Slag-Metal Reactions in the Electroslag Process
Slag-metal reactions are fundamentally controlled by the thermodynamic driving force, the physical nature of the reaction products, and mass transport in the liquid slag and metal phases
BY B. M. PATCHETT AND D. R. MILNER
ABSTRACT. Slag-metal reactions in these boundary layers that controls of requirements such as arc stability, electroslag melting have been eval the rate of reaction. The thermody protection of the weld metal from uated in terms of the fundamental namics of the reaction determines the atmosphere, weld metal deoxi- controlling parameters. Simple metal the equilibrium state established at dation and alloying, ease of slag systems using pure aluminum and the interface and hence the concen removal, etc. With present highly iron, and binary iron alloys, have tration, or activity, gradient causing evolved industrial fluxes this results been melted in inert fluxes to which diffusion across the boundary layer. in many chemical reactions oc known quantities of reactive constitu A highly favorable reaction curring simultaneously under the ent were added. Ingots, fluxes, slags produces a high concentration complex physical conditions created and electrode tips were, analyzed gradient and hence a high rate of by the electric arc, so that the funda chemically and metallographically, reaction. The nature of the reaction mental analysis of the reactions and slag bath temperatures were product also influences the rate of taking place is very difficult. The measured with sheathed thermo reaction. Liquid miscible reaction usual approach is to assume that a couples. Aluminum was also products readily diffuse and are state of thermodynamic equilibrium melted in transparent silica crucibles, transported away. Gaseous reaction is attained, on the basis that the and motion picture films of the melt products formed at the interface high temperatures and high surface- ing process provided information on slow the rate of reaction by inter to-volume ratio counteract the short the velocity of the slag motion and fering with the diffusion of reactants time available for reactions to be 1 its influence on metal transfer and through the slag boundary layer, and completed. The chemical composi droplet size. Bath velocities of about solid reaction products slow the re tions of the reaction products in the 100 cm/sec circulate the slag past action rate by blocking off part of the metal and slag phase are then the electrode tip and detach some of area available for reaction. The rate analyzed in terms of thermodynamic the molten metal as showers of of reaction is particularly high where data extrapolated from steel-making small droplets. the reactant in the slag is soluble in and an "effective reaction tempera Most of the reaction occurs at the the metal and is transported into, ture" is arrived at. electrode tip. This is due to the high and reacts with elements in, the bulk The temperatures so calculated for surface area-to-volume ratio of the of the metal, e.g., the oxidation of covered electrode and submerged melting metal and the intense silicon and carbon in steel by iron arc welding vary between 1500 and stirring in the reacting slag and oxide in the slag. 2100 C.2"6 This variation in effective metal phases. The fluid motion is reaction temperature, together with caused by the action of the Lorentz the fact that different elements in forces and rapidly transfers the reac Introduction the same weld, e.g., silicon and phos tive constituents towards the re Slag-metal reactions are of impor phorus, react as though their effec 5 action zone at the slag-metal inter tance in many fusion welding proc tive reaction temperatures differed, face. At the interface there are esses, but comparatively little is plus the further fact that slag and boundary layers in the flux and the known about the mechanisms which metal compositions depend on weld 6 metal, and it is diffusion across control them. The predominant rea ing speed, rate and size of droplet 7 8 son for this is that slag-metal reac transfer and mass transfer, all sug tions have been associated primarily gest that equilibrium is not attained. with the arc welding of ferrous mate The concept of an "effective reaction B M. PA TCHETT is with the Department rials, and they are therefore only a temperature" is thus an approxima of Materials. Cranfield Institute of Tech part of the difficult problem of de tion to an otherwise difficult to eval nology, Cranfield, Bedfordshire, England, veloping practically effective fluxes uate situation, in which reactions are and D. R. MILNER is with the Department attempting to attain to an equilibrium of Industrial Metallurgy, The University, and electrode coatings. Birmingham, England. Paper presented at Over a period of 70 years of largely corresponding to some higher tem the AWS 53rd Annual Meeting held in empirical development, fluxes have perature and kinetic processes deter Detroit, during April 10-14, 1972. been devised to satisfy a multiplicity mine the extent to which they occur.
WELDING RESEARCH SUPPLEMENT! 491-s By contrast the electroslag process offers the possibility of the use of Table 1 — Chemical Composition of Slags and Ingots Melted in Inert Fluxes much simpler fluxes under condi Slag Electrode tions in which slag-metal reactions composition, composition Ingot Composition can be more readily evaluated, since Base metal Inert flux % Al % Na % K % Na % K it utilizes resistance heating in the slag and does not involve problems Al 20% NaF-KCI 0.70 < .001 < .001 .055 .002 of arc stability and the other com % Fe 0 C % Si % c % Si % Ba MgF 1.08 02 .02 plexities of arc welding encountered Fe 2 <*01 <.01 — Fe BaF 1.07 02 <01 .02 <01 when using simple slags.9 Most 2 — Fe-C BaF2 1.10 20 <01 .19 <-01 .001 work on slag-metal reactions in the Fe-Si MgF2 0.91 0.85 electroslag process has been related — — — — Fe-Si BaF2 1.00 — 0.91 — 0.84 — to the use of the technique for the re fining of high quality steels where some evaluation has been made of the reactions giving rise to contam ination and inclusions.10 Loss of omits the electromagnetic forces eters. Then in the third stage sim essential elements and the presence which arise from current flow and plified industrially significant re of inclusions is related to the use of ox which are liable to be of consid actions are evaluated in terms of ides in the flux, and it has been found erable significance. (In arc welding the principles established in the first that the order of reactivity of various such forces have been shown to ex two stages. oxides is that which would be pre ert a very marked effect on mass dicted from their thermodynamic sta transport and therefore on the extent Experimental bility.10 of reaction.12) In electroslag welding the composition of the fluxes used In typical industrial systems involv The processes which determine ing multi-component metals and has been strongly influenced by the extent to which reactions take fluxes several competing reactions covered electrode and submerged place have been investigated for a are usually taking place simultane arc experience, although it has also model system using low melting ously. These bring about small but been realized that with less complex point organic and inorganic materi significant changes in the composi 11 conditions simpler principles of flux als by Crimes. In these experi tion of the deposited metal, but the design are possible.13 ments the reaction between drop fact that only limited changes occur lets and the fluid through which they The objective of the present work makes chemical, metallographic and fall has been examined. It was found has been to clarify the reaction proc theoretical analyses difficult. Thus that the reaction is controlled by dif esses in electroslag welding by ap for this research simple systems fusion through a boundary layer at proaching the problem in three have been selected involving the the reacting interface, with the rate stages. First, the physical conditions minimum of constituents in the flux controlling step