Instructions for use

Title An Investigation of the Crystallization of a Continuous Casting Mold Slag Using the Single Hot ThermocoupleTechnique

Author(s) Kashiwaya, Yoshiaki; Cicutti, Carlos E.; Cramb, Alan W.

Citation ISIJ International, 38(4), 357-365https://doi.org/10.2355/isijinternational.38.357

Issue Date 1998-04-15

Doc URL http://hdl.handle.net/2115/75701

Rights 著作権は日本鉄鋼協会にある

Type article

File Information ISIJ Int. 38(4)_ 357.pdf

Hokkaido University Collection of Scholarly and Academic Papers : HUSCAP

ISIJ International, Vol. 38 (1998), No. 4, pp. 357-365

An Investigation

Mold Slag Using

of

the

the Crystallization of a ContinuousSingle Hot ThermocoupleTechnique

Casting

Yoshiaki KASHIWAYA.Carlos E. CiCUTT11)and Alan W. CRAMB2)

Formerly Faculty of Engineering, Hokkaido University. Nowat Department of Materials Science and Engineering. CarnegieMellon University, Pittsburgh. PA. USA. 1)Formerly Departmentof Materials Science and Engineering. Carnegie MellonUniversity, Nowat Center for Industrial Research, FUDETEC,Argentina. 2) Department of Materials Science andEngineering, Carnegie Mellon University, Pittsburgh. PA. USA.

(Received on October 6. 1997, accepted in final form on December11. 1997)

Theconditions under which crystallization deveiops in a mold slag must be understood in order to select

or design a mold flux for use in the continuous casting of steels, In this paper, the crystallization of anindustrial mold slag wasquantified using a single hot thermocoupie technique which, whencombinedwith

a video camera based observation system, ailowed observation of the onset and growth of the crystals

which were precipitated from the melt. Thebeginning of crystallization wasdetermined by direct observationand the growth rate of crystals were measuredby frame by frame imageanalysis of recordings of the progressof crystal lization. Isothermal experiments were performed at different temperatures and a Time-Temperature-Transformation (TTT) diagram wasdetermined for this indust~ial mold slag. X-ray diffraction of quenchedsamples wasused to determine the type of crystalline phases that were precipitated. TheTTTdiagram wasdivided into two separate regions which corresponded to the precipitation of dicalcium silicate (Ca2Si04)at temperatures over I 050'C and of Cuspidine (Ca4Si207F2) at temperatures below I 050'C. Theevolutioncrystal fraction wasdescribed by Avrami's equation. This work indicates that industrial mold slags are easily

undercooled, that crystall ization occurs throughout the meit, that crystals grow initial ly as equiaxed dendritesand that the onset of crystallization is a function of cooling rate and must be described by either TTTorCCTcurves.

KEYWORDS:mold flux; continuous casting; crystallization; nucleation; crystal growth; silicate; slag; hotthermocouple; TTTdiagram.

1. Introduction

It is well knownthat mold slags* play a declsive role

in lubrication andheat transfer control in the continuouscasting moldl) whensteels are cast. Various studies 2,3)

have shown that the slag film formed between moldand strand contalns three layers: a glassy zone close to

the mold, a crystalline layer in the center and a liquid

film in contact with the steel. The proportion of crys-talline and glassy phases in the slag film have beenreported to help control the heat transfer rate in themold4~6); however, as yet there has been no quantifica-

tion of the speclfics of mold slag crystallization or of the

temporal development of the crystalline fraction. It is

postulated that those slags which develop a crystalline

layer at higher temperatures allow a reduction in the heattransfer rate between the mold and strand7,8) and this

design criteria has proved to be effective in the reductionof surface defects, mainly when peritectic steels arecast.7,9,10) Slags with a high crystallization tendency

are always used at lower cast speeds as these slags havebeen shown to increase the possibility of stlcking typebreakouts at higher cast speeds due to an increase in thefriction between the mold and the strand.11 ~13) Thus,the control of crystallization within a mold fiux is akey parameter in the prudent operation of a steel

continuous caster; however, as yet, the phenomenaofcrystallization of mold fluxes has been clearly described.

Avariety of methodshave been traditionally used todetermine the crystallinity of mold slags; however,currently there is no standard technique for its evalua-tion.14'15) Table I summarizesa variety of methods re-

ported by different authors in the literature. The mostcommonmethodis Differential ThermalAnalysis (DTA).In this case, the slag sample is melted in a furnace andthen cooled simultaneously with a reference substance,

and the temperature difference betweenboth is recorded.

