesium, 2852 A line, in Ar/H2-E.A.
(In experimental and theoretical coordinates.
Note, the right ordinate is AT'f-/A o
where AT and AiD are in units of sec-1
However, the same ordinate values would
result if ATr l/X, were plotted, where
AT and AD are in wavelength units.)
Fig. t.-Experimental curves of growth for Ag, Cu,
(In experimental and theoretical coordi-
One Sodium, 5890 A line, in Ar/H2-E.A.
Two Copper, 3247 A line, in Ar/H2-E.A.
Three Silver, 281 A line, in / .A.
Three Silver, 5281 A line, in Ar/H2-E.A.
I I I I
It 0102 110 104
10 10 4
was calculated for the specified conditions. The value of
the a parameter and the location of the high density
asymptote were then checked by a method described by
Behmenburg and Kohn (1,2). An illustration of this check-
ing procedure is given in Figure 5. The theoretical curves
of growth used in the checking procedure were prepared from
calculations given by van Trigt, Hollander, and Alkemade
(38). Values of the Doppler half-intensity width and the a
parameter are given in Table 4.
From the value of the solution concentration CI, at
the intersection point, the absolute atom concentration I,,
is calculated from equation 18. It should be noted that
the values of the oscillator strengths P used in these
calculations were taken from the present literature and
should be considered weighted values. The values for the
alkali and alkaline earth metals used in this study are well
established. The value of f for nickel, however, was
calculated from the gf value given in Corliss and Bozman
(5). Errors in an f value will produce direct variations
in absolute atom concentration NI; however, in most cases
in this study, variations in f are of the order of experi-
mental error for this study. The values of used in this
study are found in Table 3. The calculated values of NI
are found in Table 4.
Fig. 5.-Illustration of test for location of high
density asymptote and a parameter.
The following example illustrates the effect
of an erroneous choice for the location of high
density asymptote. A plot of an experimental
curve of growth representing the true location
of the high density asymptote and correct a
parameter value of 0.25 is given. From the
points U, V, and W at relatively low concen-
trations, one may be tempted to draw the upper
density asymptote. The asymptote drawn through
these points deviates only about 2 per cent
from the slope of 1/2, and gives an a parameter
at the intersection point of the low density
asymptote of 0.35. For the a parameter of
0.25, the logarithm of the total absorption at
which the theoretical curve approaches the
high-density asymptote equals about 0.9 and not
0.2 as is shown in the figure for erroneous
construction. The value at which the logarithm
of the total absorption for the theoretical
curve a = 0.35, should approach the high density
asymptote is about 1.2.
I I I
oM 4 CM OO Co n CV rI r' cr .-i C n n
cr) oo Mr-
mn) 0 N m'
4 Icno )-1 4C<
CM CM CM 1n m
04 C C
00 r- -I
* -4 C*
-41-Ii ClJ r
cn 0o CM4 -It Ln
Cn Cn CMi C CM
0 0 C\ cM
r-4 C4 cn -( )
C4M CM CM
1-1 in CM k.0 (O
o -o oC o
14 CM en t-n
C'M CM % CM C)
a0 in M r Co
00= co J('i
A N 4 C C
I- 1 m O
CM C M'~ ~
N e 3
oo o oo
* * a
oM O n >o ~
*4 0 *
i-< 0 i- D C
* * '^-<-\
r- CM o -*
SM CY) r
000 N -
CM4 00 O
c,4 -T i
CNOm O Mc 4 r 00
e e a .oo o o
rCl 4 C4 r4 C4 r
C* -I %
* e *l e**
* * -
* ** *
NNo < r
C* * -
C(n z CM4 C4
CM4 C4 00 C CM4
f- r i r- -t
13 -t %D" ID %.>
The effective molecular weight Ma, of the foreign
gas species producing collisional broadening represents a
weighted molecular weight calculated from all gaseous
species present in the flame, and was calculated from flame
dissociation data given by Zaer (47). The effective
molecular weight for the Ar/H2-E.A. flame was calculated
assuming complete combustion of 10 per cent of the hydrogen
introduced. Using these effective molecular weight values
(see Table 2), values of A L, ALT, and
and are found in Table 4.
Discussion of errors
Before reaching any conclusions with respect to the
experimental data determined in this study, a discussion of
errors is necessary. However, because of the indeterminate
nature of some of the errors, a detailed analysis of the
total error in each case will not be given. Only an esti-
mate of the order of magnitude of the various sources of
error will be given where possible.
