Analysis of Arc Emission Spectra of Stainless Steel Electric
Arc Furnace Slag Affected by Fluctuating Arc Voltage
Matti Aula,a* Ari Mäkinen,b Timo Fabritiusa
a
Laboratory of Process Metallurgy, Box 4300, 90014 University of Oulu, Finland
b
Electron Spectroscopy Group, Box 3000, 90014 University of Oulu, Finland
Control of chromium oxidation in the electric arc furnace (EAF) is a
significant problem in stainless steel production due to variations
of the chemical compositions in the EAF charge. One potential
method to control chromium oxidation is to analyze the emission
spectrum of the electric arc in order to find indicators of rising
chromium content in slag. The purpose of this study was to
determine if slag composition can be gained by utilizing electric arc
emission spectra in the laboratory environment, despite electric arc
voltage fluctuations and varying slag composition. The purpose of
inducing voltage fluctuation was to simulate changes in the
industrial EAF process. The slag samples were obtained from
Outokumpu Stainless Oy Tornio Works, and three different arc
currents were used. The correlation analysis showed that the
emission spectra offer numerous peak ratios with high correlations
to the X-ray fluorescence-measured slag CrOx/FeOx and MnO/SiO2
ratios. These ratios are useful in determining if the reduction agents
have been depleted in the EAF. The results suggest that analysis of
laboratory-scale electric arc emission spectra is suitable for
indicating the high CrOx or MnO content of the slag despite the
arc fluctuations. Reliable analysis of other slag components was
not successful.
Index Headings: Emission spectroscopy; Direct current arc; DC arc;
Correlation; Slag; Electric arc furnace; EAF.
INTRODUCTION
Interest in stainless steel electric arc furnace (EAF)
slag composition has increased due to the costly
process of chromium oxidation. Steel is produced in
the EAF by melting steel scrap. Chromium is the main
alloying element of stainless steel, and it is usually
charged in the form of stainless steel scrap. Losing
chromium to slag raises production costs and makes the
recycling of EAF slag more difficult.1 Chromium losses
can be prevented by addition of deoxidants, for example,
ferrosilicon and coke. The problem is that the amount of
deoxidants in scrap steel is not known, which causes
major chromium oxidation fluctuations between heats.
The scrap fluctuation problem could be solved with slag
composition analysis methods, which provide informa-
tion on current slag composition. In an ideal case, the
measurement system should provide on-line information
on slag compositions during the EAF process.
Recently, developments in laser-induced breakdown
spectroscopy (LIBS) methods have made it possible to
measure the compositions of the liquid steel and slag in
situ.2,3 Introducing an active measurement device is
Received 18 March 2013; accepted 30 August 2013.
* Author to whom correspondence should be sent. E-mail: Matti.Aula@
oulu.fi.
DOI: 10.1366/13-07079
0003-7028/14/6801-0026/0
26 Volume 68, Number 1, 2014 Q 2014 Society for Applied Spectroscopy
always a risk in the harsh process conditions of the EAF,
which calls for passive methods of on-line slag analysis.
One potential method to accomplish this is to gather and
analyze the light generated by the electric arc used to
melt the scrap.
Optical emission spectroscopy has previously been
used as a real-time process control method in the
coating industry and in laser welding.4,5 The emission
spectrum measurements in steelmaking have focused
on analyzing Linz-Donawitz (LD) and Creusot-Loire
Uddeholm (CLU) converters.6,7 Few attempts have been
made to measure the EAF emission spectra in the
metallurgical industry. Veilette and Simard8 measured
the emission spectrum of a pilot scale furnace for iron
and titanium ore smelting. They found that many of the
lines observed showed signs of self-absorption and
broadening.
Process control with emission spectroscopy is based
on principles used in analytic optical emission spectros-
copy (OES). The difference between analytic OES
equipment and an industrial EAF is the fluctuation of
the emission intensity. The changes in arc current, fume
absorption, and arc visibility cause the emission
intensity to vary. This results in fluctuating intensity of
the emission spectra. The purpose of this study was to
analyze the possibility of gaining meaningful information
about the slag composition ratios despite emission
fluctuations.
MATERIALS AND METHODS
Experimental Setup. The EAF slag emission spec-
trum was measured from a small-scale version of a
direct current (DC) arc furnace. The furnace configura-
tion is shown in Fig. 1. The arc is produced between two
graphite electrodes fixed to water-cooled copper con-
ductors. The sample is placed in a graphite crucible and
inserted into the furnace from below. Light passes
through a fused silica furnace window and is collected
with an optic fiber.
A Kemp Stick Master 1400 welding machine was used
as a power source. The welding machine enables the
control of the arc current, but the arc voltage varies
depending on the distance between the electrodes.
Selecting the arc current and changing the electrode
distance induced voltage fluctuations. The fluctuations
were simulated in the fixed arc currents of 50, 75, and
100 A.
The furnace chamber was purged with argon. The
measurements were conducted by first initiating the arc
with a short electrode distance and then increasing the
electrode distance by manual adjustment. After increas-
APPLIED SPECTROSCOPY

