© 2014 ISIJ 2000

ISIJ International, Vol. 54 (2014), No. 9, pp. 2000–2007

The Importance of Materials Properties in High-temperature Processes

Kenneth MILLS1) and Muxing GUO2)*

1) Department of Materials, Imperial College, London SW7 2AZ, UK. 2) Department of Metallurgy and MaterialsEngineering, KU Leuven, Kasteelpark Arenberg 44 - Bus 2450 BE 3001 Leuven, Belgium.

(Received on January 21, 2014; accepted on June 24, 2014)

In high temperature processes, physico-chemical properties of the phases involved have proved usefulin identifying the mechanisms responsible for process problems and product defects. However, recentlymathematical models have developed to the stage where they can now be used to identify these mech-anisms. These models require reliable values for the thermo-physical properties for the various phasesinvolved in the process. The thermo-physical properties of slags are very dependent upon slag structure.The factors affecting slag structure are outlined and the ways in which structural factors affect variousphysical properties are described. Finally, the manner in which the following properties (liquidus tempera-ture; viscosity, surface and interfacial tension; thermal conductivity and enthalpy) can be manipulated tooptimise process control and product quality is demonstrated for various industrial processes.

KEY WORDS: high temperature metallurgy; slag properties; process manipulation.

1. Introduction

Physico-chemical property measurements for materialsinvolved in many high-temperature processes have proveduseful in identifying the cause of problems with processcontrol and product quality. These materials involve metals,slags and refractories. The unofficial First and Second Lawsof High Temperatures state, respectively, “At high tempera-tures everything reacts with everything else” and “Theyreact very quickly and the situation worsens quickly as thetemperature increases”. Consequently, problems with processcontrol and product quality are frequent in high-temperatureprocesses. Property data have proved very useful in the pastin solving these problems. A classical example of this isvariable weld penetration in Tungsten Inert Gas (TIG)welding.1–3) This technique is frequently used as a roboticprocess to make thousands of repetitive welds; the optimumwelding conditions to obtain deep welds (Fig. 1(b)) areestablished in preliminary studies. However, it has beenfound that when these conditions are applied to anotherbatch of steel (fully meeting the materials specification)they produced poor weld penetration (Fig. 1(a)). This prob-lem is known as variable weld penetration.

There are several forces operating in the weld pool, name-ly, Buoyancy forces, Lorenz (electro-magnetic) forces andaerodynamic drag forces (caused by the flow of inert gasover the liquid metal surface) plus Marangoni (or thermo-capillary) forces (Fig. 2).1–3) The latter arise from surfacetension gradients resulting from steep temperature gradi-

ents; the temperatures at the centre and the edge of a weldpool are ca. 2 500 and 1 500°C, respectively, thus gradientsof 500°C mm–1 are normal. Marangoni flows always occurfrom low surface tension to high surface tension.

Surface tension measurements (Fig. 3) were carried outon more than 50 steels with known “poor” or “deep” weldpenetration.2,3) Poor weld penetration correlated with nega-tive values of dγ /dT, high values of γ and S contents of <30 ppm. Deep weld penetration correlated with positive gra-dients (dγ /dT), lower values of γ and S contents > 60 ppm.Consequently, it was proposed that Marangoni forces weredominant in the weld pool and that variable weld penetra-tion was caused by surface tension gradients arising fromthe temperature gradients on the surface of the weld pool.1)

For low S steels the surface flow is radially outward andtakes hot metal from the centre to the edge of the weld poolwhere melt-back results in a wide, shallow weld (Fig.4(a)).1–3) In contrast, for high S steels the Marangoni flowresults in a radially- inward and then downward flow of hotmetal in the centre of the weld (Fig. 4(b)) so that melt-backoccurs at the bottom of the pool creating a deep weld (Fig.4(b)).1–3) Thus measurements of surface tension were able toshow that variable weld penetration resulted from differenc-es in surface tension gradients in the weld pool which, inturn, arose from minor (ppm level) differences in the sul-phur content of the steel.

