Growth of Small Alumina Clusters during Ladle Deoxidation

Mats Söder, Par Jönsson and Lage Jonsson

Division of Applied Process Metallurgy, Department of Materials Science and Engineering,

KTH, SE-100 44 Stockholm, Sweden

(Received October 23,2009)

ABSTRACT today for steel grades of low oxygen content,

deoxidation with aluminium has become a standard
For the first time, the size distribution of small procedure. Aluminium is chosen since it is a strong
clusters was determined in a systematic manner from deoxidant and relatively inexpensive. The deoxidation
samples taken during deoxidaiion with aluminium in an product, alumina inclusions, are solid oxides. By
induction-stirred ladle. Chemical and microscopic collision, the inclusions grow in a branched, loose
analyses of samples were performed. The results manner and become clusters.
indicated rapid cluster growth throughout the Much work has been done to mathematically
experiment and at the end of the experiment a majority describe the deoxidation process. Lindborg and Torsell
of the small inclusions were found to be bound in /l' conducted ground-breaking work and mathematically
clusters. The cluster growth was concluded to be by described a laboratory deoxidation experiment which
collision. Growth-model calculations were carried out to used silicon as a deoxidant. Inclusion growth due to
verify the experimental findings. In order to perform collision promoted by buoyancy differences (Stokes)
these calculations, a method first needed to be and turbulent gradients (Saffmann and Turner) was
developed for conversion of the experimental data modelled. As the formed silica is liquid, the radius of
determined per unit area to data given per unit volume. the collision product, >, was set to rk = (/·,3+ 73)1/3 in
Next, the cluster collision diameter was modelled using their model.
a theory for fractal aggregates. The results from the Since then several researchers have also modelled
growth-model calculations, simulating deoxidation with deoxidation with aluminium. Models of ladles'2-57, RH-
aluminium, were found to agree well with the degassers /6, 7/, and tundishes /7-9/ exist. A common
experimental results. characteristic of the different models is that the size of
growth products is calculated according to the Lindborg
and Torsells III model that used silicon. The branched,
Keywords: Deoxidation, Alumina clusters, Growth, more voluminous, cluster behaviour of alumina is
Collisions, Inclusions, Ladle therefore mostly not taken into account. In some models
/7, 9/, however, it has been considered in modelling the
flotation of clusters. In one previous case, a tundish
1. INTRODUCTION model by Tozawa et al. /10/, inclusion growth which
considered cluster size based on the theory of fractal
Deoxidation is an important unit operation in aggregates was calculated.
secondary metallurgy. Because of the high demand

347
l'ol 28, ΛΌ. 6. 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

Commonly when the subject of clusters is dealt with 2. EXPERIMENTAL PROCEDURE

in the literature, it is with respect to very large clusters,
which often appear late in the steel production process A full-scale experiment was carried out at the ladle
(usually in tundishes and during casting). This study, furnace station at Ovako Steel AB's steel plant in
however, focused on the growth of inclusions and Hofors, Sweden. The mill is a scrap-based special steel
formation of initial small clusters through examination mill, where about 100 tonnes of scrap is melted in an
of samples taken during the initial ten minutes of a electric arc furnace (EAF). The molten steel is tapped
deoxidation experiment in an induction-stirred plant into the ladle and transferred to the slag removal unit,
ladle. The experimental procedure employed is where the EAF slag is removed by raking. Thereafter,
described in the following section of the report. Next, the ladle is transferred to the ASEA-SKF ladle furnace
the counting or tallying of inclusions and clusters on the station, where aluminium for deoxidation, other alloys
examined polished samples with an optical microscope and synthetic slag are added. After the initial ladle
and the importance of using the appropriate treatment, the steel is vacuum degassed in order to
magnification (found to be 500x) is presented. For the remove, for example, sulphur and hydrogen. Finally,
counting, a technique based on a modification of the some additional stirring is carried out to promote
Swedish standard SS 11 11 16 /I I/ was used. The inclusion removal before the steel is cast using up-hill
composition of the steel samples was determined, teaming. The final bearing-steel product usually
specifically the aluminium and total-oxygen contents, contains an average total oxygen amount of 5 ppm /12/.
and from these data the dissolved oxygen content was
calculated. Later, growth calculations were carried out,
2.2 Plant Trials
in which the decrease in the calculated dissolved oxygen
content was partly used as input data to the model. The The full-scale experiment was conducted during the
oxygen bound to inclusions was calculated as the initial ladle treatment. Aluminium bars placed on the
difference between total and dissolved oxygen and was bottom of the empty ladle were used for pre-
compared with oxygen found in inclusions from the deoxidation. After tapping of the steel from the electric
microscopic examination. arc furnace, the remaining EAF slag was first
Information on inclusion and cluster growth has thoroughly removed at the slag raking station. New
been of interest to the steelmaking community for a synthetic slag was then added and melted at the ladle
long time, as mentioned earlier, and is the reason the furnace. Thereafter, the ladle was moved to a nearby
study included calculations of growth by turbulent location where the experiment was carried out. This
collision. In order to calculate cluster growth, it was point in time is equal to the initial time in the
necessary to know the size of the cluster, i.e. collision deoxidation process when the ladle has just arrived at
area. Equations based on a theory for fractal aggregates the ladle furnace. Upwards stirring by induction at 1000
were therefore formulated. The model predictions were A was used during the whole sampling procedure.
then compared with the results from the microscopic Sampling was started with an initial sample in order to
examination of the samples. Note that in order to obtain a reference value. Thereafter aluminium wire (0
compare, the growth-calculation results with the 15 mm) was fed into the steel. A separate wire feeder
microscopic results, conversion of the experimental data than the one used in regular operation was used due to
from number per area to number per volume was the special experimental location. In total, about 65 kg
required. This was a highly complex task and was at this aluminium (-130 m) was fed into the steel during the
point only developed to consider t2 clusters on the experiment. During feeding of the aluminium wire, four
sample plane. Finally, four collision-growth calculations samples were taken. After the aluminium wire feeding,
based on different assumptions were performed and the two more samples were taken. Thus, iriTotal there -were
results compared to the experimental-findings. seven sampling. The sampling protocol for the
experiment is given in Table 1 below.

348
Mats S der, Par Jonsson and Lage Jonsson High Temperature Materials and Processes

Table 1 3. SIZE DISTRIBUTION

Sampling Protocol.
A microscope is the primary instrument employed
Time Activity Al (m)/(kg) for determining the amount of inclusions in a steel for
quality classification and production control. As the
0.00 First sampling before AI 0/0
microscopic methods using standards like ASTME4S
feeding
/I4/, Stahl Eisen's standard SEP 1570-71 /15/ and SS 11
2.03 Sampling 42.9/21.5
11 16 /I I/, have been developed for finished steel
3.37 Sampling 73.7/36.9 products they are adapted to deformed steels. During
deformation, alumina clusters break up and form
4.29 Al feeding stopped due to
elongated strings, which are characterised by length and
problem with feeding
width. For the undeformed steel samples taken from the
4.44 Al feeding started again
ladle in this study, a different method to describe their
5.09 Sampling 98.1/49 surface representation was necessary to use. It should be
6.42 Sampling noted that observed clusters on the sample surface are
seen as random plane cuts through three-dimensional
7.03 Al stopped 129.9/65 structures. The number of inclusions visible in an
8.16 Sampling identified cluster is therefore not necessarily the total
number that actually comprises the cluster. This will be
9.47 Sampling further discussed later in the paper.