Errors associated with calculation of the a para-
meter.-The relative error in measurement of the spectral band
width was less than 10 per cent. A temperature variation of
1000 K (a deviation of about 4 per cent) is needed to produce
a noticeable variation in the Doppler half-intensity widths.
Probably the largest source of error is in drawing the
asymptotes for the curve of growth to find the intersection
point. Hinnov and Kohn (18,19) have commented upon this
error, and concluded that the error limit for an experi-
mentally determined a value due to the choice of the
location of the asymptotes should not exceed 10 per cent.
However, the high density asymptote depends heavily on the
absorption values at the highest concentrations, and so this
is the region of largest variation of these values. Depen-
ding on the location of the intersection point, deviations
as high as 20 per cent are possible, and so the maximum
relative error expected for the a parameter in this study
is about 25 per cent.
Errors associated with calculation of NI and 3i.-The
magnitude of the error introduced in the location of the
intersection point will manifest itself as a similar error
in CI. The difficulty in determining the effect of this
error on NI, lies in estimating the deviation in the effec-
tive flame path length, L. It is felt that variations as
high as 50 per cent may be encountered. Similarly, the
calculation of the atom formation efficiency factor P5,
will be expected to show relatively large deviations due
to the compounding of errors from previously calculated
parameters. The relative error in estimating the aspiration
efficiency ( is probably 25 per cont. Therefore, NI and
PI, will have relative errors of the ordcr of two-fold.
Errors associated with calculation o ootce- nara-
meters.-Random errors were reduced whenever practical by
repetitive measurement.' Variations in A)L, and AJT can be
associated with variations expected in the a paramtor.
The effective collisional cross sectionc-L, can be expected
to show a somewhat larger variation than is found in the a
parameter due to addition of an indeterminate error in the
effective molecule weight of the perturbing gas species.
Comparison of results with literature data
Because of the dependence of the measured parameters
upon specific flame characteristics, it is impossible to
compare directly literature data to the results of this
study. The experimental data determined in this study were
used primarily to prepare the curves of growth by an
absorption technique and to calculate the a parameter.
Interpretations of other calculated data should be made with
reference to the absolute magnitude of the values given.
The values given in Table 5 for these parameters represent
(nearly in total) those which are available in the litera-
VALUES OF SPECTRAL AND FLAME COMPOSITIONAL PARAMETERS
TAKEN FROM LITERATURE
Hollander, Alkemade (38)
a NxlO-13cc. 2
/cc. TL (A )
.29a 5.1 18
.45a 5.2 27
.38a 11.7 26
.33a 12.6 25
.41b 10.1 30
.78a 5.3 31
(18, 19, 20)
a o-L(A )
aCalculated in a CO/air flame. The different values given
for some elements represents different mixtures of CO and air.
bCalculated in a C2H2/air flame.
Calculated in a C2H2/air flame.
dCalculated in a C2H2/air NO flame
eCplculated in a C2H2/air flame using a chamber-type burner.
When preparing curves of growth and calculating a
parameters, the test developed by Behmenburg and Kohn (1,
2) is very sensitive to small deviations from the correct
a parameter. Any errors in calculating a or in preparing
the stock solutions appeared as obvious deviations from the
correct slopes of the working curve (curve of growth).
Perhaps the largest single source of error which produced
deviations in the working curve was the spectral band width
of the monochromator. Only when the slit width was in-
creased to a relatively large width would the curve of
growth show the correct slope in the high density region.
This apparent failure of total absorption to be independent
of spectral band width cannot be explained. Hinnov and
Kohn (18,19) used large spectral band widths in their
studies but offer no explanation for their use. One
noticeable effect of the large spectral band width was the
decrease in sensitivity in the low concentration region;
this was expected, and did not affect the establishment of
low density asymptote.
When a parameter values are compared, a good cor-
relation (see Tables 4 and 5) within experimental error of
this study is found for Li and K, with the data given by
van Trigt, Hollander, and Alkemade (38); and for Li, Na, K,
Cu, Ag, Ca, Sr, and Ni with data given by Kohn and co-
workers (18,19,20) for flames of comparable composition.
This is, indeed, promising for the method.
The BI and NI factors given in Table 4 and those
found in Table 5 are not readily comparable because of the
basic difference between the aspiration mechanism of the
two types of aspirator burners used. However, the PI
values given in Table 4 are considerably smaller than the B
values obtained by de Galan and Winefordner (9) (see Table
5) for the same elements in an air-acetylene flame of low
burning velocity produced by a chamber type aspirator burner.