However, in the past decade a new demand for propertydata has arisen, namely, as input data in mathematical mod-els of processes. In recent years mathematical modelling hasproved to be a valuable tool in minimising process prob-lems. However, these models require accurate values for theproperties of the materials involved. One problem continu-

* Corresponding author: E-mail: muxing.guo@mtm.kuleuven.beDOI: http://dx.doi.org/10.2355/isijinternational.54.2000

ISIJ International, Vol. 54 (2014), No. 9

2001 © 2014 ISIJ

ally encountered in high-temperature processes is that youcan not see into the vessel/mould. Consequently, we mustrely on data from sensors to deduce the mechanisms under-lying the problem. However, mathematical models havenow developed to the state where they can provide us withinsights into the mechanisms involved in high temperatureprocesses; examples of model predictions are shown below

in Fig. 10.If you were designing a new process you would use var-

ious mathematical models to check the viability of the pro-cess, namely

(i) A thermodynamic model to determine the viabilityof the process and the optimum conditions.

(ii) A kinetic model to show productivity will be highenough.

(iii) A fluid flow and heat transfer model to anticipateprocess and product-quality problems.

(iv) Economic and Environmental models to study thefinancial and environmental impact of the process.

In this paper the principal focus will be on (iii) above, i.e.the properties involved in fluid flow and heat transfer andcase histories are given below to illustrate the importance ofslag properties in optimising process control and productquality. The principal focus is on the optimisation of the pro-cess including slag processing for valorisation.

2. Properties of Slags

Many high-temperature processes involve metal, slag andrefractory phases. In this study our main focus will be on theproperties of the slag phases.

2.1. Structure of SlagsIt should be noted that slags are ionic and slag-metal reac-

tions such as Eq. (1) are better expressed in the form of Eq.(2) where the underline denotes dissolved in the metal.

Fig. 1. Photographs showing variable weld penetration and (a) poor weld penetration and (b) good weld penetration.

Fig. 2. Schematic diagram showing the fluid flows resulting fromdifferent mechanisms acting in the weld pool (a) Marangoniforces (+) = (dγ/dT = positive); Marangoni forces (–) = (dγ/dT =negative) (b) E = Lorenz electro-magnetic forces (c) B =Buoyancy forces (d) A = aerodynamic drag forces.

Fig. 3. Surface tension (γ) and (dγ/dT) of steels with 30 and 80 ppm Sas a function of temperature.2,3)

Fig. 4. Schematic diagrams showing how fluid flows in weld poolfor (a) low S (< 50 ppm) steel and (b) high S (> 50 ppm)steel are affected by surface tension gradients (dγ/dT).1–3)

© 2014 ISIJ 2002

ISIJ International, Vol. 54 (2014), No. 9

........ (1)

.................... (2)

2.1.1. Role of the Slag PhaseSlags play a vital role in many pyro-metallurgical pro-

cesses, they carry out in the following functions:(i) They seal off the metal phase from the atmosphere

and prevent oxidation of the metal.(ii) They provide thermal insulation which prevents the

metal from freezing.(iii) Slag/metal reactions remove deleterious elements

(e.g. S) from the metal.(iv) They aid the removal of non-metallic inclusions

from the metal.

2.1.2. Slag StructureThe properties of silicate slags and glasses are very

dependent upon the structure. The basic building block insilicates is a Si4+ ion surrounded by 4 O2– ions in a tetrahe-dral array (Fig. 5). Each of these O2– ions allows one Si4+

tetrahedron to link to another tetrahedron. Thus in pure SiO2

the structure consists of Si4+ tetrahedra joined via O2– ionsat the corners to other tetrahedra in a 3-dimensional array;these O2– ions are known as bridging oxygens (BO). ThusSi is said to exhibit 4-fold coordination and the bonding iscovalent.4)

When cations such as Na+ or Ca2+ are added they tend tobreak the network and form O––Na+ bonds (Fig. 5) or O––Ca2+–O– bonds, respectively; thus the cations are referred toas network-breakers and the bond formed as Non-BridgingOxygens (NBO). These cations (e.g. Na+ or Ca2+) arearranged in octahedral geometry and exhibit, in general, 6-fold coordination.4)

When Al2O3 is added to a silicate slag the Al3+ ions areincorporated into the Si4+ network but the Al3+ needs a cation(Na+ or half of Ca2+) in close proximity for charge-balancingi.e. to form (NaAl)4+ (Fig. 6).4) Cations on charge-balancingduties are not available for network-breaking. Thus Al3+

tends to show 4-fold coordination (denoted (4)Al). With largeadditions of Al2O3 there is the possibility of (5)Al and (6)Alformation.5,7)

The parameter NBO/T is often used to represent the

degree of de-polymerisation of the melt (Eq. (3))

.......................................... (3)

where X is the mole fraction and XMO = XMgO + XCaO +XBaO + XFeO + XMnO + ... and .