FIG. 1. Schematic picture of the laboratory DC arc furnace.
ing the electrode distance, the sample was raised to the
arc. The measurement device and methods are de-
scribed in detail by Mäkinen et al.9
A commercial Czerny-Turner spectrometer Andor
Shamrock i500 was used in the measurements. The
Shamrock i500 offers multiple grating choices; from
these the grating with the widest spectral range (245–610
nm) was chosen. The selected grating had 150 lines/mm
with 300 nm blaze, and it was used with a 10 lm slit. This
configuration provided a pixel-to-pixel resolution of
0.1782 nm and an optical resolution of 0.525 nm. The
150 lines/mm grating was chosen since gaining infor-
mation on as many peaks as possible was deemed more
important than the accuracy of a single peak. The
spectrometer was used in high-capacity readout mode
with a readout rate of 50 kHz at 16 bit and an exposure
time of 0.0021 s.
Electric Arc Furnace (EAF) Slag Samples. The main
role of slag in stainless steel EAF production is to collect
the impurities in steel scrap. The liquid slag is formed in
the EAF when SiO2 in the furnace is dissolved by the
added CaO. Usually, MgO is also added in the form of
doloma to reduce refractory wear.10 The fourth major
EAF slag component is Al2O3, which enters the EAF as
metallic aluminum mixed in the scrap steel. Aluminum
has a high affinity for oxygen, which means that, at the
end of the heat, virtually all metallic aluminum in the
scrap ends up in the slag as Al2O3. The content of MnO
and Cr2O3 in the slag may also rise depending on the
amount of deoxidants in the EAF.11
The EAF slag samples were obtained from the EAFs at
Outokumpu Stainless Oy, Tornio Works. The samples
were taken from the ladle after the steel and the slag had
been poured out of the EAF and crushed to maximum
grain size of 10 mm.
The particle size and mineralogy of the samples
varied significantly. Some samples had shards of nearly
1 cm in length, and some were the size of very fine sand,
as can be seen in Fig. 2.
Five of the samples were obtained from production
line 1 and the rest from line 2. Table I shows the X-ray
FIG. 2. Two types of slag sample grain sizes: (a) coarse, (b) fine.
fluorescence (XRF) analyzed slag compositions. All
oxidization states of iron and chromium oxides are
summed in Fe2O3 and Cr2O3 content; thus, they are
referred to as FeOx and CrOy. Minor slag components
have been removed from the analysis.
THEORY AND CALCULATION
Correction of Spectral Data. The spectrum obtained
by the spectrometer contains aberrations caused by the
quantum efficiency of the charge-coupled device (CCD),
the efficiency of the grating, and the transmission of the
optic cable. The height of the curve is enhanced by
thermal radiation, which is proportional to the heat of the
matter near the electric arc. Also, due to the dark
current, a certain background intensity can be observed
even if there is no light coming to the CCD.12
Measuring the spectrum with the aperture closed and
removing the reference spectrum from the actual
measured spectrum of the electric arc allowed for the
removal of the dark current. Quantum efficiency of the
CCD, the efficiency of the grating, and the transmission
of the optical cable were corrected according to the data
obtained from the manufacturers. The effect of thermal
radiation was removed by approximating it with a fit
curve and subtracting the curve from the spectrum. The
photon flux caused by thermal radiation can be
approximated with the equation:
2c
Mð; TÞ ¼ 4 ð1Þ
 ½ech=kT  1
where M is the photon flux, T is the temperature in
degrees Kelvin, h is the Planck constant, c is the speed
of light,  is the wavelength in nm, and k is the
Boltzmann constant.13
The exact temperature near the electric arc was not
known, so the photon flux curves of different tempera-
tures were calculated according to Eq. 1, and the best
fitting curve was selected. The temperature interval of
the calculated curves was 5 K. The selection of the best
fit was performed by proportioning the curve to the local
minimum near 509 nm and selecting the fit curve with the
lowest sum of square differences of minimums in the
wavelength ranges 465–482 nm, 425–445 nm, and 515–
526 nm. Figure 3 shows an example of a raw spectrum
for 75 A measurements and its photon flux fit. The
average of the last 15 corrected frames was chosen to
represent each sample in the correlation analysis.
APPLIED SPECTROSCOPY 27
TABLE I. XRF analyzed compositions of 18 slag samples in mass percentages.