Whenan endothermic or an exothermic reaction takes

place in the sample, a temperature difference can bedetected. The onset temperature of the exothermic peak

* Mold slags are the llquids that form upon the melting of the mold powders or mold fluxes that are added to the mold of a continuous caster

used in the processing of liquid steel. The mold slag composition can be slightly different in composition from the mold powder or fiux due to

absorption of Inclusions or reaction with the liquid steel.

357 O1998 ISIJ

ISIJ Internationai, Vol. 38 (1998). No. 4

Table l. Methods and conditions employed for studyingcrystallization phenomenain mold slag.

MethodemployedCooling rate

("C/s) ('C/min)Reference

DTADTADTADTADTADTA(heating)

DTA(heating)

Cooling curveHot microscopeHot thermocouple

Casting in

metallic mold

0.0 17O. 17-33.3

0.42

0.33O. 17l0.08--0.67

O. i7- 0.08

0.33

15

ll0-2 OOO

2520

lO60540lO

20900

17)

18), 19)

20)

21)

22)

23)

24)8)

12), 13)

16), 13),

18), 22),

25)-28)

is usually called the "Crystallization Ternperature" ofthe mold slag. Fundamentally the onset of crystalliza-tion must be a function of the cooling rate in theseeasily glass forming slags; however, as shownin Tablel, there is no unique criteria to define the cooling rateemployed in the experiments. Several studies have con-sidered the effect of cooling rate on the crystalllzationtemperaturel8,19~22) but only a few have attempted toconstruct Time Temperature Transformation (TTT) di-

agrams or Continuous Cooling Transformation (CCT)diagrams which would unambiguouslydescribe the crys-tallization phenomena.

There have been numerousother techniques to deter-

mine crystallinity reported in the literature. For ex-ample, slag samples were first melted and rapidlycooled to room temperature in order to keep them in

a glassy state.23,24) After that. DTAanalysis wasperformed during heating of the sample and the tem-perature at which crystallization appears was recorded.This kind of experiment is usually called a "Recrys-tallization test".

Aslightly different methodwasapplied by Kyodenetal.8) to determine slag crystallinity. In this case, theslag sample was melted and then cooled down in afurnace. Temperatureof the samplewasrecorded. Whena deviation in the cooling curve was observed, it wasattributed to the beginning of crystallization. Anothermethodconsisted of fusion of the slag sample in a hightemperature microscope and observation of its behaviorduring cooling,12,13,16) The temperature at which theslag lost transparency wasconsidered to be the crystalli-

zation temperature.Finally, a technique that was employed to better

simulate a continuous casting mold consists of meltingthe mold slag sample, increasing its temperature to 1550'Cand then quenching it by pouring the molten slag into

a metallic mold.13,18,22,25-28) After that, the samplecan be inspected, at room temperature, by a numberofdifferent techniques. For example, the proportion ofcrystalline and glassy phases can be evaluated using

an optical microscope.13'18.25,26) In other studies, thefracture surface of the slag wasexaminedto estimate thepercentage of crystallized area.2s) X-ray diffraction can

Q1998 ISIJ 358

also be performed to determine the structure andchemistry of the crystalline phases. 13' 18) Obviously, thesemethodsare useful to define a "crystallinity index" that

can be used to compare, qualitatively, the tendencyof different slags to give glassy or crystalline phases.However, they do not provide a quantitative measure-mentthat can characterize the slag behavlor in the mold.

In actual continuous casting operation, once the moldpowder is melted on the top of the liquid steel and theliquid slag infiltrates between the mold and the strandsurface, the liquid slag is exposed to different coolingpaths which can promote or prevent crystallization. Infact it is commonto see glassy parts of the slag in areasof high cooling rates with crystalline fractions at lowercooling rates. So, a proper characterization of the moldpowder shou]d inc]ude a complete description of thetransformation kinetics. As mentioned before, somestudies have shownthe effect of the cooling rate on thecrystallization temperature of mold powders.18,19,22)

However, few attempts have been madeto develop aTTTcurve, which provides a more fundamental under-standing of the transformation process. Therefore, themain objective of this work was to develop a techniquewhich would unambiguouslyallow the determination of

a TTTcurve for an industrial mold powder.In a previous study,29) the double and single hot

thermocouple techniques (DHTTand SHTT)were de-veloped and the details of temperature measurementand in situ observation were shown. Themain advantageof these techniques are that crystallization occurring inthe samplecan be observed in situ under defined thermalconditions andthe growth rate of crystals and the changesin the crystallized fraction can be measured. Also, dueto the low thermal inertia of the system, high heatingand cooling rates can be easily obtained. This is veryuseful for the determination of TTTdiagrams where afast initial cooling rate is required.

