Slag-Metal Reactions During Welding: Part II. Theory

Slag-Metal Reactions during Welding: Part II. Theory

U. MITRA and T.W. EAGAR

A kinetic model is developed to describe the transfer of alloying elements between the slag and the metal during flux-shielded welding. The model accounts for changes in alloy recovery based on the geometry of the resulting weld bead. It also distinguishes compositional differences between single-pass and multiple-pass weld beads. It is further shown that the final weld metal oxygen content is directly related to the weld solidification time as well as the type of flux used.

I. INTRODUCTION also indicates that oxygen is transferred in this region. The results presented in Table VIII of Part IL11indicate INPart I of this series,[ll the previous theories of slag- that in multiple-pass welds, the top weld contains more metal reactions during flux-shielded welding were re- oxygen than the bottom layer. This is consistent with a viewed. Experiments demonstrated that the widely held mechanism of oxygen transfer in the droplet zone. droplet reaction theory cannot explain the transfer of al- Table VIII of Part IL1l indicates that for some welds, loying elements between the slag and the metal. In this although the weld metal gains oxygen, it loses silicon paper, a new theory is presented to explain these chem- and manganese. If, in addition to manganese and silicon ical interactions. In Part 111,[411the theory is tested using transfer, the oxidation of iron is also considered, an ox- data from submerged arc welding (SAW). ygen balance indicates that the final amount of oxygen It is proposed that chemical interaction between the transferred to the weld metal during the slag-metal re- slag and the metal occurs in three zones, as indicated in actions is much lower than the amount of oxygen trans- Figure 1: ferred at an intermediate stage of the reaction. In fact, (1) the zone of droplet reactions, in many cases, an oxygen balance based solely on slag- (2) the zone of dilution and weld pool reactions, and metal reactions indicates that the weld metal should have (3) the zone of cooling and solidifying weld pool. a lower oxygen content than the electrode/baseplate used, which is contrary to experimental observation made dur- 11. ZONE OF DROPLET REACTIONS ing the course of this study as well as that reported in the literat~re.['~-~~-~~] In this region, the droplet forms at the electrode tip The oxygen present in the arc plasma which is re- and then travels through the arc column, as shown sche- sponsible for the plasma-metal reactions in this zone has matically in Figure 1. The entire process occurs in a few two possible sources: millisecond^,[^-^-^^ and the temperature of the droplets is (1) decomposition of flux constituents into suboxides and very high: in the range of 2000 OC to 2500 0C.151Due to oxygen and the high temperatures, it is thermodynamically possible (2) contamination from the atmosphere. for several chemical reactions to occur. However, results Both sources have been considered earlier by Eagar,[6'121 of some preliminary experiments presented earliefill show Chai and Eagar,L71 and more recently by L~U.[~]The de- that there is a negligible amount of alloy transfer composition of flux constituents into suboxides and ox- (Si, Mn, Cr) in this region. ygen seems to be the primary source of oxygen, since Although the alloying elements Si, Mn, and Cr are not different fluxes produce different oxygen levels in the transferred in this zone, the results of our investigation, weld metal, depending on the stability of the flux con- as well as data from several other researchers, indi- stituent~.[~-~,~~-~~] cate that oxygen is transferred to the metal in this The analysis of Chai and Eagar on binary oxide- zone.[4,6-81The strongest evidence comes from the results calcium fluoride shows that even oxides stable of Lau,f81who determined the oxygen content in the elec- under steelmaking temperatures (such as MgO) may de- trode tips, in the droplets after their flight through the arc column, and in the weld pool. He also found that compose to gaseous suboxides or vapors and oxygen in the arc plasma and lead to the transfer of considerable changing the welding parameters did not significantly in- amounts of oxygen into the weld metal. Contamination fluence the oxygen content in the droplets. The obser- by oxygen from the atmosphere plays a much smaller vation of pores and inclusions in electrode tips and droplets r0le[~-*3~*9'~1but cannot be totally neglected as a source by other researcher~[~,~]as well as in the present work of oxygen. The fact that oxygen is transferred into the droplets in this zone, whereas there is little exchange of the alloying U. MITRA, Project Leader and Senior Member, Research Staff, is elements, may not be very surprising, if the analysis by with the Thin Film Materials Department, Philips Laboratories, North Richardson[I6] on the decarburization of levitated iron American Philips Corporation, Briarcliff Manor, NY 10510. T.W. droplets is examined. Richardson showed that in the early EAGAR, Richard P. Simmons Professor of Metallurgy, Leaders for anufacturing Professor of Materials Engineering, is with the stages of decarburization of levitated iron droplets, the department of Materials Science and Engineering, Massachusetts reaction Institute of Technology, Cambridge, MA 02139. Manuscript submitted September 18, 1989.