Production Sample
line number Cr2O3 Fe2O3 Al2O3 MgO CaO SiO2 MnO TiO2

1 1 7.8 1.5 3.8 6.3 44.4 28.1 3.0 1.7

1 2 8.6 1.5 4.6 4.1 39.7 32.5 4.4 2.2
1 3 5.2 1.7 9.4 9.8 30.4 31.4 2.7 2.9
1 4 7.6 1.6 7.8 8.6 31.8 31.1 4.4 2.1
1 5 5.4 1.3 4.4 6.3 42.5 30.5 2.9 2.1
2 6 2.1 1.3 10.0 4.4 53.5 23.4 0.7 2.9
2 7 2.0 1.4 10.1 4.3 53.4 23.6 0.6 2.4
2 8 2.3 1.2 14.4 5.3 43.3 25.3 1.5 2.7
2 9 8.6 1.3 14.3 3.9 39.6 23.5 3.3 2.2
2 10 8.5 1.4 15.8 3.7 39.8 22.7 3.1 2.2
2 11 8.0 1.4 15.4 3.9 39.6 23.5 3.0 2.3
2 12 5.6 1.5 10.5 4.6 44.7 24.4 2.6 2.0
2 13 7.9 1.6 9.1 4.8 46.1 23.3 2.3 1.9
2 14 2.5 1.2 9.5 5.5 47.9 25.9 1.4 1.9
2 15 2.5 1.3 16.8 6.1 40.7 24.5 1.4 2.2
2 16 5.0 1.5 9.5 8.0 40.3 26.1 1.8 1.8
2 17 4.7 1.3 8.4 7.7 41.3 27.0 2.3 2.0
2 18 2.2 1.3 11.8 9.1 37.4 27.2 1.5 3.1