2. Experimental

Experiments were carried out using a single hotthermocouple technique (SHTT) and the details of theexperimental apparatus were shown in a previous pa-per.29) A slag sample was mountedon a Pt-Rh ther-

mocouple(B Type) which was heated directly while its

temperature was measured simultaneously. The ther-

mocouple was located below a microscope equippedwith a video camera, which sent the image to a com-puter andaVCR(Video Cassette Recorder). Heating andcooling of the thermocouple wascontrolled by meansofacomputerprogramso that different thermal cycles couldbe imposed.

Experiments were carried out using an industrial moldslag whosecharacteristics are shown in Table 2. It is

a commercial mold powder normally employed in thecontinuous casting of stainless steels. In this case, thepowderwasprepared by the supplier without carbon in

order to facilitate the experiments.Figure I shows the typical thermal cycle employed in

this study. First of all, a small sample (lOmg) of moldpowderwas melted on the tip of thermocouple. In all

ISIJ International, Vol. 38 (1998), No. 4

Table 2. Chemical composition and physical properties of industrial mold slag.

Chemical composition (mass"/.)

T.CaO Si02 Al203 Na20 T.F MgO K20 Fe203

Basicity

CaOISi02

Viscosity

at 1300'C(Pa ' s)

37.8 33.3 7.9 8.2 8.4 0.6 0.6 l .7 l.14 0.2

O

~O

aEQH

4-5min at 1500'C

Quenching( * 80 'C/s)

Testing temperature

in Steady State

~!~.~.

_.VsiaemWpol~

sample

(5 - 30min)

Sample

,rt\~T.C.

(a)

Fig.

Timel. Thermal cycle of experiment.

dE~

r~:!1~:1/Quenchtn9

Temperatu re distributionin steady state jrlIIl~l

Heatln9

1248

1184

1120

t\

RunA

(a) RunBo 40 80 120

Oo=,U*,D~ECD

H

T.C.

dE~

T.C.

dE~

1600

1500

1400

1300

1200

11oo

tL

B:E:

Beginning of crystallizationEnd of crystallization

B E

Run

(b) --

B E Run

A

B

Fig. 2.

100 150 200-50 5OOTime [s l

Temperature variations of sample frorn quenching to

steady state

cases bubble formation wasobserved during the meltingprocess.29) The sample washeld at 1500'C for 4-5 minto eliminate the bubbles and homogenizeits compositionand temperature. After that, the sample was rapidlycooled to a given temperature and held for 560min.

The sample images were recorded on a VCRandtemperaturetime curves were stored in a computer file.

Imageanalysis of the video tape was used to determinethe time for the beginning of crystallization and to

measure the rate of crystal growth and the fraction ofcrystal transformed.

Fig.

(b) (c)

3. Convection of sample in steady state and unsteadystate during cooling and heating.

3. Results and Discussion

3.1. Temperature Profile and Convection in the SampleFigure 2 shows the temperature profiles of experi-

ments. In this experiment, the temperature wascontrolledby decreasing or increasing the electric current. Heat loss

from the samplewasby conduction to the thermocoupleand by convection heat transfer to the ambient gases (air

atmosphere). The profile of temperature in the sampleshould vary with the current; however there can be atime lag between the change in current and the attain-

ment of isothermal conditions. Twocases are shownin

Fig. 2(b). In RunA, at a higher testing temperature, asteady state is reached quickly and at the lower testing

temperature (Run B) it takes longer to reach a steadystate. Convection within the melt leads to temperatureoscillations in the sample. Illustrations of convection in

melt are shown in Fig. 3 (side view of hot thermo-couple). The detail of the hot thermocouple electrodein the top view was described in a previous paper.29)

At steady state, the sampie temperature near the ther-

mocouple is highest and there is a small temperaturegradient from the tip of the thermocouple to the left

hand side. The direction of the convection in steady

state is shownin Fig.3 (a) (Due to the thermal gradient,

Marangoni convection will dominate fluid flow in thesystem). Whenthe samplewasquenched, the temperatureprofile inside the sample would correspond to tl in Fig.

