Acta, 1055, Vol. 7, pp. 108 to 117.

Spectrochimiea Pergamon Press Ltd., London

The application of atomic absorption spectra to

chemical analysis
A. WALSH
Chemical Physics Section, Division of Industrial Chemistry,
Commonwealth Scientific and Industrial Research Organization, Melbourne, Australia
(Received 18 January 1955)

Summary-The theoretical factors governing the relationship between atomic absorption and
atomic concentration are examined and the experimental problems involved in recording atomic
absorption spectra are discussed. On the basis of the discussions, it is shown that such spectra
provide a promising method of chemical analysis with vital advantages over emission methods,
particularly from the viewpoint of absolute analysis. It is also suggested that the absorption
method offers the possibility of providing a simple means of isotopic analysis.

1.Introduction
The application of atomic spectra to chemical analysis has proved so successful
over such a wide field that there is a tendency to overlook some of the basic
limitations of existing methods. In spite of the remarkable advances in technique
which have resulted in press-button analyses of high precision at fant’ast’ic speeds,
there has been practically no progress whatsoever in solving the fundamental
problem of devising an absolute method, i.e., a method which will provide an analysis
without comparison with chemically analyzed standards or synthetic samples of
known composition. In routine analysis for production cont,rol this problem is of
little consequence, since it is only necessary to have a limited number of standards,
and in such work modern direct-reading methods leave little to be desired, except
on the score of complexity of equipment and associated expense. When analyses of
miscellaneous materials are required, the task of providing the required range of
standards becomes insurmountable and the spectrochemical method then loses
its accuracy, since accurate analyses generally necessitate the use of standards
which are closely similar in composition to the sample for analysis. In some
analyses it is also essential that the sample and standards be similar as regards
physical condition. For example, the intensity of the spectrum of a metal or alloy
may vary with the metallurgical history of the sample. This difficulty may be
overcome by taking the sample into solution, but accurate and sensitive methods
of analyzing solutions are only available for the limited range of elements, having
a low excitation potential, which can be estimated by flame photometry. In this
method, also, it is necessary to use standard solutions having compositions closely
similar to that of the test solution.
The possibility of adapting any of the existing methods to absolute analysis
does not appear t’o be promising. In the first place, there seems to be little prospect
of deGeloping a light source which is such that the emission spectrum of a given
element is not affected by the presence in the atomic vapour of atoms of other
elements. Secondly, even if these interelement effects were eliminated, there
remains the problem of absolute intensity measurement and the associated problem

108
 The application of atomic absorption spectra to chemical analysis

of det,ermining the distribution of atoms over the various energy states. In practice
there is the third difficulty that electrical discharges of the type now in use do not
give a stable output of radiation; modern methods conceal these erratic variations
in output by integrating, photographically or photoelectrically, the radiation over
a period of several seconds. Finally, there are other problems arising from self-
absorption and self-reversal, and from the fact that the processes of vaporization
and excitation are not isolat.ed from each other. A review of these and other
aspects of source behaviour has been given elsewhere [l].
At t,he present stage of its development there is no doubt that the major
obstacle to further progress in the technique of spectrochemical analysis is the
occurrence of interelement effects, since if these could be eliminated it would be
possible to use the same set of standards for the determination of any one element
in any mat)erial. With existing emission methods the intensity of a given spectrum
line due to one particular concentration of an element in different materials varies
greatly. For example, PROKOF’EV [2] reports that for the same concentration of
silicon the intensity of the silicon lines in the spark spectrum of steel is eight times
as great as in brass, and in duralumin the intensity is even less. Many other
examples of interelement effects have been published.
No satisfactory explanation of these effects has yet been given, nor can one be
expected, since the phenomena occurring in the arc and spark discharges used as
light sources are far too complex to permit of any theoretical analysis, and the
approach to the subject seems likely to remain essentially empirical. It seems
possible, however, to arrive at certain broad conclusions. Since interelement
effects are usually of the same order of magnitude for different spectrum lines of a
given element, corresponding to transitions between different energy states, they
probably arise from changes in the concentration of atomic vapour rather than
changes in the excitation conditions. However, such changes cannot occur in
sources in thermal equilibrium at a constant temperature. Thus, if thermal
equilibrium is assumed, then interelement effects must necessarily be due to a
change in the temperature of the atomic vapour. Whilst the behaviour of the
electrical discharges used in spectrochemical analysis shows that it is not justifiable
to assume thermal equilibrium or to ascribe a temperature to the discharge, it is
instructive to consider the effects of changes in temperature on a mass of atomic
vapour in thermal equilibrium.
Consider the emission of a spectrum line due to the transition from an excited
state j, of excitation energy Ej, to a ground state of energy E, = 0. Then if Pj
and P, are the statistical weights for the excited state and ground state respec-
tively, the number of atoms in the excited state, Nj, is related to the number of
atoms in the ground state, N,, by the relation