in the boundary of melting, fragmentation and trans and metal, while the first reactions layer of either the simulated metal or fer of electrode material, slag motion considered have involved massive the slag phase, depending on the cir and temperature are considered. In and therefore readily detectable cumstances. The bulk of reaction oc the second stage simple (non-indus changes in composition. curred during droplet transfer rather trial) systems involving massive reac Two metal systems, based on than during droplet formation or at tions have been analyzed to deter aluminum and iron, have been ex the "ingot". Such a model, however, mine the major controlling param- amined by melting in an inert flux to which a reactive constituent has been added. With the lower melting point aluminum system it was pos sible to observe the metal transfer ELECTRODE and slag motion by melting in trans parent silica crucibles. A massive reaction with iron was examined by putting nickel oxide in the flux to give WATER transfer of nickel into the iron. With rr±±r aluminum two reactions involving massive transfer of an alloying element from the slag to the metal were investigated by putting copper chloride or copper oxide in the flux as the active constituent. In all three V systems the element freed from the slag is completely miscible in the metal and forms readily analyzed al COPPER loys. CRUCIBLE The industrial type reactions investigated were based on iron and WATER JACKET BASE PLATE steel. They were of two types: first, reactions involving transfer of silicon or manganese into the molten metal from Si02 or Mn02containing slags; i second, the removal of silicon and carbon from electrode wires by slags containing various oxides. WATER The inert fluxes were chosen on Fig. 1 — Diagram of the electroslag apparatus for ingot deposition the basis of three criteria: suitable
492-s | OCTOBER 1 972 was water-clear when melted on its own, it quickly became opaque when Table 2 —Chemical Composition of Experimental Electrodes, % aluminum was melted into it. To obtain a visual record of the melting Constituent Al Fe Fe-C Fe-C Fe-Si process it was therefore necessary C — 0.02 0.12 0.19 0.01 to limit the depth of slag between Mg 0.025 — — — — the crucible wall and the electrode Cu 0.05 — — — — tip by off-setting the electrode to the Zn 0.01 — — — — camera side of the crucible after Si 0.10 <01 <01 <01 <91 Na <.001 — — — — melting conditions were established. K <.001 — — — — Color motion picture films at 18 and Other <01 <01 <01 <-01 400 frames/sec showed that the slag bath flowed downwards past the electrode tip at 50-100 cm/sec in the inert slag, and then upward at the crucible walls giving a continu melting and boiling points, high ther given in Table 2. Fluxes were mixed ous circulation of the slag bath as in modynamic stability, and lack of sol from laboratory-grade chemicals and ubility of any of the constituents in sintered in the solid state before the liquid or solid metals. On this crushing and sieving, to yield granu basis the 20 wt% NaF-KCI eutectic lar material with a particle size was chosen for aluminum. For the between + 300w and -2800,u (+52 ELECTRODE iron and iron alloys two alkaline mesh to -6 mesh). Slag bath tem earth fluorides — MgF2 and BaF2 — peratures were measured with were used; these have the maximum sheathed thermocouples; platinum- difference in density of the suitable platinum/rhodium couples were fluorides and therefore enabled this used for aluminum and tungsten- variable to be examined. Analysis of tungsten/rhenium for iron. Metallo ingots deposited from the various graphic and chemical analyses of the electrode materials using these fluxes, slags, ingots and electrode fluxes showed only a slight rise in tips were used to trace the progres the sodium level in the aluminum sion of mass transfer between the ingots, from <0.001 wt% to 0.055 slag and the metal phases. wt%, and a drop in silicon content of the iron-silicon alloy from 0.91 wt% Physical Conditions During Ingot to0.85wt% —Table 1. Deposition The experimental electroslag The crucible was filled with granu ingots were cast by initiating melting lated flux before starting deposition under a layer of granulated flux with of an ingot, and more flux was added a high frequency spark between the as melting took place to keep the pointed electrode tip and a base crucible nearly full. The solidified metal of electrode composition. A slag yield was 25-30 cc for each water-cooled copper crucible of 28 ingot cast. Deposition occupied 30- mm diam by 75 mm depth was used 45 sec per ingot depending on the for the bulk of the experimental work electrode feed rate used. To assess — Fig. 1. A transparent silica cru the extent to which solidification oc cible of similar dimensions was used curred during deposition, iron rods to observe and photograph the elec were plunged into several aluminum troslag deposition process with alu ingots at the instant when the melting minum electrodes. current was switched off. The rods The molten slag and metal were penetrated within 5mm of the protected from the atmosphere by a bottom of the ingots showing that vir flow of argon until solidification had tually the whole ingot was molten finished. Deposition conditions were until the end of deposition. Tungsten standardized as far as possible for rods plunged into iron ingots gave each metal to facilitate comparisons the same result. among the different reactive systems. The aluminum electrode Temperatures in the slag baths feed rate was 184 m/hr (121 ipm), were recorded by thermocouples im giving a melting condition of 50-60v mersed between the electrode and at 100-150 amp, and the iron elec the crucible wall. Slag temperatures trode feed rate was 114 m/hr (92 were 1000 + 50 C during alumi ipm), with a potential of 50-60 v and num deposition and 1900 + 100 C a current of 400-450 amp. Flux during iron deposition. Moving the composition had some effect on thermocouples around in the slags these values as did, of course, experi did not give rise to any observable ments in which the influence of melt changes in temperature, showing ing rate was examined. Any devia that the slag bath temperature was tions are given in the text. Direct cur essentially uniform. Fig. 2 — Diagram of slag motion and metal transfer observed with the melting rent with electrode positive polarity Slag Motion and Metal Transfer. was used for all experiments. of aluminum, (a) with an inert flux, show Aluminum was melted into the inert ing form of slag motion, (b) with a reac The electrodes were 3.2 mm diam NaF-KCI flux and also the same flux tive flux giving rise to a gaseous reaction by 2 m long; the chemical composi with a highly reactive addition of 25 product and submerged arcing and ir tions of the various alloys used are wt% CuCI2. Although the inert slag regular droplet transfer
WELDING RESEARCH SUPPLEMENT! 493-s o - 0 - CuO _^CuCI2 -10 - -10 •