METALLURGICAL TRANSACTIONS B VOLUME 22B. FEBRUARY 1991-73 rent, the volume of the weld metal is increased or the Coolln and Electrode ratio of the width to cross-sectional area is decreased. Solidifying Weld Pool These changes in weld geometry affect the kinetics of metal transfer. Based on these observations, a quantitative model was formulated to predict the amount alloying elements in the weld metal for any combination of welding consumables and process parameters.

A. The Kinetic Model The model considers the slag and the metal to be two immiscible stirred liquids with an alloying element M being transferred at the slag-metal interface. Then, for an interface reaction, such as Fig. 1 -The three reaction zones which control the chemical composition of the weld metal during SAW. to proceed, three events have to take place:[l9] (1) reaction species (the relevant ions in the oxide MOx) have to move between the bulk slag and the slag-metal interface; does not proceed forward (the superscripts b and s indicate bulk and surface concentration, respectively). The (2) chemical Reaction [2] has to occur at the slag-metal oxygen rapidly builds up at the surface, and this surface- interface; and active oxygen prevents the carbon from reaching the (3) the alloying element M has to move from the slag- interface and reacting. A similar phenomenon may be metal interface to the bulk metal. occurring inside the arc cavity during the process of SAW, The kinetics of slag-metal interactions may therefore be with the surface-active oxygen keeping out the other ele- controlled or affected by any of these three steps. The ments during the few milliseconds in which the drops three steps may be represented schematically by an form at the electrode tip and fall through the arc cavity. activity-distance diagram (Figure 3).

B . Assumptions 111. ZONE OF DILUTION The model assumes that: AND WELD POOL REACTIONS ^"-^' (1) A neutral point (NP) exists for each welding flux. Tt. In this zone, the falling droplets become "diluted" with slag and metal are at an effective equilibrium only when molten metal from the baseplate (Figure 1). The high thenominal composition of the weld (i.e., the total com- temperature and the large convective forces in this re- position due to the simple mixing of metal from the elec- gion lead to intimate mixing of the molten metal and trode and workpiece in the absence of chemical reactions) result in vigorous chemical reactions at the slag-metal is the same as the NP; that is, no transfer of the alloying interface near the arc. The results presented earlier in element takes place at the NP. Furthermore, the NP is Part I['] indicate that slag-metal reactions do occur in this not affected by variations in the process parameters for region. fluxes free of ferroalloys or other elemental additions. Previous researchers had found that increases in volt- This assumption is based on the results of Chai and age (DC and AC) or decreases in current result in an Eagar[I41and Chai[201and the experimental data of Thier.['ll increase in the amount of alloying elements transferred (2) The equation of continuity is valid for transfer of the between the slag and the metal and had suggested dif- alloying element. That is, the mass flux (J) of alloying ferent mechanisms of element transfer to explain this element M which flows from the bulk slag to the slag- phenomen~n.[~.~*~-~~-~~]The results of our earlier experi- metal interface is equal to the mass flux of M passing ment~,[~]when the welding conditions were varied, agree through the chemical reaction stage and is also equal to with the experimental observations of these researchers. that which is passing from the interface to the bulk metal; However, since the electrodes were virtually free of Mn, however, the amount of this mass flux (J)can and will Si, and Cr, the influence of voltage and current on the change with time. transfer of these elements is clearly due to the influence (3) The mass transfer coefficients of the alloying ele- of voltage and current on the kinetics of slag-metal re- ment M in the slag (k,) and the metal (km)are indepen- actions in the weld pool. Figures 2(a) through (c), which are plots of the amount of alloying element transferred dent of one another and of the activities of the reacting between the slag and the metal against the ratio of the species. width to transverse cross-sectional area of the weld metal, (4) The initial oxygen concentration in the weld pool qualitatively explain the effect of voltage and current on (and not the final weld metal oxygen content) is depen- the kinetics of slag-metal reactions. As voltage is in- dent only on flux composition. This assumption may seem creased, the arc broadens and the area of the slag-metal surprising considering that the transfer of the alloyin interface is increased or the ratio of the width to cross- element M also involves the simultaneous transfer of 01,^ sectional area is increased, whereas by increasing cur- ygen by Reaction [2]. However, in the weld pool,