2.2 Sampling Equipment 3.1 Inclusion Counting Method

The samples were taken with an automatic sampler Before any counting could be performed, the
at a position 0.6 m below the steel surface and a little bit samples were carefully ground and polished. After this,
beyond the centre, as viewed from the inductive stirrer. an area of 400 mm2 on the samples was examined with a
A Rescon disc-pin type of sampler, model DPS light optical microscope (LOM) at 500x magnification.
1100/900 /13/, was used. This is a specially designed Individual inclusions were assessed using the Swedish
sampler with an extra long splash protection shield standard SS 11 11 16 (JK Scale II) method /I I/.
covering the paper tube. The sample mould was Clusters observed on the surface were classified by the
protected from slag and steel inlet during immersion by number and size of the inclusions of which they were
purging argon gas through the mould. At the right comprised.
sampling depth, the gas was switched off and the mould For the round type of inclusions (i.e. "undeformable
was filled by suction. This procedure ensured that the inclusions, D type") that are of interest here, the SS 11
steel sample was taken at the targeted position. 11 16 method /I I/ classifies size by thin (T), medium
(M), heavy (H) and particular (P). According to the
standard, the sizes are "instrument dependent". In this
2.3 Chemical Analysis Methods investigation, the size ranges used were 1.7-3.5 μπι, 3.5-
Total oxygen content was determined using the 6.9 μπι, 6.9-13.8 μιη and >13.8 μηι. However, for
inert-gas fusion method with a Str hlein ON-MAT inclusions in clusters, inclusions smaller than 1.7 μιη
8500. Each sample was analysed two times. The lower were counted and the smallest size was 1 μπι. The
of the two values obtained was chosen to represent the number of inclusions in a visual field is described in the
total oxygen content. Chemical composition was SS 11 11 16 standard with an index, x, that gives the
determined by optical emission spectroscopy with a number of inclusions according to 2 X ~'. For example the
Bausch & Lomb ARL OES 3560 instrument. code T3 means that four Τ inclusions are observed in
the visual field. For clusters a different notation was

349
Vol. 28, ΛΌ. 6. 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

used. Individual clusters were instead classified by clusters as a function of process time could easily be
observed inclusion sizes and number of inclusions. For observed. There was a great increase in these numbers
example, the code M1T4 indicates that the cluster is from the sample taken before the aluminium wire
made up of one Μ and four Τ inclusions. For clusters, addition to the first sample taken after the start of
the microscope employed was equipped with a digital deoxidation. At this early stage, the increase in the
camera and pictures were taken of some selected number of small clusters was substantial. Later, as the
clusters. deoxidation experiment proceeded, the increase in the
number of inclusions included in the clusters was
pronounced. In Figure 2 some pictures illustrating the
3.2 The Role of Magnification in Viewing Clusters
cluster growth are presented.
The magnification used in examining the samples
was 500x, though 200x is more commonly used in the
SS 11 11 16 method. The difference between 200x and
500x magnification can be seen in Figure 1. At 200x
magnification in the left picture (a), the cluster appears
to contain two inclusions. When magnification is
increased to 500x (picture b), the cluster instead is
observed as containing four inclusions on the surface.
Thus, it is important to view inclusions using different
magnification in order to get a more complete
description of the clusters.

a) b)

b)

c)

ΙΟμιρ

Fig. 1: Cluster viewed at 200x (a) and 500x(b)

magnification.
* ΙΟμπι
3.3 Cluster Evolution during Deoxidation
Fig. 2: Clusters of different sizes seen on the surface of
Using the modified counting technique, the number
a polished steel sample. Photo a) t2-cluster, b) t4
inclusions and clusters were counted using the
cluster and c) tl 1 cluster.
microscope. An increase in the number and size of

350
Mats Söder, Par Jönsson and Lage Jonsson High Temperature Materials and Processes

Note that the clusters seen on the surface can easily as the number observed. Clusters containing more
contain more inclusions since the cluster is not two, but inclusions than four were summed up and presented in
three dimensional. The number of inclusions in a cluster the >t4 group. It can be seen that single inclusions (T)
is presented as the number Observed on the sample first increase rapidly when the Al feeding is initiated.
surface' (2D). This can be expected since many inclusions are formed
There was a significant increase in the number of when the dissolved oxygen content decreases (further
inclusions when the aluminium feeding began. The discussed in section 4). Also, an increase in the number
number of single small t inclusions increased almost of all small clusters can be observed during the whole
threefold and t2 clusters more than fivefold. This experiment. Noteworthy, is the relatively rapid increase
increase is plotted in Figure 3. In order to make a in the amount of clusters larger than t4 after the stop of
comprehensible presentation of this information, the aluminium addition (the two last sampling time points).
number of inclusions of size T, t2, t3 and t4 is presented
1.4 -

4 6 10
Time (min)

Fig. 3: No. of single small inclusions (T) and small clusters as function of time. Numbers determined with microscope.
Dotted vertical line indicates end of aluminium addition (start at time zero)

Here, it is important to note that the diameter of a The single T inclusions and t2 clusters seem to be
cluster influences the likelihood of it being observed on connected to the aluminium feeding as they increase in
the sample surface. This relationship between number of number during feeding and thereafter decrease at the
inclusions of a certain size in the steel volume Nv, and time point of the last sampling. During the deoxidation
the number visible on the surface NA, can be better process there is a shift towards larger size clusters. In
understood by studying Equation l /16/, where the Figure 4 it can be seen that the percentage of t2 clusters
diameter is taken to be the external diameter of the decreases in relation to the total amount of clusters
cluster, D,, i.e. the diameter of a sphere that surrounds during and after the aluminium feeding. When the
the whole cluster. If the number of clusters in two size aluminium feeding is stopped, the relative number of
groups increase and the relation between them per unit inclusions bound to clusters containing five or more
volume stays the same, the number of clusters in the inclusions increases. Similarly, Feng /17/ observed in
surface will increase for the larger sized as the quotient laboratory deoxidation experiments an increase in the
between the diameters of the clusters in the larger and size of clusters.
the smaller size group.

N - (1)
" ~ D.

351
Vol. 28, No. 6. 2009 Growth of Small Alumina Clusters during Ladle Deoxidalion

-»-t2 clusters (l/mm2)

Percent t-inclusions in t2
clusters of total cluster
amount

- Percent t-inclusions in
clusters that sits in
clusters containing five
or more inclusions

10

Fig. 4: No. oft?, clusters and percentage t inclusions in clusters compared to total cluster amount.

2.5 τ—

T
M
Tic
Mic

Fig. 5: Number of inclusions bound and unbound to clusters. T and M denote the single inclusions of size T (1.7-3.5
μπι) and Μ (3.5-7 μιη). Tic and Mic are the number of T- and M-size inclusions bound to a cluster.