No attempt will be made to correlate o-L values given
in Tables 4 and 5, nor will an attempt be made to calculate
values for this parameter from any of the present collisional
theories (e.g., Weisskopf-Lindholm Impact Theory). The
readers attention is called to the excellent articles by
Behmenburg (1), and van Trigt, Hollander, and Alkemade (38)
for discussion of the effects of variations in the nature
of the perturbing species on the collisional cross section.
The variations of the parameter found in this study from
element to element and from flame type to flame type are
attributed to variations in the concentration and type of
perturbing species found in these flames. For this reason,
these parameters are not corrected for the effects of
hyperfine structure due to nuclear spin and isotopic shift
and line broadening effects due to quenching collisions.
Data from this study can be used to indicate the
range of variation of the a parameter, effective collisional
cross section, and total line width for conditions found
in this type of flame. The a parameter will have a value
less than 1.5, an effective collisional cross section less
than 100 A and a total half-intensity line width on the
order of 10 sec (i.e., approximately 0.01-0.1 A for a
hypothetical line at 5000 A wavelength). This is in line
with the conclusions reached by Parsons, McCarthy, and
In conclusion, the inherent advantages of preparing
working curves (curves of growth) in absorption, coupled
with the predictable shape of these curves not only
establishes the validity of the method, but should be of
real use to the analyst. In addition, preparation of
curves of growth in absorption should make this method a
useful means for the study of many fundamental flame
EXPERIMENTAL MEASUREMENTS DESIGNED TO EXTEND ANALYTICAL
APPLICATIONS OF THE CONTINUOUS SOURCE
Atomic absorption flame spectrometry (using a line
source) has enjoyed great popularity in recent years. This
is primarily due to the broad applications of this technique
to the trace analysis of metals. Complete commercial
instruments are now available for performing absorption
analyses; also a multitude of commercial components are
available and these can be assembled to meet the analyst's
specific needs. However, with any instrument, the analyst
soon discovers limitations in his system, and these are:
(1) Choice of fuel and oxidant because of burner
(2) Restriction to performing only flame absorption
analyses and not flame emission as well as flame absorption
(3) And most important, the time and expense involved
in using line sources of radiation for nearly every element
In connection with limitation (3) above, more speci-
fic problems may be encountered with a line source, some
of which are:
(a) time involved in optically positioning the lamp
for each elemental analysis;
(b) difficulty encountered in finding lamps of satis-
factory intensity, stability, and lifetime for
all elements of interest to the analyst; and,
(c) the inability to carry out simultaneous qualita-
tive and quantitative analyses (in most cases)
for more than one element.
With these limitations in mind, this investigation was
carried out to develop an experimental method which would
have maximum versatility without significant loss in sensi-
tivity of measurement for performing absorption measurements.
The instrumental set-up used to make the flame ab-
sorption measurements is shown in schematic form in Figure
6. The specific components used in the experimental set-up
are given in Table 6.
As shown, the solid angle of radiation is first
reduced by the housing around the xenon lamp, or by the
baffle placed in front of the tungsten lamp. The radiation
is chopped and passes through collimating lens L1, where it
enters the tube and is passed through it by a process in-
volving multiple reflection (14). The resultant radiation
Q C ,C
P I CDr
^.^ ^/ ; \ s ^-
^~C cs Q 3 *-|
a~C E^) u ?
<~ 1 e J __ ^
*p^ -^ j. -. L-
5 i ^ ^
\^_ ^^^ _J
Y <" ^v \ (-
i S-^K^ -'c
\L V '3*
EXPERIMENTAL SET-UP FOR DETERMINING LIMITS OF
DETECTION WITH A CONTINUOUS SOURCE
1. Spectral Continua.
Xenon arc, 150-watt (Eglehard,
Hanovia, Newark, N. J.) powered by
a regulated a.c. supply (Sola
Electric Co., Chicago, J1l.) used
for range of 2700-6000A. The
jacket surrounding the xenon lamp
has a connector for cooling air
and a baffle hole of 5/32" diameter.
Tungsten filament lamp (No. 2505,
Beckman instruments Inc., Fullerton,
Calif.) powered by a 6 volt storage
battery and used for range of 3500-
8500 A. A baffle with a 5/32" hole
was placed in front of the tungsten
Single pass, chopped at 320 cps.