74 -VOLUME 22B, FEBRUARY 199 1 METALLURGICAL TRANSACTIONS B ' Ã‘Ã

Fig. 2-(a) Effect of the ratio of weld width to transverse cross-sectional area (w/a)on the loss of chromium from the metal to the slag. Flux Fx-1, stainless baseplate, and A-7 electrodes used for all welds. (b)Effect of the parameter (w/a)on the loss of silicon from the metal to the slag. Flux Fx-2, HiSi baseplate and A-7 electrode used for all welds. (c) Effect of the parameter (w/a)on the transfer of manganese from the slag to the metal. Flux Fx-1, 1008 baseplate, and A-7 electrodes were used. transfer of several elements takes place simultaneously. generally proceeds only forward. Also, as explained ear- The reactions lier, the oxygen content of the weld metal is controlled by plasma-metal reactions involving the liquid-metal -Si + 20 = (SOz) Pal droplets. Thus, in the region of weld pool reactions, al- and though some metal oxygen transfer may occur from the weld metal to the slag, more oxygen is introduced in the -Mn + 0 = (MnO) [2bl form of falling droplets and possibly through the weld pool surface directly under the arc. The net result is that may either proceed forward or backward, but the oxi- the weld pool oxygen level remains almost constant. The tion of iron existence of a singular NP, as demonstrated by Chai and Fe + Q = (FeO) [2cl Eagar[l4]and which also may be observed from Thier's

METALLURGICAL TRANSACTIONS B VOLUME 22B, FEBRUARY 1991-75 where the subscript eq denotes the equilibrium composition. Neglecting the effect of interaction terms on the activity of the metal,[19,221this may be written

Since the flux and metal are at equilibrium only at the NP ([pet MIeq = NP),DOlthen

where m is the partition coefficient of the alloying element M between the metal and the slag. From assumptions (4) and (3,it follows that m is a constant. Even if assumption (5) is not strictly correct, the variation in D i stance m will still be small, since a decrease in temperature increases the value of K but decreases the value of [ao]. activity of oxide in bulk slag a: Let dq be the amount of alloying element M trans- a : activity of oxide at slag-metal interface a1f. ferred between the metal and the slag at any instant. Then %M : activity of element in bulk metal (neglecting mb interaction terms)

%M . : activity of element at slag-metal interface If (neglecting interaction terms) where J = the mass flux of alloying element M (wt pet Fig. 3-Activity-distance diagram for oxide (MO,) in the slag and alloying element M in the metal at the slag-metal interface. cm/s); Vm= the volume of the weld metal (cm3); As/m= the area of the slag-metal interface (cm2);n strongly supports the current assumption. dM = the differential change in the concentrati~ Nonetheless, it should be noted that unlike the initial weld of the alloying element M in the metal (in pool oxygen content, the final amount of oxygen in the weight percent); and weld metal does depend on welding process parameters dt = the differential change in time (s). and electrode/baseplate composition, in addition to flux Equation [7] may also be written as composition. This is due to a mechanism of inclusion formation, growth, and separation, which will be discussed subsequently. (5) The temperature range over which weld pool reactions occur near the arc is small. The fact that the Since the equation of continuity is valid (assump- arc-metal interface is maintained at temperatures up to a tion (2)), the mass flux J is also given by the following maximum of 2500 0C151provides some justification for equation~:[~~,~~1 this assumption. The validity of this assumption will be justified later by considering the data of other researchers. (6) The chemical reaction rate does not influence the kinetics of metal transfer. The high temperatures of the weld pool justify this assumption, since at high temper- where k, and km = the mass transfer coefficients in the atures mass transfer rather than chemical reaction con- slag and the metal, respectively; trols the kinetics.[1y1However, it is shown later that the aSband a,$ = the activities of the chemical oxide model would still be valid even if chemical reactions were MOx in the bulk slag and at the slag- rate-controlling, as long as the reactions were first-order metal interface, respectively, at any reactions. instant; and Mb and My = the concentrations of the element M C . Derivation in weight percent in the bulk metal and at the metal-slag interface, re- If Reaction [2] is at equilibrium, then the equilibrium spectively. Since interaction terms are constant may be written being neglected, Mb and Myalso con respond to the activities of the el ment M in the bulk metal and the metal-slag interface.