and, neglecting self-absorption and induced emission, the intensity of the emitted
line is proport,ional to Nj.
In order to illustrate the magnitude of N,/N,, the calculated values for resonance
lines of various elements at different temperatures are given in Table 1.

109
 A. WALSH

Table 1. Values of Nj/No.for various resonance linea

-

Remnance
Transition pjIpO
line
T = 2,OOO”K T = 3,OOO”K T = 4,OOO”K 1T = 5,OOO”K
- -.

Cs 8521 A 2s l/2 -
2P 212 2 4.44 X lo-’ 7.24 X lo-9 2.98 X 1O-2 j 6.82 X 1O-2

Na 6890 A 25 l/2 - 2p2/2

2 9.86 x IO-6 6.88 x 10-4 4.44 X 10-s 1.51 X 10-2

Ca, 4227 A ‘S, - ‘P, 3 1.21 X lo-’ 3.69 X 10-h 6.03 X 10-4 3.33 X 10-s

A IS, - ‘P, 3 7.29 x lo-15 5.58 X lo-‘0 1.48 X lo-’ 4.32 k 1O-6

-
Zn 2139
- - -

It will be seen that in nearly all cases the number of atoms in the first excited
state is only a small fraction of the numbers of atoms in the ground state. The
fraction only becomes appreciable at high temperatures for states of low energy.
Since most elements have their strongest resonance line at wavelengths below
6,000 A, and since we shall be concerned mainly with flames or furnaces having
temperatures below 3,00O”K, we may regard N, as negligible compared to No.
The fraction of atoms in higher excited states is much less than those given above,
and thus CN, is also negligible compared with N,, and the latter can be considered
as equal to the total number of atoms, N.
Thus, whilst the number of excited atoms varies exponentially with tempera-
ture, the number of atoms in the ground state remains virtually constant and
therefore the integrated absorption J K, dv due to transitions from the ground state
is independent of temperature. (This discussion only applies to atoms having a
ground state well removed from the lowest excited state; the case of atoms
having a multiplet ground state is discussed in Section 2.)
On the basis of the above discussion it would appear that atomic absorption
spectra would have important advantages over emission spectra as a means of
chemical analysis. It is therefore surprising to note that the research in this field
has been devoted almost exclusively to emission spectra; the annual review [3] of
progress in spectrochemical analysis is, in fact, given under the general title of
“emission spectroscopy.” Apart from the special case of estimating the contamina-
tion of room and laboratory atmospheres by ‘mercury vapour, the application of
atomic absorpt,ion spectra to chemical analysis appears to have been confined to
astrophysical work on the determination of the composition of the solar and stellar
atmospheres.
The purpose of this paper is to examine the theoretical factors governing the
relationship between atomic absorption and atomic concentration and to discuss
the experimental problems involved in recording atomic absorption spectra. On
the basis of these discussions, it is shown that such spectra, provide a promising
method of chemical analysis with vital advantages over emission methods,
particularly from the viewpoint of absolute analysis. It is also suggested that the
absorption method offers the possibility of providing a simple means of isotopic
analysis.
110
 The application of atomic absorption spectra to chemical analysis