Analysis of Spectral Data. The two main methods
used to analyze the spectral peak intensity are the
calculations of peak height and area. The use of peak
height is a simple and fast way to analyze peak intensity,
but it is more sensitive to measurement noise.14 In this
study, speed is a more important factor since the aim is
to analyze the spectral data in real time.
Correlation Statistics. The effect of the variation of
the excitation energy and the visibility conditions can be
reduced by using peak intensity ratios instead of
absolute intensities. In this work, the best reference
peak for each component is sought using correlation
analysis, which shows the peak ratios least affected by
the variations of the arc voltage.
The advantage of using the correlation method is that
it allows simultaneous analysis of the usefulness of
multiple peaks. Correlation analysis disregards the
effects of the spectral peak overlap and self-absorption,
which decreases the usefulness of some peaks. There-
fore, correlation analysis can only be used in identifying
the component ratios not requiring the corrections of
overlap or self-absorption. On the other hand, analyzing
each observed peak for overlap and self-absorption
would require a vast amount of effort since the spectrum
has over 200 peaks.
FIG. 3. Example of fitting the thermal radiation photon flux curve to the
emission spectrum.
When the composition of the material is described with
ratios of its components, the reference peak and the
peak of the studied component cannot be separated in
the correlation analysis. The correlation between the
measured peak ratios and the XRF analyzed slag
composition ratios can be analyzed by calculating their
Pearson product momentum correlations using the
equation:
X
ðx  xÞðy  yÞ
r¼X X ð2Þ
ðx  xÞ2 ðy  yÞ2
where r is the Pearson product momentum correlation, x
is the value of the sample population 1, x is the mean
value of the sample population 1, y is the value of the
sample population 2, and y is the mean value of the
sample population 2.
The weakness of the Pearson correlation coefficient is
that it can only find linear correlations. The basic
correlation of the spectral peak to line intensity is linear,
but it can be obscured by self-absorption.15 This results
in decreased correlation coefficients compared to Spear-
man’s rank correlation coefficients, which also considers
non-linear correlations. On the other hand, the prelim-
inary calculation of the Spearman correlations showed
that the lines with the high correlations to the XRF data
did not show signs of self-absorption.
Correlation Significance. Significance of correlations
can be tested with the false discovery rate method. This
method is less conservative in choosing the significant
correlations than the classical family-wise error rate
methods.16 This is especially useful when the number of
correlation tests is as large as in this study.
The tested null hypothesis was the appearance of
correlation between the samples when there is no
correlation between the populations. Variance of corre-
lation has been found to agree with t-distribution when
there is no correlation between the populations. In this
case, the t value for a single correlation coefficient can

28 Volume 68, Number 1, 2014

TABLE II. Amount of peaks and peak ratios used in the correlation
analysis along with the amount of resulting correlations.

Measurement arc current

50 A 75 A 100 A

Number of peaks 218 256 238

Number of peak ratios 46 656 65 536 56 644
Calculated correlations 2 600 640 3 655 680 3 158 736

be calculated using the equation:

pffiffiffiffiffiffiffiffiffiffiffi FIG. 4. Peak ratio giving the highest correlation coefficient to FeOx/
r12 n  2 MnO.
t¼ ð3Þ
1  r212