We prefer to use the parameter, Q (Eq. (4)) which is ameasure of the degree of polymerisation and which is cal-culated via Eq. (4).

.......................... (4)

It might be expected that TiO2 would behave like SiO2

and would be incorporated into the silicate network but theCaO–SiO2–TiO2 system exhibits a large miscibility gap;although there are some Si–O–Ti bonds formed there is apreference for Ti–O–Ti and Si–O–Si bond formation whichaffects both the structure and the properties. It might also beexpected that Cr2O3 and Fe2O3 would behave like Al2O3.

This is largely true for Cr2O3 but Fe2O3 acts partially as anetwork former and partially as a network-breaker.

Some silicates exist in both crystalline and glassy forms.In crystals the atoms or ions are in fixed positions whereasthere is much more disorder in the rapidly-quenched glassyphase. Thus the entropy of the glassy phase is higher thanthat of the crystal (Sgl>Scryst) and the additional entropy inthe glass (arising from the disorder) is referred to as config-urational entropy.

When a glass or crystal is heated the disorder tends toincrease. At lower temperatures the thermal vibrations areinsufficient to cause much disorder in the quenched glassbut when the glass reaches the glass transition temperature(Tg) there is a rapid increase in disorder as the glass is trans-formed into a supercooled liquid (scl). This increase inconfigurational entropy (ΔSTg

config) is accompanied by astep-increase in heat capacity and 3-fold increase in thermalexpansion coefficient (α).This transition does not occur forcrystals so Cp and S values for the crystal phase are lowerthan those for the scl. However at higher temperatures crys-tals exhibit an entropy of fusion (ΔSfus) at Tliq which isabsent for the scl.

The nature of the cations can also affect the structure andthe properties of silicate slags7) but their effect is muchsmaller than the degree of polymerisation. Cations canaffect the structure of the melt, e.g. reaction (5) is promotedby cations with high field strength (i.e. high values of z/r2)

Fig. 5. Schematic drawings showing (a) Si4+ tetrahedron and join-ing of tetrahedral through O2– ions and at bottom free O’sformed in de-polymerised melts (ca. Q < 1 where Q isdefined in Eq. (4)).

4 3 3 22Al SiO Si Al Osteel slag steel 2 3slag+ = +

4 3 3 44 3Al Si Si Al+ = ++ +

Fig. 6. Schematic diagram showing incorporation of Al3+ into Si4+

network and showing charge-balancing cations.

NBO T 2 X X X / X 2XMO M O Al O SiO Al O2 2 3 2 2 3/ = + −( ) +( )∑∑

X X X XM O Li O Na O K O2 2 2 2= + +

Q NBO T= − ( )4 /

ISIJ International, Vol. 54 (2014), No. 9

2003 © 2014 ISIJ

which, in turn, affects the properties.

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

Alternatively, some properties are affected by the size of thecations (i.e. by their radius, r).

2.2. Effect of Structure on Properties2.2.1. Viscosity (η), Electrical Resistivity (R), Conductiv-

ity (κ) and Diffusion Coefficient (D)Viscosity (η) is a measure of the resistance to the move-

ment of one layer of molecules over another layer of mole-cules. The viscosity increases with increasing degree ofpolymerisation (Q, Fig. 7(a)) and decreasing temperature,which is usually represented by the Arrhenius relationshown in Eq. (6):

......................... (6)

where η is in dPas, T is in K, AA is the pre-exponential termand BA represents the activation energy (E) term (B = E/R*where R* is the universal Gas constant. The temperaturedependence in the scl phase is much stronger and is repre-sented by the Vogel-Fulcher-Tamman relation.