Chopper consists of a disc with
eleven equally spaced holes rotated
by an 1800 rpm synchronous motor
(Bodine Electric Co., Chicago, Ill.).
L1 and L2 are quartz lenses. L1 is
a collimating lens, 8,0 cm. focal
length. Flame tube is thick walled
Vycor, 12 inches long with a 1.6 cm.
O.D. and 1,0 cm. I.D. (Englehard
Industries Inc., Hillside, N. J.).
Tube was held in place by a labora-
tory clamp. The tube holder was
fitted with connectors to direct air
at the flame tube to prevent burnout
and to extend the lifetime of the
flame tube. A small hood was posi-
tioned over the flame tube and
holder to remove flame gases and to
cool the unit.
Jarrell-Ash, Model 82000, 0.5 meter
Ebert Mount, grating spectrometer
(Jarrell-Ash Co., Waltham, Mass.).
Grating is ruled for 1250 lines/mm.
and is blazed at 5000 A.o Reciprocal
linear dispersion is 16 A/mm. in
the first order.
6. Gas Pressure and
EMI 9558 QB photomultiplier (1650-
9000 A). Regulated d.c. power
supply (No. 418 A, Fluke Mfg. Co.,
A.C. photomultiplier output signal
was fed into an 0.R.N.L. Model No.
7 a.c. (23) amplifier tuned to
320 cps. The plate and filament
voltages were taken from a dual
power supply (No. R 100 B, Phil-
brick Researches, Boston, Mass.).
The a.c. signal was rectified and
the d.c. signal was recorded on a
potentiometric recorder (Model TR,
E. H. Sargent and Co., Chicago,
Tank pressure is reduced by appropri-
ate two stage high pressure regula-
tors (The Matheson Co., Inc., East
Rutherford, N. J.). The gas flow
is then further regulated by a
Beckman High Precision regulation
unit (No. 9220, Beckman Industries
Inc., Fullerton, Calif.). The re-
sultant flow is monitored by rotam-
meters (No. 4-15-2 Ace Glass Co.,
Inc., Vineland, N. J.).
Total-Consumption type. Medium
bore (No. 4020, Beckman Instruments
Inc., Fullerton, Calif.).
is focused by lens L2, on the slits of the monochromator
where it enters and is dispersed. The aspirator-burner and
holder are mounted on a ring stand at a 450 angle to the
optical axis of the flame tube with the tip of the burner
less than one centimeter from the tube opening. The solu-
tion pipe of the aspirator burner was extended with a small
piece of tubing to allow aspiration in a horizontal posi-
The flow rate of aspirating gas was 2.25 1./min.
for argon and 2.75 1./min.' for air. The resultant solution
flow rate was then 1.0 ml./min. The flow rate of hydrogen
was optimized for each element and is given in Table 7.
The slit width used throughout this investigation
was 10 microns. This corresponds to a theoretical spectral
band width of 0.16 A in the first order. The spectral band
width, however, was measured by the method described in
the previous section, and was found to be 0.32 A.
Near the limit of detectability, a scale expansion
technique was used to increase the accuracy of measurement.
Scale expansion was accomplished by proper adjustment of
the zero suppress and by using a lower scale setting (i.e.,
12.5 mV. to 4.0, 2.5, or 1.25 mV.) on the recorder. The
phototube voltage used during the determination of each
element is given in Table 7.
Lr) Lr) LQ CY) LO
0 0 0 N 0 L
00000000000 10 1 O-'00 o
CYn CY) l r-I CY) %D V) i~
OOOOOO00 1oo 1
0000000000 o o -zO C,4
ooooo ooo oo in o oooe
OOOOO M0 00tt000f 0
r0LC0 fr) 0 0-1 C4 r0 1 7-
oC oC '4C'J
O ~~~o~o~~o ~
0 o0~ ~ o~~oo ~
*Cr CT CI 00 a 0
Sc r o (r 0-
Tag I o o-4
U 3*0 a u C u a
C< 4 L r in D r-c oo
r-4 -4 r-4 r-4 r-4 r-I r4
0 Ln C'4
0 C) 0
< 00 I'D
0co C: r
Stock solutions of 1000 ppm for each element were
prepared by dissolving the appropriate salt in aqueous, or
the metal in acidified-aqueous solution. More dilute
solutions were made by successive dilution of this stock
solution. In the case of the solutions prepared by dis-
solving the metal in acid, a small but constant concentra-
tion of acid was maintained in each dilution. The specific
salts used to prepare the stock solutions are given in Table
The wavelength used for the atomic absorption measure-
ments of each element is selected by aspirating a 50 ppm
solution of the element of interest and then manually scan-
ning to locate the wavelength of maximum absorption. Scan-
ning the wavelength range is useful for the following
(1) The optimum wavelength for analysis can be
determined if more than one wavelength is available (e.g.,
the transition metals).