359 C 1998 ISIJ

ISIJ International. Vol. 38 (1998), No. 4

Fig. 4. Comparisonofcrystal image betweensimulated moldslag and industrial mold slag.

3(b) owing to the domlnant heat conduction from thesample to the thermocouple. After that, the temperatureprofile will return to the steady state condition (12 orFig. 3(a)). This phenomenonwould correspond to theresults of Run(A) and Run(B). In Run (A), since the

temperature difference between testing temperature andmelt temperature before quenching wassmaller than thatof Run(B), the time to return to steady state (from Fig.

3(b) to Fig. 3(a)) would be shorter than that of Run(B).

Thus the direction of convection in sample B, at thebeginning of crystallization, changed between Run (A)and Run(B). This change in convection direction in thesamplehas beenconfirmed by viewing the crysta] move-ment, after precipitation, in the sample (Fig. 5 (RunB) and Fig.6 (Run A)). The crystals movedfrom the tip

of thermocouple to the left hand side in the case of Run(A) in Fig.6. Onthe other hand, the crystals movedfromthe left hand side to the tip of thermocouple in Run(B)(Fig. 5).

Furthermore, as shown in Figs. 3(b) and 3(c), thedirection of convection differed during quenching andheating. The convection in the sample was enhancedduring heating, as the thermal gradient increased fromthe tip of thermocouple to the left hand side (Fig. 3(c)).

3.2. Results of in Situ Observation

Figure 4 shows the comparison of crystal irnageprecipitated in the undercooled sample between an in-

dustrial mold slag (a) and a simulated mold slag (b),

(3.8masso/oNa20-6.8masso/oAl20344.7masso/oSi02-44.7massoloCaO,C/S= l). Simulated mold slag experi-

mentswere carried out using the DHTT(Fig. 4(b)). For

O1998 ISIJ 360

Fig. 5. Crystallization behavior of industrial mold slag.

(1 142*C, RunB)

simulated slags the melt was transparent, the precipitat-ed crystals were clearly vlslble and measurementof thenumber of nucleation events and the rate of crystalgrowth was unambiguoususing automated computerimage anaiysis. On the other hand, trials with an in-

dustrial mold slag revealed an image that was trans-lucent (probably due to the iron oxide content ofthe flux) and more difficult to analyze by automatedcomputer techniques (Fig. 4(a)).

In Fig. 4(a) crystallization in the industrial slag canbe seen to occur by nucleation and growth of individualcrystals in an undercooled liquid rather than by planefront solidification indicating that this phenomenamustbe treated using the techniques developed for classical

nuc]eation theory. This industrial mold slag can under-cool significantly below its equilibrium liquidus tem-perature and the phase dlagram (if one existed) wouldnot be useful in the determination of the conditionsunder which crystallization will occur. Additionally,this resuit suggests that nucleants could be used tomodify the crystallization behavior of fluxes.

Figure 5shows the sequence of crystal precipitationand growth In the sample at I 142'C. The beginning ofcrystallizatlon was 20s from the start of cooling. Figs.5(a), 5(b) and 5(c) were 16.32, 20.79 and 49.18s fromthe beginning of crystalll7_ation, respectively. Thecrystals

ISIJ lnternational, Vol. 38 (1 998), No, 4

Fig. 6. Crystanization behavior of industrial mold slag.

(1 215'C, RunA)

began to appear from the left hand side of the sample,

as mentioned before. Also It is interesting that the crystal

morphologywasnot spherical but irregular andangular.Figure 6 shows the sequence of crystallization at

1215'C. The crystals were appeared spherical and thebeginning of crystal]ization occurred 65 s after the startof the isothermal hold. Becauseof the long Incubationtime of crystallization, the convection in the samplehadreturned to a steady state (Fig. 3(a)) and the crystals

began to be observed from the tip of thermocouple.Whenthe experimenta] temperature decreased to

1009'C, the size of crystals were very small and it wasdifficult to measure the initial crystal size accurately(Fig. 7). To examine the morphologyof the crystals, thesample wasquenchedfrom 1200'C after crystallization

had initiated. Figure 8 shows the results of SEMob-servation. The crysta]s were dendrites even though aspherical crystal wasobserved macroscopically.

3.3. TTTDiagram and Mechanismof Crystal Growth

Anumberof experiments were perfonlled over a widerange of temperatures in order to construct a TTTdi-

agramfor the moid fiux using the SHTT.Thebeginning

361

Fig. 7.

iC! '

Crystallization behavior of industrial mold slag.