.................... (7)

It can be seen from Fig. 7 that both the viscosity and theactivation energy term, B, are sensitive functions of thedegree of polymerisation, Q.

The electrical conductivity (κ) represents the ability ofthe ions to move under an applied electrical field; the cat-

ions are the mobile species since the anions (silicate ions)are too large and sluggish. The electrical resistivity (R = 1/κ)is the resistance to the cation movement and is therefore theequivalent of viscosity. There are two factors affecting theresistivity, namely, (i) the resistance to the movement of cat-ions and (ii) the number (n) and mobility of the cations; noteNa2O produces 2 cations whereas CaO only produces 1which accounts for the differences between the curves forM2O and MO in Figs. 8(a) and 8(b).

The factors affecting the diffusion coefficient (D) are verysimilar to those for the electrical conductivity and thus a plotof (1/D) versus Q looks very similar to Fig. 8.

2.2.2. Density (ρ) and Thermal Expansion (α , β)Thermal expansion in slags results from the asymmetry of

the thermal vibrations produced when a slag is heated. Thethermal expansion coefficient (α) decreases with increasingpolymerisation (Q) and decreasing field strength (i.e. z/r2).7)

However, for glassy samples there is a large increase in αat the glass transition temperature and ρcrys>ρ scl for temper-atures between Tg and Tliq; however, this is compensated bya change in density for the crystalline phase at Tliq which isabsent for the scl phase.

The density of the crystalline phase is slightly higher thanthat of the glass because of the better packing in the crystalphase.

Fig. 7. (a) Viscosity (ln η1 900 K) and (b) activation energy parame-ter, Bη , for MO–SiO2 and M2O–SiO2 systems for viscosity(upper curve = MO; lower curve = M2O),8) in which MO =MgO, CaO, BaO, FeO and MnO and M2O = Li2O, Na2O,K2O.

2 1 1Q Q Qn n n= +− +

η = ( )A exp B TA A /

η = −( )( )A exp B T TV Vo/

Fig. 8. (a) Electrical resistivity (ln R1 900 K) and (b) activation energyterm , BR, of MO–SiO2 and M2O–SiO2 systems as a functionof the parameter Q (upper curve = MO and lower curve =M2O); + = LS; x = NS; * = KS; o = CS; ▲ = MS; ◇ = SrS;● = BS; ▲ = FS; ■ = MnS; ◆ = MFS where B = BaO; C =CaO; F = FeO; K = K2O; L = Li2O; M = MgO; Mn = MnO;Sr = SrO; N = Na2O and S = SiO2.8)

© 2014 ISIJ 2004

ISIJ International, Vol. 54 (2014), No. 9

2.2.3. Surface and Interfacial Tension (γsl, γmsl)Surface tension (γ ) is a surface property and not a bulk

property. Certain constituents with low surface tension(known as surfactants) tend to inhabit the surface layer ofthe melt and cause both a marked decrease in surface ten-sion (γ ) and cause (dγ /dT) to become progressively morepositive. Typical surfactants in metals are soluble sulphurand oxygen (i.e. not sulphides and oxides); e.g. 50 ppm S inFe causes a decrease in γm of 25%. The surfactants in slagsdo not have as dramatic effect as S and O in Fe but do influ-ence the slag surface tension; typical surfactants in slags arethe fluxes used in most slags, B2O3, CaF2, K2O and Na2O.

The interfacial tension between metal and slag (γmsl) isgiven by Eq. (8) where ϕ is an interaction coefficient:9)

.................. (8)

The principal factor affecting γmsl is γm since for mostmetals γm ≈ 4 γsl and thus γmsl is greatly dependent upon thesoluble S and O contents of the metals.

2.2.4. Thermal Conductivity (k) and Diffusivity (α)Heat transfer across a slag layer at high temperatures is an

exceedingly complex process since it involves several differ-ent heat transfer mechanisms, namely, thermal conduction,radiation and for melts, convection. In semi-transparentmedia, like glasses and slags, the measured (or effective)thermal conductivity (keff) (Eq. (9)) contains contributionsfrom both the lattice conductivity (klat) and the radiation con-ductivity (kR) which involves absorption and re-emission ofIR radiation by each successive layer in the sample. Valuesof kR were found to increase with increasing sample thick-ness (d) until it reaches a critical point; the sample is denot-ed as optically-thick at this critical point which occurs whenα*d > 3 where α* is the absorption coefficient) and there-upon keff remains constant.10) Values of kR can be calculatedfor optically-thick conditions using Eq. (10) and the depen-dence on TK

3 ensures that kR increases dramatically at hightemperatures.