2. Variation of atomic absorption with atomic concentration

The relationships between atomic absorption and atomic concentration, under
various conditions, are fully discussed in several papers, mainly in astrophysical
journals, and in standard reference books [4-s]. For the purpose of this discussion
it is sufficient to consider only the most fundamental of these relationships. In
this section the discussion will be further restricted by assuming that atomic
absorption lines possess no fine structure; the case of isotopic hyperfine structure
is discussed later in Section 4.
Consider a parallel beam of radiation of intensity,I,, at frequency v incident on
an atomic gas or vapour of thickness 1 cm. Then if I, is the intensity of the trans-
mitted beam, the absorption coefficient K, of the vapour at frequency v is defined
by
I, = I,” emKJ (2)
The dependence of K, on v, i.e., the shape of the absorption line, is determined
by the nature of the transition involved in the absorption and on the physical
conditions such as temperature, pressure, and electrical fields, to which the atoms
are subjected during the measurement.
According to classical dispersion theory, the relationship between absorption
and concentration is given by

I K,dv = n$=N,ff

where e is the electronic charge, m the electronic mass, c the velocity of light,
N,, the number of atoms per cm3 which are capable of.absorbing in the range v to
Y + dv, and f, the oscillator strength, is the average number of electrons per atom
which can be excited by the incident radiation. Equation (3) is not valid for strong
absorption lines, since it assumes that the refractivejndex is of the order of 1 over
the breadth of the absorption line.
For a resonance line due to a transition from a ground state which is well
separated from the’lowest excited state, N, can be considered as equal to N, the
total number of atoms per cm3 (see Table 1). If, however, the transition does not
originate in the ground state, or if there is a multiplet ground state, then the number
of atoms capable of absorbing is given by

where i denotes the initial state involved in the transition, and the summation in
the denominator extends over all possible energy states. In practice, of course, the
sunimation can be restricted to the low-lying levels.
In terms of transition probabilities the equation corresponding to equation (3)
is

where I is the wavelength

s K,dv =z.
r 1
NJ,, (1 -

at the centre of the absorption

$$)
I
(5)
line, Pi and P, are the

111
 A. WALSH

statistical weights of the lower and upper states, respectively, involved in the
transition, and Aji is the Einstein coefficient of spontaneous emission for the
j -+ i transition. N, is the number of atoms in the upper electronic state, and will
generally be negligible compared with N,, and in this case equation (5) reduces to
the well-known relation:

K, dv = s, N,A,, (6)
s 97 ,

In the general case N, is given by (4), but for resonance lines is equal to N. Similarly,
if the transition from the jth to ith state is the only one which can occur, as in the
case for resonance lines, Aj, = l/r, where 7 is the mean life of an atom in the
excited state j.
Equations (3) and (6) provide simple relationships between absorption and
concentration, and it is now necessary to investigate whether they can be applied
to a practical met(hod of spectrochemical analysis. Since the intensities of spectrum
lines are usually expressed in terms of oscillator strengths, it is convenient to
consider equation (3).
Firstly, it is necessary to know whether suitable absorption lines occur in
regions of the spectrum which are amenable to measurement. In terms of sensi-
tivity, it is obviously desirable to use the strongest resonance lines, and in general
these will correspond to the strongest lines occurring in emission spectra. These
are listed in a paper by MEGQERS [7], and reference to this shows that with the excep-
tion of the rare gases, hydrogen, mercury, the halogens and the metalloids, all
elements have their most sensitive lines in the region 2,000-9,000 ip. Thus the lines
for all the more common elements all lie in regions of the spectrum where measure-
ments are simple to make.
The oscillator strengths of some of these lines have been determined, and are
listed in Table 2. Theoretical calculation [S-14] of f-values is possible for atoms
having simple electronic structures, but has not yet been carried out for heavy
atoms having a complex structure. BATES and DAMGAARD [15] have described a
simplified theoretical method and have published tables from which the absolute
strengths can be rapidly obtained. The method has been shown to give accurate
results for all transitions in the lighter simple systems, but there are insufficient
experimental data to enable one to judge to what extent the method can be
applied to the more complex electronic structures.
It is interesting to note that the oscillator strength for the strongest‘copper
line is approximately the same as for the alkali metals, in spite of the fact that the
closed 3d shell of the copper atom is not nearly so tightly bound as the inner
electron shells of the alkali metals. By comparison with the f-values of other
elements in Group 1, it seems probable that rubidium, silver, and gold will have
f-values of the order of 0.7 for their st.rongest resonance lines. Similarly, the Group
2 elements Zn and Sr may be expected to have f-values of the order of 2 for their
strongest lines.
It‘ does not appear possible at the present time to make any corresponding
estimates of the oscillator strengths of spectral lines of other elements, particularly
those such as iron and cobalt, with complex electronic structures and multiplet