The number of inclusions bound to a cluster 4. ALUMINIUM AND TOTAL OXYGEN

increased significantly throughout the experiment. In CONTENTS
Figure 5 this is quite clear for T-size inclusions and can
also be seen as the case for M-size inclusions. The amount of alumina inclusions that are formed
Interestingly the increase in cluster-bound Τ inclusions during deoxidation is controlled by the amount of
(Tic) seems to stop after aluminium feeding ceases. At aluminium and oxygen in the steel. These amounts are
the same time, the number of single Τ inclusions thereby connected to the number of inclusions and
decreases. These two observations can be interpreted as clusters that can be found on a sample surface. The
effects of the collision growth of inclusions in the steel. dissolved oxygen amount was calculated using a mass

352
Mats Söder. Par Jönsson and Lage Jonsson High Temperature Materials and Processes

balance between aluminium and total-oxygen As previously mentioned, the steel was first
concentrations determined by chemical analysis. The deoxidised upon tapping of the electric arc furnace with
amount of oxygen bound to inclusions calculated as the aluminium bars placed on the bottom of the empty ladle.
difference between total and dissolved oxygen was The added aluminium resulted in a total aluminium
compared with the amount of oxygen calculated based content of 0.007 % by weight at the start of the
on inclusions from the microscopic examination. Later, experiment (Figure 6). Thereafter, a total of 65 kg
in the growth-model calculations, the decrease in the aluminium wire was fed into the steel for seven minutes,
dissolved oxygen content was partly used as input data which resulted in a total aluminium content of 0.057 %.
to the model. This corresponds to an aluminium yield of 77 %.

0.06 0.0045

0.05 -

0.04 -
Total Al
Iu 0.03 - Total oxygen
• Oxygen equilibrium
•a 0.02 -
I 0.01 -

0 2 4 6 8 10
Time (min)
Fig. 6: Total aluminium and oxygen contents determined from chemical analysis and calculated dissolved oxygen
content plotted as a function of time (% by weight). Plotted total oxygen content is lowest of two values
determined from each sample.

One effect of the deoxidation is seen in plotting the The activity of oxygen and aluminium was
results for total oxygen content. As shown in Figure 6, calculated based on the interaction with elements in the
the oxygen content is initially 40 ppm, a result of the steel. The element concentrations are provided in Table
first deoxidation upon tapping of the EAF. The total 2. The interaction parameters were taken from Elliott et
oxygen then drops to 19 ppm at the end of the al. /19/. The dissolved aluminium content was
aluminium-wire feeding. This value corresponds very iteratively calculated from the determined total
well to the average total oxygen value of 19 ppm aluminium content to balance with the dissolved oxygen
reported by Huet et al. /12/ for samples taken three content. A dilute-solution approximation was used in
minutes after the end of aluminium feeding at Ovako the calculations, where the activity of alumina was
Steel AB. Thus, the experimental heat is representative assumed to be one and the steel temperature was
of normal production conditions. assumed to be 1600°C.
The dissolved oxygen content was calculated using
the following equation: Table 2
Steel Chemical Composition (percent by weight)
2AJ + 30 = A12O3 (s) (2)
c Si Mn Cr Mo Ti
AG° = -1202050 + 386.3T (J/mol)/18/ (3) 0.61 0.28 0.27 0.41 0.03 0.0008

353
Vol. 28. ΛΌ. 6, 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

An observation in Figure 6 that clearly stands out is inclusions on a sample surface. The amount of bound
the initial rapid drop in total oxygen. This corresponds oxygen was calculated using Equation 4 for the
to a rapid removal of alumina deoxidation products samples taken in the study and compared with the
from the steel. Calculated as alumina, the oxygen drop difference between the determined total oxygen content
is equivalent to 3.4 kg for the whole ladle - or as seen on and calculated dissolved oxygen content in Figure 7.
a surface: 12.8/mm2 as 3 μιη inclusions. The mechanism
behind this initial inclusion removal is not fully known
by the authors but some possible explanations are ΟΛ (4)
suggested later in the report.
The oxygen amount that is equal to the difference
between the determined total oxygen content -and where ΜΛΠΛ/Λ« is the molar weight of oxygen in
calculated dissolved oxygen content is bound to alumina, ρΑηο3 is the density of alumina (3990 g/m3)
inclusions. This bound oxygen amount is seen as /20/, F,Ki is the area fraction of inclusions, MAno3 is the
inclusions on the polished samples. This means that it molar weight of alumina and p/.-e is the steel density
can be calculated from the observed fraction of (7100kg/m3)/20/.

-·- Oxygen difference

between total and
dissolved amount (ppm)
* Oxygen in inclusions as
A12O3 (ppm)

2 4 6 10
Time (min)

Fig. 7: Difference between total oxygen and dissolved oxygen amounts shown in Fig. 6 compared with oxygen amount
calculated to be contained in A12O3 inclusions.

The curve for the calculated oxygen in inclusions (as area fraction. It is dependent on the size and number of
A12O3) shows quite good agreement with the difference inclusions. The very large inclusions of size P are
between total and dissolved oxygen. The trends are defined by their specific size and individually
declining in both curves although not so pronounced for substantially contribute to the area fraction. The
the difference between total and dissolved oxygen. The percentage contribution to the area fraction from the
curve for the calculated oxygen amount in inclusions different size classes is plotted in Figure 8. It can be
should lie under the curve for the difference between seen that large P-type inclusions give a large
total and dissolved oxygen as it is reasonable to believe contribution initially and this steadily decreases through
that not all inclusions existing on a sample surface will the duration of the experiment. Further it is seen that the
be identified. This will be discussed further below. contribution from small inclusions increases during the
It should also be noted that the calculated amount of aluminium feeding.
oxygen in inclusions as alumina is based on the surface-

354
Mats S der, Par J nsson and Lage Jonsson High Temperature Materials and Processes

a? 90 :
80

0 « D T , 1.7-3.5 μπι
60 -
ο + DM, 3.5-6.9 μπι
50 --
···* DH, 6.9-13.8 μπι
DP, >13.8 μπι

2 4 6 10
Time (min)

Fig. 8: Proportion of the different size classes to surface-area fraction of inclusions. Inclusions in clusters individually
counted.

The calculated oxygen amount in inclusions is the amount and size of clusters in the studied part of the
composition dependent. Composition affects the deoxidation period was seen. Thus, it is vital that this
stoichiometric and the density of the oxides. In the type of growth must be taken into account in any growth
microscope it was observed that many of the large P model for alumina inclusions. With this in mind, model
inclusions were perfectly round. This means that they calculations were performed in order to see if similar
were liquid and, based on the current experimental results to what was found in sample examination by
conditions, were most likely composed of a high microscope would be obtained. In order to do this, it
amount of silicon oxides. Huet et al. /12/ reported that was first necessary to determine cluster size regarding
Al2O3-SiO2-MnO inclusions of size 6 to 40 micrometers growth by collision. It was also essential to be able to
that originate from predeoxidation with FeSi are present convert the results from the microscopic analysis, which
during this part of ladle treatment. Silica inclusions are determined in number per unit area, to number per
would render a lower oxygen amount in the inclusions unit volume as used in the calculations of the growth
due to the low density of these inclusions. This would model.
probably lead to the first part of the curve for the
calculated oxygen amount in inclusions being under the
5.1 Turbulent Growth
curve for the difference between total and dissolved
oxygen in Figure 7 since the proportion of P inclusions Due to the induction stirring of the steel in the ladle,
in the area fraction is initially large (Figure 8). the fluid flow is turbulent. The turbulence causes
velocity gradients
r
which give rise to collisions between
the inclusions. The number of turbulent collisions may
S. CALCULATION OF CLUSTER GROWTH be calculated based on Saffman and Turners I2\l
equation:
As mentioned in the introduction, many researchers
have developed models of inclusion growth, but only a dny
few have taken cluster growth into consideration in their (5)
~dt
calculations. In the results of the microscopic
examination of samples in this study a clear increase in

355
Vol. 28, No. 6, 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

where α is the collision efficiency, /·, and r, are the radii is the same in the equation, the number of inclusions
of the. two colliding particles, ε is the local turbulent that comprise a cluster is determined by the prefactor ke.
energy dissipation of the steel in the ladle and v^c is the The prefactor then describes the number density of the
kinematic viscosity cluster given a certain diameter.