where r12 is the tested correlation coefficient, and n is the
sample size.17 The significance of the correlations can
then be evaluated by calculating the P values and testing
if they satisfy the criterion proposed by Benjamini and
Hochberg16:
i 
PðiÞ  q ð4Þ
m
where P(i) is the P value corresponding to the order
number i, i is the order number, m is the number of the
correlations, and q* is the decided confidence level.
Correlations of Slag Compositions. The calculation
of correlation coefficients is based on the assumption
that the tested species are independent. This is not true
in compositional data because every component is
related through the sum of compositions being 1.
Components can also have correlations caused by the
EAF processing practice and the chemical reactions in
the EAF. These relations were tested by calculating the
correlation matrix of the slag compositions. The signif-
icant correlations found with Bonferroni correction were
between the slag components CrOx and MnO and also
between SiO2 and Al2O3. The apparent high correlation
between SiO2 and Al2O3 can be explained by the
difference in their content in the slags of different
production lines, which can be observed in Table I. The
correlation of CrOx and MnO content is plausible since
the amount of both CrOx and MnO increase in the slag
when the reducing agents run out during EAF process-
ing. The high correlation of CrOx and MnO content in the
slag samples means that the correlation analysis cannot
discriminate between the peaks of CrOx from MnO and
comparison to atomic spectra databases is needed.
RESULTS AND DISCUSSION
Various peak amounts were observed with different
arc currents. The number of peaks also varied from
sample to sample. Only the spectral peaks, which were
found in all but one sample, were selected for the
correlation analysis. Table II shows the amount of peak
ratios used in the correlation analysis and the amount of
correlations between the peak ratios and the composi-
tion ratios. The amount of peaks for the 100 A current
was slightly reduced due to saturation.
The criterion presented in Eq. 4 was used to find the
limits of correlation significance. The null hypothesis
was rejected for the correlations passing the criterion,
and the correlations were deemed significant. The
rejection region for 50 A measurements was over
0.703, for 75 A over 0.690, and for 100 A over 0.626 with
the confidence level of 0.95.
The best reference component for each analyzed
component would be the component ratio giving the
highest correlations between the spectral peak ratios
and the component ratios. Simply relying on the
correlation coefficients for finding the best reference
was found to be insufficient, which can be seen from Fig.
4 showing the peak ratio having a correlation of 0.98 with
the composition ratio of FeOx/MnO. The correlation is
clearly enhanced by the lack of data with the high FeOx/
MnO ratio.
This behavior was observed only for the component
ratios with MnO as a reference. The reason for the high
correlation shown in Fig. 4 is the small amount of
samples with low manganese content. When these two
data points were removed, the R2 of the linear fit was
reduced to 0.72. The effect was removed by disregarding

TABLE III. Composition ratios having the highest correlations with observed peak ratios.

50 A measurements 75 A measurements 100 A measurements

Analyzed
component Composition ratio Correlation Composition ratio Correlation Composition ratio Correlation

CrOx CrOx/FeOx 0.916 CrOx/FeOx 0.961 CrOx/FeOx 0.961

FeOx FeOx/MnO 0.945 FeOx/MnO 0.926 FeOx/CrOx 0.955
Al2O3 – – – – – –
MgO MgO/MnO 0.929 MgO/MnO 0.922 – –
CaO CaO/CrOx 0.929 CaO/MnO 0.902 CaO/CrOx 0.970
SiO2 SiO2/MnO 0.947 SiO2/MnO 0.955 SiO2/MnO 0.947
MnO MnO/FeOx 0.957 MnO/SiO2 0.956 MnO/SiO2 0.958
TiO2 TiO2/CrOx 0.942 – – TiO2/MnO 0.917