............................. (9)

...................... (10)

where σ = Stefan-Boltzmann constant and n = refractiveindex. These kR contributions can be reduced by (i) absorp-tion of IR radiation by FeO, NiO etc in the slag and (ii) scat-tering of IR radiation by crystals.

The lattice conductivity (klat) is affected by several fac-tors:

(i) whether the sample is crystalline or glassy sincekcrys ≈ 2 kgl due to the closer packing of atoms inthe crystal (cf. glass) and which results in higherconductivity values.

(ii) Porosity in the sample, klat decreasing with increas-ing porosity.

There is a paradox associated with the heat transfer in aslag layer at high temperatures, namely, that crystallites inthe slag reduce the heat flux (q) despite the fact that kcrys ≈2 kgl; this is due to the much larger reduction in the kR dueto the scattering of the IR radiation by the crystals.

One further complication has been identified recently. Ithas been found that the thermal conductivity values for slags

of almost identical composition are very different whenmeasured using the two most popular techniques (i.e. laserpulse (LP) and transient hot wire (THW)). It can be seenfrom Fig. 9 that (i) the k-T relations for kTHW and kLP arevery different and (ii) for the liquid, kLP ≈ 7 kTHW. These dif-ferences have been ascribed to (i) kR contributions in the LPmethod and (ii) electrical leakage in THW measurements,but at the present time the problem is unresolved.

3. Manipulation of Slag Properties for Process Optimi-sation

In this section we show how slag properties can bemanipulated to optimise process control and product quality.The same techniques could be used in valorisation activities.

3.1. Manipulation of Melting CharacteristicsThere are cases where the combined attack of slag and

metal on the refractory has proved so severe that otherwiseviable processes have to be abandoned. One method used toextend refractory life is to distribute the molten slag aroundthe slag walls to create a “freeze lining” of slag in a water-cooled vessel; this sacrificial lining then protects the refrac-tory from further attack. Freeze linings are used in variousprocesses e.g. the Lurgi coal combustion and Basic OxygenSteelmaking (BOS) processes. In the LD version of the BOSprocess, the carbon is removed from the molten mixture of

γ γ γ ϕ γ γmsl m sl m sl= + − ( )20 5.

k k keff lat R= +

k n TR K= 16 32 3σ α/ *

Fig. 9. Thermal conductivity as a function of temperature for (a)Mould fluxes; line, ● = kTHW results and ◇,o = kLP resultsand (b) Na2O–CaO–SiO2 slags of similar composition usingthe laser pulse (LP) and transient hot wire (THW) methods.

ISIJ International, Vol. 54 (2014), No. 9

2005 © 2014 ISIJ

pig iron (from the blast furnace) and scrap by blowing oxy-gen at supersonic speeds. A slag phase is formed by SiO2

(from oxidation of Si in the pig-iron), FeO and added CaO.Wear of the refractories is significant because of the greatturbulence of the gas, slag and metal phases. Thus there isconsiderable “downtime” in the process to replace refracto-ries. “Slag splashing” was developed in the USA and Chinato extend refractory life and minimise “downtime”. In thisprocess when the metal has been drained from the vessel,the lance is lowered and the slag is splashed around thewalls using a gas jet, thereby creating a freeze lining. Sev-eral companies tried slag splashing but found the benefitswere outweighed by a rapid loss of the freeze lining andcontamination of the metal. Some companies persisted andoptimised the conditions for slag formation. Ad-hoc researchshowed that optimum conditions required that slag containsca 13% FeO and >8% MgO. It was subsequently shown thatthese levels of FeO and MgO avoided formation of low-melting ferrites and ensured that high-melting MgOFe2O3

was formed.11) Refractory lives of >50 000 heats have beenrecorded with slag splashing. Thus manipulation of the slagcomposition to provide high-melting phases resulted in sig-nificant valorisation for the BOS process.

