Effect of Arc Welding Electrode Temperature on Vapor and Fume Composition

N. T. Jenkins, Ph.D.Massachusetts General Hospital, Charlestown, Massachusetts, USA

P. F. Mendez, Ph.D.Colorado School of Mines, Golden, Colorado, USA

T. W. Eagar, Sc.D.Massachusetts Institute of Technology, Cambridge, Massachusetts, USA

Abstract

Welder exposure to potentially toxic compounds in fume can bereduced through the control of the welding process. Suchefforts can be improved by better understanding how processparameters affect the elemental composition of welding fume.The vapor emitted from a consumable electrode is the dominantsource of arc welding fume; for bare wire electrodes, the metalcontent of arc welding fume is essentially the same as that of thevapor. Many researchers have measured fume composition bymany methods, but only a few have tried to use thermodynamicequilibrium calculations to predict the composition of the vaporthat forms the fume. Attempts to do so have been flawed byassumptions that the temperature of the electrode droplet doesnot vary throughout the droplet, or with welding parameters.The temperature of the droplet surface, not of its bulk,determines the elemental composition of the vapor. Becauseevaporation is strongly dependent on temperature, a smallchange in surface temperature will cause a large change invapor composition. This can be illustrated with a vapor-liquidbinary phase diagram. Thus the vapor composition andtherefore the fume composition are affected by changes inwelding parameters that control droplet surface temperature. Forexample, because the surface temperature of the electrodedroplet during spray mode is less than it is during globularmode, fume generated during spray mode has a greater fractionof volatile metals than fume formed during globular mode.

Fume Composition

Welder exposure to potentially toxic compounds in fume can bereduced through the control of the welding process. Suchefforts can be improved by better understanding how processparameters affect the elemental composition of welding fume.Researchers have measured welding fume chemistry by manymethods, but only a few have tried to predict the composition ofthe original vapor from thermodynamic calculations (1). Theresults of such modeling have been mixed, primarily because ofincomplete assumptions about the temperature. Thetemperature is the primary factor that determines vaporcomposition (2).

Most fume researchers have assumed a single value for the gasmetal arc welding (GMAW) droplet temperature to use in theirthermodynamic calculations. Gray, et al., (3) and Podgaetskii etal., (4) reported thermodynamic equilibria without explicitly

stating the temperature they used. Hewitt & Hirst (5) consideredfluxes in their calculations and Buki & Feldman (6) includedoxides, but they too did not report the temperature. McAllister& Bosworth (7) assumed droplet temperatures were 1800–2200K and were the first to include the effect of various shieldinggases. Eagar, et al. (8) calculated gaseous equilibria forhexavalent chromium using varied shielding gases, at 2673 K.

None of these researchers considered the change in droplettemperature with time or how it changes with weldingparameters. They also equated the surface temperature, whichdetermines the composition of the evolved vapor and the fumeformation rate, to the average temperature across the entireelectrode droplet. They also did not consider that the surfacetemperature of the welding droplet during globular transfer wassubstantially different from that of a droplet during spraytransfer (9). This can lead to misleading conclusions.

Electrode Temperature

Fume researchers have used simple approximations for theelectrode temperature probably because determining thetemperature of a material while welding is difficult. Thetransitory nature and extreme heat of welding prevents the useof thermocouples; calorimetric methods are also complicated toset up in situ. The use of pyrometric techniques is problematicbecause the intense radiation from the arc overwhelms theinfrared emissions of heated surfaces. Therefore many weldingresearchers rely on calculations of the weld metal temperatures.

The weld pool temperature is relatively simple to calculate. Thishas been performed several times, notably by Block-Bolten &Eagar (10) who found that because of evaporative heat losses,an upper limit of 2300K to 2800K exists for the temperature ofthe weld pool of steel gas tungsten arc welding (GTAW).However, for consumable electrodes in GMAW, the temperatureof the welding droplet is more complicated to determine,because of an input of higher energy density, because of a moretransitory nature (droplet detachment), and becausemeasurements with which to compare the calculations areharder to obtain. Generally, values for the bulk temperature ofthe electrode are reported close to 2900K (11).