5.2.2 External Diameter and Collision Area of Small

5.2 Derivation of Cluster Size Model
Clusters
In Equation 5 the radii of the colliding particles It is of vital importance in modelling growth by
affects the collision rate in a cubic relation. It is collision that the size of the clusters be known since the
therefore necessary for accurate calculation of the collision rate is dependent on the collision area. The size
collision growth of inclusions and clusters to know the of clusters in the calculation is controlled by both the
geometrical relationship between the number of power-law prefactor, ke, and the fractal dimension, Df.
inclusions in a cluster and its size. It should also be In the work of Tozawa et al./10, 221 it was
noted that cluster-removal models developed in the concluded that the fractal dimension should have a
future would need to consider the geometry of clusters. value of 1.8. Note, that the fractal dimension has not
Removal was however not dealt with in this work. been measured for alumina clusters. However, it seems
reasonable that the fractal dimension would be
5.2.1 Alumina Clusters as Fractal Aggregates approximately 1.8 since both simulations 724, 257 and
In nature and in industrial applications, fine particles experiments 726-287 with systems resembling the
and similar systems are known to form fractal alumina system have produced values of around 1.8.
aggregates22. Alumina inclusions in steel seem to be no The actual size of the clusters should be derived
exception. Some researchers have considered this when from measurements of real clusters. For small clusters,
describing the behaviour of clusters 79, 10, 237. In the which are the focus of this research, an expression for
theory of fractal aggregates the size-number relationship the external diameter was derived from the form of the
is: t2 cluster. Setting the fractal dimension to 1.8, the
prefactor ke then becomes 0.57 for the case where the
cluster inclusions are the same size (Equation 6). Based
(6) on the two-inclusion cluster case, and rearranging the
equation for fractals, the external diameter can then be
written as:
where Ν is the number of inclusions, ke is the power law
prefactor or numerical density, Kc is the external radius,
r, is the inclusion radius and Dfis the fractal dimension. (7)
The external radius, /?e, is the radius of a sphere that
surrounds the whole cluster. The fractal dimension has a
value between 1 and 3. However, this diameter cannot be used in a collision
The number of inclusions in the cluster is described model since it would lead to overestimation of the
by the fractal dimension and the ke p'efactor. In the cluster collision area. The degree of overestimation can
physical system, the number of inclusions is determined be exemplified by the two-inclusion cluster case. In
by the clustering behaviour of alumina inclusions. For Figure 9 the collision area is shown for two different
example, the size of the individual inclusions, cuts through the two-inclusion cluster. In the χ , y and ζ
coordination numbers, deformability of formed cluster directions, the dimension varies from d, to "id,. A
structure, sintering, collision mechanisms, etc. all affect collision diameter can then be defined using the average
the size-number relationship of a cluster. Thus, it is diameter in the three directions:
understood that clusters can have the same fractal
dimension but different numbers of inclusions. When D

356
Mats Soder, Par Jonsson and Lage Jonsson High Temperature Materials and Processes

from these pictures. The number of inclusions in a

cluster seen on the polished surface was then
recalculated to the number of inclusions in the whole
cluster as follows:

(10)
Fig. 9: The collision dimensions in the Χ, Υ and Ζ Id.
directions.
The equation uses the relation between the number
LM 4 of inclusions on a surface and in the volume similar to
(8)
3 ' 3 ' what was presented in Equation 1. The number of
inclusions in a cluster is determined based on the cluster
This diameter can subsequently be used to formulate an area and the volume defined by the cluster diameter, De.
expression for the collision diameter: It is also assumed in using the equation that i) the
inclusions are of the same size, ii) the cluster is cut
through the centre, and iii) the inclusion number density
(9) seen on the surface is the same for the entire cluster.
In Figure 10 is the recalculated number of
inclusions in the cluster plotted as a function of the
The A,, prefactor for this case is equal to 1.19 when Ν = observed external diameter. It can be seen that for the
2 and Re = 4/3r, in Equation 6. smaller clusters, the number of inclusions is higher than
A comparison was made in order to examine how the line for the external diameter plotted using
the theoretically derived diameters Dt and Dc Equation 7. The equation for the collision diameter,
correspond to actual observation of cluster sizes. Some DC, describes the size of the small clusters better.
clusters were photographed, as illustrated earlier in However, the number of inclusions for the small
Figure 2. This was done mainly for the largest clusters, clusters is most likely somewhat overestimated as their
but for the sake of completeness, also for a few small shape is more two dimensional. The values determined
ones. Thereafter, the external diameter, Dn was for larger clusters lie between the two lines.
measured and the number of inclusions, nA, was counted

1000 .. ::

10 100 1000
External diameter (μιτι)

Fig. 10: Number of inclusions seen in cluster on polished sample surface converted to number of inclusions in 3D
derived from Nv=NA/d, relationship. Upper line defined by Dc and lower (bold) by De.

357
Vol. 28, No. 6. 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

problem and can be compensated for as suggested by,

5.3 Conversion of Data Per Unit Area To Data Per
for example, Underwood16.
Unit Volume
Up to this point in the report, determination of
cluster size distribution discussed has been based on
results from counting what could actually be observed
on the two dimensional sample surfaces (section 3.1).
However, these results from sample examination by
microscope cannot be used directly as starting values in
a growth model, as they are given in number per square
mm (I/mm 2 ). They need to be converted to number per Fig. 11: A "false" diameter viewed on the surface of a
unit volume (1/m3). In addition, the surface polished steel sample.
representation of clusters must be considered. This
section presents how a derivation of the surface When it comes to counting alumina inclusions that
representation of t2 clusters can be done and how have collided before the steel solidified, i.e. formed
convertion of the data given per unit volume can be clusters, the problem is further complicated because not
performed. It should be noted that a clear understanding all clusters on the surface appear as clusters. Thus, in
of surface representation of clusters is vital when order to obtain more realistic numbers of the size
making comparisons between growth-model results and distribution of clusters in a sample based on information
microscope results. from microscopic assessment, a more reliable method of
determining cluster size distribution had to be
5.3.1 Probability of Observing Small T2 Clusters developed. In the following text, the developed method
When determining the size distribution of alumina for converting the two-dimensional information
inclusions on a polished steel sample by microscope, (I/mm 2 ) to three-dimensional (I/mm 3 ) for the case
some errors arise due to the fact that the cuts through where a cluster contains two inclusions is presented.
the inclusions are rarely made precisely through the The fact that not all clusters in the surface are seen
centre of the inclusions. As a result, a "false" diameter as clusters is illustrated in Figure 12 for clusters made
will be seen as illustrated in Figure 11. This is a known up of two inclusions, here called t2 clusters.

XXX

Fig. 12: Clusters containing two inclusions on surface of polished steel sample. Note that sample surface must be
ground down to grey area (a) for the t2 cluster to be seen.