^' - -• -^. -Jr~" ~~"~—CuCI -30 .-- "••*" z AICI3 Al 9U -70 0 ^^ 2°3 OXYG E LUQ. ATIO N / CQO ..O -90 £5 -90 . I*- ^ tD CE Kip-110 UJ "- z LU ^ 1 rr= 115,-130 LL _l LU l/> -150 -150 1 1 1 1000 1500 1000 1500 2000 TEMPERATURE TEMPERATURE °C Fig. 3 — Left, the free energy of formation of iron compounds and associated reactive slag compounds and reaction products; right, the free energy of formation of aluminum compounds and associated reactive slag compounds Fig. 2a. Occasional submerged arc attributed to the fact that three transfer with the MgF2 based fluxes ing was observed in the slag at the types of reaction product were which, having a lower density, did electrode tip. The visible metal trans formed — liquid, gaseous and not contain as much nickel per unit fer took place as intermittent show solid, with the rate of reaction de volume as the BaF2 based fluxes. ers of small droplets with a group creasing in this order. The reac Ingot compositions were substantial velocity of 50-75 cm/sec. tions are therefore considered indi ly uniform from top to bottom — In the reactive slag the flow pat vidually in these terms. (The Table 4. tern followed the same general reaction products and their state at More detailed information on the trend, but there was much more sub operating temperatures for the sys reaction mechanism came from an merged arcing and a great deal of tems considered in this paper are analysis of the droplet tips and slags. random activity at the electrode tip given in Table 3.) The droplet shown in Fig. 5a appar and disturbance of metal transfer — Liquid Reaction Product. Pure ently formed under comparatively Fig. 2b. This was probably asso iron electrodes were melted in fluxes stable conditions; it was obtained ciated with evolution of aluminum containing NiO additions to the BaF2 with a flux containing 25 wt% NiO chloride gas at the electrode tip as a and MgF2 inert carriers. The reaction and shows a surface layer rich in result of the reaction: involved is: nickel with a particularly high nickel concentration at the neck of the (CuCI2) + AI %Cu+ [AICI3] (NiO) + Fe % Ni -t-(FeO) drop. Electron probe microanalysis The solidified slags from these ex The reaction will proceed strongly showed that the nickel rich regions periments and also slags from the to the right at 1900 C since nickel at the surface and neck of the drop deposition of iron ingots contained oxide is much less stable than iron contained 50 to 65 wt% nickel, while approximately 1 wt% of metallic oxide — Fig. 3a. There was a very the body of the drop had a compara droplets (see Figs. 6, 10 and 13). large transfer of nickel from the slag tively constant nickel content of 15- to the metal, with virtually all of the 20 wt%. The droplet in Fig. 5b has Systems Showing Massive Reaction nickel removed from the slag (Table been displaced. As a result, the It was found that the rate of re 4) so that the ingot composition was surface layer has broken up and action varied considerably for the a linear function of the vol % NiO in parts of it have been swept in with three systems examined. This was the flux — Fig. 4. There was less traces of high nickel content ma- Table 3 — Physical State of Reaction Products During Electroslag Melting Reactive Melting Boiling Physical state Electrode Slag bath slag Reaction point, point. during metal temperature, C compound product C C reaction Al 1050 Oxide Al203 2027 >2500 Solid Al Chloride AICI3 — 180 Gas Fe 1900 Oxide FeO 1368 >2500 Liquid Fe Fluoride FeF2 1102 1827 Gas Fe Chloride FeCI2 677 1026 Gas Fe-C 1900 Oxide CO <25 <25 Gas Fe-Si 1900 Oxide sio2 1610 2500 Liquid Fe-Si Fluoride SiF4 <25 <25 Gas 494-s | OCTOBER 1972 terial showing the pattern of fluid to the ingot as before (Table 5) al flow; in this case the body of the though the nickel content of the drop contained 10-15 wt% and the ingot was lower, since a smaller pro ingot 22-24 wt% nickel. portion of the flux above the ingot All of the slags contained metallic melted than for the standard depo droplets and dendrites which were sition conditions. rich in nickel (Fig. 6), while the slag Gaseous Reaction Product. A surrounding the droplets was vir high driving force giving massive re tually devoid of nickel, but contained action, but with a gaseous reaction large quantities of iron. Thus the re product, was involved when melting sidual nickel content of the slags can pure aluminum with fluxes contain be largely attributed to small droplets ing cupric chloride. The reaction is of trapped from the metal transfer proc the form: ess, while iron is present in two forms — as metallic iron in the drop 3(CuCI2) + 2AI <^H> 3 wt% Cu + lets and as iron oxide in the slag, 2[AICI3] which was not completely miscible As shown by Fig. 3b, this reaction in the solid state with the BaF 2 will proceed strongly to the right. At carrier. the standard electrode feed speed of An increase in the electrode feed 183 m/hr (120 ipm) for aluminum, 1 speed by a factor of 2 /2 was used to the operating voltage was 65 v and o MgF2 BASE assess the effect of a higher melting the current about 100 amp. Consid rate using the 25 wt% NiO-BaF2 flux. erable fuming and some submerged 0 10 20 The new deposition conditions were: arcing occurred during deposition, VOL"/. NiO IN FLUX 284 m/hr (186 ipm) electrode feed especially in slags containing large speed 30-35 v, 650 amp. At the amounts of CuCI2. The amount of Fig. 4 — Ingot composition vs. flux compo higher melting rate virtually all of the copper transferred to the ingots was sition for pure iron melted in N/'O-bearing available nickel was still transferred again a linear function of the flux slags 2 3 / X y • x 10-5 •Biff, POSITION %Hj_ 1 0 2 10-5 3 22-5 L 4 24 5 27-5 6 52 7 60 8 13 9 36 X100 10 6 Fig. 5 — The pattern of Ni absorption by pure Fe at the electrode tip obtained with a flux containing 25 wt% NiO: left, showing the high concentration of Ni at the metal surface and at the neck of the droplet; right, showing the macroscopic mixing of alloyed surface layers by stirring forces in the molten metal WELDING RESEARCH SUPPLEMENT! 495-s composition — Fig. 7. But the overall reaction rate was much lower in that only about half of the copper in the slag was transferred to the ingot — Fig. 8. In the melted areas of the droplet tips, the copper level and microstruc ture were similar to that of the ingot, showing that the reaction took place virtually entirely at the electrode tip. There was some inhomogeneity in the droplet and evidence for a some what higher copper concentration around the neck of the drop, but not Fig. 6 — Left, typical example of metal dendrites and particles in a NiO-BaF2 slag (XlOO); on the scale observed with nickel right, electron probe analyses of a selected metal dendrite and particle showing high transfer; there was no evidence of a nickel content, absence of nickel in the slag and the lack of m/scibility between the FeO copper-rich surface layer — Fig. 9. reaction product and the inert BaF' slag (X600J. Photomicrographs reduced 54% This suggests that in this reaction the transfer of the copper reaction product into the droplet is more rapid than its formation at the drop surface. Table 4 — Nickel Trar isfer from Oxide Slags to Pure Iron The slags contained areas which Ingot composition, were rich in copper and poor in alu Flux composition Flux content, Slag composition. wt' % Ni minum, and others that contained wt% NiO wt% Ni wt% Ni Top Bottom some aluminum but virtually no copper. This indicated that the reac 5 NiO-BaF2 2.9 001 4.4 4.8 tion process was subjected to spas 10 NiO-BaF 2 5.6 0.05 15.2 15.6 modic interference, presumably re 15 NiO-BaF2 10.8 0.01 13.6 14.0 lated to the evolution of the gaseous 20 NiO-BaF2 12.4 0.09 16.6 14.8 aluminum chloride and the accom 25 NiO-BaF2 17.2 0.19 34.8 34.8 panying intermittent submerged arc 20 NiO-MgF2 11.8 0.50 4.2 4.6 35 NiO-MgF2 25.6 1.21 22.1 23.7 ing. The slags also contained metal lic droplets which were virtually pure copper, despite the fact that alumi num was being deposited — Fig. 10. Thus small aluminum droplets cre Table 5 — The Effect of Electrode Feed Rate on Ni Transfer to Fe ated in the metal transfer process apparently react completely with the Electrode Feed Rate, Slag composition Ingot composition, slag to produce pure copper droplets Flux m/hr wt% Ni wt% Ni and aluminum chloride, which dis 25 NiO-BaF, 114 0 19 34.8 perses. 