(It should be noted that for GMAW, welding fume is dominatedby evaporation from the electrode rather than from the weldpool. Heile & Hill (12) and Sreekanthan (13) both compared thecomposition of fume collected while GMA welding with an

Figure 1: Heat transfer through GMAW electrodes is controlled by electrode droplet size (17). Shadowgraphs are from2%O2-Ar shielded GMAW mild steel electrode (19).

electrode of a different composition than that of the base plateand both found that the fume composition was determinedalmost entirely by the electrode. The authors performed asimilar study of hexavalent chromium. GMAW fume wascollected while welding under spray conditions with E308Lstainless steel welding wire on mild steel pipe and on stainlesssteel pipe at 31V, 175A with a 2%O2-Ar shielding gas. ER70S-6mild steel wire, which does not contain chromium, was used toweld at 33 V, 215A on a 304 stainless steel plate, also with spraytransfer and with the same shielding gas. Fume was collected at2.25 L/min for 30 seconds. The concentration of hexavalentchromium in fume created from the mild steel electrode was <0.001 wt%. For the stainless steel electrode, it was 0.1 wt%,regardless of which base plate was used. Because mild steeldoes not contain chromium, the source of the hexavalentchromium in the fume must have been the electrode. It is clearthat GMAW fume composition is dominated by vaporizationfrom the electrode, not from the weld pool. The reason for thisis because the weld pool is much cooler than the electrode.)

Haidar (14–16), using a Cray supercomputer, performed acomplete fundamental calculation of electrode droplettemperature and how it changes with time and weldingparameters. The use of a supercomputer can be avoided if thedroplet size is measured and a semi-empirical model created,like those in papers by Mendez, et al., (17) and Bosworth &Deam (18). A summary of such models can be seen in Figure 1.

Smaller droplets, such as those formed during spray or pulsedtransfer, have cooler surface temperatures than do largerdroplets, like those from globular transfer, because there is asmaller barrier to heat transfer from the arc spot to the liquid-solid interface of the electrode. (However, smaller droplets mayhave greater bulk temperatures.) Therefore vapor of acomposition enriched in the more volatile elements is evolvedand further, less vapor forms. This can be observed in thechange in fume composition with welding parameters, such asshielding gas oxidation potential (Figure 2 to Figure 5) orcurrent (Figure 6). Both of these welding parameters change thesize of the welding droplet and therefore its vaporization rate.Conditions that create fume containing the greatest fractions ofvolatile metals (e.g., manganese) are the same that create thesmallest electrode droplet. Such conditions also minimize totalfume formation. (Thus, efforts to decrease fume formation ratesmay also increase the fraction of manganese in the fume.Manganese emission rates therefore may or may not changesignificantly.)

Superheat

Based on an energy balance of a GMAW electrode, the surfacetemperature of the electrode droplet may exceed the boilingpoint (17). Haidar (16) also calculated the surface temperatureof a mild steel 1.2 mm GMAW electrode at 183A and comparedit with the value measured pyrometrically by Villeminot (23) at160A. Both found that the surface temperature at the tip of theelectrode droplet was approximately 3100K. This is very closeto, if not above, the boiling point of pure iron (values in theliterature include 3025K (16), 3273K (24), 3121K (25)); it is

certainly above the vapor-liquid transition for a typical ironalloy used in welding (see Figure 7).

Like other phase changes, boiling can occur through nucleation.Vapor bubbles usually form heterogeneously in crevices of thecontainer that holds the boiling liquid. (Consider a boiling potof water on a kitchen stove; the bubbles form on the bottom ofthe pan.) For bubbles to form, superheating must be present inorder to overcome the activation energy required to create newsurfaces. Because heat is usually added at the solid-liquidinterface and lost at the gas-liquid interface, the amount ofsuperheating depends on crevice size where vapor bubblesheterogeneously nucleate (26).

However, when heat is added by a high-energy-density source,i.e. electron condensation in arc welding or laser radiation, theheat input is at the gas-liquid interface. (For an experimentalstudy of surface superheats from laser processing, see Craciun& Craciun (27).) The solid-liquid interface is where heat is lostand is therefore the coolest part of liquid. Thus vapor bubbleswill not nucleate at any solid crevices. Instead, the vapor phasemust form bubbles completely in the liquid. Another difficultyfor forming bubbles when intense heat is introduced at theliquid surface is that only a slender boundary layer will be hotenough for bubbles to nucleate in it. Under globular conditions,a 1.2 mm GMAW electrode droplet will have a thermalboundary layer approximately the size of 1/3 of the droplet(17). If the convective core of the droplet were at the boilingpoint, the surface temperature of a steel electrode droplet wouldhave to be near 3300K for a bubble smaller than 0.1 mm indiameter to form, according to classical nucleation theory. Thesurface temperature would naturally have to be much greaterthan 3300K for even smaller bubbles to form.

