KIRUNA

GEOPHYSICAL
INSTITUTE

A NEW WAY OF ORGANIZING SPECTRAL LINE

INTENSITY RATIO FLUCTUATIONS
OF DIFFERENT ELEMENTS
B. Thelin
KGI PREPRINT 068
FEBRUARY 1986

.Kiruna Sweden
A NEh WAY GF ORGANIZING SPECTRAL LINE INTENSITY RATIO
FLUCTUATIONS OF DIFFERENT ELEMENTS

by

B. Thelin
Kiruna Geophysical Institute, P.O. Box 704,
S-981 27 Kiruna, Sweden

KGI Preprint 08S

February 1986

Printed in Sweden
Kiruna Geophysical Institute
Kiruna 1986
ISSN 0349-2656
 h new way of organizing spectral line intensity ratio
fluctuations of different elements

by

Bo Thelin
Kiruna Geophysical Institute
P. 0. 3ox 704, S-981 27 Kiruna, Sweden

Abstract

A new analysis method of organizing spectral line intensity

ratio fluctuations for different elements has been developed
to see which line pair of analysis is more or less stable at
repeated measurements. A linear graph was obtained for spec-
tral lines from different elements when the spectral line
ratio fluctuations v/er.i plotted versus the photon energies and
the ionization energies of these lines. When these fluctua-
tions were plotted versus the upper energies instead, no line-
ar structure at all was obtained. This means that there was no
linearity at all with the conventional intensity formula with
the Boltzmann exponent, but there was if exp[-(hvmn+J)/kT]
was used instead.

The same exponent is supported by a method using absolute

intensities from standard tables, where 17 different elements
have been studied using direct current arc measurements with
photographic registration.

The light sources used in this work were an inductively

coupled plasma (ICP) with a slot type nebulizer: system, a
double chamber hollow cathode lamp (DCHC) for tubular metal
samples, and a hollow cathode lamp (FSHC) tor flat metal
samples. The analyses were carried out by using a computerised
image dissector ecnelle spectrometer system (IMS).

Keywords: Inductively coupled plasma, intensity formula,

fluctuations, échelle spectrometry.
1 • Introduction

The basis of modern analytical atomic spectroscopy was intro-

duced by Gerlach (1) in 1925 who started the internal standard
technique. The principle of the internal standard technique is
based on relating the intensity ratio of selected spectral
lines of the analytical and internal standard elements to con-
centration ratios.

In many papers following Gerlach's technique there was a

tendency to exclude some "bad" line pairs from the analysis
without giving any satisfactory explanation. However, no
satisfactory explanation of some big discrepancies observed in
this analysis method has been presented so far.

In a theoretical paper by Barnett et al. (2) the principles

that affect the choice of analysis and internal standard lines
in analytical atomic spectroscopy have been investigated with
an ICP. In that paper the effect of excitation energy, ioniza-
tion energy, partition functions, and electron density on the
analytical line pair intensity ratio was studied.

In a paper by Suckewer (3) the essential sources of spectro-

analytical errors connected with the determination of the
density ratios of atoms in plasmas are described. Together
with a powerful pulsed discharge system the effect of the
distribution function of the atom density for different
plasma radii and of deviations from local thermal equilibrium
(LTE) in the plasma is studied. These authors and others
assume that the line intensity from an atomic transition
K m +E n is proportional to the Boltzmann factor
exp(-Em/kT). Ejj! is here the energy of the upper energy
level, and nvmn^m'^n is the energy of the emitted
photon. Thus the conventional intensity formula is written

ia = c a (' -1
(D
mn mn •
where I is the intensity of light with the photon energy
hv from sample constituent "a". C is here built by fac-
mn mn
tors which are derived from transition probability, electron
and atomic density, quenching, absorption effects, sample
geometry and apparatus constants.
Y m -L/kT is here the partition sum over all energy levels of
r i i
an atom, and is the statistical weight of an energy
level.

In this paper the new method of organizing spectral line ratio

fluctuations (4,5 and 6) has been used for different
elements. This means that fluctuations of simultaneously
measured line intensity ratios have been organized into a
linear graph of each an inductively coupled plasma (ICP) and
two different kinds of hollow cathode lamps. In this way it
has been shown that it is possible to see which line pair of
analysis is more or less stable at repeated measurements. The
theoretical basis of these results has been developed (7) and
newer theoretical results have recently been presented by Dr.
S. Yngström at the institute (8). The experimental results in
this paper are also strongly supported by a recently published
paper by the author and Dr. S. Yngström, where absolute inten-
sities from 17 elements taken from standard tables have been
studied (9).