284 0.50 13.5 These observations suggested that, at the electrode tip, insufficient time was available to react all of the available copper chloride with alumi Table 6 — Voltage and Current Levels for Varying Aluminum Electrode Feed Rates num. To determine the effect of reac tion time two further experiments Electrode feed rate, Voltage, were carried out. In the first, the m/hr Current, amp electrode feed rate was altered to 114 65 50 change the time available for reac 183 65 100 tion of the electrode tip; in the sec 284 65 200 ond, an attempt to reach an equilib rium state was made by melting a mass of aluminum under a molten slag using an inert tungsten elec Table 7 — Nickel and Copper Transfer to Iron trode and long reaction times. The changes in the electrode feed Flux composition. Ingot composition, Slag composition, Flux wt% wt% wt% rate were made with the 25 wt% CuCI flux to maximize any effect on 25 NiF -BaF 12.14 Ni 0.91 Ni 2 2 2 12.48 Ni the copper transfer. The operating 20 NiCI2-MgF2 10.00 Ni 4.13 Ni 8.66 Ni 25 CuO-BaF, 18.34 Cu 24.58 Cu 0.92 Cu current increased with feed rate, but there was no change in the oper ating voltage — Table 6. Ingot con centrations decreased with in creasing feed rate (Fig. 11), showing Table 8 — Mn Transfer from Mn02 Slags to Pure Iron the extent of reaction to be kineti- cally determined. In the inert elec Flux composition, Ingot composition. Slag composition, trode experiments an arc was struck wt% oxide wt% Mn wt% Mn wt% Fe on a 50 gm piece of aluminum under an argon shield and flux was then 100 MnO2 0.70 51.2 23.4 added until the arc was extinguished 20 Mn02-BaF2 038 9.84 6.60 and electroslag conditions were achieved. These melts differed from 496-s | OCTOBER 1 972 18 o Cu IN Cu 0 SLAG CuCI FLUX 0 2 • AL IN Cu 0 SLAG 16 • Cu IN Cu C l2 SLAG • AL IN CuCI2 SLAG 14 12 O 0 / o •7,10 0 / Z / U o / j* 6 / • 10 15 20 5 // WT% CU IN FLUX U Fig. 7 — Ingot composition vs. flux composition for pure AI • A melted into CuC/2 containing slags which form a gaseous reac / a 0 g-- Q—-- — t/on product and Cu 0 bearing slags which give rise to a solid / a 2 2 reaction product 1 0 —L- i i u 1 • Fig. 8 — Slag composition vs. flux composition 10 15 20 2 5 30 for CuCI2 and Cu2 0 bearing fluxes WT"/. Cu IN FLUX the consumable electrode melts in quid close to its boiling point — The slag composition showed a two ways: first, electrode negative Table 3. In all cases the reaction is corresponding loss of copper and polarity was used, and second, the thermodynamically highly favorable. pick-up of aluminum — Fig. 8. melting was carried out in the trans It was found (Table 7) that with the Alterations in the electrode feed rate parent crucibles to avoid excessive fluoride and oxide additions there had virtually no effect on the copper heat loss with the water cooled cru was a very high level of transfer of level in the ingots -- Fig. 11. These cibles. reactant to the ingot, as with NiO in ingots were cast in a flux containing Fluxes containing 5 and 25 wt.% the flux. On the other hand, with the 12.7 wt% Cu, which was very nearly chloride addition the transfer was the same as the copper content of CuCI2 were used with an operating current of 100 amp for a 3 min much reduced as with the transfer of the chloride flux for which the effect period (longer melting times brought copper to aluminum. of melting rate was determined. The on a collapse of the silica crucibles). Solid Reaction Product. A high low overall invariant rate of copper A condition close to the reactive limit driving force with a solid reaction pick-up with the oxide flux shows that the barrier to mass transfer was was reached with the 5 wt% CuCI2 product was obtained by melting alu effective and rapidly formed during flux, since the slag contained only minum with Cu20 in the flux — Fig. melting. 0.10 wt% Cu; with the 25 wt% CuCI2 3b. The reaction is of the form: flux most of the copper was again Microscopic examination of melted 3(Cu 0) + 2AI = 6 wt% Cu + WELDING RESEARCH SUPPLEMENTl 497-s 7. i X50 I t " .* "' ".„ •••••*' X17-5 !** ''W 1 / X100 Fig. 10 — Top, metallic droplets formed in CuCI2 and Cu20 slags F/ 141- elements from the slag to the metal 12 phase because of the high thermo dynamic driving force. Results have shown that the extent to which re action takes place is dependent on the nature of the reaction product, 10 i.e., whether it is liquid, gaseous or solid, and the barrier it therefore o creates to further reaction. Reactions in industrial systems do ?8 not involve massive transfer since the primary aim is protection of the weld metal, not extensive oxidation- reduction processes. Also reactions O take place not only giving rise to transfer of elements from the slag to the metal, but also from the metal to the slag. In this section much more limited reactions are considered, involving the transfer of elements typically found in industrial systems based on iron and steel, i.e., silicon, manganese, and carbon and the • CuCI2 FLUX effect of putting various oxides in the oCu20 FLUX flux. In some cases these reactions have been enhanced by going well beyond the concentrations used in dustrially, so as to make their effects 0 100 200 300 more obvious. WIRE FEED SPEED- M/HR The Transfer of Si from a Si02- Fig. 11 — Ingot composition vs. electrode feed rate for pure AI melted in Containing Flux. In these experi CuCI2and Cu20 fluxes ments SiO2 additions of up to 30 498-s I OCTOBER 1 972 wt% were made to the inert BaF2 and MgF2 fluxes, and these were used for melts with pure iron and iron-0.91 wt% silicon electrodes. The potential reaction is the absorp tion of Si by iron; (Si02) + 2xFe ^ %Si + 2(FexO) where the reaction product, FeO, entering the slag is liquid and the freed silicon is completely miscible with iron in the liquid state. With ingots deposited from pure iron electrodes chemical analysis showed that very low levels of sili con transfer took place, since even with 30 wt% Si02 in the BaF2 base the maximum reached was 0.37 wt% Si. Silicon transfer increased rapidly VOL.7. Sf02 IN FLUX with increase in Si02 content of the flux at low levels, but the rate of in crease dropped off as the flux Si0 Fig. 12 — Ingot composition vs. flux composition for pure Fe melted in Si02 2 bearing fluxes content increased — Fig. 12. On a wt% Si02 basis the MgF2 based fluxes again transferred less material to the ingot, but on a volume-percent basis the Si transfer no commercial significance at this to MnO and the excess oxygen was similar in both cases — Fig. 12. concentration, provided data for a oxidizes the iron to FeO which goes Increasing the electrode feed rate theoretical analysis of the reaction into solution in the slag, then 97% of from 114 to 284 m/hr (92 to 186 process (see II under Appendix). Man the iron content of the slag is ac ganese oxides have similar thermo counted for. ipm) with the 30 wt% Si02-BaF2 flux did not produce a measurable dynamic stability to Si02 along with Absorption of Elements from change in the resulting Si content of similar melting points and solubility Mixed Oxides in the Flux. The effect the ingot which was 0.30 wt%, i.e., behavior, but they have variable of having both high and low ther- well within the range for ingots cast oxidation states; as obtained and modynamically favorable reactions under standard conditions. Depo added to the flux mixture at low tem occurring at the same time was ex perature the oxide exists as Mn0 sition of an ingot with the iron—0.91 2 amined by combining NiO and Si02 whereas at high temperature in the wt% Si using the 30 wt% Si02 in a BaF2 base. The fluxes were containing flux showed that the equi slag it exists as MnO. The reaction made up with high and low levels of librium Si level of the iron had not under consideration is thus: both reactive compounds combined been reached. This was because (MnO) + Fe *=* %Mn+(FeO) to make one oxide dominant (fluxes there was still Si transfer from the With iron electrodes deposited 1 and 2, Table 9) and with the oxides flux to the metal, the Si increasing under standard conditions the trans approximately balanced (fluxes 3 and from 0.91 to 1.07 wt%, although the fer of Mn to the ingot was small 4). Chemical analysis of the ingots rate of Si transfer was appreciably (less than 1 wt%), but somewhat higher produced under standard conditions lower at this higher Si level. than the transfer of silicon, (Table showed that Ni transfer occurred at 8). Large amounts of iron were the levels expected, but that very The Si02 content of the slags picked up by the slags. If the Mn02 is little Si transfer took place with any dropped markedly, in quantities far regarded as decomposing completely of the fluxes — Table 9. too large to be accounted for by transfer to the ingot; for instance, the slag from the 30 wt% Si02 flux contained 40% less Si02. An acrid vapor evolved during melting and was probably SiF4 which could account for this drop in slag silica level. Neither melted electrode tips nor droplets trapped in the slag re vealed any information about the Si transfer, because the Si levels were too low to be observable metal lographically or with the electron probe. The unusual slag structure (Fig. 13b) proved to be a result of the association of the Si02 in the slag with the FeO reaction product to form some type of iron silicate which was not completely miscible in the inert fluoride base. The Absorption of Manganese into Iron. Two fluxes were made to evaluate this reaction. One was 20 Fig. 13 — The structure of SiO2 slags: left, the duplex phase structure (XI60); right, the wt% Mn02 in a BaF2 base; the other electron probe analysis revealing the association of Si02 and FeO in the slag (X1000). was pure Mn02 which, although of Reduced 42% WELDING RESEARCH SUPPLEMENT! 