APPLIED SPECTROSCOPY 29
TABLE IV. NIST wavelengths of the validated atomic emission content of the slag are easiest to observe. Moreover, the
peaks found in the peak ratios having high correlations with the ratios mostly reflect changes in the amount of MnO or
slag component ratios.18
CrOx since other components have smaller percentage
differences between the slag samples. An exception to
Current Species Wavelength (nm)
this behavior is Al2O3, but peak ratios with high
50 A Cr I 363.98 correlations to the slag composition ratios including
397.67 Al2O3 were not observed. The lack of Al2O correlations is
579.10 explained by the low number of suitable atomic emission
50 A Mn I 307.0
314.818 lines. The lines at 394.4 and 396.1 nm were merged to the
323.678 lines of Ca II, and the lines near 309.3 nm were merged
325.614 to the lines of Mg I. Other significant Al I lines have high
326.471 excitation energy, and they did not appear at all in some
353.212
354.780
samples.
367.052 The peak ratios of the highest correlations were
398.524 validated by comparing them to the National Institute of
449.890 Standards and Technology (NIST) database for atomic
50 A Mg I 383.829 emission lines. The validation was performed by
517.268
518.360 checking if the peaks were related to the same slag
75 A Cr I 297.548 components according to both the correlations and the
355.852 NIST database. Table IV shows the NIST validated
363.980 peaks.18 Most matches were found for Mn I and Cr I,
364.900
365.626 and some to Mg I. Also, many of the peaks related to
400.144 MnO according to the correlations actually matched to Cr
416.552 I in the NIST database. This is explained by the
435.963 correlation of MnO and CrOx content of the slag.
452.647
453.572
Many potential peak ratios with high correlations to
488.701 ratios of CrOx and MnO content were observed with all
75 A Mn I 291.460 three arc currents. The ratios CrOx/FeOx, MnO/FeOx, and
304.457 MnO/SiO2 have especially high correlations to the peak
307.027 ratios. Percentage differences in CrOx and MnO content
326.471
329.688 of the slag samples were high, which makes it easier to
353.212 perceive the changes in their peak heights despite the
354.803 arc fluctuations. The feasibility of analyzing MnO ratios
356.949 is supported by the successful analysis of manganese
75 A Mg I 517.268
100 A Cr I 297.111 from steel converter hot spot emission spectra.7
398.391 The rest of the peaks attributed to the components
420.359 MgO, FeOx, CaO, SiO2, and TiO2 did not match their
433.945 NIST-documented peaks. Instead, 11 of these peaks
434.451
451.190
matched NIST documented Mg I peaks and two to Mg II
456.551 peaks with a tolerance below the optical resolution of
457.168 0.525 nm. This indicates that the found peak ratios are
459.139 not reliable in observing changes in the amount of FeOx,
464.617 CaO, SiO2, and TiO2. One possible explanation is that the
492.227
526.415 peaks of Mg I and Mg II describe the arc fluctuations
100 A Mn I 324.852 better than the actual MgO content of the slag. This
371.893 would also explain the lack of high correlations to MgO
417.660 content of the slag. The correlations to other components
425.766
are likely caused by the overlapping of their peaks with
Mg I.
The possible causes for a lack of convincing correla-
the samples with low MnO content in the calculations tions for CaO are broadening and self-absorption. These
with MnO as a reference. effects reduce the calculated correlations. CaO content
The correlations were filtered so that only the of the slag is more than 37%, which makes Ca I and Ca II
maximum correlation coefficient of each peak ratio was lines susceptible to self-absorption. The lines of Ca II at
further analyzed. This negates the effect of correlations 393.36 and 363.84 nm could be observed from the
appearing for the different composition ratios containing measurements, but they were not found among the high
the same integer or denominator as the highest correlations. The absence of the most prominent Si I
correlation. The five highest correlations for each slag peak at 288.16 nm does not support analysis of Si ratios.
component were included in the further analysis. Table Also, distinct peaks of Fe I or Ti I were not observed from
III shows the reference components giving the highest the emission spectra.
correlation for each analyzed component. The accuracy of the validated peak ratios was studied
According to Table III, all ratios include either MnO or by approximating the CrOx/FeOx ratio of the slag from
CrOx. This indicates that changes in MnO and CrOx the observed peak ratios. Figure 5 shows the best linear