3.2. Manipulation of the Slag ViscosityIn high-temperature processes such as Coal Gasification

or in smelting operations it is important to drain the slagfrom the vessel. The drainage rate is inversely related to theslag viscosity and viscosities of ca. 1 Pas are needed toachieve reasonable drainage times. In the case of coal gas-ification the mineral matter (or ash) forms a highly-viscousslag (with Q > 3). We have seen above (Fig. 7) that viscosityis primarily affected by the degree of polymerisation Q, andit is necessary to reduce Q to obtain the required viscosity.CaO additions reduce both Q (and hence the viscosity) andthe Tliq for the slag. The low melting regions of the CaO–SiO2 and CaO–Al2O3 systems lie in a narrow compositionalrange XCaO = (0.4–0.5) and (0.5–0.7) respectively. Thus theCaO additions should be carefully adjusted to ensure theyfall within these limits since the presence of any solid par-ticles significantly increases the viscosity.

Disasters (where the slag solidifies in the vessel) haveoccurred when a fixed amount of CaO has been added to theslag, for instance, when the normal coal has been replacedby a low-ash coal; since the latter creates a much lower slagvolume, the fixed CaO addition results in the formation of

a high-melting, CaO-rich slag.It should be noted that when the slag is drained from a

vessel it leaves a thin layer of slag on the vessel wall (“ladleglaze”) this slag can become a source of slag inclusions inthe next heat;12) where un-contaminated metal is required itis necessary to “wash” the vessel to minimise the contami-nation.

3.3. Surface and Interfacial Tension3.3.1. Entrainment of Slag and Metal

The entrapment of metal droplets in the slag and slagdroplets in the metal are serious problems in metal process-ing since:

(iii) in the production of precious metals and copper etc.metal droplets trapped in the slag represent a signif-icant loss of valuable metal.

(iv) in the continuous casting of steel entrapped slagcauses significant deterioration in the mechanicalproperties of the steel.

Metal entrapment and slag entrapment are caused by thedifferences in velocities of the metal and slag phases andtend to be accentuated by high velocities in the two phases.Take for instance, the continuous casting of steel, high pro-duction rates involve using high flow rates of molten steelin the nozzle (SEN) and more importantly, in the mould.These high flow rates give rise to turbulent metal flows inthe mould resulting in both the formation of standing waves(Fig. 10(a)) and vortices at the metal/slag interface (Fig.10(b)).13)

The flow patterns in the mould are shown in Fig. 11(a).The radially-inward flow of metal is much higher than thatof the slag since the viscosity of the slag is ca. 50 timeshigher than that of the slag. These differences in the flowvelocities result in “necking” of the slag filament and even-tually “detachment” and “capture” of the slag droplet (Fig.11(b), steps 3 and 4 respectively). This mechanism wasidentified by water-modelling studies.14,15) However, recent-ly a mathematical model has been proposed13) which incor-porates the fluid flow (metal and slag), heat transfer andsolidification of the steel; this model satisfactorily predictsthe formation of standing waves, vortices and slag neckingand detachment (Fig. 10).13)

Tsutsumi et al.15) performed water modelling studies andfound that the mass of entrapped slag (mentrap) caused byKarman vortices (Eq. (11)) was affected by both slag vis-cosity (ηsl in Pas) and interfacial tension (γmsl). Increased

Fig. 10. Mathematical model predictions for the formation of (a) standing wave (b) Karmen vortex near the nozzle and (c)necking and detachment of slag in metal.13)

© 2014 ISIJ 2006

ISIJ International, Vol. 54 (2014), No. 9

interfacial tension can best be achieved by lowering the Sand O contents of the metal but decreasing the B2O3, K2Oand CaF2 in the slag would also help in reducing entrap-ment.

........ (11)

It would seem probable that similar mechanisms wouldbe responsible for metal entrapment in the slag (e.g. fromvortex and standing wave formation) especially where rota-ry furnaces are involved. The mass of entrapped slag can bereduced by increasing the interfacial tension. However, thetime taken for the metal droplets to sink back into the metalphase would be increased with increasing slag viscosity.