499-s The Loss of Silicon from 0.91 experiments the electrode tips re wt% Si-Iron Melted in an Oxidizing tained some melted material and pro Flux. The experimental work up to vided more detailed information on this point has considered only the the progress of the reaction. Fig transfer of elements from the slag to ure 17 shows three important fea the metal. An equally important reac tures of the carbon oxidation pro tion, especially in practical welding cess: situations, is the loss of alloying 1. The slag-metal interface is very elements in the metal to reactive irregular. slags. A simple reaction of this type 2. Porosity is present at the solid was investigated by melting 0.91 metal-liquid metal interface. wt% Si-Fe alloy in fluxes containing 3. The melted metal is apparently Fe203, where the reaction is: devoid of carbon and full of oxide in clusions. 40 60 80 100 %Si + 2(Fe 0)< — x •2 Fe+(Si02) These facts show that decarburiza tion was taking place throughout the This reaction has a high molten metal and that the resultant Fig. 14 — Ingot composition vs. flux thermodynamic driving force and li gas evolution fragmented the elec composition for a 0.91 wt% Si-Fe alloy quid reaction products. trode tip continuously during melt melted into Fe203 -bearing fluxes ing. The ingot carbon level in this With small amounts of Fe203 in the flux there was marked loss of Si particular case was 0.023 wt%, cast from the metal which dropped from from a 0.12 wt% C electrode. In ap the original level of 0.91 to 0.1 wt%, proximate terms, the loss of 0.10 wt% Fig. 14. With further increases in C to form CO at 1900 C produced 120 cm3 of gas for each cm3 of iron, Fe203, even to 100%, there was little further change. This loss of Si to cause the porosity and fragmen from the metal to the slag was much tation observed. The oxide inclusions greater than the reverse reaction of are the result of the increase in absorption from the slag into the oxygen level, which rose from metal. Increasing the electrode feed 0.0015 wt% in the 0.12 wt% C iron rate to 284 m/hr (186 ipm) had no to 0.0187 wt% in the ingot. effect on the residual Si level. There In the slags a larger amount than was evidence for considerable usual of iron particles was found due excess oxygen in the electrode tips to the explosive nature of the trans and the ingot structure which ap fer process. The evolution of CO gas Fig. 15 — Oxide inclusions showing trans peared as inclusions on solidification continued in small metal droplets in fer of oxygen into the bulk metal in an -Fig. 15. the slag, although there was little ingot produced from 0.91 wt% Si-Fe alloy Carbon Loss in Iron Oxide evidence of carbon remaining in the melted in a pure Fe202 flux. XlOO (re metal — Fig. 18. This is consistent duced 45%) Fluxes. Iron-carbon alloys at three levels of carbon content (0.19, 0.12 with the observations in the copper and 0.02 wt%) were melted into the chloride-aluminum experiments where the reaction continued in • 19 WT-/. C ELECTRODE Fe203 - containing fluxes. The reac tion under consideration is: small metal droplets suspended in * .12 WT*/. C ELECTRODE the slag until all of the aluminum ° 02 WT7. C ELE CTRODE (FexO) + % C = x Fe = [CO] was consumed and copper remained. where the reaction product CO is Carbon Loss with' Fluxes gaseous. There was a large reduc Containing Various Oxides. Further tion in the carbon level with small experiments on the oxidation of car 16 additions of oxide to the flux — Fig. bon were made with the 0.19 wt% C- 16. This was similar to the silicon Fe alloy using a series of fluxes loss observed with the iron-silicon which contained oxide additives vary alloy. The low carbon (0.02 wt% C- ing in reactivity from CuO to the Fe) alloy was, however, apparently most stable available oxide, CaO. In unaffected by the melting operation. some cases the fluxes would not The effect of an increased electrode support the standard voltage and cur feed rate was investigated with the rent (50-60 v, 400-450 amp) at the 0.19 wt% C-Fe alloy and two fluxes 114 m/hr (92 ipm) feed rate; the — pure Fe203 flux and the 5 vol % electrode feed rate was then main Fe203-BaF2 flux. The increase in tained and the electrical power electrode feed rate caused some loss values allowed to adjust to the slag of control over the voltage and cur characteristics. rent with voltages fluctuating It is apparent that the degree of -•06 between 10 and 40 v and the current carbon oxidation is related to the between 500 and 1000 amp. How thermodynamic stability of the oxide ever, there was no apparent change and, correspondingly, so is the in the carbon content of the ingots amount of alloy element transferred within the range of experimental to the ingot — Table 11. With any scatter observed with the standard oxide in the flux there was some loss feed rate of 114 m/hr (92 ipm) — of carbon; only by melting in pure 0 20 40 60 80 100 Table 10. fluoride was all the carbon retained. V0L"/.Ft0 IN FLUX The electrode tips of iron-carbon al It must be pointed out, however, that this carbon loss would not occur Fig. 16 — Ingot composition vs. flux loys melted into the pure oxide flux rarely contained any melted material. with commercial fluxes which are composition for iron-carbon alloys melted fused in graphite crucibles and there- into Fe2 03 -bearing fluxes However, in the oxide-fluoride flux 500-s I OCTOBER 1 972 by pick up carbon which transfers to the deposited metal; the fluxes used Table 9 — Ni and Si Transfer from Combined Driving Force Fluxes in the present work were made by solid state sintering, specifically to Ingot composition, avoid contamination by crucible ma lux No. NiO, wt% SiOj, wt% BaF2, wt% wt% Ni wt% S terials. 1 25 2 0 73 22.5 0.02 2 5 20 75 4 7 0.01 Discussion 3 25 15 60 22.0 0.06 4 5 2 5 92.5 3.9 0.02 The results of the present work show that in electroslag melting it is possible, with an inert flux and inert gas coverage, to deposit metal of the Table 10 — Carbon Oxidation at Increased Electrode Feed Rates same composition as the electrode. Alternatively, it is possible to some Electrode Ingot extent to control the composition of feed rate, Voltage, Current, composition the melt by adjusting the compo Flux m/hr V amp wt% C sition of the flux. These attributes of 114 30 400 .026-050 the electroslag system are not 229 15 1000 .02 readily attainable with arc welding, 5 vol% Fe203-BaF2 114 50 400 .03 -.07 because arc welding cannot be car 183 30 600 .02 -.05 ried out with inert fluxes or with 284 20 800 .02 -.04 fluxes designed simply to obtain compositional control of the weld metal. Arc welding fluxes require constituents that promote arc stabil ity, adequate penetration, ease of • ^' * • -v slag removal, etc.; they do not there 7'77 7:7:7,7:77: fore have the flexibility of electroslag - fluxes. The authors tried to investi . - gate slag-metal reactions in arc welding with a range of flux composi tions in a similar way to the present investigation. In the vast majority of cases, however, no control of the system was possible due to arc in stability. The Processes of Reaction The reactions that take place in electroslag welding do so primarily at the electrode tip. There are two reasons for this — the high surface- to-volume