76- VOLUME 22B, FEBRUARY 1991 METALLURGICAL TRANSACTIONS B Also, since chemical reactions are not rate-controlling, Substituting Eq. [8] for J, we get the slag and the metal are always at equilibrium at the slag-metal interface so

During the course of the welding process, both the metal Solving this differential equation from t = 0 to t = t and and the slag compositions change. The activity of the using the initial condition at t = 0, Mb = Mi (nominal weld composition) gives a solution of the form oxide in the bulk slag is therefore not exactly the same as that in the flux. However, this change is small NP + JSIMiI - [1 + PI [Mfl due to mass balance considerations. For example, a change in metal composition from 0.1 pet Mn and 0.1 pet = [NP - Mil .exp [(-A+/VJ (1 + P) (a)] [I91 Si to 1 pet Mn and 0.5 pet Si will change a 30 pet where MnO-30 pet Si02 flux composition to approximately 29 pet MnO-29 pet SO2, or a change in metal composition from 1.0 pet Mn and 0.1 pet Si to 0.5 pet Mn and 0.5 pet Si will change a slag composition of 0.1 pet MnO-50 pet Si02to about 1 pet MnO-49 pet Si02. Thus, and Mf is the final weld metal composition. the slag can be considered to have a greater "capacity" In the above derivation, it was assumed that the chem- as a source or sink of alloying elements than can the ical reaction rate does not influence the kinetic mecha- metal. nism. It will now be shown that even if the chemical reaction step is the rate-controlling step, Eq. [19] will still be valid as long as the reaction is first order. Let kl and ki be the forward and backward reaction rates for the chemical reaction represented by Eq. [2]. where y = the activity coefficient of the oxide (MOX)in Then, the mass flux J across the slag-metal interface is the flux; given by q = a mass balance factor 7 = (mol wt of MOx)/ (at. wt of M) (wt of metal)/(wt of slag); J = ktasif- k2Mf [211 Mi= the initial or nominal concentration of alloy- Also, at equilibrium (and only at equilibrium) ing element M in weight percent; and kidsif = k2Mif 1221 which, combining with Eq. [6], gives where Me, and Map are the alloy content of the electrode and baseplate, respectively, and d is the dilution factor. Combining Eq. [6] with Eq. [I I], we get Combining Eqs. [2I] and [22a],

Combining Eqs. [13] and [9], we get Adding Eq. [2la] to Eqs. [lo] and [I61 and substituting Eq. [8] for J, we get a modified form of Eq. [18]: or letting a factor

Solving this differential equation, as indicated earlier, still produces the original form of Eq. [19], with only a being redefined as Equation [lo] may be written as

that is, with the extra 1/k2 term as compared with Eq. [20]. Adding Eqs. [lo] and [16] and noting that My = masy It should be noted from Eq. [20a] that the value of a from Eq. [IOa], we get depends not only on the mass transfer coefficients, km and ks, and the chemical reaction rate constant, k2, but also on the partition coefficient, m. The value of m is given by Eq. [6], where m depends on the value of the equilibrium constant K of the reaction being considered,

METALLURGICAL TRANSACTIONS B VOLUME 22B, FEBRUARY 1991-77 as well as on the initial oxygen content of the pool. As Equation [19b] may also be written as oxygen content in the weld pool increases the value of m, the partition coefficient decreases. Thus, the value of m for highly oxidizing fluxes, such as the acid silicate fluxes, should be lower than for the less oxidizing fluxes, such as the basic silicate or the fluoride-alumina-base where AM is the amount of alloying element transferrt- fluxes. Studies in iron and ~teelmaking[~~,~~]indicate that from the slag to the metal or vice versa. Since the factor often the rate-controlling step in slag-metal reactions is [l - exp (- aAslm/V,,,)]is always positive, the sign of the diffusion of the alloying element in the slag boundary AM will depend on the sign of (NP - Mi). That is, layer at the slag-metal interface. From Eq. [20a], it can whether the alloying element M is transferred from the be seen that if the transfer of an alloying element is con- slag to the metal or the reverse depends only on the trolled in the slag phase, then a significant difference thermodynamic factor (NP - Mi). Kinetic consider- should be observed in the rate at which the alloying ele- ations (welding process parameters) may change the ment is transferred between the slag and the metal when magnitude of AM, but they cannot determine its sign. using highly oxidizing fluxes as compared to less oxi- This excludes the effect of the welding process param- dizing fluxes. Furthermore, this should apply whether eters on dilution and the consequent effect on (NP - Mi). the alloying element is being transferred from the slag This is consistent with the experimental results of all re- to the metal or the reverse. Examination of data available searchers who have studied the effect of welding process in literature shows this is indeed the ~ase.1~~1 parameters on weld composition.[2-4-8-10-15~17~251In the For most welding fluxes, it can be shown that for the derivation of Eqs. [18] and [19], factors such as surface transfer of the Mn and Si[143241 renewal were not considered. However, it has been shown that the error caused by neglecting this in models where phase resistances are added (such as the one derived in this section) is low. L19-26-27]