3.3.2. Other Processes Affected by Surface and InterfacialTension

Surface and interfacial tensions are involved in a numberof pyro-metallurgical processes e.g.

(i) the removal of inclusions from the metal by attach-ment to bubbles.

(ii) the formation of scales on the surface of continuously-cast slabs where fluxes rich in Na2O, K2O and Li2Otend to adhere strongly to the steel furnace and donot fall off in the spray chambers; the SiO2 in thesefluxes reacts with FeO (formed by oxidation of thesteel surface) to form FeSiO3 which forms the scale.

(iii) the separation of droplets of Fe, Cu and preciousmetals from the slag for the recovery of these metalsin slag valorisation processes.

3.4. Manipulation of Heat Transfer Through a Slag“Freeze Lining”

It was pointed out above (in the section on thermal con-ductivity) that heat transfer process in a slag layer at hightemperatures is exceedingly complex. The primary purposeof a freeze lining is to protect the refractory but it also pro-vides valuable thermal insulation. In the continuous castingof steel, liquid slag penetrates between the shell and themould (Fig. 12) and solidifies to form a “slag film” (i.e. afreeze lining) which consists of a solid layer (ca. 2 mmthick) and a liquid layer (0.1 mm thick). The thickness ofthe solid layer (ds) determines the heat transfer and thethickness of the liquid layer (dl) the level of lubrication.13)

Medium–carbon (MC) steels are prone to longitudinalcracking and the remedy is to ensure the steel shell is both

thin and uniform. This is achieved by (i) minimising (ds/keff)and (ii) reducing radiation conductivity (kR) by designing aslag film which develops a high crystalline fraction to scat-ter the radiated heat. This is achieved by selecting the slagfilm to have a high Tliq (to give a thick ds) and a high basicity(i.e. %CaO/%SiO2) to ensure there is significant crystallisa-tion. In contrast, high-carbon (HC) steels are prone to “stickerbreakouts” where steel escapes from the mould because theHC shells are weak. The solution is to create a thick shell(i.e. the opposite of that for MC steels) and this is achievedby using a slag with (i) a low Tliq (to minimise ds) and (ii)a low basicity to promote high heat fluxes via.kR. Thus thecomposition can be manipulated to obtain the required levelof heat transfer.

3.4.1. Enthalpy of Slag at High TemperaturesEnergy costs constitute about 30% of the total costs in

steelmaking. Although several improvements to energy effi-ciency have been made in recent years, the energy costsremain high. One source of heat wastage is the heat tied upin molten slag phase which is just poured off. A few yearsago, Mr. R. Tata, the chairman of Tata Steel was visiting theBOS plant in the steelworks at Jamshedpur and noticed a

Fig. 11. Schematic diagrams showing (a) the fluid flow in the metal in the continuous casting mould (b) the mechanismresponsible for necking and detachment and (c) the location of slag inclusions on the steel slab.

mentrap sl msl= × ( ) ( )− − −1 06 10 7 0 255 2 18

. .. .η γ

Fig. 12. Schematic diagram showing half-section of continuouscasting mould; (1) air; (2), (3) & (5) slag liquid ; (4) &(10) steel shell; (6) mould; (7) solid slag layer; (8) moltensteel; (9) liquid slag film.13)

ISIJ International, Vol. 54 (2014), No. 9

2007 © 2014 ISIJ

flame emanating from the slag pit where molten slag wasbeing cooled (and granulated) with water. He enquired aboutthe cause of the flame and was told it was hydrogen burning;the hydrogen being formed by the decomposition of thecooling water. He then suggested that they should try to har-vest the hydrogen and a research programme was initiated.The major problem in hydrogen harvesting lies in the sepa-ration of oxygen (from the decomposition and the atmo-sphere) from the hydrogen. Tata Steel has successfullyproduced a gas containing > 70% hydrogen with the remain-ing gas consisting of nitrogen and CO2. The proof of con-cept was then demonstrated using a pilot LD plant and isnow being applied in a plant used for the production of fer-ro-alloys. It has been calculated that if this technology issuccessful, the savings in energy for a steelworks producing6 million tonne of steel p.a. would be 30 million dollars p.a.The hydrogen formed can be used either to produce cleanenergy or for the direct reduction of iron ore; it will be usedinitially as a heat source in the ferro-alloy plant.