ratio and the rapid trans port processes. Spherical droplets of 1 to 4 mm diam (which embraces the observed range of droplet sizes) have surface-to-volume ratios of 15 to 60, whereas 50 to 100 gm ingots of 2.5 mm diam have ratios between 0.35 and 0.7. There is thus an over whelming increase in surface area for a given volume at the electrode tip. Current enters the slag through the constricted area of the electrode tip and then broadens out to take the path of least resistance. It was shown by Maecker,14 and the concept applied to welding systems by other workers, 12<15 that where a current path through a fluid broadens out for a constriction there occurs a Lorentz force-induced high velocity motion of the fluid away from the constriction. Applying Maecker's formula for the maximum C / • velocity of flow gives a value of about X400 50 cm/sec for the NaF-KCI slag used with the aluminum melts (see I Fig. 1 7 — The melted electrode tip of a 0.12 wt% C Fe alloy melted in an 8 vol-% FeJJ3- under Appendix), compared with the BaF2 flux, showing the explosive effect of CO evolution at the slag-metal and liquid metal- experimentally measured value of 50- solid metal interfaces (X 20). The details show (a) metal droplets expelled at the start of 65 cm/sec. Higher theoretical values oxidation; (bi oxide inclusions and the absence of carbon in the oxidized molten metal; and (c) the iron-carbon transformation microstructure near the liquid metal-solid metal are obtained for the steel slags — interface. Reduced 32% WELDING RESEARCH SUPPLEMENT! 501-s 115 cm/sec for MgF2 and 90 electroslag deposited metal is, there completely denuded of nickel — cm/sec for BaF2. There is thus rapid fore, less under control. Table 4. However, with 25 wt% NiO transport of reacting constituents in Reaction at the Electrode Tip. in the flux solidified droplets show the slag. The active constituent in the slag is that a high level of Ni transfer was Similarly, although it is not pos swept, by the Lorentz force-induced still taking place at the end of the sible to estimate its magnitude, there motion, towards the reacting surface melt despite a low concentration in will be fluid motion within the of the molten electrode tip. Within the flux. The droplet in Fig. 5a molten droplet due to the broadening the metal drop the Lorentz force is shows the existence of a nickel rich out of the current path from where it also acting to produce fluid motion surface boundary layer of about 2 x enters the slag through the con that distributes the reaction product 10-3 cm thickness. This is consistent stricted area of the solid electrode at throughout the bulk of the drop. The with a high velocity slag motion that the slag surface. At the ingot, current evidence from the present work, gives rise to a thin boundary layer on will enter the surface at something together with established knowledge the slag side of the reaction inter approaching uniform current density. of the type of motion produced by the face, and a lower degree of fluid mo Taking aluminum melts made at 150 Lorentz force, suggests that at the tion within the metal droplet, and amp this would give a current slag-metal interface the two motions therefore a thicker boundary layer on density of 30 amps/cm2, compared will oppose each other as in Fig. 19. the metal side of the interface. Diffu with 2,000 amp/cm2 at the elec As the slag approaches close to the sion across the metal boundary layer trode. This low current density can interface, the velocity goes through a then becomes the rate controlling not give rise to any significant transition from that in the bulk of the process. degree of motion so that there will slag to that of the interface. The slag be only comparatively slow transport and metal at the interface thus move Since the transfer of Ni into Fe is a of constituents in the ingot. together (or would be stationary if thermodynamically highly favorable the two motions exactly balanced). reaction there will be a virtually 100% A similar intense reaction at the The slag motion does not, therefore, concentration of nickel at the metal electrode tip occurs between the gas carry the reacting constituent right surface (see II under Appendix). phase and the molten metal in arc up to the interface. There is then a high concentration welding, although the reaction tem gradient in the metal boundary layer perature is considerably higher. But At the interface itself the reaction causing diffusion of nickel into the whereas there is also considerable denudes the slag of reacting ma bulk of the droplet where it is taken reaction with the deposited metal in terial, and the supply of further react- up by bulk transport processes and arc welding, there is much less re ant is then determined by diffusion distributed comparatively uniformly action in the electroslag system. This through the boundary layer from the throughout the drop. A calculation of is because, in arc welding, there is high concentration in the rapidly re the concentration of Ni built up in firstly a high reaction temperature plenished bulk of the slag. The react the drop, due to diffusion through the where the arc impinges on the ing constituent in the metal is like boundary layer during the period of surface of the weld pool; second, wise transported towards the inter growth of the droplet on the end of there are rapid, Lorentz force- face by the bulk, Lorentz force- the electrode, gives a result of 10 to induced, transport processes within induced motion and then has to 20 wt% nickel (see Appendix III) the the weld pool; and, third, there is a pass, by diffusion, through a range being primarily due to the diffi high degree of super-heat of the boundary layer to arrive at the inter culty in assigning values to the diffu molten metal. face. The products of reaction are dis sion coefficient. This compares with persed by the reverse processes of the experimentally determined con All of these are absent in diffusion away from the reaction centration of nickel in the droplet of electroslag deposition. Thus elec interface through the boundary about 15 wt%. troslag welds tend to be much more layers into the high velocity regions free from compositional and asso In addition to the diffusion of where there is rapid dispersal into ciated defects than are arc welds. nickel through the periphery of the the bulk material. It is the rate of dif Where reactions do occur, how drop there is also an important con fusion of reactants to the interface, ever, they do so relatively rapidly so tribution from the reaction taking or of reaction products away from it, that in the electroslag system any place by a similar diffusion process that determines the rate of reaction. reacting constituents in the flux are through the layer of molten metal soon used up and are only replaced A convenient system to analyze in formed on the tapering electrode on 'topping up'. With arc welding, on more detail is that of the transfer of above the drop — Fig. 19. Here there the other hand, there is a continuous nickel, from NiO containing fluxes, is a very high surface area-to-volume supply of fresh flux. From this point into iron. Clearly the rate of transfer ratio and thus a very high concentra of view the chemical composition of is high so that the flux is virtually tion of nickel as shown by analyses Table 11 — Carbon Oxidation by Oxide Fluxes Flux composition, Operating conditions, Ingot composition, wt% oxide amp wt% C wt% X 25 CuO-BaF2 45 300 .031 29.3 Cu 25 NiO-BaF2 55 400 02 35.8 Ni 25 Cr203-MgF2 60 250 024 2.59 Cr 25 Mn02-BaF2 50 400 .05 0.46 Mn 35 Si02-BaF2 50 400 07 0.70 Si 30 Ti02-BaF2 60 200 .12 0.01 Ti 20 MgO-MgF2 55 350 .13-1 6 — 30 AI203-BaF 70 200 .13 0.02 Al 50 BaO-MgF2 25 400 12 — 25 CaO-BaF2 65 400 .16 — Fig. 18 — FeC droplet disintegrating in BaF, 55 400 19 — an iron oxide slag due to CO evolution. X400 (reduced 45%) 502-s | OCTOBER 1972 Diffusion 100 Boundary layer thicknes Nickel Slag motion ag motion Silicon and Mang Distance into metal Fig. 20 — The effect of the thermodynamic driving force on Fig. 19 — Schematic diagram of