Richardson (28) states that classical nucleation predictshomogeneous superheats much greater than what is observed,but this is probably because the experimental measurementswere made in liquids heated through the solid-liquid interfacewhere crevices caused heterogeneous nucleation. This is alsoprobably the case for molten steel where vapor bubbles canpresumably nucleate heterogeneously at pre-existing bubbles ofgas, like carbon monoxide, coming out of solution. If this werethe case, it would explain why lower carbon content in steelelectrodes is linked with lower fume formation rates (29-32).

When gas-liquid interfaces are heated, cooling by non-nucleatesurface evaporation normally prevents superheating. However,with high-energy-density heat sources and where heat loss isphysically limited (like with the thin GMAW electrodes), it ispossible to inject energy into the system faster than it can beremoved by evaporation, as calculated by the Langmuirequation with even the most conservative estimates (33). Notethat the Langmuir rate is for evaporation into a vacuum. Whensurrounded by an inert gas at atmospheric pressure, metalsurfaces evaporate more slowly than the Langmuir rate, becausethe metal vapor must diffuse away from the evaporating surface.Thus evaporative cooling during welding is limited even morethan that predicted by the Langmuir rate. If an energy balance isto be maintained, the surface temperature, which controls

Figure 3: Fume composition versus shielding gasoxidation potential in GMAW with 1.2 mm E308 stainlesselectrode (13).

Figure 4: Electrode droplet period / droplet size versusshielding gas composition in 1.6mm mild steel GMAW forvarious currents (20).

Figure 5: Fume formation rate versus shielding gas with1.2 mm mild steel GMAW at 250 amp (12, 21).

Figure 6a: Fume generation rate and the nickel : manganeseratio versus current for argon shielded GMAW with 1.2 mmAWS ER307 Si electrode (22).

Figure 6b: Fume generation rate and the nickel :manganese ratio versus current for argon shielded GMAWwith 1.2 mm ER308L Si electrode (22).

Figure 2: Oxidation potential effect on composition (ICPMS) offume from single-pulse (100 amps for 500 ms without dropletdetachment) GMAW of ER308 stainless steel. GTAW fumecomposition also plotted. (Silicon is minutely present, but notmeasureable by ICPMS, so molar fractions of smaller magnitude(i.e. nickel), may have error)

vaporization and thus evaporative cooling, must increase untilthe conditions for balance are met. If the surface temperatureexceeds the boiling point, superheating occurs. This causes theformation of vapor bubbles that burst.

The superheat, or amount by which the boiling point isexceeded, will determine the chemical composition of amulticomponent vapor, similar to how the supercoolingdetermines the composition of a condensate of an alloy. SeeFigure 7, the iron-manganese phase diagram at 0.3 atmospheres(the approximate partial pressure of metal vapor in a GMAWarc (1)). One can see that even a small change in superheat willeffect a large change in vapor composition, given the sameinitial composition of the liquid. This explains how researchershave calculated that the vapor in a mild steel gas metal arcwelding arc may be much as half manganese, even though thewire contains less than 2 wt% Mn (30). In Figure 7 it is shownhow an increase of a mere 20K causes the vapor composition todecrease from 0.18 mole fraction of manganese to 0.08 whenthe liquid is initially at 0.015 mole fraction. The difference indroplet surface temperature between spray and globular transfer,or between pulsing and straight current can easily be greaterthan 100K (16).

Figure 7: Iron-manganese vapor-liquid phase diagram at 0.3atmosphere, adapted to show the effect of superheat on vaporcomposition (25).

This is naturally a simple example of what can becomecomplicated if one considers multicomponent alloys, the effectof halides on metal volatility (34), the presence of surface activeelements (35), the effect of a plasma on vaporization (36), ormixing of the metal vapor with the surrounding gas (37).

