Fundamentals of the Chemical Behavior of Select Fluxes

Changes in weld metal Mn levels as a function of electrochemical reactions in the are investigated

BY A. POLAR, J. E. INDACOCHEA AND M. BLANDER

ABSTRACT. This investigation evaluates new alloys. These new materials are pro­ a comprehensive analysis of weld chem­ the relative effects of thermochemical and duced, usually, with the consideration of istry. electrochemical reactions on the trans­ meeting the most stringent mechanical It is widely accepted that the chemical port of elements, particularly manganese, and chemical requirements, and with little composition of the welding flux affects from the flux to the weld metal in sub­ or no attention to the of the the final weld metal chemistry (Refs. 12, merged . The experimental alloy. 13); however, the mechanisms responsi­ fluxes used were silica-calcium oxide- Many of the new advanced alloys have ble for this elemental transfer (flux to weld based, containing 20 wt-% MnO, 15 wt-% their designed physical properties only metal and vice versa) have not been CaF2, and SiC>2 to CaO ratios that varied within a confined compositional and mi­ clearly described. Attempts have been from 5.50 to 1.16. The slags formed show crostructural range. This limitation imposes made to develop a purely thermodynamic good detachability, and the welds pro­ a need for tight control of the weld model (Refs. 9, 12, 14) on the assumption duced have good bead morphology. The chemistry and consequently of the weld that equilibrium is achieved (or at least ap­ dilution effect was eliminated by drawing microstructure. It is therefore important to proached) between the weld pool and the welding wire from the same materials understand and, to some extent, control the , despite the short time that the as the base plate. The welding parameters the different mechanisms operating in arc weld pool remains liquid. Chai, et al. (Ref. were held constant during weld produc­ welding, which affect the composition of 9), observed that elemental transfer be­ tion and two polarities were used. The arc the weld metal. tween the slag and the weld pool could be was found to be stable, but was more so Systematic studies have been made on explained thermodynamically for the case for -positive (reverse) polarity fluxes, focusing of Mn and Cr, but their model has dis­ welds. on the reactions of these fluxes with the crepancies when it is applied to the trans­ ference of Si. The results of chemical analyses of molten metal during welding (Refs. 1-12). fluxes, slags and welds are consistent with Some of this information may be helpful to Lately, it has been proposed and ex­ the following three mechanisms working deduce an elaborate conceptual structure perimentally supported that, in addition to in parallel that all affect the compositions for predicting chemical variations in the thermochemistry, electrochemical effects (and the apparent compositions) of weld weld metal, for explaining the introduc­ influence weld metal compositions (Refs. metal: 1) the pyrochemical reactions be­ tion of impurities from the flux, or for es­ 15-19) of submerged arc welds. In this in­ tween the slag and the metal; 2) electro­ tablishing the application of a particular vestigation, we examine the role of elec­ chemical reactions at the anodes and welding flux for a given type of weld. trochemistry in the manganese transport cathodes (oxidation and reduction, re­ Studies of pyrometallurgical reactions be­ from the flux to the weld metal by select­ spectively); and 3) occlusion of slag or tween flux and weld metal, in general, are ing a Si02-CaO-CaF2-MnO flux containing solid products of reactions in the weld deficient in that fundamental concepts of 20 wt-% MnO, 15 wt-% CaF2, and Si02 to pool. the thermodynamic properties of slags CaO ratios that varied from 5.50 to 1.16. and metals are not linked with kinetic and The changes in the levels of manganese in electrochemical considerations to deduce the weld metal and of manganese oxide in Introduction the slag are interpreted based on thermo­ One of the primary concerns in the dynamic and electrochemical mecha­ welding of metals is to produce welds nisms. with physical and mechanical properties similar to or better than those of the ma­ KEY WORDS terials joined. This concern for the quality Chemical Behavior Experimental Work of the weld has become more pressing in Welding Fluxes view of the design and development of Thermochemical Reaction Materials Electrochemical Reaction The Si02-CaO-CaF2-MnO flux system Element Transport was selected for this investigation with A. POLAR and). E. INDACOCHEA are with the Manganese CaF and MnO fixed at 15 and 20 wt-%, Civil Engineering, Mechanics and 2 SAW respectively; the Si02 varied from 35 to 15 Department, University of Illinois at Chicago, Slag Chicago, III. M. BLANDER is with the Chemical wt-% and CaO made up the remainder, as Technology Division/Materials Science Pro­ Welds shown in Table 1. These fluxes were pro­ gram, Argonne National Laboratory, Argonne, Weld Metal cessed in the laboratory using reagent 111. grade powders, accurately weighed,