Experimenta1

The experimental part concerning the spectral line ratio fluc-

tuations was done at the Swedish Institute for Metals Research
in Stockholm, where a "side-on" mounted inductively coupled
plasma (ICP) light source was used in combination with an
image dissector échelle spectrometer (IDES) system. The same
kind of experimental work was also done with the double
chamber hollow cathode la-np (DCHC) (10) and the flat sample
lamp (FSHC) (11) using the same spectrometer system at the
same institute.
The ICP unit used in this work was a Plasma Therm (2.5 kW)
with an automatic tuning system. The optimised experimental
parameters of this ICP in combination with a new type of slot
type nebulizer and a cyclone spray chamber (12,13), are given
in Table I. The sample used was a diluted steel sample (NBS
363) with the weighings 1g and 5g/100 ml.

A new excitation source for emission spectrochemical analysis

of trace elements in steel, the double chamber hollow cathode
lamp (DCHC), has been used in these studies. The light source
is a water-cooled demountable hollow cathode lamp with a
tubular sample as cathode. Another new excitation source for
spectrochemical analysis of trace elements in steel, the flat
metal sample hollow cathode lamp (FSHC), has also been used in
these studies. The experimental parameters are given in Table
I.

As registration system an image dissector échelle spectrometer

(IDES) system was used (14). This system, which works with
photon counting, is very versatile and consists of an échelle
spectrograph with high resolution, an image dissector tube,
and a minicomputer system.

In the measurements and analyses using the IDES-system, a

computer program for peak measurements was used. In this
program every spectral line is measured for 0.1 s on the top
at each integration. By integrating 15 times, each spectral
line is measured for 1.5 s, which is time long enough to get
accurate results with this spectrometer system.

Results and discussion

In the previous paper a new method of analysis was presented

(6), where fluctuations of ratios between the intensities of
two simultaneously measured spectral lines were studied, using
ICP-liyht sources ("end-on" and "side-on") and iDES-spectro-
meter systems. By forming the ratio between the intensities of
two simultaneously measured lines from sample constituents "a"
and "b" respectively and by using logarithmic differentiation
of equation (1), one obtains
 dT (E
.a b ?a
T nfEm"E

E is here the Boltzmann mean value and is equal to

(^9 -E-e i )/£ge i ) and is estimated from the term-
i i
scheme of each element (15).
The absolute values of the logarithmic derivatives have been
identified as maximum relative deviations observed in the set
of data obtained in a serie of repeated simultaneous measure-
ments of line intensities. By forming ratios of pairs of
simultaneously measured line intensities, the average value of
each line pair ratio is determined. The maximum relative
deviation from the average value of a line pair ratio will be
identified with the absolute value of the left hand side of
equation (2). By using "R-value" as an abbreviation of maximum
relative deviation and by substituting

a
D(E) = |E
I m - Em ,-ET+E
1 (3]