Conclusion

It has been found that GMAW fume created during globulartransfer contains less volatile metals than fume created duringspray transfer. This is what prompted the proposal of theexistence of welding fume that did not form from vapor, butfrom ejected liquid microdroplets, which would be similar in

composition to that of the electrode (38). It has been claimedthat globular fume has a greater portion of these coarse“unfractionated” particles and because of this, had acomposition closer to the original electrode composition.Although such particles have been found to exist in GMAWfume, they do not form at a rate that significantly affects theoverall composition of the fume, which is dominated byparticles that condense from vapor (39). In addition, althoughoverall fume formation rates increase, the fraction of coarseparticles found in welding fume does not change from spraytransfer to globular transfer, meaning the change in compositionbetween the two modes is not caused by the existence of morecoarse particles (1). Instead it must be due to a change in thevapor composition, which is caused by a change in surfacetemperature. Other welding process parameters that changedroplet size and temperature, such as pulsing, will presumablyalso change fume composition accordingly.

It has been shown how sensitive vapor composition is totemperature. Therefore, it is important to use appropriate valuesfor the surface temperature of the liquid metal vapor source thatone is trying to model. In this way, welding fume compositioncan be linked with welding parameters, thus granting weldingmanufacturers and users predictive capability.

Acknowledgements

The funding for this project was provided by a grant from theU.S. Navy, Office of Naval Research.

References

1. N. T. Jenkins, Chemistry of Airborne Particles fromMetallurgical Processing, Ph.D. Thesis, MassachusettsInstitute of Technology, Cambridge, MA, (2003)

2. P. A. A. Khan and T. DebRoy, Alloying elementvaporization and weld pool temperature during laser-welding of AISI 202 stainless-steel, Metall. Trans. B, 15,641–644 (1984)

3. C. N. Gray, P. J. Hewitt, and R. Hicks, The prediction offume compositions in stainless steel metal inert gaswelding, Proceedings of the Weld Pool Chemistry andMetallurgy International Conference , N. Bailey, ed.Abington Hall, Cambridge, UK (1980)

4. V. Podgaetskii, A. Golovatyuk, and O. Levchenko,Mechanism of formation of welding aerosol and predictionof its composition in CO2 welding. Paton Weld. J., 1,561–564 (1989)

5. P. J. Hewitt, and A. A. Hirst, Development and validationof a model to predict the metallic composition of flux-coredarc-welding fumes, Ann. Occup. Hyg., 35, 223–232 (1991)

6. A. A. Buki and A. M. Feldman, Prediction of compositionof aerosol formed in welding in shielding gases, Weld.Prod., 27, 8–12 (1980)

7. T. McAllister, and M. Bosworth, Thermal mechanisms forthe production of Cr(VI) [hexavalent chromium] in weldingfume [in arc welding of stainless steels], Australas. Weld.J., 44, 42–46 (1999)

8. T. W. Eagar, P. Sreekanthan, N. T. Jenkins, G. G. K.Murthy, J. M. Antonini, and J. D. Brain, Study ofchromium in gas metal arc welding fume, Proceedings ofthe Trends in Welding, American Welding Society,Warrendale, PA (1998)

9. J. Ma, and R. L. Apps, MIG transfer discovery ofimportance to industry, W eld. Met. Fabr., 9, 307–316(1982)

10. A. Block-Bolton and T. W. Eagar, Metal vaporization fromweld pools, Metall. Trans. B, 15B, 461–469 (1984)

11. O. Levchenko, Processes of welding fume formation(review), Paton Weld. J., 8, 210–215 (1996)

12. R. F. Heile, and D. C. Hill, Particulate fume generation inarc welding processes, Weld. J., 54, 201s–210s (1975)

13. P. Sreekanthan, Study of Chromium in Welding Fume, M.S.Thesis, Massachusetts Institute of Technology, Cambridge,Massachusetts (1997)

14. J. Haidar, A theoretical model for gas metal arc weldingand gas tungsten arc welding, J. Appl. Phys., 84,3518–3529 (1998)

15. J. Haidar, Predictions of metal droplet formation in gasmetal arc welding. II, J. Appl. Phys., 84, 3530–3540 (1998)

16. J. Haidar, An analysis of heat transfer and fume productionin gas metal arc welding. III, J. Appl. Phys., 85, 3448–3459(1999)