Physically, Eq. [23] implies that there is little change D. The Electrical Analog in the composition or the chemical potential of manganese and silicon in the slag during the welding A simple way of looking at the process of element process, as discussed earlier using mass balance transfer during SAW is by using the electrical analog considerations. shown in Figure 5. The slag may be considered as a bat- Equation [19] then reduces to tery with a potential corresponding to the NP. The weld metal is equivalent to a capacitor (C,,,) charged initially (NP - Mf) = (NP - Mi) exp (-aSlm/Vm) [19a] to a potential Mi (initial nominal weld composition) whiq discharges to Mr (final composition) after welding. Ti. If we consider the approximation treatment of the slag as a component of high capacity has been described previously. A battery has a very high capacitance as compared with most electrical capacitors. The metal, on the other hand, has a relatively low ca- since as mentioned earlier the flux composition does not pacity for alloying elements. A small change in slag change significantly during the welding process, then composition does not significantly change the NP* so Eq. [19a] may be directly obtained by substituting Eq. [24] for Eq. [ll] in the derivation process. This is true if /3 = 0 in our analysis. When varies significantly If the terms in Eq. [19a] are rearranged, then an from zero, the full expression, Eq. [19], must be used, and this simple electrical analogy fails. expression is obtained for predicting weld metal - - composition: the slag is like a battery at a constant potential, but the metal is like a capacitor of variable potential. The resistance to the transfer of the element due to mass transport at the slag interface, nonequilibria of the In the right-hand side of this expression [Eq. 19b], chemical reactions at the interface, or mass transport from the first term (NP) is the neutral or equilibrium point the metal interface may be considered analogous to three reached if, and only if, the slag and metal are at electrical resistances in series. equilibrium. The second term [(NP - Mi) Comparing the solution of the kinetic model (Eq. [19]) exp (-aAs/m/V,,,)] is the deviation from the NP. The . to that of the simple RC circuit (Eq. [25]), many simi- term (NP - Mi) is the thermodynamic driving force, and larities may be observed: the term exp (-aAslm/Vm) incorporates the kinetic considerations which include the effect of variations in the (NP - Mf) = (NP - Mi) exp (-a *Aslm/Vm) [19b] welding process parameters. The welding current, volt- - age, travel speed, and groove geometry will each influ- ["battery - ^capacitor (after discharge)] ence weld geometry and consequently will have an effect - - [Â£'batter - ^capacitor (initial)]' exP (- t/RC) I251 on (ASlm/V). Figure 4 (from Reference 42) vividly il- lustrates the effect of the key welding parameters, namely, The value of A,;,,, in the kinetic model is inversely pr current, voltage, and travel speed, on weld pool geom- portional to R, since the area of the slag-metal inte& ?-> etry. Further details on quantitative estimation of (ASlm/V) increases as the resistance decreases. The value of Vm 111 are given in Part IIIL411of this series. the kinetic model is directly proportional to C, since as