(2) The noise on the background signal adjacent to
the wavelength of interest can be measured.
(3) The presence of contaminating elements in solu-
tion which may interfere with the element under investigation
can be determined.
The flame tube is flushed after each measurement by
continuously aspirating water. This is necessary to extend
the lifetime of the flame tube by preventing burn out and
salt fusion, and to reduce the noise on the background
signal. Periodic flushing (between use in analysis) of the
tube with dilute HNO3 minimizes the decrease in reflectivity
of the flame tube due to salt fusion.
Additional discussion of the experimental technique
and problems encountered when using the flame tube may be
found in articles by Fuwa and Vallee (14), and Koirtyohann
and Feldman (24).
Limits of detectability for twenty-one elements
measured using the experimental technique described above
are given in Table 7. The limit of detectability (45) is
defined as that solution concentration resulting in a
signal-to-noise ratio of 2.0. Noise is defined as peak-to-
peak divided by 2. The limits of detectability given in
Table 7 represent an average of nine determinations having
about a 50 per cent relative standard deviation. It should
be noted that the values given in Table 7 represent the
lowest obtainable using the experimental set-up described
in the apparatus section of this chapter, and various
combinations of fuel and oxidant. In each case for each
element, various combinations of C2H2, H2, 02, and air as
well as argon were tested to see which combination gave tho
lowest limits of detectability.
Relative standard deviations wore calculator orn the
basis of nine runs for 100 ppm and at a ton-fold concentra-
tion higher than that determined for the limit of detecta-
bility as well as at the limit of detectability. For
concentrations ten-fold higher than the limit of c-.o.-'cility,
the relative standard deviation was never larger than 1.0 per
cent for all elements run; whereas at 100 ppm, the relative
standard deviation was between 2.0 and 8.5 per cent, de-
pending on the particular element.' in Figure 7, typical
working curves for three elements are given. The data for
these three elemtns are treated in the conventional manner
(i.e., plots of absorbance vs. concentration) to illustrate
the similarities between the two types of absorbance working
curves (i.e., curves plotting relative absorption and
absorbance vs. concentration). As discussed in a previous
section (see Figure 4 for fraction of intensity vs. concen-
tration working curves for the same three elements), the
shapes of the working curves are characterized by a linear _
portion in the low concentration range which extends to
about a 10 per cent absorption. From Figure 7 for the flame
tube study, it can be seen that the working curve for silver
is linear from 0.01 ppm to about 100 ppm, whereas the copper
and sodium curves are linear from about 0.02 to about 30 ppm.
All other elements give curves similar to those of sodium
Due to the extreme curvature of the working curves at
high concentrations, a working curve must be prepared prior
to performing quantitative measurements. Fassel and co-
workers (12) have described the signal corrections which
must be considered when preparing absorbance versus solution
concentration working curves. From a comparison of the two
types of working curves (see Figures 4 and 7), it should be
noted that when it is necessary to work with high solution
concentrations falling on the nonlinear portion of the
absorbance curve, the experimental procedure described in
the Introduction would be preferred because of the more
favorable slope (proportional to concentration) in this
To meet the requirements imposed upon the experimental
system, all components chosen for the analysis system are
characterized by a wide range of operating conditions in
order to meet the varying needs of the analyst.
A spectral continuum as a source of radiation
certainly eliminates many of the restrictions placed on
atomic absorption by the use of line sources of radiation.
The added versatility of the qualitative aspect of atomic
absorption analyses eliminates this present inherent limita-
tion of the line source. In this study, a qualitative
investigation and quantitative estimate of the sensitivity
of the major atomic absorption lines of each element was
used to determine the absorption lines for measurement.