the physical conditions at the elec the concentration gradient in the metal phase boundary trode tip during slag-metal reactions layer for slag-to-metal transfer of the material around the neck of would again appear to be compatible neck of the drop so that the copper the drop of around 60 wt% Ni. All the with a ratio of about 25% of the equi concentrations of the droplet and the molten metal is transferred across librium silicon level at the metal ingot are almost the same. into the ingot where it mixes to give surface. The reduction of carbon in iron by a mean composition of around 22- Where a gaseous reaction product oxides in the flux, involving as it 24 wt%. At lower levels of NiO in the is formed, as with the reaction be does a gaseous reaction product, flux the same processes of nickel tween NiCl2 in the flux and iron, or invites comparison with the CuCI2- transfer will be rate controlling until with the absorption of copper into Al system. But there is an essential the concentration of Ni in the slag is aluminum from a slag containing difference between the two cases. so reduced that the rate of diffusion CuCI2, the bubbles forming at the With the AI-CuCI2 system the gase across the slag boundary layer is interface interfere with the diffusion ous reaction product is formed at the less than the rate of diffusion across of reactant through the slag bound slag-metal interface. In the FeO-C the metal boundary layer. ary layer. A typical aluminum ingot system on the other hand, oxygen is The mechanism of transfer of man would absorb 5 wt% copper. For a first absorbed into the iron and this ganese and silicon from the slag into drop of 1 gm with a diameter of 0.35 circulates and comes into contact the iron should be the same as the cm forming on the end of the elec with the carbon within the body of transfer of nickel, with the same trode this would give rise to the the drop giving rise to gas formation 3 boundary layer characteristics, except evolution of 55 cm of aluminum and fragmentation of the drop. This that the concentration gradient is chloride gas, forming bubbles and further increases the surface area much lower. A driving force provided dispersing at the interface during the and the rate of reaction, so that the by a concentration of 100% nickel at period of existence of the droplet. carbon is rapidly removed. Only com the slag-metal interface resulted in a In applying the analysis of diffu paratively small quantities of FeO in concentration of nickel in the ingot sion across a boundary layer (see III the slag are required to remove most of about 25% of this level. Thermody under Appendix) to diffusion across of the carbon because of the high namic calculations (see II under activity of FeO at low concentrations the slag layer, very approximate cal 16 Appendix) show that for a slag of culations, limited because of inade in fluoride slags. pure manganese oxide the equilib quate knowledge of diffusion coeffi The removal of silicon from iron rium concentration of manganese cients, suggest an "effective thick with slags containing FeO occurred at the metal surface would be about ness" of the slag boundary layer two to a much greater extent that the re 2.2 wt%; there is thus a much lower to three times that of the metal verse process of absorption of silicon driving force (Fig. 20), but the ingot boundary layer. Thus for these sys from the slag. This is because the composition was about the same pro tems diffusion through the slag situation is similar to the removal of portion of the concentration at the boundary layer becomes rate con carbon in that oxygen is taken into metal surface, i.e., 30%, or 0.7 wt%. trolling and there is no evidence of a the body of the metal droplet and For the flux containing 20 wt% man copper rich boundary layer in the oxidation of silicon occurs through ganese oxide the ratio of manganese metallographs of the aluminum drop out the bulk of the droplet as well as in the ingot to that at the metal in Fig. 9. at the slag metal interface. The effec surface works out at about 27%. Another feature of the aluminum tiveness of small amounts of FeO in The data for silicon are more diffi system is that melting takes place al silicon removal is again due to the cult to interpret since the silica and most perpendicular to the electrode, enhanced activity of low concentra iron oxide form a complex silicate — i.e., the electrode does not become tions of FeO in fluoride slags. Fig. 13. This makes it difficult to assign tapered. There is thus little active When an insoluble reaction prod values to the crucial parameters of molten layer on the melting elec uct is formed, as with the Cu20-Al activity of the silicon and iron in the trode where intense reaction occurs system, some of the reaction slag. However, the experimentally and only a small amount of high interface is blocked off. This limits observed silicon levels in the ingot copper concentration material at the the rate of reaction. The marginal in- WELDING RESEARCH SUPPLEMENT! 503-s crease in the ingot Cu levels with This occurs with the oxidation by 14. Maecker, H., Z. Physik, vo. 141, p, large increases in Cu content of the FeO in the slag of silicon and carbon 198, 1955. slags indicates that a relatively con in iron, whereby oxygen enters the 15. Salter, G. R., and Milner, D. R., stant reduction of area is achieved iron and reacts with the silicon and "Gas Absorption from Arc Atmospheres," and that the reaction proceeds only carbon throughout the molten drop British Welding Journal, 1 (2), 89 to 100, 1960. if the oxide film is then removed or let. 16. Davies, M* W„ "The Chemistry of disturbed. Such a disruption will 6. Complete equilibrium is not CaF2-based Slags," Internationa! Sym occur every time a droplet is re achieved in slag-metal reactions. For posium of Metallurgical Chemistry, Shef moved from the electrode tip; it will reactions occurring at the interface field, 1971. probably also occur when irregular under the conditions investigated in 17. Basic Open Hearth Steelmaking, ities caused by submerged arcing, the present work, about 30% of the 3rd Ed., Chap. 1 6, AIME, 1 964. itself brought about by the presence equilibrium level concentration of 18. Edwards, J. B., et al, Metals and of the oxide film, take place. reaction product was attained in the Materials, 2(1), p. 1, 1968. deposited metal provided there was no prior depletion of the supply of re Appendix actant from the slag. For reactions Conclusions occurring throughout the bulk of the I. Slag Velocity Near the 1. Slag-metal reactions take place molten droplets equilibrium was Electrode Tip primarily at the molten electrode tip much more closely approached. The velocity of flow in a plasma due to the high surface-to-volume jet created by a constriction in the ratio of the molten metal and the in current path is:13 tense stirring forces. Reactions con tinue in small droplets contained in Acknowledgements The authors wish to thank Professor 2|2 the slag and created by the metal V E.C. Rollason, Head of the Department of ma P rr a2 transfer process. Reactions at the Industrial Metallurgy at the University of t ingot surface are limited because of Birmingham, for his interest and encour where I = absolute amperes, the low surface-to-volume ratio and agement, and the Ministry of Defence P = fluid density, and a = radius much weaker stirring forces. (Navy Dept.) for financial sponsorship of of constriction. 2. Slag motion is caused by the this project. We also wish to thank Dr. P. J. In this case, the current is 150 Lorentz force as a result of the drop Emerson and Mr. M. Burke of B.C.I.R.A. amp (1 5 amp absolute), the fluid den in current density between the elec for their generous assistance on the elec tron microprobe work. sity is that of the NaF-KCI eutectic trode and the conductive slag bath, (approx. 