If the cluster in Figure 12a is to be viewable as a t2 from number of t2 clusters seen on the surface to
cluster in a plane cut, the sample must be ground down number of t2 clusters per unit volume is to be made, it is
to the grey area. Note that the height of the grey area is absolutely necessary to know the proportion of t2
a little bit less than the inclusion overlap in the grinding clusters that should be visible on the surface. The
direction since there is a minimum size of inclusions probability that an observed cluster is a t2 cluster is
that can be seen through a microscope. If a conversion calculated below in converting a number actually seen

358
Mats Soder, Par Jonsson and Lage Jonsson High Temperature Materials and Processes

on a sample (per mm2) to a corresponding number per simulated or substantiated by making an integration
volume (per mm3). In doing this, the following calculation. However, here it is of interest to convert the
assumptions were made: results of what is actually seen on the surface to what
can be found in a volume (2D to 3D or I/mm 2 to
1. The inclusions are spherical I/mm 3 ) and vice versa. As there is a minimum diameter
2. The two inclusions in the t2 cluster have the same for when an inclusion can clearly be observed, a reliable
radius length convertion should take a visible diameter into
3. The inclusions are randomly orientated in the steel consideration. Furthermore it should be related to a
matrix certain depth in order to relate what is seen on the
4. There is a minimum diameter that can be seen with surface to what is under the surface. The likelihood of
the microscope seeing a t2 cluster on the surface with the microscope
was here defined as equivalent to the quotient of two
The validity of these assumptions is discussed later sums. The first sum represents the sum of all heights, hn
in the report. at all positions on a sphere that gives a visual t2 cluster
By studying one inclusion as a sphere and assuming for all angles (δ,). The second sum is the sum of the
that a connected inclusion of the same size can sit on the total height, //,„„ (or depth if related to the polished
sphere at any position, the average angle to the surface surface) equal to a circum-ambient diameter of the t2
of the polished sample can be determined to have a cluster, i.e. here 4r and (later called De) for all positions
value of 30°. This angle is illustrated in Figure 13. At a around the sphere. The number of connection positions
30° angle, the distance in the z direction of one of the for the inclusions at a certain angle is related to the
inclusions is equivalent to half the radius length. This circumference of the sphere, C, perpendicular to the ζ
average contact point is located as indicated by the axis. The summation is done by integrating the product
dashed line where the area Al equals A2 on the half- of the heights and the circumference with respect to the
sphere of the lower inclusion. angle, δ. The nomenclature and geometric relationships
used in the derivation are defined in Figure 14.
As mentioned above, the angle of the t2 cluster must
be large enough to produce a segment diameter for both
inclusions that is greater than the smallest viewable
diameter (2<z). This angle, δη can be expressed as

δ ν = arcsin (α//·)δ (Π)

It was estimated that the smallest diameter that can be

seen with the microscope is 1 μπι in length and the
average size of a T inclusion is 3 μπι. For this case the δ
Fig. 13: Average angle t2-cluster radii form with „ angle is 20°. The smallest possible δ,. value is taken
sample surface. into account in the derivation below.
As seen in Figure 14, the height hv can be expressed
The likelihood that both inclusions of a t2 cluster on as:
the surface have been cut through can be calculated
from the geometric relationship illustrated in Figure 13 = 2l-2rcoso -a2 -2rcoso (12)
between the height A and total height (P,2 = 1/3). The
probability that only one of the inclusions in the t2 and the circumference at the contact point perpendicular
cluster has been cut can also be calculated from the to the ζ axis is
figure (/>,/ = 2/3). This can also quite easily be

359
Vol. 28, ΛΌ. 6, 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

I Grinding direction

ΧΓΧ X XX X X

"C"
Fig. 14: Relationships used in calculating probability that a t2 cluster will be visible.

Case 1 Case 2
1.4 -r 1.4 τ

tl
12

4 6 10 2 4 6 10
Time (min) Time (min)

Case 3 Case 4
1.4 -,

tl -tl
12 12

4 6 10 2 4 6 10
Time (min) Time (min)

Fig. 15: Model calculations, cases 1-4.

360
Mats Soder. Par J nsson and Lage Jonsson High Temperature Materials and Processes

(13) all. The former case is illustrated by Figure 12d and the
latter by Figure I2e.
Thus, the following integral is obtained:
5.3.2 Conversion from 2D to 3D
"A Knowing the probability of how the t2 clusters are
J47zrvr 2 -a2 sinS-4πτ2 represented in the surface is not enough for the sought
after conversion of data given per area to data given per
(14) volume. Thus, an equation using the number of
cos2<5 inclusions, size and probability was needed.
Conversion of the information determined from the
two-dimensional sample surface to data expressed per
Similarly for the total height, the following expression unit volume for spherical inclusions is a standard
can be formed: procedure for those who work with data from
microscopic analysis. It can be done using the following
"A
relationship/16/:
J&Z7·2 sin&iS = [-8ΛΤ2 cos<?J (15)

(18)
which leads to the following expression of probability:

where Ny is the number of inclusions per unit volume,

n
-a
Γ
12 ~ (16) NA is the number of clusters per unit area, and d, is the
inclusion diameter. For t2 clusters the equation is:

For a viewable diameter (2α) of Ιμπι and an inclusion

or Nv = (19)
size of 3μηι, the chance of seeing a t2 cluster as a t2 Pl22d,
cluster is 0.22, or 22%. This probability is represented
by the case seen in Figure 12b. For larger clusters the theoretical representation in a
sample surface is much more complicated than for t2
Similarly, the probability of seeing the t2 cluster as a clusters. As the inclusions grow larger they extend in all
single Τ inclusion is 0.50, calculated with: directions in a random manner. Cuts through this
random structure will show it to be comprised of fewer
|- 8Wr 2 -a 2 cos δ j," + [- 2πτ2 cos 2<?]£ inclusions than it actually is. Some may even be seen as
t2 clusters in the surface. The proportion of larger
clusters taken to be t2 clusters, based on what can be
observed on a surface, is unknown. The task of
However, single inclusions were counted strictly
determining this is significantly more complex and
according to the SS 11 11 16 standard and the smallest
requires a greater amount of knowledge on the
inclusions counted were 1.7 μηι. Using this diameter
stochastic geometrical form of clusters and thus was
and thus setting a to 0.85 reduces the probability to
beyond the scope of this study.
0.48.This is represented by the case seen in Figure 12c.
The likelihood that a 12 cluster is not visible at all,
i.e. that it lies at a depth of 4r, corresponds to the 5.4 Growth Calculations
remaining 30%. Some inclusions that are not visible In order to find out if the increase in the number of
have been cut such that their diameter is less than the clusters as seen by microscope could be predicted with
smallest viewable diameter. Others are simply not cut at the growth model presented in Equation 5, four