WELDING RESEARCH SUPPLEMENT 115-s with acetone, and the fluxes were stored Si02/CaO ratio (based on nominal wt-% overnight in an oven at about 300°C chemistries) in the flux, as shown in Fig. 2. (572°F). The welds were single pass, flat, It is observed that the amount of manga­ bead-on-plate, and automatically pro­ nese transferred decreases as the Si02/ cessed. The same electrode extension and CaO ratio increases. This behavior can be electrode-to-workpiece distances were interpreted thermodynamically since the maintained for all welds. The welding pa­ activity of the manganese oxide decreases 3HME =5RMS rameters utilized were programmed in a with an increase in the Si02/CaO ratios microprocessor that controlled the power within this composition range. It should be supply. The weld parameters, shown in noticed, however, that there is a differ­ Table 2, were kept constant during weld ence in the amount of manganese trans­ production for both polarities. The stabil­ ferred between the reverse and straight ity of the arc was monitored for every polarity welds for Si02/CaO ratios less weld, as shown in Fig. 1. It can be seen that than 2.25. This effect is consistent with a RRM3 the arc was fairly stable and constant in large electrochemical component in the mechanism for manganese transfer, since -L.II=:..-':-^;i ;--^8'-~5j every flux with electrode-positive (EP) po­ vfMMH for EP polarity the weld pool becomes the L :: r : r ; : 1 larity; however, in the case of electrode- T--Z~ —el~-" F7i. - r "r"- E3 ; " .;- . J. !. s; cathode, implying that the MnO in the slag negative (EN) polarity, there was a de­ fU f-Up at the cathodic weld pool is reduced; this crease in arc stability, although the aver­ .._,; . . ., • |\;-i-i ,- results in the electrodeposition of manga­ age voltage remained constant. nese at the weld pool. SRM4 Chemical Analyses The differences in the amount of man­ ganese transferred due to the change in Oxygen analyses of the weld metal polarity can be understood by considering were done in a LECO-R016 analyzer, the kinetic factors that control the elec­ while atomic absorption photometry was trodeposition rate. In the case of welding used to determine the manganese and sil­ with EP polarity, manganese is electrode- icon levels of the weld metal. Analyses of posited in the cathodic weld metal. As a 5RMS RPM5 the flux and slags were also performed result, the MnO concentration at the . •±'.; g jf^''^ -^^t^—-— using atomic absorption photometry. Care slag/metal interface is lowered consider­ |inMNliifi*dUn*i~- was taken to separate the unfused flux ably, and the transfer of MnO from the from the slag. All metal samples were slag to the interface becomes diffusion carefully cleaned prior to the chemical • controlled. On the other hand, the chem­ analyses. ical potential (n) of MnO for a fixed con­ Fig. 1 — Arc stability data for all flux samples. The centration is much lower in the high-silica R and S in the charts stand for reverse and Results and Discussion than in the low-silica flux; consequently, straight polarity, respectively. the diffusive driving force,