The total-consumption aspirator-burner was chosen
primarily because of the simplicity of sample introduction,
and because of the wide range of fuels and oxidants which
can be used with it. However, one major disadvantage of the
total-consumption aspirator-burner for atomic absorption
studies is the very short absorption path length of the
flame even under very fuel-rich conditions. This is a
serious handicap when sensitivity is desired. Several at-
tempts were made at finding a means for lengthening the path
length of the radiation through this flame. Multiple pass
of radiation from mirrors through the flame, multiple
burners (in line), and multiple pass of radiation through
multiple burners were just a few of the designs tried. How-
ever, it was found that the flame tube, when combined with
the argon/hydrogen-entrained air flame (Ar/H2-E.A.) gave
results superior to all other designs. The flame tube not
only extends the path length of the radiation through the
flame, but in addition, because of its cage effect (it
directs the flame in the radiation path and prevents outer
diffusion of atoms), it extends the residence time of atoms
in the light path.
Fuwa and Vallee (14) were the first to suggest the
use of the flame tube for atomic absorption analyses; how-
ever, they used line sources of radiation. Later
Koirtyohann and Feldman (24), and more recently, Koirtyohann
and Pickett (25,26,27) have determined limits of detectability
and discussed possible sources of spectral interference
using the flame tube. It is interesting to note that al-
though their application involved use of a line source of
radiation, a spectral continuum source was suggested as a
means of correcting for spectral interference.
The flame tube has the additional advantage that
misalignment of the aspirator-burner or incorrect position-
ing of the flame in the light path is not as critical as
alignment of the line source and acetylene-air flame used
with most chamber type aspirator-burners. Of course, align-
ment of the flame tube in the light path is critical.
The Ar/H2-E.A. flame was used in these studies be-
cause of the success of this flame in atomic fluorescence
(39) and atomic emission (46) studies. The total-consumption
aspirator-burner combined with the Ar/H2-E.A. flame is an
efficient method of producing atoms. This is evident from
data given in Table 7.
The good sensitivity of the Ar/H2-E.A. flame is a
result of two factors, namely: the Ar/H2-E.A. flame has
a very low background which results in greater sensitivity
(39,45); and the total-consumption aspirator-burner in con-
junction with the Ar/H2-E.A. flame and the flame tube
results in greater efficiency of atomization than is ob-
tained for most other aspirator-burner-flame systems. The
longer residence time of atoms in the flame tube and the
reducing characteristics of the flame gases decrease compound
formation, and,therefore, increase efficiency of atomization.
The requirements of the monochromator when using a
spectral continuum of radiation are, of course, more criti-
cal than when using a line source of radiation. For the
case where the line source is used, Walsh (41) concluded
that because the decrease in the peak intensity of the line
was measured, only a low resolution monochromator capable
of isolating the spectral line from the source was required.
For the case where the spectral continuum is used, Winefordner
(42) has shown that as the spectral band width of the mono-
chromator approaches the absorption line width of the atoms
in the flame, the sensitivity of measurement will increase
almost linearly with decrease in the spectral band width.
This effect undoubtedly accounts in part for the increase
in sensitivity found for elements in this study. These re-
sults confirm the statement by Gibson, Grossman, and Cooke
(16) that a medium resolution, large aperture bench-sized
monochromator combined with a scale expansion technique
should give good sensitivity; the monochromator used in
their study and in this one were identical.
One additional comment about the monochromator. If
the analyst wishes to analyze for elements which have widely
varying resonance wavelengths (e.g., cesium at 8500 A and
zinc at 2100 A), some of the difficulties encountered with
the transmission range of the monochromator can be overcome
by choosing a grating with a long wavelength blaze (e.g.,
7500 A in first order). In the lower wavelength region
where intensity in the first order is low (for a 7500 A
blazed grating), higher orders of the radiation can be used
In conclusion, using the experimental system previ-
ously described, the limits of detectability are comparable
to or greater than the best values listed in the literature
for a line source, acetylene-air, chamber-type flame system
and the values listed by Fassel and co-workers (12) for the
continuous source, fuel-rich acetylene-oxygen, total-
consumption system (see Table 7). This certainly indicates
that the spectral continuum (in conjunction with the flame
tube and the Ar/H2-E.A. flame system) as a source of excita-
tion should be competitive to the use of the line source and
a typical air-acetylene flame and chamber-type aspirator-
burner measurements in addition to providing added versa-
tility to the system, which the analyst can always use.
The work presented in this dissertation demonstrates
the feasibility of using a continuous source in atomic
absorption flame analyses. However, during this investi-
gation, other areas for research became apparent, primarily
as a result of the interpretation of data concerr-n, ve
measured a parameters. Research is definitely needed to:
(1) Establish the reasons for the wide divergence
in atom formation efficiency factors found for chamber and
total consumption-type burners; an approach utilizing the
measurement of absolute emission intensity and total ab-
sorption for elements atomized in both types of burners
could be used to determine the absolute atom concentration.