30 Volume 68, Number 1, 2014

FIG. 5. Relation between the XRF analyzed CrOx/FeOx ratio and the observed peak ratios with different arc currents: (a) 100 A, (b) 75 A, (c) 50 A.
fit curves to the data points with CrOx/FeOx ratio as y
value and peak ratio as x value. The CrOx/FeOx ratio of
the sample could be estimated from the linear fit with an
average relative error of 11.0% for 100 A measurements,
11.4% for 75 A measurements, and 14.3% for 50 A
measurements. Lower accuracy for 50 A measurements
is likely caused by the higher effect of voltage fluctua-
tions on the spectrum.
CrOx/FeOx and MnO/FeOx ratios are especially useful
in analyzing the process conditions of EAF. The iron
oxide content of the slag is relatively constant, so an
increase of ratio indicates an increase of CrOx and MnO
in the slag. This in turn indicates chromium oxidation,
which should be prevented with deoxidant additions. A
relative error of over 10% would be acceptable,
considering there is currently no information on slag
compositions during the EAF processing.
Measuring emission spectra from an industrial EAF
will be considerably more difficult than from a laboratory
furnace. The electric arc is sometimes completely
obscured by the foaming slag in industrial furnaces.
Fortunately, this is less of a problem in stainless steel
EAF because of the low foaming tendency of high CrOx
slag. Also, variation in the melt height will make aiming
the optic fiber difficult.
A completely different challenge is the harsh process
conditions of EAF. The optic fiber has to be protected
from the splashes of liquid slag and flames coming out of
the furnace. The success of analyzing slag composition
ratios in industrial stainless steel EAFs depends on
developing a reliable system for gathering the optic
emission of the electric arc.
CONCLUSIONS
Despite fluctuation of arc voltage and varying sample
compositions, the arc emission spectra provided numer-
ous peak ratios correlating well to CrxOy/FeOx and MnO/
SiO2 ratios of the slag. Numerous Cr I and Mn I lines
related to high correlations were also documented by
NIST. This shows that analyzing CrxOy/FeOx and MnO/
SiO2 ratios holds the most promise for finding indicators
of excessive chromium and manganese oxidation in
industrial EAFs. Some peak ratios correlating to MgO/
MnO could also be validated. Correlating peak ratios
were found also for CaO, SiO2, and TiO2, but comparison
to the NIST database proved these peaks unreliable, as
many of their peaks matched Mg I and Mg II lines.
Convincing peak ratios correlating with the amount of
Al2O3 in the slag were not found. Analyzing other ratios
besides MnO/FeOx, CrxOy/FeOx, and MnO/SiO2 would
require analysis of both self-absorption and broadening,
and possibly spectrometer configuration with a better
optic resolution.
The results show that correlation analysis is a useful
method in providing potential peak ratios for component
ratio analysis. Due to dependent parameters like MnO and
CrxOy content in slag, the correlation method cannot be
used as a sole tool for peak selection, but requires
comparison with the NIST atomic emission spectrum
database to separate the peaks of dependent components.
ACKNOWLEDGMENTS
This research is part of the ‘‘Fluorescence Measurement of EAF’’
project. It is coordinated by the University of Oulu Innovation Centre
and the Laboratory of Process Metallurgy. The Finnish Funding Agency
for Technology and Innovation (TEKES) is acknowledged for funding
this work. Mr. Juho Kunelius and Dr. Paavo Hooli from Outokumpu
Stainless Oy Tornio Works are acknowledged for providing the EAF
samples and the XRF analysis results. The personnel from the Energy
Efficiency and Lifecycle Efficient Metal Processes (ELEMET) research