16-s | JANUARY 1991 Table 2—Welding Parameters

Voltage 33 V Current 600 A 5 0.35 Weld Speed 12.5 in./min (5.3 mm/s) Wire Speed 75.0 in./min (31.75 mm/s) Heat Input 95.0 kl/in. (3.74 k|/mm) 0.25 - S left in the weld metal. This effect parallels the tendency during electrodeposition to Fig. 2 —Manganese emplace more manganese the lower the transferred from the silica content of the slag. The polarity of flux to the weld the current will also influence the extent of metal as a function the back reaction. In the case of EP polar­ 0.05 of nominal flux ity, the weld pool is the cathode and composition and manganese is electrodeposited there, polarity of the current. where rapid convection currents mix it Si02/CaO (w/o) and decrease the surface concentration. This leads to a lesser back reaction than with EN polarity, where the manganese is electrodeposited on the welding wire that production by chemically reducing the deposition of manganese on the metal is detached as a droplet. In the case of EN manganese oxide with iron must also be (weld metal and weld wire) polarity, manganese can be lost from the considered. The chemical reduction and that a significant fraction of the man­ surface of the droplet according to the would proceed according to the reaction ganese back reacts thermochemically with chemical reaction shown in Equation 1, as Si02 to replace some of the manganese the metal drop falls through the slag, or by MnO(S|ag) +-r Fere(mela„ i) = rvinMn,(met, ai) +• oxide lost from the slag. vaporization as it drops through the FeO(s|ag) (2) An examination of the amount of iron plasma. Finally, as the droplet touches the which has a strong tendency to go to the oxide in the slag indicates levels that are weld pool, it comes in contact with a layer right because of no initial iron oxide in the too large and have the wrong depen­ of an electrochemically generated iron original flux. The iron oxide content of the dence on composition to be produced by oxide that reacts with surficial manganese slag has been plotted as a function of the the oxidation of metallic iron by MnO in to form the more stable manganese oxide. Si0 /CaO ratio, as shown in Fig. 4. These the slag. Oxidation of iron by silica in the The driving forces that affect the back re­ 2 results, when compared to those of Fig. 2 slag can take place at metal-slag interfaces actions are dependent on the amount of and 3, show inconsistencies, with a ther­ by the reaction silica present in the slag. Thus, for high sil­ mochemical mechanism being more than ica, the back reaction is more complete Si°2(J«j + 2Fe(metal) = 2FeO | a minor factor in transfer of Mn to weld (s ag) the larger the amount of silica in the slag, metal since the iron oxide content of the + Si(metal) (3) and it is greater for EN polarity than for EP polarity. These conclusions are consistent slag decreases as the manganese in the where this reaction goes to the right at with the results shown in Figs. 2 and 3. weld metal increases and the manganese high temperatures (>2000 K) as has been oxide content of the slag decreases, a discussed previously (Ref. 12). This reac­ We have primarily discussed the man­ trend that is opposite to that predicted by tion is driven partly by the low activity co­ ganese transfer from the flux to the weld the chemical reaction (Equation 2). There­ efficients of Si in iron. Since the activities of metal in terms of an electrochemical mech­ fore, it is likely that an electrochemical silica are higher and the activity coeffi­ anism, but the possibility of manganese mechanism is largely responsible for the cients of FeO are somewhat lower the

CaF2 = 15 w/o MnO = 20 w/o A EP o EN

\ A

\ \ \ \ o A o 0.5

o ~~ __ 0

1.0 •I.I) 6.0

Si02/CaO l w/o) Si02'CaO i w/o) Fig. 3 — Manganese oxide consumed as a function of nominal flux com­ Fig. 4 - Iron oxide content in the slag as a function of nominal flux chem­ position and current polarity. istry and current polarity.