361
Vol. 28. No. 6, 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

calculations were made. The populations of Case 2: Starting with a total number of inclusions of
inclusions/clusters from tl to t20 were calculated. 1.2/mm2 and assuming inclusion size to be 3 μπι.
However, due to the uncertainty of how to estimate the The inclusions were assumed to initially be
number of multi-inclusion clusters in a surface, only tl distributed as tl and t2 inclusions in accordance with
and t2 inclusions were plotted for comparison with the the microscope results for the first sample. The
microscopic results. Note that only growth was alumina content was assumed to be constant.
considered and the removal of inclusions was thus
neglected. This will, of course, need to be considered in As cases 1 and 2 do not consider the aluminium
a future more extensive study on the growth and addition in the experiment, they will not reflect what
removal of small clusters during deoxidation. was seen in the results of the production experiment. In
In the calculations it was assumed that the collision the microscopic analysis results, presented in Figure 3,
efficiency was 0.3, the energy dissipation was 0.01 it can be seen that the number of inclusions increases
m2/s3, the viscosity was 0.005 kg/sm and the single rapidly in the first sample after aluminium wire feeding.
inclusion size was 3 μηι. Furthermore, the radius of Simultaneously, the total oxygen content and calculated
clusters containing a certain number of inclusions was dissolved oxygen content decreased rapidly, which is
set according to the collision diameter, Dc (Equation 9). illustrated in Figure 6. The formation of small
Inclusions were added or subtracted according to the inclusions and the drop in dissolved oxygen content are
collision rate considering the population balance. most likely related to each other. Assuming that the
Cluster size in calculating the number of clusters per formed alumina is reflected in the drop in dissolved
mm", NA, was set as the external diameter, De. The oxygen content, this decrease could be used as input to
conversion from Λ^ to the number of t2 clusters per unit the growth calculation. However, the amount of alumina
volume, Nv, and vice versa was done using Equation formed in relation to the drop in dissolved oxygen is
19. much greater than what is seen as the increase in small
For the initial values and presentation in 2D inclusions in the microscope on the samples. Also in
(I/mm 2 ), the probabilities as described in section 5.3 Figure 6, the decrease in total oxygen corresponds to a
were used. As single inclusions were strictly counted removal of oxygen in alumina inclusions (this is
according to the SS 11 11 16 method, the smallest discussed further later). The amount of oxygen bound in
inclusions counted were 1.7 μηι. This meant that for 3 oxide (the difference between total and dissolved
μιη inclusions, only 82% of the inclusions on the oxygen) seems however more or less constant over time
surface would be counted (calculated (1.52- and not related to what is seen to be occurring with the
0.852)l/2/1.5). In the growth calculation this probability population of small inclusions. It is for these reasons not
was considered. correct to directly use the drop in dissolved oxygen.
In order to show how different assumptions and Instead, it was found that 11% of the decrease in
input data influence predictions on cluster growth, four dissolved oxygen corresponded to an increase in small 3
separate calculations are presented. First, two simple μιη inclusions.
calculations were done in order to show how only Based on the above discussion, two more
growth without any aluminium addition during the calculations were done using the data pertaining to the
stirring influences the growth predictions: decrease in dissolved oxygen:

Case 1: Starting with a total number of inclusions of Case 3: Starting with the same amount of tl and t2
1.2/mm2 evenly distributed as 3 μιη single inclusions as detected by microscope on the surface
inclusions. The initial value 1.2/mm2 was chosen of the first initial sample (Figure 3), then calculated
based on the number of single Τ inclusions found in with an addition of alumina as the amount of
Figure 3. The alumina content was assumed to be "formed alumina" corresponding to 11 % of the
constant. decrease in dissolved oxygen content (Figure 6).

362
Mats Söder, Par Jönsson and Lage Jonsson High Temperature Materials and Processes

The alumina was distributed as tl inclusions and the clusters, would have a higher chance of being cut such
values updated at each sampling occasion. that they are mistaken for two-inclusion clusters. The
Case 4: Similar to case 3, but alumina was distributed actual probability of this in numerical terms is not
between tl and t2 inclusions as the distribution seen known to the authors and would be an appropriate
on the first sample (Figure 3). subject for future research. However, to get an idea
regarding what effect this might have on a more realistic
In Figure 15 the results from the different prediction, the number of 12 clusters was modified to
calculations are shown. incorporate the assumed influence of the number of t3-
The calculations show that tl inclusions decrease in t!2 clusters on the number of t2 clusters. In Figure 16,
number quite a lot - by about one third - due to growth half of the t3-t4, one third of the t5-t6, one fourth of the
into larger inclusions. Also, the number of t2 clusters t7-t8, one fifth of the t9-tlO and one sixth of the tl l-t!2
tends to decrease if there are not enough tl inclusions or clusters were added to the number of t2 clusters. It can
"added" t2 clusters to compensate for the loss due to be seen that the predicted change in the t2 cluster
collision growth. population with stirring time agrees much better with
It is obvious that the calculations where all alumina the results in Figure 3 when those additional
was added initially, cases 1 and 2, do not render results assumptions were made. This indicates that any future
that show agreement with the microscope results in investigation aimed at better prediction of the evolution
Figure 3. This is because they do not consider the of 12 clusters with stirring time needs to include the
continuous addition of the deoxidants. It is, however, influence of clusters consisting of more than two
interesting to note that the number of t2 clusters does inclusions on the probability of finding a t2 inclusion on
not increase in case 2. In case 1, the amount of t2 a surface.
clusters increases and, although hard to see in the figure,
starts to decrease at the end of the calculation. In cases 3
and 4, the number of t2 inclusions eventually begins to
-•-tl
decrease because the number of tl inclusions that can ·»- 12
grow to become t2 inclusions is not high enough. —12 (13-14 addition)
·«-· 12 (t3-t!2 addition)
Instead, the decrease in t2 inclusions due to collision
growth has a greater effect on the final number of t2
inclusions.
The predictions for case 4, where an addition of t2
clusters due to aluminium feeding is included, show the
best agreement with the results from microscopic Fig. 16: Calculation case 4: results modified with
analysis in Figure 3. More specifically, the predicted assumed t2 addition from the number of t3-
peak value is almost I.I/mm 2 for the tl inclusions. This tI2 clusters.
is a slightly lower value than the experimental value of
In Figure 15, showing the cluster-growth calculation
1.2/mm2, which is taken from the microscopic results in
Figure 3. However, the agreement between predicted results for the four cases, only tl and t2 clusters are
and experimental data is quite good. shown since these can be more accurately presented as
The experimental values from the microscopic study seen on a polished surface and therefore be compared
pertaining to t2 inclusions are higher than the predicted with the microscopic examination results. As a
values in Figure 14. One possible reason for the comparison, the predicted values for the clusters per
deviation is that some of the larger inclusions on a unit volume from calculation case 4 are plotted in
sample surface can be cut such that they are actually Figure 17. Interestingly enough, the curves for t3, t4
and >t4 clusters resemble the ones seen in Figure 6 of
seen as t2 inclusions. Also, smaller clusters, for example
the microscopic results although they have a bit higher
t3 - 112, that are present in greater numbers than larger
values here. It is also worth noting how the group