program are acknowledged for their ideas and groundwork for this
study. Juha Roininen is acknowledged for his support of this study. The
Electron Spectroscopy Research Group at the University of Oulu is
acknowledged for loaning the measuring device. Mr. Henri Tikkala and
Paavo Turunen are acknowledged for their support in conducting the
measurements.
1. V.Y. Risonarta, T. Echterhof, H.P. Jung, C. Beiler, S. Lenz, M.
Kirschen, H. Pfeifer. ‘‘Application of an Off-Gas Analysing System to
Control Oxidation during Stainless Steelmaking in an EAF’’. Steel.
Res. Int. 2010. 81(9): 778-783.
2. J. Gruber, J. Heitz , N. Arnold, D. Däuerle, N. Ramaseder, W. Meyer,
J. Hochörtler, F. Koch. ‘‘In situ Analysis of Metal Melts in
Metallurgic Vacuum Device by Laser-Induced Breakdown Spec-
troscopy’’. Appl. Spectrosc. 2004. 58(4): 457-462.
3. V. Sturm, H.-U. Schmitz, T. Reuter, R. Fleige, R. Noll. ‘‘Fast Vacuum
Slag Analysis in a Steel Works by Laser-Induced Breakdown
Spectroscopy’’. Spectrochim. Acta, Part B. 2008. 63(10): 1167-1170.
4. S. Stojadinović. J. Jovović, M. Petković, R. Vasilić, N. Konjević.
‘‘Spectroscopic and Real-Time Imaging Investigation of Tantalum
Plasma Electrolytic Oxidation’’. Surf. Coat. Technol. 2011. 205(23–
24): 5406-5413.
5. A.R. Konuk, R.G.K.M. Aarts, A. J. Huis in ‘t Veldt, T. Sibillano, D.
Rizzi, A. Ancona. ‘‘Process Control of Stainless Steel Laser Welding
Using an Optical Spectroscopy Sensor’’. Phys. Procedia. 2011.
12(Part A): 744-751.
6. W. Wendt, W. Persson. ‘‘Optical Spectroscopy for the Character-
ization of an Alloy Steel Converting Process’’. Appl. Spectrosc.
1990. 44(6): 987-991.
7. K. Chiba, A. Ono, M. Saeki, M. Yamauchi, M. Kanamoto, T. Ohno.
‘‘Development of Direct Analysis Method for Molten Iron in
Converter: Hotspot Radiation Spectrometry’’. Ironmak. Steelmak.
1993. 20(3): 215-220.
8. P. Veilette, D. Simard. ‘‘Electric Arc Characterization Using High
Speed Cinematography and Spectroscopy’’. 42nd Electr. Furn.
Conf. Proc. Toronto; December 4–7, 1984.
APPLIED SPECTROSCOPY 31
 9. A. Mäkinen, J. Niskanen, H. Tikkala, H. Aksela. ‘‘Optical Emission
from a Small Scale Model Electric Arc Furnace in 250–600 nm
Region’’. Rev. Sci. Instrum. 2013. 84(4): 043111.
10. E.B. Pretorius, R.C. Nunnington. ‘‘Stainless Steel Slag Fundamen-
tals: From Furnace to Tundish’’. Ironmak. Steelmak. 2002. 29(2):
133-139.
11. T.S. Kho, D.R. Swinbourne, B. Blanpain, S. Arnout, D. Langberg.
‘‘Understanding Stainless Steelmaking through Computational
Thermodynamics Part 1: Electric Arc Furnace Melting’’. Miner.
Process. Extr. Metall. Rev. 2010. 119(1): 1-8.
12. J.C. Dunlap, M.M. Blouke, E. Bodegom, R. Widenhorn. ‘‘Modeling
Nonlinear Dark Current Behavior in CCDs’’. IEEE Trans. Electron
Devices. 2012. 59(4): 1144-1122.
13. E.L. Dereniak, D.G. Crowe. Optical Radiation Detectors. New York:
John Wiley and Sons, 1984.
32 Volume 68, Number 1, 2014
14. U. Kale, E. Voigtman. ‘‘Signal Processing of Transient Atomic
Absorption Signals’’. Spectrochim. Acta. Part B. 1995. 50(12): 1531-
1541.
15. M. Miller. ‘‘Basic Concepts in Atomic Emission Spectroscopy’’. In:
A. Montaser, D.W. Golightly, editors. Inductively Coupled Plasma in
Analytic Atomic Spectrometry. VHC Publishers Inc., 1987. Pp 17-68.
16. Y. Benjamini, Y. Hochberg. ‘‘Controlling the False Discovery Rate:
A Practical and Powerful Approach to Multiple Testing’’. J. R.
Statist. Soc. B. 1995. 57(1): 289-300.
17. C.H. Goulden. Methods of Statistical Analysis. New York: John
Wiley and Sons Inc., 1952. 2nd ed. Pp. 130.
18. Y. Ralchenko. A. Kramida, J. Reader, NIST ASD Team. ‘‘NIST
Atomic Spectra Database (version 5.0)’’. 2012. http://physics.nist.
gov/asd. Gaithersburg, MD: National Institute of Standards and
Technology [accessed Jan 15 2013].