(1 009'C, Crystallized region wasemphasizedby imageprocessing.)

Fig. 8. SEMobservation of crystallized part which sample

wasquenchedfrom 1200'C.

of crystallization which is defined whenthere is 0.5 vololo

of crystallization, could be determined by imageanalysis(to.005) and the time of 500/0 volume fraction (t0.5) wereplotted against temperature and the TTTdiagram ob-

(") 1998 ISIJ

ISIJ International, Vol. 38 (1998), No. 4

1300Meltin oint 1265•C

Oo5,5*oaE,D

H

1200 -

11oo

1OOO

900 -

800 -

t0.005

o

oo

to. 5C]

e

,

e

o

I

I

e I I

eto

. 5e

to.oo5

Oa2Si04

1050'c

Ca4Si207F2

::

ooa,*~~>'I'),

o,o

sot

7001o loo

Time[s JFig. 9. TTTdiag,'am orindustrial mold slag obtained.

tained is shown in Fig. 9. The diagram was dividedinto two regions above and below I 050'C as two dis-

tinct curves were detected which suggested two sepa-rate nucleation events. Each of the curves in the TTTdiagram had a C-shape and the observed crystal mor-phology drastically changedat I 050'C.

Theminimumtime before the crystalline phase couldbe determined was approximately lOs at I 070' and950'C. The critical cooling rates for glass formationwould be 30'C/s or higher for thls industrial mold slag.

The necessity of considering crystallization as functionof time and temperature is clearly seen in Fig. 9. Alsoclear from this figure is that there is no specific crys-tallization temperature for a mold flux but a range of

temperatures defined by experimental conditions.

To clarify the chemistry and structure of the crystal

phase. X-ray diffraction of the sample were carried outusing quenchedsamples and a high temperature X-raydiffraction technique. Themaximumtemperature of high

temperature X-ray diffraction is I 150'C and samples

were prepared by quenching from 1450'C. The sample

wasconfirmed by ordinary X-ray diffraction as a total-

ly amorphousphase before high temperature X-raydiffraction.

Figure 10 shows the high temperature X-ray diffrac-

tion patterns below I 048'C. It was confirmed that thecrystal phasebelow I048'C wasCuspidine (Ca4Si207F2,ASTM41-1474).The height of peaks decreased withincreasing temperature, which means that the crystal

becameunstable close to I 048'C.Because of the upper temperature limit of the high

temperature X-ray diffraction device, ordinary X-raydiffraction was carried out to characterize the higher

temperature phases using quenchedsamples which weremelted on the hot thermocouple at each temperature,held until crystallization finished and then quenchedby

a gas jet (cooling rate ~~ 150'C/s). Figure ll shows the

comparison of X-ray diffraction patterns between high

>,,,

=,D

=

c

cc

f

c

1048'c

@1998 ISIJ

1ooo'c

C: Cuspidine(Ca4Si207F2' 41-1474)

cc c c ccccc

900'c

830'c

362

20 30 40 50 60 70

20Fig, lO. Comparison of high temperature X-ray diffraction

patterns of industrial mold slag in the range from 830to I050'C. (Cu-Ka:)

>:!::,b

ca,

E

D

DD: 2CaOSi02 (ASTM33-302)

D Si(1 OO)~t.

D quenched \DDDDD DD from 1150'C

~,------ 2cao si02 (ASTM31-298)

at 1150'C

20 50 60 7030 4020

Fig, Il. x-ray diffraction patterns between quenchedsampleand sample at high temperature. (Cu-Ka)

temperature X-ray diffraction at I 150'C and ordinaryX-ray diffraction of sample quenched from 1150'C. It

wasfound that the phasewas2CaO• Si02 in both cases.Figure 12 shows the X-ray diffraction patterns of

samples quenched from I 053' to I 190'C using the

SHTT.Dicalcium silicate (Ca2Si04) was the dominantphase measuredin this temperature range. TheseX-raydiffraction results (Figs. 10 and 12) allow the two curvesin the TTTdiagram to be identified.

3.4. Kinetics of Crystal Growth in High TemperatureRegion

Imageanalyses were performed to obtain the rate ofcrystal growth and to evaluate the volume fraction ofcrystals evolved as a function of time. As mentionedbefore, the crystal imagewasnot as clear as a simulated

ISIJ International. Vol. 38 (1998), No. 4

>u,

coc

DDD D

D: b-Ca2Si04

DDDD

( AsTM33-302)

1190'C

DDD D D 1105'c

1090'c

1053'c

e

,

3.8s

20 30 40 50 60 702e

Fig. 12. ComparisonofX-raydiffractionpatternsofindustrial

mold slag from I053 to 1190'C. (Cu-Kc()

16. 14s

15.0s

20.09s

32.39s

e!'Fig. 13. Variation ofcrystallized area in industrial mold slag.