363
Jo/ 28. No 6,2009 Growth of Small Alumina Clusters during Ladle Deoxidation

representing clusters made up of at least four inclusions and 8. These calculations show good agreement in the
(>t4) increases with stirring time. First, while there are difference between the total oxygen content and the
few smaller clusters, the increase is small. Later, when dissolved oxygen content.
smaller clusters - the "bricks" that large clusters are The microscopic results in Figure 3 show that
made up of- have formed, the growth rate increases. alumina inclusions form clusters during deoxidation. It
should be noted that the observed clusters are mostly
350 ·
made up of two, three and four alumina inclusions.
300 ·
Thus, these clusters are very small compared to what
ε
α 250 -
tl has previously been discussed regarding clusters in the
200 - 12
t3
open literature. These small clusters are present during
S 150 - ·· t4 the entire ΙΟ-minute deoxidation experiment. The larger
- >t4
C clusters (three or more inclusions per cluster) increase in
number and si/e throughout the deoxidation experiment.
It should be noted that the increase, especially for the
4 6 t4-clusters, continues after the stop of aluminium
Time (min)
addition after 7 minutes. This is taken as evidence that
cluster formation/growth is due to collisions in the steel
Fig. 17: Results of model calculation case 4 with no. of
bulk and not primarily a result of aluminium addition.
inclusions per unit volume.
As implied by the proposed theory for determining
the probability of actually seeing a t2-cluster on the
surface of a steel sample, far from all clusters are seen
6. DISCUSSION
in their true nature. This is because they are three
dimensional and can be oriented and cut in any
6.1 Cluster Size Distribution direction. The number of larger clusters will for this
reason be underestimated. Thus, in reality, a number of
There seems to be no contribution from secondary
observed t2 clusters on a sample will actually be larger
alumina inclusions (due to the decrease in oxygen
clusters. The proportion of larger clusters mistaken for
solubility upon solidification of the sample) to the size
t2 clusters is not known by the authors at this time, but
distribution of inclusions in the steel. Especially, the
it seems likely that for example t3 and t4 clusters would
initial sample (before AI wire addition) should be used
appear as t2 clusters than t3 and t4 clusters on the
as a reference and compared with the other samples.
sample surface.
Although the dissolved oxygen content is much higher
As the single inclusions grow to form small clusters
in the first sample, the numbers of inclusions seen on
they decrease in number as observed on the sample
the sample surface in the microscopic analysis are much
surface. Simultaneously, the number of large clusters
fewer. If secondary alumina inclusions would exist, they
increases. Growth of clusters may therefore be wrongly
would be expected to be present as small inclusions
interpreted as removal if microscopic examination is
which would be formed from the oxygen dissolved in
performed at 200x magnification. It is thus important to
the steel. Also, the use of 500x magnification should
count all inclusions in the clusters to get the full picture.
have exposed any significant contribution of secondary
The amount of removed inclusions in this investigation
inclusions. Thus, it is most likely that any formed
does not seem to be very high. The cluster growth is
secondary alumina inclusions may therefore be expected
though much more obvious.
to be substantially smaller than 1 μπι. Another
indication that no contribution of secondary inclusions
to the si/.e distribution of inclusions existed, is the 6.2 Aluminium and Oxygen Equilibrium
calculation results of the oxygen amount found in
The loss of aluminium during wire feeding is much
inclusions in the microscopic examination, Figures 7
larger than what corresponds to the decrease in the

364
Mats Soder. Par Jonsson and Lage Jonsson High Temperature Materials and Processes

oxygen content determined by chemical analysis minor errors in the calculations must result from
(Figure 6). The loss is therefore not due to rapid applying this assumption. Assumption numbers three
deoxidation and removal of deoxidation products. and four, that the inclusions are randomly oriented in
Instead, it is more reasonable to believe that due to the the steel matrix and that there is a minimum diameter
slow feeding speed some of the aluminium is burnt off that can be seen in the microscope, respectively, can
at the surface of the ladle or in the slag. The loss, both be considered correct based on the observations in
corrected for the removed alumina, is estimated to this study. In addition to these possible sources of error,
correspond to a 0.8 mm rind around the 15 mm wire. there will also be some larger inclusions/clusters that are
The total oxygen analysis results show a drastic cut such that their visible diameter can be small and in
decrease after the start of aluminium addition. This the Τ size range. Their proportion in the number of
would represent a very fast removal of oxygen through observed inclusions/clusters depends on the size
alumina inclusions. This could be explained if the distribution of the single inclusions and is further
oxygen is removed by the formation of very large complicated by the random cuts through clusters
inclusions that have very high buoyancy that almost containing more inclusions than two.
float directly up to the surface. The fluid flow towards As the extent to which larger clusters can be
the surface could also assist in the transport of these mistaken for observed t2 clusters on a sample surface is
large inclusions. Mo increase in very large inclusions not known, the accuracy of the latter part of the curve in
was observed in any of the samples that were examined Figure 3 from the microscopic examination is therefore
by microscope. However, if the inclusions were about somewhat questionable regarding what inclusions and
ΙΟΟμηι or larger they would be so rare that it is unlikely clusters were actually seen on the surface. This
that even one such inclusion would have been observed especially makes any comparison between single tl
on the examined surface. Another, maybe more far- inclusions and 12 clusters using the probability of seeing
reaching explanation might be the precipitation of small t2 clusters a bit unsure for this part of the curve.
inclusions directly to the ladle wall and top slag. It was However, the initial values pertaining to the early part
earlier calculated that the rapid drop in total oxygen of the deoxidation are less affected by collision growth
corresponds to 3.4 kg alumina. If this amount was and considered less likely to be influenced by the
evenly distributed on the inside of the ladle, a 45 μηι presence of larger clusters. For the initial values in the
layer would form. growth calculations, this would have a minor effect, but
it was neglected in this work.
Depending on what starting data is used in the
6.3 Clusters Growth
calculations, very different results were obtained. It was
In the derivation of the conversion of number of t2 found that a method for taking the aluminium addition
inclusions seen on a surface to number per volume, four into consideration is necessary for more accurate
assumptions were made. A few of these assumptions modelling. The calculations also showed that the
are, of course, a simplification of what really occurs in alumina addition both in amount and what size, in terms
order to make calculation possible. The first of number of individual inclusions in the clusters, has a
assumption, that the inclusions are spherical, is not clear affect on the modelling results. A conclusion
totally correct for alumina as the inclusions are solid in drawn when comparing the growth calculations for
the steel and have a shape formed by crystallisation cases 3 and 4, is that the high ratio between t2 and tl
factors. However, the shapes of the majority of the inclusions can not be explained by the collisions of
observed small alumina inclusions are quite well single tl inclusions. In case 4, there is an increase from
extended in all three dimensions. Thus, the assumption formed alumina being distributed both to tl and t2
of a spherical shape in the calculations can be seen as inclusions. Thus, a greater resemblance to the t2 curve
fairly good. Assumption number two, that the two in Figure 3 from the microscopic examination is
inclusions forming a cluster are of the same size, is a obtained. This implies that clusters are not only formed
generalisation that stands for the most. However, some