78 -VOLUME 22B, FEBRUARY 1991 METALLURGICAL TRANSACTIONS B I Speed, No. mm/min 2 250 3 350 4 450 5 550 6 650

Fig. 4-Effect of welding parameters on weld pool

Transport in Metal Chemical Reaction Transport in Slag where

+J. and d is the amount of dilution. Since F < 1 and d < 1, Capacitance = Vm Initial Charge =Mi for large n, (Fd)" -+ 0. So for large n, a steady-state T After Discharge = Mi solution is obtained: ? This steady-state solution indicates that after a large Fig. 5-Electrical analog of the kinetic model. number of passes (typically six to ten), the weld composition will remain constant, providing the welding the volume of the weld pool increases, the change in the conditions are not changed. (If the welding parameters concentration (chemical potential) of the alloying ele- are changed, the values of F and d, and consequently ments decreases. Mn, will be changed.) During this steady-state condition, the extra amount of alloying element entering the E. Multipass Welding weld pool as drops from the molten electrode is exactly Equation [19c] applies only to single-pass welds. When equal to the amount of alloying element lost to the slag more than one pass is made, then the composition of the by weld pool reactions or vice versa. This steady-state subsequent passes may be calculated by repeated use of condition has been observed by different research- Eq. [19a].1241In such a calculation, the value of Mi to be er~,~~~-~~-~~~but they have incorrectly interpreted the ex- used for computing the composition of the n-th layer is istence of this condition as evidence that the transfer of calculated from Eq. [12] with the value of Mf in the alloying elements is controlled at the droplet stage. (n - 1)-th layer substituted for Mgp in Eq. [12], since Equation [28] may also be written as the (n - 1)-th layer acts like the baseplate for the n-th layer. If the welding conditions are kept constant, then the composition of the n-th layer (M^) is given by Eq. [26]. Comparison with Eq. [19c] indicates that for multipass welding (after six or more passes), the thermodynamic driving force for the amount of alloying element M transferred is (NP - A?,.,) as compared to (NP - Mi) for single-pass welds. Also, it should be noted that even when

METALLURGICAL TRANSACTIONS B VOLUME 22B. FEBRUARY 1991 -79 Mi= Md, AM from Eq. [19c] will be lower than AMN arate. Table VIII of Part I of this seriesL1]shows that from Eq. [28a] by a factor of (1 - Fd). This has been deoxidants (Mn and Si) present both in the electrode and observed in practice,[281but so far researchers could not the baseplate lower oxygen content in single-pass layers, explain this difference between AM and AMN. The greater which supports the above statement. Furthermorn value of AMN as compared to AM is not surprising con- thermodynamic considerations indicate that deoxida sidering that equilibrium is not normally reached during (Si and Mn) present in the electrode cannot prevent ox- the welding process, and in multipass welding, due to idation in the droplet stage. Rather, the deoxidants in the repetitive nature of the process, equilibrium values both the baseplate and in the electrode reduce the oxygen are more closely approached. However, the steady-state content of the metal due to inclusion formation and sep- condition Mu should not be confused with the equilib- aration during the cooling cycle of the weld pool. rium condition (NP). The value of Mv, unlike that of The time taken for a simply shaped casting to solidify NP, changes with the welding conditions, as indicated is given by Chvorinov's by Eqs. [28] and [27] and as demonstrated by the experimental results of other researchers.[2-4~17~18~21^2g1 However, under certain welding conditions (when F is where t, = solidification rate; low), the steady-state composition may rapidly approach V = volume of the casting; the equilibrium composition. S = surface area of the mold-metal interface in the casting; and IV. ZONE OF COOLING C = a constant. AND SOLIDIFYING WELD POOL A weld can be considered as a simple casting, and In this region, the molten weld pool behind the elec- Eq. [30] may be applied to it with V being the volume trode starts to cool and solidify as the electrode moves of the weld metal and S the area of contact between the away from it. Christensen and Chipman[301found that weld metal and the workpiece. (Heat loss through the there is a small drop in the manganese content near the insulating slag layer is neglected.) It should be noted that slag-metal interface in manual metal arc welds. They at- due to the presence of a traveling heat source, different tributed this interfacial phenomenon to the increasing points along the same longitudinal line (parallel to the stability of oxides at lower temperature and the conse- direction of travel of the electrode) will solidify at dif- quent shifting of the equation ferent times. However, the time for solidification (t,) for each such point will still be the same. The ratio (V/S) Q = (MnO) 1291 + in Eq. [30] may be substituted by the ratio (a/s), where to the right. Later, Christen~en[~l]and North[251observed a is the area of the transverse cross-sectional area off^^ similar manganese depletion zones in submerged arc weld and s is the length of the fusion line, as indica welds. Depletion of chromium and silicon in addition to in Figure 6, so manganese has been observed during SAW of alloy steels with fluxes containing chromium (111) oxide, but this has been attributed to interfacial reactions resulting in the As mentioned earlier, a decrease in solidification time reduction of trivalent chromium oxide to lower oxide will increase oxygen. Thus, if the ratio (s/a) is in- forms.[311This zone of alloy depletion in submerged arc creased, the oxygen content should increase. Also, as welds extends only up to a maximum depth of 0.5 mm (s/a) is increased, the average inclusion size should de- from the surface and does not affect the overall alloy crease. Both the above predictions may be observed from content of the metal.[10~2s~30~321However, interfacial re- the experimental results of Chai and Eagar.[151Later (in actions are not the only reactions which occur in this Part 1111411of this series), this hypothesis is verified by zone of cooling and solidifying weld pool. More im- directly changing the solidification time without varying portant changes take place within the molten pool itself. welding parameters and by applying the analysis pre- The increased stability of oxides at lower temperatures sented here to the data of other researchers. Before con- and the high amounts of oxygen present in the weld pool cluding this section, it should be noted again that the will result in the formation of inclusions or deoxidation products inside the molten metal.[33-361Furthermore, these inclusions may grow with time by coalescence, and the Insulating Slag larger ones are more likely to separate from the liquid \ metal into the slag.[37,38,391Even in this zone of the cooling weld pool, fairly large convective forces may be present; consequently, it is unlikely that Stokes' law is applicable. However, the convective forces would help the small inclusions to collide and form larger inclusions. If the time for the weld pool to cool and solidify increases, the inclusions will have a larger amount of r vV/ time to grow and separate out. Also, if the weld pool Cross-Section \ contains larger amounts of alloying elements, oxygen n Area of Weld '\'.-. ,A supersaturation and consequent inclusion formation will (a) Length of Fusion Line , , occur earlier, at higher temperature^,[^^-^^,^^] and again, these inclusions will have more time to grow and sep- Fig. 6-Schematic figure illustrating the parameter (S/a)