(1 142'C, RunB)

mold slag which madeautomatic imageanalysis difficult

as computerized edge detection was not possible. Theportion of crystals which could be clearly recognized

were outlined in each frame manually and their areas

were counted by computer (Figs. 13 and 14). The par-ticles were treated as spherical in cross-section and the

volumewascalculated using the thickness of the sample.Although the detail of crystal outline could not re-

produced exactly, a total of 200OOOpixels were usedfor calculation and the error was within 5o/o for sever-al calculations. Also each of the crystals were extracted

and volumes were calculated from the area (V=4/3 -

?c(A/1c)3/2. V: cm3, A: cm2) and results are shown in

Fig. 15. The variation of crystal volume fraction withtirne wascomparedwith the volume fraction of the cry-stal in the whole sample in Fig. 16. In each method, thevariation of volume fractions F(t) was the sameandcould be represented by Avrami Equation.32~42)

363

Fig, 14.

16.1 4s

17.1 1s

18.1ls

23.00s

24.22s

32.39S

34.1 6s

Fig. 15.

Variation of crystallized area in industrial mold slag.

(1 215'C, RunA)

NoI No2 No3 N0.4 N0.5

' ' llt • 'b

,

~IL

1' Ib

, ,'1'

,

b

Variation of crystal size

(1 142'C, RunB)

precipitated in sample.

F(t)=1-exp(-(K't)")........

..........(1)

wheren represents the mechanismof crystal growth andK is a coefficient corresponding to the nucleation andgrowth mechanism. For a constant volume nucleation

rate N, n is equal to 4 and the equation is equal toJohnson-Mehl-Avrami equation.32)

F(t) = I -exp( -(7T

' N• u3/3)•t4) ......,......(2)

where u is growth rate. Thevalue of n which is equal to

4 meansthat the rate of reaction is quite fast and anon-diffusional transport mechanismis dominant. This

transport mechanismof the crystal growth in an in-

dustrial mold slag during the experiment is probablydue to convection.

@1998 ISIJ

ISIJ Internationa[, Vol. 38 (1998). No. 4

a:

'a

~o~-o=-o'OLLco

~'a)

E=~>

1

0.8

0.6

0.4

0.2

-o.

o

2

Run B,

1142•C

Volumefraction ofO [] each particle

Volumefraction of~e~whole sample

O 20 40 60 80 100 120Time [s l

Comparisonof variation of volume fraction betweenwhole sample and each particles.

O.

oSS'Da

EoH

1300

1200

11oo

1ooo

900

800

700

l'n in 12 -

to, oo o

o *t=6.0xl 02 2

eo e

eto,oo5

Fig. 16.

(1 600-T)7 8(T-1 050)o

1050"C

t*8.0xl O1 5

(125O-T)5. (T- 65O)6

e

1.2

1

0.8

0.6

0.4

0.2

o

-0.2

Fig. 18.

u'>

o=-o'OLL

OE

7;

>

10 100Incubation time, t [s l

Comparison of calculated incubationestimated one from to 005 and /o

5'

time with

F=1-exp(-K.(t-t')4 )

Run B,

1142 ='C

K=1.6xl O~ 6

t'=20

6~

c]

d:l

Run A,1215'CK=2.5xl O~ 6

t'=55

20 40 60 1201OO80Time [s l

Comparisonof volume fractions of crystal betweenobservations and calculation.

o. oool

1

U) 1~~

1

Fig. 17.

Figure 17 showsa comparison of calculated and ob-served crystal growth rate at 1215 and I 142'C. Thecalculation of crystal growth rate at 1215'C wasin goodagreementwith the observation until 70 o/o of the volumehad transformed. However, because of temperaturedecrease (Fig. 2RunA), the rate of crystal growth wasretarded at this time. In the case of 1142'C, the tem-perature in the initial stage of crystal growth ( s,

Fig. 2) was higher (~~ I 150'C) than the average tem-perature, so that the deviation of measured crystal

growth wasobserved in the upper side of the calculation.