WELDING RESEARCH SUPPLEMENT 117-s higher the silica content of the slag, the in solution in the slag. Since activity coef­ structure. driving force for reaction (Equation 3) is ficients of FeO are somewhat lower in the The mechanism for manganese transfer larger at high-silica than at low-silica levels. high-silica slags, one would therefore ex­ to the weld metal involves both the pyro­ The dependence of the FeO content on pect more FeO to dissolve in the high-sil­ chemical and electrochemical mecha­ the slag composition is consistent with this ica slag than in the low-silica slag. Since sil­ nisms. Manganese can be transferred to difference in the chemical driving force. icon, which has been deposited at the the metal by Equation 2 reaction up to the The weight-percent of FeO in the slag is cathode, and the oxide that remains at the point where it can be oxidized by Si02- considerably greater than that of silicon in anode (weld pool or droplet) mix in the The driving force (i.e., the affinity— Ref. the metal. However, the equivalence ra­ weld pool, the back reaction would lead 22) for this reaction is considerably larger tios of Si/FeO are crudely consistent with to the same type of distribution of the at low-silica contents than at high-silica the equivalence ratios of the Mn trans­ compositions of FeO in the slag and Si in contents. In addition, at low enough FeO ferred to weld metal to MnO lost from the the weld pool. In fact, comparison of the contents, Mn can be oxidized by Si02 by slag, so that these ratios indicate the silicon with the oxygen contents of the Equation 1 reaction. The electrochemical probability that the number of moles of Si metal (Figs. 5 and 6) reveals near equiva­ mechanism is deduced from the fact that or Mn transferred to the metal is approx­ lency (with analyses spanning O/Si molar at the lower weight ratios of Si02/CaO imately equal to the number of moles of ratios of about 1.4-2.2), indicating that (<2), the manganese transferred to the Si02 or MnO lost from the slag. There­ most of these two elements could be metal (Figs. 1 and 2) is much larger in EP fore, it is possible that FeO was emplaced present as occluded Si02, either as the re­ polarity than in EN polarity. Thus in EP, Mn in the slag by the Equation 3 reaction. action product of Si and FeO in the metal is deposited at the weld pool and swept or as occluded slag. On the other hand, into the weld pool before it can back re­ An electrochemical mechanism is also the Mn content of the weld metal pro­ act with the silica. On the other hand, the possible. In this mechanism, oxides of the duced with the three lower Si02/CaO ra­ Mn deposited on the droplets formed at anode metals (largely Fe) are produced at tios in the flux is much too high and the the cathodic welding wire in EN fall through the slag/metal interface. These oxides oxygen content too low for the Mn to be the slag and are partly oxidized by S1O2, or tend to dissolve in the slag and in the liq­ largely present as MnO in occluded sili­ Mn falls through the plasma where it can uid metal. For EP polarity, oxide formation cates. be volatilized at relatively high tempera­ is driven by the high current densities at tures. At high Si0 /CaO ratios (>2), the the anodic welding wire, while the oxide 2 driving force (affinity for reaction 1) is so tends to partly dissolve in the molten slag. Summary large (because of the activity of Si02 and When a droplet separates from the weld­ low activities of MnO and FeO) that most ing wire, oxidation ceases, but the oxide Our results are consistent with three of the electrodeposited Mn is back re­ continues to dissolve in the slag. With EN possible mechanisms for fixing the chem­ acted. Relatively little Mn is transferred to polarity, the weld pool is the anode, the istries of weld metal in SAW. One mech­ the metal for both polarities at the higher current density is lower, and dissolution of anism is by pyrochemical reactions be­ Si02/CaO ratios. the oxide takes place from a larger area tween the slag and the metal. A second is (but for a shorter time because of the by electrochemical reactions at the an­ Iron oxide can be produced chemically rapid stirring of the weld pool by eddy odes (oxidation) and cathodes (reduction). by Equation 3 reaction and electrochem- currents) than for the case of EP polarity. Electrode-positive (EP) polarity is most ically at the anodes (welding wire in EP, If this mechanism is correct, the net result commonly used in SAW, with the welding weld pool in EN). There appears to be a (Fig. 4) appears to be a somewhat greater wire being anodic and the weld pool ca­ significant difference in the content of dissolution of FeO from the welding wire thodic. The third mechanism that alters FeO in the slag for EP and EN at the two and droplet in EP polarity than from the the apparent chemistry of weld metal is highest values of the Si02/CaO ratio — weld pool in EN polarity. The driving force the occlusion of slag or solid products of Fig. 4. The higher FeO content in the two controlling the rate of dissolution is the reactions in the metal. Such solid inclusions slags from welding with EP could have re­ difference between the high chemical can provide nucleation sites for alloy crys­ sulted from dissolution of the surface ox­ potential of FeO at the interface and that tallization and thus influence the micro- ide during the fall of the welding wire