we can thus write equation (2) in the form

(4)
••..•Am'n'*
This means that according to equation (4) it should be mathe-
matically possible to obtain a straight line with the direc-
tion coefficient (1/kT) • R(T) in a diagram when plotting
R(I /I > -) versus D(E) for all studied line pairs of the
elements a and b because equation (4) is a linear differential
equation.
By using very stable ICP light sources when making repeated
intensity measurements of different iron lines with the IDES-
system, the present author discovered that no linear
structure was obtained with D(E) = IE -E ,I in equation (4).
i m mi
An example of this lack of linearity is shown by the circular
points in the RD-graph in Fig (1) of a "side-on" ICK These
points form a cloud with no linear structure at all. The
author then discovered that, if the R-values of the intensity
ratios were plotted versus D(E) = Ihv -hv * ,\ instead, a
i mn m ni
linear structure (triangular points) in Fig (1) was obtained.
Similar results have earlier been presented by the author (6),
using "end-on" and "side-on" ICP:s in combination with
different IDES-spectrometer systems. These results were very
obvious and were performed from experiments that included
much more data than Fig (1), which will serve more as an
example of the Fe-Fe-combination. One might then question if
the temperature and energy dependent exponential factor in the
convential intensity formula in eq (1) is really exp(-E /kT).
Concerning this fluctuation analysis, a more detailed descrip-
tion of the statistical questions is discussed in the previous
paper (6).
By studying RD-graphs of different element combinations, the
present author discovered that the element combinations from
different elements (analysis line/analysis line or analysis
line/iron line) have bigger fluctuations than the element
combinations from the same elements (iron line/iron line). The
author and Dr.S. Yngström at the institute then discovered
that the different element combinations were organized
linearly, if D(E) = |J -J + hv - hv * *| was used in equa-
tion (4), where J is the ionization energy for the different
elements. This can be seen in Fig (2) from an ICP-measurement
("side-on"), where R ( I m n / I m ' n ') is
plotted versus D(E) for
spectral line pairs from various element combinations. Notice
that the slope of the Fe-lines (J -J =0) is the same between
the curves in Figs (1) and (2). To illustrate the influence of
J -J in equation (4) a comparison between Figs (2) and (3)
has been made for different element combinations. In Fig (3)
J -J has been put equal to zero for all element combinations.
As is shown in Fig (3), the circular points do not represent
an ordered pattern as in Fig (2), where the linear pattern is
obvious with a very good correlation coefficient (r=0.90).
Theoretical motivation for introducing the ionization energy
here was proposed by Dr. S. Yngström (7,8).
liv using D(E)=|Ea-E ,-Ea+E I according to the standard inten-
sity formula (1) instead, no linear relationship at all was
obtained (r=-0.17), which is seen in Fig (4) for ICP. To
illustrate the separation between the points in Fig 3, four
examples (points A, B, C, and D) in these plots were chosen.
Point A represents the combination between the line pair
Co Cr
Co3453A - Cr4254Ä with J -J =1.11 eV, and point B represents
the combination between Cr4254Ä-Fe4271Å with J r-JFe=-1.10 eV.
These two examples illustrate big positive translations along
the D(E)-axis, where the same element (Cr) has been combined
with different elements. Points C and D represent the combina-
tion Ni3415A-Fe5167Ä and rio3864A-Fe5167A respectively and
illustrate negative translations along the D(E)-axis, where
J -J =-0.23 eV and J -J =-0.48 eV.

The points in Fig 2 were obtained from 10 repeated determina-

tions of NbS 363 at different occations with the ICP-light
source at the salt concentrations 1 g and 5 g/100 ml (n=20).

This means that each point in Fig 2 represents two salt con-
centrations. In this way the D(E) dependence is more pro-
nounced because the intensity ratio fluctuations are "col-
lected" from two different samples with different salt con-
centrations and the meanvalue of K(I /I n ' n ') *s calculated
and plotted versus D(E). This means that each point in the
graph has "collected" intensity ratio fluctuations from 20
repeated measurements. From this data set it is possible to
compare the two exponents in the graphs. By doing such a com-
parison one must seriously ask oneself how to obtain such
results without changing the standard intensity formula in
equation (1).

Plots similar to the one in Fig (2) of the ICP can be seen in
Figs (5) and (6) of a double chamber hollow cathode lamp
(DCHC) and in Fig (7) of a hollow cathode lamp for flat metal
samples (FSHC). These plots seem to follow as nice a linear
relationship as Fig (2) of the ICP, and they prove that it is
possible to obtain such plots with other types of plasmas,
having different temperatures and plasma densitites than the
ICP. It is also interesting to compare Figs (5) and (6) for
the DCHC-lamp, which clearly demonstrates the linearisation
effect for two different powers (170 W and 250 W) in the lamp.
These graphs clearly demonstrate the temperature dependence

between the slopes (*-=T * R ( T ) ) °f these graphs, where Fig (6)

has less slope depending on the higher temperature compared to
Fig (5).

By using D(E) = |E -E * - I +E | according to the standard

intensity formula (1) instead, no linear relationships at all
were achieved for any of these lamps, in the same way as in
Fig 4, of the ICP.

In the DCHC-lamp a hollow steel sample is sputtered from two

sides, which will produce a long stable plas/na along the
optical axis. In the FSHC-lainp a polished flat steel sample is
sputtered together with a conical graphite electrode. There-
fore, a very stable plasma is achieved. Anyhow, the standing
plasma wave system is still more stable in the DCHC than in
the FSHC and can be operated stably at higher power in the
DCHC (250 W) than in the FSHC (120 W ) .