112
 The application of atomic absorption spectra to chemical analysis

ground states. The possibility of determining approximate f-values for such

elements by a simple experimental method is discussed in Section 4.
For those elements for which the oscillator strengths are known, the atomic
concentration can be determined from the integrated absorption coefficient, using
equation (3); the experimental problems involved in measuring such coefficients
are discussed below.
Table 2. List of f-values

r-
- -
Resonance line Transition f Reference

L -

Li 6708 A 2s l/2 - 2p2/2

0.50 PI

Na 5890 25l/2 - 2P 312 0.70 Pel

K 7065 25l/2 - 2P 212 0.64 [171

cu 3247 2sl/2 - 2P 3/2 0.62 [181

cs 8521 2sIla - 2P 312 0.66 [I91

Be 2349 lS() -'P, 1.82+t [ll, 13, la]

Mg 2852 ‘Se - ‘P, 1.74* [13, 141

Ca 4227 ‘S, T lP, 2.28* 7 [12, 13, 141

Cd 2288 ‘S, - ‘P, 1.20 VW

Ba 5535 IS, - lP, 2.10 WI

Hg 1849 ‘Se - lP, 1.19 c221

Tl 2769 2P l/2 - 2*2/2 0.20 ~231

Cr 4254 '52 - ‘Pi 0.084 ~241

Ni 3415 3D, - 3F!j 0.02 ~251

Fe 3720

* Theoretical values.
-
a6D4 - z5F,
- 0.013
-!_
rw

7 BIERMANN and TREFFTZ [13] state that these values should be corrected according to the method
described in ref. 14. See also the values quoted in ref. 30.
J’ooootnote.
Since the purpose of this table is to indicate the order of thef-values for various lines, the
values obtained by different observers are not included. They are discussed by KORFF and BREIT [26]
and by MITCHELL and ZEMANSKY [a].

3, Experimental determination of atomic absorption coefkients

The shape of an atomic absorption line is determined by (a) the natural width of the
line due to the finite lifetime of the excited state; (b) the Doppler contour due to
the motions of the atoms relative to the observer; (c) pressure broadening, either
by atoms of the same kind giving rise to resonance broadening or to foreign gases;

113
 A. WALSE

and (d) Stark broadening due to external electric fields or to neighbouring charged
particles. The natural width of an atomic spectral line is of the order of lo4 A,
and for the purposes of this discussion is negligible compared to the width due to
other causes.
The Doppler width of a line is given by

where R is the universal gas constant and M is the atomic weight. Typical values
of D, are given in Table 3.
Table 3. Valuer, of DA at various ternperaturea
- -
DA
Element 1 M
1,OOO”K j 2,OOO”K 1 3,OOO”K
- -____ i
I
Na 6890 22.3 0.028 A o.oxa A O-048 A

cu 3247 63.0 0.0092 A 0.0013A 0.016A

Zll 2139 66.4 0~0060 A / 0.0085 A o-010 A

- -I I I
If we assume that. a temperature of 2,OOO”Kis required to produce sufficient
v&pour, then the Doppler width is of the order of 0.01 A. The accurate measure-
ment of the profile of such a line would require a resolution of about 500,000,
which is beyond the performance of most spectrographs. In addition, if it is
desired to use photoelectric methods of intensity measurement, then it is scarcely
feasible to use a continuous source, since the energy emitted over such a small
spectral slit-width would be too small to give a high enough signal/noise ratio.
In the past this difficulty has often been overcome by using the method of total
absorption, in which the energy removed from the incident beam is measured.
This method has the advantage that the measurement is independent of the resolu-
tion of the monochromator, but suffers from the disadvantage of giving a compli-
cated relation between N and f, according to the region of the curve of growth in
which the measurement is made. However, if the absorption is so strong that it
is not possible to make an accurate measurement of the absorption coefficient, a8
in astrophysical work, then the curve-of-growth method is the only one available.
The method has been successfully applied to the measurement of oscillator strength
from furnace absorption spectra by KING [18] and by ESTABROOK[24, 251.
From the point of view of spectrochemical analysis, a more attractive method
appears to be to measure the absorption coefficient at the centre of the line, using
a sharp-line source which emits lines having a much smaller half-width than the
absorption line. If the shape of the latter is determined entirely by Doppler
broadening, we have [27]
212 G-2 TV9
K max = - J -- . a Nf (8)
DA ,r
114
 The application of atomic absorption spectra to chemical 8ndysia