0.20 CaF2 = 15 w/o MnO = = 20 w/o A EP r 0 EN CaF2 = 15 w/o MnO =- 20 w;o A EP "o 0 EN £ 0 "c3 0.16 - CD J&-~"~"—' 1400 ^ t. 2 "aS £ A A ^* s 0.12 •0 A 1000 0 £,

00

0.04 1.0 2.0 3.0 4.0 200 1 1 .0 2.0 3.0 4.0 6.0

Si02/CaO 1 w/o 1 Si02'CaO (w/o 1 Fig. 5 —Silicon transferred from the flux to the weld metal as a function Fig. 6 — Oxygen transferred from the flux to the weld metal as a function of nominal flux composition and polarity of the current. of nominal flux composition and polarity of the current.

18-s I JANUARY 1991 droplets through the slag. This difference edge the financial support by the U.S. In­ 12. Indacochea, J. E., Blander, M., Christen­ is indicative of an electrochemical effect formation Agency and the Fullbright sen, N., and Olson, D. L. 1985. Chemical reac­ superimposed on the effect of a pyro- LASPAU Faculty Development on behalf tions during SAW with FeO-MnO-Si02 fluxes. Metallurgical Transactions B 16B, pp. 237-245. chemical reaction. of Mr. Polar. Inland Steel Company is also 13. Blander, M., and Olson, D. L. 1984. It appears that an explanation of weld gratefully acknowledged for supplying Thermodynamic and kinetic factors in pyro- metal chemistry requires the integration of materials for this work. The authors also chemistry of submerged arc flux welding of the different possible effects that influ­ appreciate the assistance of Mr. Rod iron-based alloys. Second International Sympo­ ence elemental transfer. Thus far, it has Hawkins, with the Indiana Harbor Works sium on Metallurgical Slags and Fluxes, H. A. Fine been necessary to consider pyrochemical of Inland Steel, in the chemical analysis of and D. R. Gaskell, eds., TMS-AIME, pp. 271- and electrochemical effects as well as slag the weld samples. 277. occlusion (Refs. 9-19). Possible mecha­ 14. Mitra, U., and Eagar, T. W. 1984. Slag- metal reactions during submerged arc welding nisms involving the plasma have been References of alloy steels. Metallurgical Transactions A, considered. The difficulties involved in 1. Christensen, N., and Chipman, ). 1953. 15A, pp. 217-227. deducing the mechanisms of elemental Slag-Metal Interaction in Submerged Arc Weld­ 15. Blander, M., and Olson, D. L. 1986. Elec­ transfer from or through the plasma stem ing. WRC Bulletin, Series No. 15, WRC. trochemical effects on weld pool chemistry in from a lack of knowledge of the compo­ 2. Babcock, D. E. 1941. The fundamental submerged arc and DC . sition and physical state of the plasma. Be­ nature of welding: Part V —the physical chem­ Advances in welding science and technology, S. cause of the relatively high voltage and istry of the arc welding process. Welding lour­ A. David, ed., Proceedings of the International the necessary presence of positive ions to nal 20(4):189-s to 197-s. Conference on Trends in Welding Research, achieve charge neutrality, a significant, if 3. Bischof, F. 1943. The solubility of N in liq­ Gatlinburg, pp. 363-366. not major, mechanism for elemental trans­ uid welding metal during the arc welding of 16. Shah, S., Blander, M., and Indacochea, J. fer via the plasma is probably electro­ bare, differently alloyed electrodes. Electro- E. 1987. Submerged arc flux welding with schweissung 14, pp. 63-66. CaF -CaO-Si0 fluxes: possible electrochemi­ chemical. 2 2 4. Christensen, N. 1949. Metallurgical As­ cal effects on weld metal. Proceedings of the pects of Welding Mild Steel. Welding Journal Joint International Symposium on Molten Salts, Conclusions 28(4):373-s to 380-s. 172nd Electrochemical Society Meeting, Hono­ 5. Claussen, C. E. 1949. Metallurgy of cov­ lulu, Hawaii, Vol. 87-7, pp. 916-927. 