(2) Establish the relationship between the spectral
band width required to produce a satisfactory curve of
growth and the half-intensity width of the line under
investigation; as the first step, an instrument of high
resolving power (interferometer) could be used to establish
the profiles of the lines produced in this type of flame
(see Appendix I).
(3) Determine the nature of the perturbing species
in the flame produced from a total consumption-type burner;
the effect of the large quantity of water introduced into
this flame should be established.'
The absorption technique used in this study should
be used to determine fundamental atom parameters (such as
those investigated in this study) in flames of low burning
velocity from a chamber-type burner. These values will be
compared with those previously determined by one of the
The major contributions made in this dissortation
(1) Application of theoretical relationships derived
in astrophysics to atomic absorption measurements ''th a
(2) Derivation of a relationship between fraction of
radiation absorbed cq, and absorber concentration and useful-
ness of this relationship for preparation of experimental
working curves over wide ranges of absorber concentration
(working curves are useful over a wide concentration range
and any interference can be readily detected by a change
in the characteristic shape of such a working curve).
(3) Measurement of spectral parameters such as
damping constants and collisional cross-sections of lines
of atoms in flames of high burning velocity.
(4) Demonstration of the usefulness of the continuous
source and a Ar/H2-E.A. flame in a quartz tube for the de-
tection of low concentrations of a number of elements by
atomic absorption flame spectrometry.
The advantages of using a continuous source and a
Ar/H2-E.A. flame are;
(1) The low limits of detection obtained are
comparable with similar values obtained using line sources.
(2) The use of one experimental system for the
qualitative and quantitative determination of a large
number of elements in a variety of sample matrices.
The only serious disadvantage of such a system is
that an intense continuum and a medium resolution rather
than a low resolution monochrcmator must be usc. hv;-
ever, such a system is still considerably cheaper and more
versatile than an atomic absorption spectrometer using line
If the spectral band width, s, is of the same order
of magnitude as the absorption line half-iZ-tezsity wdlth
of atoms in the flame gases, for example, as for alkali
atoms at high atom concentrations in flames, then the
measured value of ca, will not be the same as Aj/s as de-
fined by equation 12. A correction factor relating the
measured ac to AT/s can be determined by assuming a certain
slit function distribution, for example, if a triangular
slit function g. is assumed then
aid if the above expression is evaluated, then
A C (I + -), (24)
T S S
where the term in parenthesis is a correction factor to
convert the measured value of ac to the defined value of ac
(equation 12). As can be noted for the above case, little
error results as long as the measured values of ac are less
than 0.l.' However, for the alkali metals, ac is greater
than O.1 unless slit widths greater than 200 microns are
1. Behmenburg, W., J.g.S.R.T. 4, i77 (1964).
2. Behmenburg, W., Kohn, H., J.~.S.R.2. 4. 1~ (19., ).
3. Bockris, J. 0'M., White, J. L., MacKenzie, J. D.,
"Physiochemical Measurements at High Temperatures,"
Butterworths, London, 1959.
4. Broida, H. P., Lalos, G. T., J. Chem. Phos. 20, 1466
5. Corliss, C. H., Bozman, W. R,, "Experimental Trransition
Probabilities for Spectral Lines of Seventy Elements,"
N.B.S. Monograph No. 53, Washington, D.C., 1962.
6. Crosswhite, H. M., "The Spectrum of Iron I," John
Hopkins Spectroscopic Rept. No. 13, August, 1958.
7. de Galan, L., Personal Communication.
8. de Galan L. Winefordner, J. D., Anal. Chem., in
9. de Galan, L., Winefordner, J. D., Submitted 1966.
10. de Vos, J. C., Physica 20, 690 (1954).
11. Fassel, V. A., Mossotti, V. G., Anal. Chem. 35, 252
12. Fassel, V. A., Mossotti, V. G., Grossman, W. E. L.,
Kniseley, R. A., Soectrochim. Acta. 22, 347 (1966).
13. Fery, Ch., Comptes Renders 137, 909 (1903).
14. Fuwa, K., Vallee, B. L., Anal. Chem. 35, 942 (1963).
15. Gaydon, A. G., Wolfhard, H. G., "Flames, Their Struc-
ture, Radiation and Temperature," Chapman and Hall
Ltd., London, 1960.