where D is the Doppler width. Thus in this case also there is a linear relation
between absorption and concentration.
If such a sharp-line source is used, it is now no longer necessary to use a spectral
slit-width of the same order as the half-width of the absorption line. The require-
ment now is the ability to isolate a selected line from other lines emitted by the
source. Thus spectrographs having the same resolution as those used in conven-
tional emission methods are adequate. In many cases it may be sufficient to use
filters.
Various methods of producing such sharp-line sources are available. In our
work it has been found convenient to use hollow-cathode discharge tubes, and it
has proved possible to make sealed-off tubes about the size of a photomultiplier
tube.
There is one other experimental difficulty. In many cases vaporization of the
sample will result in the emission of radiation at exactly the wavelength where it is
desired to make the absorption measurement. This difficulty can be overcome by
modulating the incident radiation before it reaches the atomic vapour and amplify-
ing the output of the detector by an amplifier tuned to this modulation frequency.
Thus the radiation emitted by the atomic vapour, which is not modulated, produces
no signal at the output of the amplifier.
So far it has been assumed that the line shape is determined solely by Doppler
broadening, and this is sensibly true if the vapour is produced by a vacuum
furnace, such as that used by KING, and if the vapour pressure is so small that
resonance broadening is negligible. Another convenient method of vaporizing
the sample is to atomize a solution of the sample into the air supply of a Meker
burner, as in emission methods of flame photometry. In this case there is broad-
ening due to foreign gases, and, although we have not made accurate measurements
for the flame we have used, its magnitude is probably of the same order as that
due to thermal motions. Once the broadening due to pressure is known, the
corresponding correction to equation (8) can be applied, using the tables published
by ZEMANSKY [28]. As an example, if the pressure broadening width is equal to
the Doppler width, the maximum absorption is 43 per cent of that due to Doppler
broadening alone.
IV. Discussion
The above discussion of atomic absorption spectra has indicated the attractive
possibilities of using them for chemical analysis and of developing a method which
will provide a useful complement to emission methods and in many cases may well
supersede them. One of the main attractions of this absorption method is that,
theoretically, it is expected to be much less susceptible to interelement effects.
In so far as effects observed in emission are due to variations in the distribution
of atoms over the tarious excited states, they would have no counterpart in
absorption where this is due to a transition from the ground state. Similarly,
absorption will not be critically dependent on the temperature of the atomic
v&pour, since the Doppler width only varies as Tlj2, whereas small changes in
temperature produce large changes in the intensity of the emitted radiation.
In addition, the integrated absorption coefficient is independent of wavelength, in
marked contrast to the emission intensity, which will vary according to Planck’s
law.
115
 A. WALSH