80Ã‘VOLUM 22B, FEBRUARY 1991 METALLURGICAL TRANSACTIONS B main factor controlling weld metal oxygen is flux com- 7. C.S. Chai and T.W. Eagar: Weld. J., 1982, vol. 61 (7), position, since flux composition controls the oxygen po- pp. 229s-232s. tential in the arc cavity (in the zone of droplet reactions). 8. T. Lau: Ph.D. Thesis, University of Toronto, Toronto, ON, Canada, 1983. '-(owever, the welding parameters and the alloy content 9. T.H. North, H.B. Bell, A. Nowicki, and I. Craig: Weld. J., 1978, the electrode and baseplate also significantly influ- vol. 57 (3), pp. 63s-65s. ence final weld metal oxygen content, due to a mecha- 10. G.R. Belton, T.J. Moore, and E.S. Tankins: Weld. J., 1963, nism of inclusion formation and removal in the cooling vol. 47 (7), pp. 289s-297s. 11, N. Christensen: Contract DA-91-591 EVC 3455, U.S. Dept. of weld pool. The effect of the welding parameters on weld Army, European Research Office, Trondheim, Norway, metal oxygen is particularly significant on welding with Nov. 1965. highly oxidizing fluxes. On welding with these fluxes, 12. T.W. Eagar: Weld. J., 1978, vol. 57 (3), pp. 76s-80s. a change in the process parameters may change the ox- 13. S.S. Tuliani, T. Boniszewski, and N.F. Eaton: Weld. Met. Fabr., ygen content of the weld metal by as much as 1000 ppm. 1969, VO~.37 (8), pp. 327-29. 14. C.S. Chai and T.W. Eagar: Metall. Trans. B, 1981, vol. 12B, On welding with the less oxidizing or basic fluxes, changes pp. 539-47. in the process parameters usually do not change the weld 15. C.S. Chai and T.W. Eagar: Weld. J., 1980, vol. 59 (3), metal oxygen content by more than 400 ppm.[151 pp. 93s-98s. 16. F.D. Richardson: Proc. Int. Symp. on Chem. Metall. of Iron and Steel, 1973, pp. 82-92. V. CONCLUSIONS 1.1. Frumin: iitom. Svarka, 1977, vol. 6, pp. 1-4. N.N. Potapov and K.V. Lyubavski. Svar. Proizvod., 1970, no. 7, The slag-metal reactions which occur during flux- pp. 4-5. shielded welding can be understood by dividing the pro- F.D. Richardson: Physical Chemistry of Melts in Metallurgy, Academic Press, New York, NY, 1974, vol. 2. cess into three stages comprising reactions within the C.S. Chai: Ph.D. Thesis, Massachusetts Institute of Technology, droplets, within the diluted weld pool, and within the Cambridge, MA, 1980. solidifying weld pool. In this paper, a kinetic model for H. Thier: Proc. Conf. on Weld Pool Chemistry and Metallurgy, the second region has been developed which is capable The Welding Institute, London, Apr. 1980, pp. 271-78. G.K. Sigworth and J.F. Elliott: Met. Sci., 1974, vol. 8 (9), of explaining the compositional differences between both pp. 298-310. single-pass and multiple-pass