16. Gibson, J. H., Grossman, W. E. L., Cooke, W. D.
"Analytical Chemistry" (proc. Feigl Anniv Sym.) pg. 296,
17. Hermann, R., Alkemade, C. Th. J. "Chemical Analysis
by Flame Photometry" (trar.liatdi by,, T. G-lbert),
John Wiley and Sons, New York, 1965.
18.' Hinnov, E., J. O0t, Soc. Am. 47, 151 (1957).
19. Hinnov, E., Kohn, H., J. Opt. Soc. Am. 47, 156 (1957).
20. Hoffman, F. W., Kohn, J., J. Opt. Soc. Am. 51, 512
21. Hollander, Tj., Thesis, Utrecht (1964).
22. Ivanov, N. P., Kozireva, N. A., Zh. Analit. Ki. 19,
23. Jones, H. C., Fisher, J. D3., Kelley, K. T., "Fifth Con-
ference on Analytical Chemistry in Nuclear Reactor
Technology," Gatlinburg, Tennessee, October, 1961.
24. Koirtyohann, S. R., Feldman, C. "Developments in
Applied Spectroscopy," Vol. 3, J. E. Forrette ed.
Plenum Press, New York, 1964.
25. Koirtyohann, S. R., Pickett, E. E., Anal. Chem. 37,
26. Koirtyohann, S. R., Pickett, E. E., Anal. Chem. 38,
27. Koirtyohann, S. R., Pickett, E. E., Anal. Chem. -38,
28. Ladenburg, R., Reiche, P., Ann. d. Phys. 42, 181 (1913).
29. Minkowski, R., Z. f. Phys. 36, 839 (1926).
30. Mitchell, A. C. G., Zemanski, M. V., "Resonance Radi-
ation and Excited Atoms," Cambridge University Press,
31. Parsons, M. L., Winefordner, J. D., Appl. Spect. in
32. Parsons, M. L., McCarthy, W. J., Winefordner, J. D.,
Anal. Chem. in press, 1966.
33. Penner, S. S., "Quantitative Molecular Spectroscopy
and Gas Emissivities," Addison-Wesley, Reading, Mass.,
34. Rubeska, I., SvobodaV., Anal. Chi-n. Acta 32, 253 (1965).
35. Slavin, W.,.Atomic Absorption Newsletter, No. 24,
36. Unsold, A., "Physic der Sternatmosphoren," Springcr,
37. van der Held E F. M., Ornstein, S., Z. f. Phys.
77, 459 (193 ).
38. van Trigt, C., Hollander, Tj., Alkemade, C. Th. J.,
J..S.R.T. 5, 813 (1965).
39. Veillon, C., Mansfield, J. M., Parsons, M. L.
Winefordner, J. D., Anal. Chem. 38, 204 (1966).
40. Vickers, T. J., Remington, L. D., Winefordner, J. D.,
Anal. Chim. Acta in press, 1966.
41. Walsh, A., Spectrochim. Acta. 7, 108 (1955).
42. Winefordner, J. D., Aopl. Soect. 17, 109 (1963).
4r. Wineforiner, J. D., Mansfield, C. T., Vickers, T. J.,
Anal. Chem. 35, 67 (1963).
44. Winefordner, J. D., Vickers, T. J., Anal. Chem. 36,
45. Winefordner, J. D., Vickers, T., Anal. Chem. 36, 1947
46.' Zacha, K., Winefordner, J. D., Anal. Chem. in press,
47. Zaer, R. A., Thesis, Paris (1935).
BIOGRAPHICAL S NETCH
William Walter McGee III was born June 27, 1939,
in Toledo, Ohio. In June, 1957, he was graduated from
Jesup W. Scott High School, Toledo, Ohio. In June, 1961, he
received the degree of Bachelor of Science from the Uni-
versity of Toledo, Toledo, Ohio. In June, 1962, he received
the degree of Master of Science from the University of
Toledo, Toledo, Ohio. From September, 1963, until the
present time, he has pursued his studies toward the degree
of Doctor of Philosophy.
He is married to the former Mary Ann Nietz of
This dissertation was prepared under the direction
of the chairman of the candidate's supervisory committee
and has been approved by all members of that committee. It
was submitted to the Dean of the College of Arts and Sciences
and to the Graduate Council, and was approved as partial
fulfillment of the requirements for the degree of Doctor of
December 17, 1966 /
Dean, Col ge,/of Arts and Sciences
Dean, Graduate School