2 gm/cc), and the radius of and the Lorentz force also causes the constriction is the radius of the circulative mixing within the liquid References electrode (1.6 mm). Thus: metal formed on the electrode tip. 1. Fitch, I. C, "Whither Electrode These stirring forces act to distribute Design?", British Welding Journal, 1 (10), reaction products from the slag- 600 to 604, 1 960. /2(1 61(15) Vma - 53 cm/sec metal interface into the bulk of the 2. Babcock, D. E., "The Fundamental T 2ir (0.1 6)2 slag and metal phases. Nature of Welding, Part V — The Physical Chemistry of the Arc-Welding Process," In the MgF and BaF slags, the 3. The Lorentz force induced Welding Journal, 20 (4), Research Suppl.. 2 2 motion carries the reactants in the 189-s to 197-s, 1941. current is 400 amp (40 amp abso slag and in the metal towards the 3. Bennett, A. P., and Stanley, P. J., lute) and the densities approximately slag-metal interface. At the interface "Fluxes for submerged-arc welding of 3 gm/cc and 4.5 gm/cc respectively. itself the motion is the same for both Q.T.35 steel," British Welding Journal. This gives velocities of 1 1 5 cm/sec slag and metal, so that the final 13(2), 59 to 66, 1966. in MgF, and 90 cm/sec in BaF2 . transport to the interface is by diffu 4. Belton, G. R., Moore, T. J., and Tan- sion through boundary layers in the kins, E. S., "Slag-Metal Reactions in Sub merged Arc Welding," Welding Journal. II. Calculations of the Equilibrium slag and metal. For most reactions 42 (7), Research Suppl., 289-s to 297-s, State at the Slag Metal Interface diffusion across the metal boundary 1963. A number of the reactions which layer is rate controlling. 5. Claussen, G. E., "The Metallurgy of were investigated with iron elec 4. The concentration, or activity, the Covered Electrode Weld," Ibid., 28 trodes are evaluated thermody- gradient determining the extent of (1), 12 to 24, 1949. namically to illustrate the principles diffusion across the boundary layer 6. Chipman, J., and Christensen, N., involved. The temperature at the re is a function of the thermodynamics "Slag-Metal Interaction in Arc Welding," action interface is assumed to ap of the reaction. For a thermodynam- Welding Research Council Bulletin No. proximate to the slag temperature, ically highly favorable reaction there 15, 1953. i.e., 1900 C and thermodynamic is a high concentration of the reac 7. Bagryanskii, K. V., and Lavrik, P. F., Welding Production, 1 1 (5), p. 72, 1964. data17 have been extrapolated from tion product at the boundary layer 8. Baryshnikov, A. P., and Petrov, G. L., steelmaking temperatures to this and hence a high rate of diffusion. Welding Production. 1 5 (10), p. 9, 1 968. temperature. Conversely, a low equilibrium level 9. Hazlett, T. H., "Coating Ingredients at the boundary layer gives rise to a Influence on Surface Tension, Arc Sta Pure Fe reaction with NiO slags. low concentration gradient and a bility and Bead Shape," Welding Journal, This is an example of a high rate of low rate of transfer. 36(1),ResearchSuppl., 18-sto22-s, 1957. reaction with liquid reaction prod ucts. Nickel is transferred to the iron 5. The rate of reaction is reduced 10. Duckworth, W. E., and Hoyle, G., when the reaction product formed at Electroslag Refining. Chap. 3, Chapman according to: the interface is gaseous because it & Hall, London, 1969. (NiO) + Fe^ %Nj_ + (FeO) interferes with the transport of ma 11. Crimes, P. B., "Rates of Mass To obtain the free energy of forma terial to the interface and, when a Transfer in Electro-slag Remelting—Part tion: solid reaction product is formed, it 1: Model Studies," BISRA Report Mg/A/72/64, 1964. blocks off areas of the interface from 12. Woods, R. A., and Milner, D. R., (NiO) t=i %Ni + %0; reaction. The rate of reaction is par "Motion in the Weld Pool in Arc Weld AG = +27,350 - 35.72T ticularly high where the reactant in ing," Welding Journal. 50 (4), Research the slag is soluble in the metal, and Suppl., 163-s to 173-s, 1972. it is transported into, and reacts with 13. Paton, B E., Electroslag Welding. Fe + %0 — (FeO); elements in, the bulk of the metal. Chap. 3, AWS, 1962. AG = +28,914 - 12.51T 504-s | OCTOBER 1 972 Adding ac ~ 1 where R is the rate of melting of the Fe electrode. (NiO) + Fe — %Ni+ (FeO); a C is the concentration at the sothat%Si = 1.1 x iQ-2. Si°2 0 AG = +56,264 - 48.23T surface, taken as the equilibrium con KeO>2 centration. C, is the concentration For the equilibrium constant K: within the drop — this is taken as -AG -12,300 The difficulty here is that a sio2 constant as the calculation shows it log K = = + 10.55 and a Fe0 are known not to behave to vary from 0 to about 0.1 Cb. Thus: 4.575T T ideally, but the values of a sio2 and a Fe0 are not known. 1 /3 D (C0- C.) (47T) (3Rr T = 1900 C = 2173 K It is clear from the value of the con Q log K = 10.55 - 5.67 = 4.88 stant that the concentration of Si at the metal surface will be more akin %Ni (aFe0) K = 7.6 x 104 = = to that of Mn than Ni. The experi mental fact that Si was absorbed 'a\ into iron already containing 0.9 wt% aFe can be considered as approxi Si, but at a comparatively slow rate, J: mately equal to 1. Therefore: suggests an equilibrium Si level of around 1.5 - 2 wt%. % Ni = 7.5 x 104 fjw Integrating and substituting the 3FeO Fe-C Reaction with FeO Slags. final drop diameter, r, for the time of This reaction occurs throughout the drop formation, t, according to aFe0 cannot be greater than 1, so bulk of the molten drop which has a 4 7rr3 that even for small values of aNi0 the temperature of the order of 1600 - t = surface layer of the metal will be 1 700 C, say 1 650 C. The reaction is: 3R pure Ni until a Ni0 becomes too low 2 for the rate of diffusion through the (FeO) + %C ^ Fe + [CO]; 755 Dr boundary layer in the flux to sustain (Ce-C,) -.. (3.1! AG = +23,600 -22.OT Rd equilibrium. At = 1650 C: Pure Fe Reaction with MnO x For the transfer of nickel into iron Slags. This is an example of a low (a, ( Pc considered in the Discussion R = rate of reaction with liquid reaction 2 1.35 x 10 = 0.254 cc/sec, d = 210"3 cms., r2 = products. At elevated temperatures 2 (a Fe0 ) (%C_) 0.063 cm , (C - CJ =i C = 1Q0 the reaction is: G 0 wt%. The difficulty is to assign a val (MnO) + Fe ±^ %Mn + (FeO); ue to D. From data provided by Ed For virtually pure Fe and pure FeO, wards18 D is probably 1 to 210 "4 AG = +29,500 - 13.47T 2 Fp = a F„n =1 and for CO evolved cm /sec. From this: at 1 atm pressure Pco 1 4 Cf = 9.5for D =1.10" At 1 900 C and withaFe^ 1, this gives 4 %C 0.007 to 19 for D = 2.10" a MnO K K = 1.0 = a „ (%Mn) 'FeF On This is substantially lower than the experimentally determined range of a FeO %Mn = 0.01 -0.05%C. List of Symbols 3 MnO ( ) liquid component in slag A slag containing only manganese III. Kinetic Considerations phase and iron oxides is known to behave The final concentration, Cf, of [ ] gaseous component in ideally, so that the activity of each nickel in a drop as a result of slag phase oxide is equal to its mol. fraction. diffusion through a boundary layer of % X alloy component in The experiment with pure Mn02 flux thickness d in the metal surface is metal phase gave a slag containing 51.2 wt% Mn < > solid component in slag and 21.4 wt% Fe, if 2 wt% Fe is phase allowed for metal droplets in the f V slag. Calculations based on these fig where Q is the amount of nickel that ures give an equilibrium %Mn at the has diffused into the drop and V is metal surface of 2.2 wt%. If the mu the final volume of the drop. Q is tual behavior of the oxides is as given by integrating the diffusion Conversion Factors sumed tp remain ideal in the BaF2 through the surface boundary layer flux, then the 20 wt% MnO - BaF2 of the drop over the period of its de C flux gives an equilibrium %Mn = 1.6 velopment on the end of the elec =-f-(F-32) wt%. trode: giving 1100 C Si 2010 F Pure Fe Reaction with SiO 2 and 1900 C =s 3450 F Slags. For this reaction: 0 C = 273K OA (C - Ci) dt 1 m/hr = 0.66 ipm (Si0 ) + 2Fe t=i %Si + 2(FeO); 0 2 cm/sec = 23.6 ipm d 1 AG = +81,500 - 28.5T "[' 1 cmVsec = 3.9ft2/hr At 1900 Cr 1 cm =0.395 in. 1 cm2 = 0.156 in.2 2 where the surface area is a function 3 3 %^*(aFcn) of time given by: 1 cm = 0.062 in. K = 1.1 x 10" 1gm/cm3 = 3.6 x 10'2 lb/in.3 a a 2 sio, < Fe) A = (4TT)11/ /J3 (3R)/OD\2 i//J3 tt 2/3 WELDING RESEARCH SUPPLEMENT! 505-s