This means that the DCHC-lamp is probably a more stable lamp

construction than the FSHC-lamp. The fluctuations in concen-
tration between the elements studied is probably the reason
why the relative deviations of the points of the different
element combinations have shown to be higher than those of the
points of only the iron lines. These separations of the com-
binations seem to be more pronounced in the hollow cathode
lamps than in the ICP, where the points are more mixed. This
is probably due to the fact that, when using aerosols in the
ICP-plasma, higher atomization rate with smaller concentration
fluctuations is obtained (Fig 2). In the hollow cathode lamps,
on the other hand, bigger concentration fluctuations were
achieved probably because of the sputtering of solid samples.
In Fig 5 of the DCHC-lamp (170 W ) , four different steel
samples (JK 5A, 7Af 2B, and 1B) were run 5 times each, where
n=20 for each point in the graphs, which include about 200
points each. In Fig 6 (250W) of the same lamp, two different
steel samples (JK 2B and 6A) were run 10 times each, where
n=20 for each point. Fig 6 does not include so many element
combinations as Fig 5, but serves more as an example of the
observed linearisation effect at different powers. For the
FSHC-lainp two different steel samples (NBS 1261, 1262) were
run 10 times each, which gave n=20 for each point in Fig 7.
The spectral lines used in these analyses with the DCHC- and
FSHC-lamps are seen in Table II.

It should be mentioned that the same study as the one of the

different light sources in this paper has also been made of a
single chamber hollow cathode lamp (SCHC) (16) and a glow
discharge lamp (17). The results of these investigations are
exactly the same as the results of this paper, but they are
not yet ready for publication.

A conclusion from Figs 2, 5, 6, and 7 is, therefore, that in

the analytical work it is necessary to select line pairs with
D(E)-values as small as possible to get them temperature inde-
pendent. This can easily be done with the IDES-system, where
line pairs with D(E)-values in the range 0.01-0.30 eV have
been selected without any difficulty because of the high reso-
lution. In this way a very good precision has been obtained at
the work of analysis, where various light sources and the
IDES-system have been combined.

The results from this investigation really show that linear

relationships are obtained for various light sources and
elements when using the exponent exp-(hv +J)/kT. On the other
hand, if exp-(E /kT) from equation (1) is used instead, no
linear structure is obtained. These results are experimentally
obvious and unequivocal, but they contradict the basic theory
of optical emission spectroscopy.
 10

These results are also in agreement with a recently published

paper by the author and S. Yngstrom (9), where absolute inten-
sities from standard intensity tables of seventeen elements
are studied. The result of that paper gives very strong
support to an intensity formula with the above mentioned
exponential factor including the photon energy and the ioniza-
tion energy. The intensity values in these standard tables
were derived from arc- and Geissler tube measurements together
with photographic registrations.

Acknowledgement

The author thanks Dr. S. YngsLröm at KGI for discussions and

important suggestions on the theory basis and Prof. B.
Hultqvist for his kind support of this project. The author
also thanks the staff at the Swedish Institute for Metals
Research in Stockholm for providing laboratory facilities.
 11

Table I.

Optimised experimental parameters

The ICP-light source

Experimental parameters Slot-type nebuliser with

cyclone spray chamber

Power 1.2 kw
Henry frequency 27.12 MHz
Argon cooling gas flow-rate 12 1 min"1
Nebuliser flow-rate of argon 7 1 min"1
Nebuliser pressure 131 Pa
Sample uptake rate 3 ml min"1
Observation height 15 mm

The DCHC-lamp
Flush time 60 s
Pre-burn time 60 s
Pressure, Ne 1 torr
Current 1.1 A
Power 250 W
Sample, tube shape, 3 cm x 4 mm i.d.
 12

Table I.

The FSHC-lamp

Flus time 60 s
Pre-burn time 60 s
Pressure, Ne 0.4 torr
Current 0.03 A
Power 120 W
Sample, flat sample

Spectrometer

Type Echelle (IDES)

Wavelength region 200 - 800 nm
Linear dispersion 0.16 ran mm"1 at 200 nm
0.32 nm mm"1 at 400 nm

Entrance and exit slits 35 \im

Integration time 60 s
Integration time/line 1.5 s
Integration time/ 0.1 s
measured point
 13

Table II.