4. Discussion

We have seen above that property measurements can beexceedingly useful in identifying the causes of process prob-lems. We have also noted that property values are needed forinput data into mathematical models and this requirement isbecoming ever-more important. Furthermore, mathematicalmodels have proved to be capable of identifying the causesof process problems and product defects and these modelswill be used to optimise the process and may also prove use-ful in valorisation operations.

However, there is a wide-range of compositions used inhigh-temperature processes (e.g. the continuous casting pro-cess outlined above) and it is unrealistic to make all therequired measurements on such a wide range of materials.Consequently, there will be an increasing demand for mod-els to calculate property values from chemical composition,which is available on a routine basis. There are some extantmodels available to calculate properties but, in general, thewider the range of slag composition covered, the worse theprediction. This is because the properties usually reflect thestructure. The models which calculate properties for a widecompositional range try to describe the degree of polymeri-sation, using functions like Q (see Figs. 7 and 8), (NBO/T)or thermodynamic functions; they then apply corrections forcation effects. However, certain oxides (e.g. TiO2) causechanges to the slag structure (such as changes in the coor-

dination of Ti4+ cation or a preference for the company ofother Ti4+ ions rather than Si4+ ions). Such structural featuresare not included in the present models and more detailedstructural information may be needed to derive accuratethermo-physical properties of slags.

5. Conclusions

(1) The thermo-physical properties are very dependentupon slag structure.

(2) Mathematical models will be used to gain insightsinto process problems and product defects and this requiresthe need for reliable thermo-physical property values forslags.

(3) Mathematical models will be developed to estimatethe required property data from chemical composition; how-ever, to obtain highly accurate property estimations it maybe necessary to incorporate structural features (like changesin cation coordination) which are not covered at the presenttime.

(4) Knowledge of the factors affecting physical proper-ties allows one to manipulate these properties in order tooptimise these processes.

REFERENCES

1) C. A. Heiple and J. R. Roper: Weld. J., 61 (1982), 975.2) K. C. Mills, B. J. Keene, R. F. Brooks and A. Shirali: Phil. Trans.

Royal Soc, London A, 356 (1998), 911.3) K. C. Mills and B. J. Keene: Int. Mater. Rev., 35 (1990), 185.4) B. O. Mysen: Earth Sci. Rev., 27 (1990), 281.5) G. S. Henderson: Can. Mineral., 43 (2005), 1958.6) J. F. Stebbins, I. Farnan and X. Xue: Chem. Geol., 96 (1992), 371.7) K. C. Mills, L. Yuan, Z. Li, G. H. Zhang and K. C. Chou: High Temp.

Mater. Proce., 31 (2012), 301.8) K. C. Mills, L. Yuan, Z. Li and G. H. Zhang: Proc. 5th Int. Cong. of

Science and Technology of Steelmaking, The Institute of Iron andSteel Technology, Dresden, Germany, (2012).

9) T. Tanaka, M. Nakamoto and J. Lee: Proc. Metal Separation Tech-nology, Copper Mountain, Helsinki Univ. Technol., Finland, (2004),135.

10) R. Gardon: J. Amer. Ceram. Soc., 44 (1961), 305.11) K. C. Mills, Y. Su, A. B. Fox, Z. Li, R. P. Thackray and H. Tsai: ISIJ

Int., 45 (2005), 619.12) S. Riaz, K. C. Mills and K. Bain: Ironmaking Steelmaking, 29 (2002),

107.13) K. C. Mills, P. E. Ramirez-Lopez and P. D. Lee: High Temp. Mater.

Process., 31 (2012), 221.14) S. Feldbauer and A. W. Cramb: Proc. of 13th PTD Conf., Nashville,

TN, USA, (1995), 327.15) K. Tsutsumi, K. Watanabe, M. Suzuki, M. Nakada and T. Shiomi:

Proc. of VII Int. Conf. on Molten Slags, Fluxes and Salts, SouthAfrican Institute of Mining and Metallurgy, Johannesburg, (2004),803.