Whilst the discussion has been limited to the ideal case of a vapour in thermal
equilibrium, in practice there may be effects due to a sheath of cool vapour.
But whereas in emission spectroscopy this causes self-reversal which reduces the
peak intensity, in absorption it will contribute to the absorption and will in no
sense reduce the sensitivity. Incidentally the self-reversal observed in emission
spectra provides a good indication of the sensitivity of the absorption method.
Another cause of deviation from thermal equilibrium is chemiluminescence;
this may well produce a number of excited atoms, which is comparable with that
due to thermal excitation, but will not produce any significant change in the number
of unexcited atoms. Thus the effects on the absorption spectrum are negligible
compared to those on the emitted radiation.
On the experimental side, the important advantage of the absorption method
lies in the fact that the measurement of the absorption coefficient consists of
measuring the ratio of two intensities, which is much simpler to achieve than the
measurement of emission intensities in absolute units.
The possibilities of absolute analysis have been discussed and the use of a
relative absolute method may also be noted. For example, if the sample solution
is atomized into a flame, then a calibration for one element would serve to deduce
the approximate calibration of other elements, provided the oscillator strengths
are known. In this connection it would appear that, using standard solutions,
this method could be used to determine oscillator strengths, at least to within
an order of magnitude.
Finally, the absorption method may prove suitable for isotopic analysis.
If sources emitting spectra of only one isotope are used, then an analysis for this
isotope can be obtained directly, since the oscillator strength is the same for each
component of the hyperfine structures of the excited level. If no pure isotope is
available, then sources having different concentrations of isotopes may be used.
Alternatively, an isotopic filter containing the vapour of one or more isotopes
may prove satisfactory.
The results of some preliminary experiments have been in full accord with the
conclusions arrived at in this paper, and future papers by J. P. SHELTON and the
author will describe the construction of an atomic absorption spectrophotometer
and its applicat’ion to various analytical pr’oblems.

References
[l] WALSH, A.; Chapter 7 of Metal Spectroscopy by F. TWYE~AN, Griffin and Co., London, 1951.
[Z] PROKOF’EV, V. F.; C.R. Acad. Sci., U.R.S.S. 1940 29 443.
[3] MEDOERS, W. F.; Anal. Chem. 1954 26 54.
[4] MITCHELL, A. C. G. and ZEMANSKY, M. W.; Resonance Radiation and Excited Atoms,
Cambridge Press, 1934.
[5] ROSSELAND, S.; Theoretical Astrophysics, Clarendon Press, Oxford, 1936.
[6] UNS~~LD, A.; Physik der Sternatmosphliren, Springer, Berlin, 1938.
[7] MEGGERS, W. F.; J. Opt. Sot. Amer. 1941 31 39.
[8] SUGIARA, Y.; Phil. Mag. 1927 4 495.
[9] TRUMPY, B.; 2. Physik, 1930 66 720.
[lo] VINTI, J. P.; Phys. Rev. 1932 42 632.
[ll] HARTREE, D. R., and HARTREE, W.; Proc. Roy. Sot. London, 1936 A154 588and A166 45.
[12] HARTREE, D. R., and HARTREE, W.; Proc. Roy. Sot. Londan, 1938 Al04 167.

116
 The application of atomic absorption spectra to chemical analysis

[13] BIERMANN, L., and TREFFTZ, E.; 2. Aetrophya., 1949 aS 213.

[la] TREFFTZ, E.; 2. Astrophys., 1944 26 240.
[16] BATES, D. R., and DAM~AARD, A.; Phil. Tram. Roy. Sot. 1949 A%2 101.
[IS] MINKOWSKI, R.; 2. Physik, 1926 36 361.
[17] WEILER, J.; Ann. Phya. 1929 1361.
[IS] KING, R. B., and STOCKBAROER, D. C.; Astrophy8. J. 1940 91 488.
[19] MINKOWSKI, ,R. and M~~HLENBRUCH, W.; 2. Phy.9ik 1930 63 198.
[20] ZE%%ANSKY,M. W.; Z. Phyaik 1931 72 687.
[21] WESSEL, G.; 2. Physik 1949 1s 440.
[22] WOLFSOHN, G.; 2. Phytik 1930 63 634.
[23] KUHN, W.; Det. Kgl. Dmwke Videnakab. Math.-Fy8. Medd. 1926 7 12.
[24] ESTABROOK, F. B.; A8t?+Ophy8. J. 1952 115571.
[25] ESTABROOK, F. B.; Astrophys. J. 1951 113684.
[26] KORFF, S. A., and BREIT, G.; Rev. Mod. Phys. 1932 4 471.
[27] See ref. 4, p. 100.
[28] ZEMANSKY, M. W.; Phy8. Rev. 1930 36 919;seealso ref. 4, p. 329.
[29] &NC+, R. B.; Aetrophya. J. 1942 98 78.
[30] KOPFERMANN, H., and WESSEL, G.; 2. Phyaik 1951 130 100.

117