Spectral 1ines used

Line wavelength (Å)

Cu 3247.54
Ni 3414.77
Cr 4254.33
Mn 4034.49
Al 3961.52
Ål 3092.84
MO 3864.11
Co 3453.51

Fe 3859.91
Fe 3440.61
Fe 3236.22
Fe 5167.49
Fe 3570.10
Fe 4271.76
Fe 3922.91
Fe 2987.29
 14

References

1. Gerlach, W., Z. Anorg. Allgem. Chem. 142, 383 (1925).

2. Barnett, W. B., Fassel, V. A. and Kniseley, R. N.f Spec-
trochimica Acta, 23B, 643 (1968).
3. Suckewer, S., Spectrochimica Acta 26B, 515 (1971).
4. Thelin, B., Kiruna Geophysical Institute, Preprint 063
(1983).
5. Thelin, B., ICP Information Newslett. 10, 835, (1985).
6. Thelin, B., The Analyst (1986) (in press).
7. Yngström, S., and Thelin, B., Kiruna Geophysical Insti-
tute, preprint 062 (1983).
8. Yngström, S., ICP Information Newslett, 10, 818 (1985).
9. Thelin, B., and Yngström, S., Spectrochimica Acta B (1986)
(in press).
10. Berglund, B., and Thelin, B., The Analyst 107, 867 (1982).
11. Thelin, B., Jernkontorets Forskning, Serie D., nr 365
(1981).
12. Thelin, B., Jernkontorets Forskning, Serie D., nr 355
(1981).
13. Thelin, B., The Analyst, 106, 54 (1981).
14. Danielsson, A., and Lindblom, P., Appl. Spectrosc, 30,
151 (1976).
15. Meggers, W. F., Corliss, C. H. and Scribner, F., Tables of
Spectral Line Intensities, National Bureau of Standards,
Monograph 32, Part I, U.S. Government Printing Office,
Washington P.C. (1961).
16. Thelin, B., Jernkontorets Forskning, Serie D, nr 343
(1980).
17. Thelin, B., Jernkontorets Forskning, Serie D, nr 369
(1981).
 C
£
oE
a.

D(E)feV)

Figure 1 Relative fluctuations of spectral line pair

intensity ratios plotted versus D(E)=|E -E A
i in m i
(circles) and D(B)=|nv -hv , ,l (triangles) as
i mn m n i
defined in the text. This data set was obtained
from 20 repeated measurements of Fel line intensi-
ty ratios with a "side-on" ICP.
ol

D(E)(6V)
Figure 2 Relative fluctuations of spectral line pair inten-
sity ratios plotted versus D(E)=|J -J + hv -
- hv > A for a "side-on" ICP. The triangular
points are the same as in Fig. 1. The circular
points describe data from line pairs originating
from two different kinds of metal atoms. These
line pairs have been selected among spectra from
Cr, Mn, Fe, Co, Ni, Cu and Mo. The points A, B, C,
and D describe data from the line pairs Co 3453A -
Cr 4254A, Cr 4254A - Fe 4271 A, Ni 3415A -
Fe 5167A, and Mo 3864A - Fe 5167A respectively.
They represent positive and negative translations
along the D(E)-axis in comparison with Fig. i and
are described in the text.
 D(E)(eV)

Figure 3 Relative fluctuations of spectral line pair inten-

sity ratios plotted versus D ( K ) s | h v m n ~ h v i n ' n ' | w
*tn
the same spectral line data as the ones used in
Fig.2 (ICP). This plot is an illustration of the
fact that no linear relationship is obtained with-
out the atomic part J -J in D(B).
 oE
c
sE

D(E) eV

Figure 4. Kelative fluctuations of spectral line pair inten-

sity ratios plotted versus D(E) = |&^-E^.-É a +É b |.
The same spectral lines as in Fig. 2 were used
(ICP).
 12

10

•# • *. C *• ••

~ 6
. s . •* • •
of •••
a. - . t.* • • • • •
-V ' •• • •
• •

NEW
• • r*0.76

°oL 2
D(eV)

Figure b. RD-plot of the DCHC-lamp (170 W ) . This figure

should be compared with Fig. 2 with D(E) =
a b a
jj -j + hv - hv ^ .I and has been obtained from
20 repeated measurements of 4 tubular metal
samples.
 D(eV)

Figure 6. RD-plot of the DCHC-lamp (250W). This figure

should be compared with Figs. 2 and 5 with D(E) =
|j a -j b + a h v - b h v . A and has been obtained from
I rnn m n i
20 repeated measurements of two tubular metal
samples
Figure 7 RD-plot of the FSHC-lamp. This figure should be
compared with Figs. 2, 5, and 6 with D(E)=
a a
|j -J + hv - hv ' 'I and has been obtained from
i mn in n >
20 repeated measurements of two flat metal
samples.