As shownin Fig. 9, the observed results had relatively

large scatter, since the nucleation and growth in eachexperiment wasstrongly affected by temperature and the

numberof nucleation sites in the sample. Thesurface of

a liquid often provides a nucleation site; however in this

study, the crystal appeared inside the sample. Occasion-ally it was observed that the crystal grew from a smallbubble that remained in the sample.

If the nucleation rate is zero before a given incubationtime t*, Eq, (2) can be modified to Eq. (3).

F(t)=1-exp( K (t t*)4) ..........(3)

Theincubation time (t *) andcoefficient Kwerecalculated

from the meanvalues measured in each experiment.Using tirne at 0.05 and 500/0 volume fraction, to,005 and

101000 I050 1100 1150 1200 1250 1300

Temperature [ 'C JFig. 19. Variation of coefficient K in Avrami Eq, for crystal

growth of industrial mold slag,

t0,5' the following expression is obtained from Eq. (3),

l/4 to,005

= - **In(1 - 0,005)

_t

(PIn(1-0,5) ~ to.5~t

"""""(4)

Then incubation times t* were obtained for each ex-periment and plotted in Fig. 18 for each temperatureranges.

t* _(p ' t0.5 ~ to.005

,....,...,(5)

q)-l

It is useful to use an empirical expression for describingthe incubation time as a function of temperature. Thefollowing expressions give acceptable calculation results

that fit the measuredincubation time.

t* =6,0 x 1022

(1 600- T)7.8(T- I 050)0,4

1050'C T 1265'C

t* =8.0 x 1015

(1 250- T)3'o(T- 650)3'o

700'C T I050'C

The coefficients Kin each experiment were calculated

from the incubation time t* and Eq. (3) and are shown

Oo o

o~ 5 o o ooo o

oo o~)o

o

~l o

o~ 6o

(b

o

o~ 7

o

1150'c 1265'.8

C 1998 ISIJ 364

ISIJ International, Vol. 38 (1 998), No. 4

in Fig. 19. As mentioned before, nucleation and crystal

growth are strongly affected by temperature and the

numberof nucleation sites in each experiment and thederived coefficient Khas a relatively large scatter. Thecoefficients K obtained at 1215 and I 142'C by fitting

to the Eq. (3) were 2.5 x l0~6 to I .6 x l0~6, respectively,

and are plotted in Fig. 19 as the points markedas 'a'

and 'b'. Although the coefficients Kin 1215 and I 142'Cwerewithin the region of error in Fig, 18, the values could

not be predicted from the average value.

4. Conclusions

TheSingle Hot ThermocoupleTechnique (SHTT)wasemployed to study the crystallization behavior of anindustrial mold slag. This technique proved to be effec-

tive to visualize the beginning of crystallization and to

measure the variation of the crystals size and volumefraction as a function of time. Isothermal tests werecarried out at different temperatures andaTTTdiagramfor an industrial mold slag was determined. The TTTdiagram wasdivided into two regions above and belowl 050'C as two different phases were seen to nucleateand grow in these different temperature regimes. Thecrystal phases were determined by X-ray diffraction

of the sample and dicalcium silicate (Ca2Si04) andcuspidine (Ca4Si207F2) corresponded to high tempera-ture and low temperature region of the TTTdiagram,respectively.

The Avrami equation was applied for the crystal

growth at high temperature and Johnson-Mehl-Avramitype equation wasfound to represent the observed results

for crystal growth at temperatures over I 050'C.

Acknowledgement

This research was funded by the Center for lron andSteelmaking Research at Carnegie Mellon University.

1)

2)

3)

4)

5)

6)

7)

8)

REFERENCESM. Wolf: ISS Process Technology Conf., 13, (1995), 99.

K. Mills, P. Grievson. A. Osuanyaand S. Bacha: TIM-Contin-uous Casting, (1985), 571.

K. Mills: ISS Steelmaking Proc,, (1991), 121.

M. Jenkins: ISS Steelmaking Proc., (1995), 315.

A. Yamauchi, K. Sorimachi, T, Sakuraya and T. Fuji: ISIJ Int.,

33 (1993), 140.

M. Susa, K. Mills, M. Richardson, R. Taylor and D. Stewart:

lronmaking Steelaraking, 21 (1994). 279.

M. Emi: ISS Steelmaking Proc., (1991), 623.

H. Kyoden. T. Doihara and O. Nomura:ISS Steelmaking Proc.,

(1986), 153.