17. P. Mendez, N. T. Jenkins, and T. W. Eagar, Effect ofelectrode droplet size on evaporation and fume generationin GMAW, Proceedings of the Gas Metal Arc Welding forthe 21st Century, American Welding Society, Miami,Florida (2000)

18. M. R. Bosworth and R. T. Deam, Influence of GMAWdroplet size on fume formation rate, J. Phys, D Appl. Phys.,33, 2605–2610 (2000)

19. L. A. Jones, T. W. Eagar, and J. H. Lang, Images of a steelelectrode in Ar-2%O2 shielding during constant current gasmetal arc welding, Weld. J., 77, 135s–141s (1998)

20. S. Rhee, and E. Kannateyasibu, Observation of metaltransfer during gas metal arc-welding, Weld. J., 71,381s–386s (1992)

21. D. E. Hilton, and P. N. Plumridge, Particulate fumegeneration during GMAW and GTAW, Weld. Met. Fabr.,12, 555–560 (1991)

22. F. Eichhorn and T. Oldenburg, Untersuchung derScheissrauchentstehung beim Schweissen mit mittel- undhochlegierten Zusatzwerkstoffen, DVS, Duesseldorf, BRD(1986)

23. P. Villeminot, Pyrometrie photographique appliquee ausoudage. Soudage et Techniques Connexes, 21, 367 (1967)

24. Strem Chemicals, Catalog No. 18, p 418, Strem Chemicals,Newburyport, MA (1999)

25. B. Sundman, A program for performing thermodynamiccalculations, User Aspects of Phase Diagrams, F. H. Hayes,ed., Institute of Metals, London, UK (1991)

26. G. F. Hewitt, J. M. Delhaye, and N. Zuber, eds., MultiphaseScience and Technology Vol. 1. Hemisphere PublishingCorporation, Washington DC (1982)

27. V. Craciun, and D. Craciun, Does the subsurfacesuperheating effect really exist? Proceedings of theAdvances in Laser Ablation of Materials Symposium, R. K.Singh, D. H. Lowndes, D. B. Chrisey, E. Fogarassy, and J.Narayan, eds., Materials Research Society, Warrendale, PA(1998)

28. F. D. Richardson, Physical Chemistry of Melts inMetallurgy, Academic Press, London, UK (1974)

29. E. T. Turkdogan, P. Grieveson, and L. S. Darken, Theformation of iron oxide fume, J. Met, 66, 521-526 (1962)

30. O. Grong, and N. Christensen, Factors controlling MIGweld metal chemistry, Scand. J. Metall., 4, 155-165 (1983)

31. T. Suga, and M. Kobayashi, Fume generation in CO2 arcwelding by flux-cored wire, J. Japan Weld. Soc., 2, 68-75(1984)

32. S. E. Ferree, New generation of cored wires creates lessfume and spatter, Weld. J., 74, 45 (1995)

33. J. D. Cobine and E. E. Burger, Analysis of electrodephenomena in the high-current arc, J. Appl. Phys., 26,895–900 (1955)

34. D. A. Tillman, Trace Metals in Combustion Systems,Academic Press, San Diego, CA (1994)

35. P. Sahoo, M. M. Collur, and T. DebRoy, Effects of oxygenand sulfur on alloying element vaporization rates duringlaser-welding, Metall. Trans. B, 19, 967–972 (1988)

36. P. Sahoo, and T. DebRoy, Effect of low-pressure argonplasma on metal vaporization rates, Mater. Lett., 6,406–408 (1988)

37. E. T. Turkdogan, P. Grieveson, L. S. Darken, Enhancementof diffusion-limited rates of vaporization of metals, J. Phys.Chem., 67, 1647-1654 (1963)

38. C. N. Gray, P. J. Hewitt, and P. R. M. Dare, New approachwould help control welding fumes at source (MIG andMMA) part two: MIG fumes, Weld. Met. Fabr., 10,393–397 (1982)

39. A. T. Zimmer, P. A. Baron, P. Biswas, The influence ofoperating parameters on number-weighted aerosol sizedistribution generated from a gas metal arc weldingprocess, J. Aerosol Sci., 33, 519-531 (2002)