365
Vol. 28. No. 6, 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

by collision in the steel bulk but also directly at the spot parameters. For accurate modelling of the removal
of alumina addition, if the growth calculations are to be characteristics, it would be of great value to understand
considered fully accurate. There may thus be different the flotation process of alumina clusters. It would thus
growth rates and even growth mechanisms for be necessary to investigate the D/, ke and actual effective
inclusions in the vicinity of the addition spot and the inter-particle locking of steel (i.e. cluster density in the
rest of the steel bulk. steel).
At the same time as the calculated dissolved oxygen
decreases very rapidly, it was observed from the
microscopic sample analysis that there is a large 7. CONCLUSION
increase in the number of small inclusions. In cases 3
and 4 of the growth model calculations, the rate of the Inclusion growth and cluster formation have been
drop in dissolved oxygen content was used as the rate of investigated through examination of samples taken
formation of new inclusions. It was shown that the during a ten-minute deoxidation experiment in an
growth could be modelled based on this assumption. induction-stirred plant ladle. Inclusions and clusters on
This result could be of interest for a process-control polished samples were counted when viewed by
model for inclusion behaviour in stirred melts since it is microscope. Clusters were found in all samples and they
based on relatively easily measurable and attainable were very small in comparison to what is normally
values. As inclusions collide and clusters grow reported in the open literature. Based on aluminium and
continuously in the stirred steel bath, a growth model total oxygen contents determined from chemical
should also consider the "age" of the melt after analysis, the amount of corresponding dissolved oxygen
deoxidation. Furthermore in developing a deoxidation was calculated and the amount of oxygen bound in
model, it appears vital to know the initial oxygen inclusions was compared with the microscopic
content of the steel melt at the start of deoxidation. examination results. In order to make a cluster growth
Inclusion removal was not considered in the growth model, a size equation based on the theory of fractal
calculations. However, from the microscopic analysis aggregates was developed and values determined
results, no obvious effect of removal on the inclusion applying the theory compared with the microscopic
population could be observed. As the dominant results. Four growth model .calculations based on
mechanism for removal is generally considered to be turbulent collision were performed. For comparison
flotation of inclusions to the top slag, which is very with the microscopic examination results, a method was
much controlled by the buoyancy, the increase in developed for converting model data from numbers per
buoyancy for the small clusters formed could be too area to numbers per volume for randomly oriented
minor to have any pronounced effect on the inclusion clusters containing two inclusions.
population. It should also be noted that the buoyancy of The primary findings of this investigation are the
clusters has by several researchers /7, 9, 107 been following:
regarded to be limited by the increased density of the
clusters; thinking of the fluid dynamics, the steel is 1. Examining polished steel sample surfaces by
"locked" in the structure of the cluster. microscope requires a level of magnification that
It seems readily apparent that the fractal form of allows inclusions present on the surface to be
alumina clusters requires further investigation. It is viewable as separate entities. A magnification of
reasonable to believe that the fractal dimension for 500x provided the degree of resolution deemed
different steel qualities and different operational lines is necessary in this study.
the same, around 1.8. The numerical density value, ke, 2. The results from both the microscopic examination
may though vary some depending on chemical and the determination of aluminium/oxygen contents
composition, size of the individual inclusions, the lapse by chemical analysis indicated that most of the
of time after the aluminium addition and other physical alumina inclusions are formed within the first two

366
Mats Söder. Par Jönsson and Lage Jonsson High Temperature Materials and Processes

minutes. The microscopic examination results REFERENCES

showed a significant increase in the number of
inclusions and clusters at the beginning of the l.U. Lindborg and K. Torsseil, Trans. Met. Soc.
aluminium feeding. During the experiment there was AIME, 242, 94 - 102 (1968).
a shift towards the formation of larger inclusions. It 2.K. Nakanishi and J. Szekely, Trans. IS1J, 15, 522-
is obvious that there is quite a large amount of 530 (1975).
clusters in the steel during ladle refining and that a 3. M. Hallberg, En matematisk modellför desoxidation
majority of the small inclusions are bound to av en omrörd stälsmälta, Tech. lie. Royal Institute of
clusters. It was further concluded that the cluster Technology, ISBN 91-7170-949-2 (1996).
growth was due to collision. 4.D.Y. Sheng, M. Söder, P. Jönsson and L. Jonsson,
3. Both the results from the microscopic sample Scand. J. Metall, 31, 134-147 (2002).
analysis and growth calculations indicated that not 5. M. Wakoh, K. Fuchigami, K. Endoh, N. Imamura,
only single inclusions, but also larger clusters, are A. Kiyose, I. Sawada, Scanmet I, Luleä, Sweden,
formed at the point of aluminium addition. The June, 267-274 (1999).
higher concentration of both inclusions and 6.K. Shirabe and J. Szekely, Trans. ISIJ, 23, 465-474
aluminium may lead to higher growth rates for (1983).
inclusions in the vicinity of the location where the 7.Y. Miki, Y. Shimada, B.C. Thomas, A. Denisov,
aluminium is added than for those in the steel bulk l&SM, ISS, 28, 31-38 (1997).
flow. 8.A.K. Sinha and Y. Sahai, ISIJ Int., 33, pp. 556-566
4. In predicting growth by collision, the equation (1993).
formulated to express the collision diameter for 9.Y. Miki and B.G. Thomas, Materials Transactions
small clusters was found to render reliable results. B, 30B, 639-654 (1999).
Good agreement with the microscopic observations 10.H. Tozawa, Y. Kato, K Sorimachi and T. Nakanishi,
was found when 11% of the decreased dissolved ISIJ Int., 39,426-434 (1999).
oxygen content was calculated as an addition of 3 ll.Svensk Standard SSI 11116. Steel - Method for
inclusions distributed as single inclusions and Assessment of the Content of Non-metallic
clusters containing two inclusions. Inclusions. Swedish Institute for Standards, June,
5. In applying equations formulated to convert data (1987), pp. 1-16.
from number per surface area to number per volume 12.L. Huet, P. Jönsson and F. Reinholdsson, Steel
for clusters containing two inclusions lying in the Times Int., 11, 47-50 (1997).
surface it was concluded that only 1/3 of the clusters 13.H. Hägglund, Rescon-ElectroNite, P.O. Box 35, SE-
are actually oriented such that they could be 190 63 Örsundsbro, Sweden.
theoretically observed as two-inclusion clusters. 14. Steel - Determination of content of non-metallic
Moreover, these clusters can be positioned in such a inclusions - micrographic method using standard
manner that 2/3 could theoretically be mistaken for diagrams, International Standard ISO 4967- (E)
single inclusions. It was also concluded that some (1979).
larger clusters may wrongly be observed as clusters 15. Mikroskopische prufung von edelstählen auf
containing two inclusions. nichtmetallische einschlusse mit bildreihen, Stahl-
6. Although inclusion removal was not taken into Eisen-Prufblatt 1570, Verlag Stahleisen, Dusseldorf
account in this study, any future cluster-removal I,postfach8229.
model would need to consider cluster form and the 16.R.T. DeHoff and F.N. Rhines, Quantitative
steel that is in hydrodynamic aspect locked in the Microscopy, McGRaw-Hill, New York, (1968).
structure of the cluster (i.e. effective cluster density 17.Y. Feng, A Kinetic Study on the Deoxidation of
in the steel). Liquid Steel by Silicon and Aluminium, PhD Thesis,

367
Vol. 28, ΛΌ 6, 2009 Growth of Small Alumina Clusters during Ladle Deoxidation

Royal Institute of Technology, ISBN 91-7170-828-6 23. K. Tozawa, Y. Kato and T. Nakanishi, CAMP-ISIJ,
(1995). 7,276(1994).
18.G.K. Sigworth and J.F. Elliott, Met. Sei., 8, 298-310 24. P. Meakin, Phys. Rev. Lett., 51, 1119-1122 (1983).
(1974). 25. M. Kolb, R. Botet, and R. Jullien, Phys. Rev. Lett.,
19. J.F. FJIiott, M. Gleiser and V. Ramakrishna, 51, 123-1128(1983).
Thermochemistry for Steelmaking, Addison-Wesley, 26.D.A. Weitz and M. Oliveira, Phys. Rev. Lett., 52,
London, 2, 620(1963). 1433-1436(1984).
20. K. Asano and T. Nakano, Tetsu-to-Hagane, 57, 27. P. Dimon, S.K. Sinha, D.A. Weitz, C.R. Safinya,
1943-1952(1971). G.S. Smith, W.A. Varady and U.M. Lindsay, Phys.
21. P.O. Safiman and J.S. Turner, J. Fluid Mch., 1, 16-30 Rev. Lett., 57, 595-598(1986).
(1956). 28.J.E. Martin, J.P. Wilcoxon, D. Schaefer and J.
22. J. Fedcr, Fractals, Plenum, New York, 33 (1988). Odinek, Phys. Rev. A, 41, 4379-4391 (1990).

368