welds. In addition, geo- Y. Kawai and K. Mori: Trans. Iron Steel Ins?. Jpn., 1973, metric parameters of the pool which are included in this vol. 13 (3, pp. 303-17. model can be related indirectly to the practical weld pro- U. Mitra: Ph.D. Thesis, Massachusetts Institute of Technology, cess parameters, such as voltage, current, travel speed, Cambridge, MA, 1984. T.H. North: Weld. Res. Abroad, 1977, vol. 23 (I), pp. 2-40. and weld preparation geometry. In Part I11 of this se- C.J. King: AlChE J., 1964, vol. 10 (5), pp. 671-77. ries,1411this model is tested with experimental data. It J. Szekely: Chem. Eng. Sci., 1965, vol. 20, pp. 141-45. - -:I1 be seen that the model is capable of producing very H. Thier and R. Killing: Schweissen Schneiden, Sept. 1982, urate predictions of weld pool composition. pp. E174-E176. H. Thier: IIW Doc., XII-802-83, 1983. N. Christensen and J. Chipman: Weld. Res. Counc. Bull., ACKNOWLEDGMENT Jan. 1953, no. 15, pp. 1-14. U. Mitra and ~.~:~a~ar:Metall. Trans. A, 1984, vol. 15A, The authors are grateful for support of this work by pp. 217-27. the National Science Foundation under Grant No. DMR- H.J. Palm: Weld. J., 1972, vol. 51 (7), pp. 358s-360s. 8200596. E.T. Turkdogan: J. Iron Steel Inst., Jan. 1972, vol. 210 (I), pp. 21-36. M.L. Turpin and J.F. Elliott: J. Iron Steel Inst., Mar. 1966, REFERENCES vol. 204 (3), pp. 217-24. W.O. Philbrook: Int. Met. Rev., Sept. 1977, vol. 22, pp. 187-201. U. Mitra and T.W. Eagar: Metall. Trans. B, 1991, vol. 22B, J.F. Elliott and J.K. Wright: Can. Metall. Q., 1972, vol. 11 (4), pp. 65-71. pp. 573-84. I.K. Pokhodnya: Avtom. Svarku, 1964, vol. 17 (2), pp. 1-10. U. Lindborg and K. Torssell: Trans. AIME, Jan. 1968, I.K. Pokhodnya and B.A. Kostenko: Avtom. Svarka, 1965, vol. 242 (I), pp. 94-102. vol. 18 (lo), pp. 16-22. S. Linder: Scand. J. Metall., 1974, vol. 3 (4), pp. 137-50. N.N. Potapov and K.V. Lyubavski: Svar. Proizvod., 1971, no. 1, M.C. Flemings: Int. Met. Rev., Sept. 1977, vol. 22, pp. 201-07. pp. 11-13. M.C. Flemings: Solidification Processing, McGraw-Hill, New A. Block-Bolten and T.W. Eagar: in Trends in Welding Research York, NY, 1974. in the United States, ASM, Metals Park, OH, 1982, pp. 53-73. U. Mitra and T.W. Eagar: Metall. Trans. B, 1991, vol. 22B, T.W. Eagar: in Weldments: Physical Metallurgy and Failure pp. 83-100. Phenomenon, R.J. Christoffel, E.F. Nippes, and H.D. Solomon, K. Ishizaki: Int. Conf. on Weld Pool Chemistry and Metallurgy, eds., General Electric Corporation, Schenectady, NY, 1979, London, Apr. 15-17, 1980, The Welding Institute, Cambridge, pp. 31-42. United Kingdom, 1980, vol. 1, pp. 65-76.

METALLURGICAL TRANSACTIONS B VOLUME 22B. FEBRUARY 1991 -81