1) Changes in the manganese weld ered-electrode weld metal. Welding Journal 17. Kim,). H., Frost, R. H„ Olson, D. L., and metal content appear to be largely con­ 28(1):12-24. Blander, M. 1987. Electrochemical reactions at trolled by electrochemical deposition re­ 6. Wolstenholme, D. A. 1978. Intl. Conf. on the electrode in submerged arc welding. Pro­ actions with thermochemical control of Trends in Steel and Consumables for Welding. ceedings of the Joint International Symposium the back reaction of the electrodeposited p. 123, The Welding Institute, London, U.K. on Molten Salts, 172nd Electrochemical Society manganese with Si02 and FeO. 7. lackson, C. E. 1973. Fluxes and slags in Meeting, Honolulu, Hawaii, Vol. 87-7, pp. 928- welding. WRC Bulletin, Series No. 190, WRC, 938. 2) The iron oxide content of the slag December. 18. Indacochea, |. E., Blander, M., and Shah, appears to be produced electrochemi- 8. Garland,). G., and Bailey, N. 1975. Fluxes S. 1989. Submerged arc welding: evidence for cally and chemically by reaction (Equation for submerged arc welding ferritic steels: a lit­ electrochemical effects on the weld pool. 3) at the anodes. erature survey. The Welding Institute, Member Welding Journal 68(3):77-s to 83-s. 3) The occlusion of slag or solid prod­ Report M/84/75. 19. Polar, A., Indacochea, )', E., and Blander, ucts of reactions in the liquid metal appar­ 9. Chai, C. S., and Eagar, T. W. 1981. Slag- M. Electrochemically generated oxygen con­ ently alters the chemistry of the weld metal equilibrium during submerged arc weld­ tamination in submerged arc welding. 1990. metal. ing. Metallurgical Transactions B, 12B, pp. 539- Welding Journal 69(2):68-s to 74-s. 547. 20. Turkdogan, E. T. 1980. Physical Chemis­ 10. Eagar, T. W. 1978. Sources of weld try of High Temperature Technology, p. 81, metal oxygen contamination during submerged Academic Press, New York. Acknowledgments arc welding. Welding lournal 57(3):76-s to 21. Pankratz, L. B. 1982. Thermodynamic 80-s. Properties of Elements and Oxides. United The work at the Argonne National Lab­ 11. Mitra, LJ., Sutton, R. D., and Eager, T. W. States Department of the Interior, Bureau of oratory was supported by the Office of 1983. Comparison of theoretically predicted Mines Bulletin 672, Supt. of Documents, U.S. Naval Research under Navy Order No. and experimentally determined submerged arc Govt. Printing Office, Washington, D.C. N00014-87-F-0064. The authors from the weld deposit compositions. Metallurgical 22. Prigogine, I., and Defay, R. 1962. Chem­ University of Illinois at Chicago acknowl­ Transactions 14B, pp. 510-513. ical Thermodynamics. Wiley, New York, p. 69.

WRC Bulletin 352 April 1990

In October 1987, the PVRC Steering and Technical Committees on Piping Systems established a task group on independent support motion (ISM) to evaluate the technical merits of using the ISM method of spectral analysis in the design and analysis of nuclear power plant piping systems. The results of the task group evaluation culminated in a unanimous technical position that the ISM method of spectral seismic analysis provides more accurate and generally less conservative response predictions than the commonly accepted envelope response spectra (ERS) method, and are reported in this WRC Bulletin. The price of WRC Bulletin 352 is $25.00 per copy, plus $5.00 for U.S., or $10.00 for overseas, postage and handling. Orders should be sent with payment to the Welding Research Council, 345 E. 47th St., Room 1301, New York, NY 10017.

WELDING RESEARCH SUPPLEMENT 119-s