9)

lO)

l 1)

l2)

l3)

14)

l5)

16)

l7)

l8)

19)

20)

21)

22)

23)

24)

25)

26)

27)

28)

29)

30)

31)

32)

33)

34)

35)

36)

37)

38)

39)

40)

41)

T. Wada,M. Suzuki and T. Mori: I,'on Sleelnlclke,', (1987), 31.

T. Kappey: ISS Steelmaking Proc., (1992), 543.

H. Gilles. M. Byrne, T. Ruso and G. DeMasi: ISS ProcessTechnology Conf. Proc., (1990), 123.

K. Sorimachi, M. Kuga, M. Saigusa and T. Sakuraya: Fachber.Halltenp,'ax. Meta!!weiterve,'a,'b., 20 (1982), 244.

H, Nakato, T. Sakuraya, T. Nozaki, T. Emiand H. Nishikawa:ISS Steelmaking Proc., (1986), 137.

M Frazee: ISS Steelmaking Proc., (1995), 639.

R. Branion. D. Dukelow. G. Lawson, J. Schade, M. Schmidt andH. Tsai: ISS Steelmaking Proc., (1995), 647.

T. Kishi, H. Tsuboi. H. Takeuchi, T. Nakano, M. Yamamiyaand T. Ando: Nippon Stee! Tecl7. Rep., 34 (1987), 11.

R. Bommaraju: ISS Steelmaking Proc., (1991), 131.

H. Sakai, T. Kawashima.T. Shiomi. K. Watanabeand T. Iida:

Molten Slags, Fluxes and Salts '97 Conf., (1997), 787.

K. Watanabe,M. Suzuki. K. Murakami.H. Kondo.A. Miyamotoand T. Shiomi: Tc'tsu-to-Hagan~, 83 (1997), 31.

G. Fisher, A. Golloch and J. Kasajanow: Stee/ Res., 67 (1996),

479.J, Ch~vez, A. Rodriguez, R. Morales and V. Tapia: ISSSteelmaking Proc., (1995), 679.

M. Bhamra.M, Charlesworth, S. Wong,D. Sawyers-Villers andA. Cramb: ISS Electric Furnace Conf., (1996), 551.

J. Dubrawski and J. Camplin: J. The,'m. Ana/., 40 (1993), 329.

O. Afrange, M. Fonsecaand H. Polivanov: Conf, on ContinuousCasting of Steel in Developing Countries, Chlna, (1993), 623.

P. Hammerschmidand D. Janke: Stee/ Res., 9(1991), 395.

M. Lanyi and C. Rosa: Metall. Trans. B, 12 (1981), 287.

S. Chang, I. Lee, M. Kim, S. Yang, J. Choic and J. Park: Conf.

on Continuous Casting of Steel in Developing Countries, China,(1993), 832.

W.Qiang, X. Bing and C. Jinghao: Conf, on Continuous Castingof Steel in Developing Countries, China, (1993), 842.

Y. Kashiwaya, C.E. Cicutti, A. Cramband K. Ishii: ISIJ Int.,

38 (1998), 348.

D. Uhlman: J. Am. Ceranl. Soc., 66 (1983), 95.

C. Gatellier, H. Gaye. J. Lehmann,J. Pontoire and P. Riboud:Stee/ Res., 64 (1993), 87.

K. F. Kelton: Solid State Physics. Vol. 45, Crystal Nucleation in

Liquids andGlasses, AcademicPress, Inc,, NewYork, (1991), 75

,

M. Avrami: J. Chem.Phys., 7(1939), I103.

W. A. Johnsonand R. F. Mehl: TAIME,135 (1939), 4i6.

H. Rawson: Inorganic Glass-Forming System, AcademicPress,

London, (1967), 160.

G. Lucadamo,K. Barmakand C. Michaelsen: Mat. Res. Soc.

Symp.Proc., 398 (1996), 227.J. W. Cahn: Mat. Res. Soc. Symp. P,'oc., 398 (1996), 425,

P.G. Shewmon:Transformation in Metals.. McGraw-Hill, NewYork, (1969), 88,

J. Jean and T. K. Gupta: Kinetic of Cristobalite FormationBinary Borosilicate High Silica Glass Composite, CeramicTransactions. Vol. 30, (1992), Nucleation and Crystallization in

Glasses and Liquids, 347.

R. Hill: Physical Metallurgy Principles, D. Van Nostrand Co.,

(1960), 447.

A. Navrotsky: Met. Res. Soc. Symp.P,'oc., 321 (1994), 3.

365 (() 1998 ISIJ