O
pen
A
rchive
T
OULOUSE
A
rchive
O
uverte (
OATAO
)
OATAO is an open access repository that collects the work of Toulouse researchers and
makes it freely available over the web where possible.
This is an author-deposited version published in :
http://oatao.univ-toulouse.fr/
Eprints ID : 16719
To link to this article : DOI:10.1016/j.corsci.2016.02.033
URL :
http://dx.doi.org/10.1016/j.corsci.2016.02.033
To cite this version : Fabas, Aurélien and Monceau, Daniel and Josse,
Claudie and Lamesle, Pascal and Rouaix-Vande Put, Aurélie
Mechanism of metal dusting corrosion by pitting of a chromia-forming
alloy at atmospheric pressure and low gas velocity. (2016) Corrosion
Science, vol. 107. pp. 204-210. ISSN 0010-938X
Any correspondence concerning this service should be sent to the repository
administrator:
staff-oatao@listes-diff.inp-toulouse.fr
Mechanism
of
metal
dusting
corrosion
by
pitting
of
a
chromia-forming
alloy
at
atmospheric
pressure
and
low
gas
velocity
Aurélien
Fabas
a,
Daniel
Monceau
a,∗,
Claudie
Josse
b,
Pascal
Lamesle
c,
Aurélie
Rouaix-Vande
Put
aaCIRIMAT,UniversitédeToulouse,CNRS,INPT,UPS,ENSIACET,4alléeEmileMonso,BP44362,31030ToulouseCedex4,France
bUMSCastaing,UniversitédeToulouse,ECA,3rueCarolineAigle,31400Toulouse,France
cICA(InstitutClémentAder),UniversitédeToulouse,MinesAlbi,INSA,ISAE,UPS,CampusJarlard,F-81013Albicedex09,France
Keywords:
A.steel
C.hightemperaturecorrosion
C.carburization
C.internaloxidation
a
b
s
t
r
a
c
t
FeNiCrsamples(800HT)wereexposedat570◦C,1bartoa47.25CO-47.25H2-5.5H2Oatmosphere(ac=33)
flowingat18m/s.Pittingcorrosionwasobserved.Pitsshowedaflattenedmorphologyandaconstant pitdiametergrowthrate.Corrosionringsappearedsuccessivelyatthesurfaceduringpitgrowth.A four-stepmechanismisproposedwhichincludesinternaloxidationofcarbides,graphitisationandlocalised enhancedgraphitisation.Gasvelocityandthermalcyclingplaykeyrolesinpitmorphology.Thermal cyclinginducescircularcracks.Lowgasvelocityinducesthegastoevolveincrevices,duetolocaloxygen consumption.
1. Introduction
Metal dusting is a catastrophic form of corrosion affecting Ni-,Fe- and Co-based alloys when exposed tohighly carburis-ingatmospheres(ac>>1)[1,2].Thealloydisintegratesintoadust
of fine metallic particles and graphitic carbon, named “coke”. Althoughmultiple studies wereperformed to determinemetal dustingdegradationmechanism,thiscomplexprocessisstillnot wellunderstood[3,4].Thecombinedinfluenceofthecomposition ofthealloy[3,5,6],ofitsmicrostructure[7,8]andofthe environ-ment[9–12]explainsthesedifficulties.Multiplemechanismswere proposedforferritic[2,4,13,14]andaustenitic[5,15]materials.It isusuallyagreedthat acementitescale formsatthesurface of low-alloyedferriticmaterials.Hochman[1,2]andGrabke[10,16]
proposeditdecomposesintoironandcarbon,leadingtothescale andalloydegradation,whileYoungetal.[4]proposedthegraphite nucleationatthesurfaceandatthedefectsofthiscementitescale leadstometaldusting.Howevertheauthorscorrelatedthemetal dusting degradation rateof low-alloyed austenitic materialsto theepitaxialrelationshipbetweengraphiteandaustenite[17–19]. EarlyworksbyGrabkeandcoworkers[19,20],Pippeletal.[17]and Chunetal.[18]proposedthatmetaldustingoccursbytheinward growthofgraphitewhosebasalplanesareapproximatelynormal
∗ Correspondingauthor.
E-mailaddress:daniel.monceau@ensiacet.fr(D.Monceau).
tothesamplesurface.SeveralyearslaterZengandNatesan[15]
putforwardthatgraphitenucleationoccurredinside the metal-licalloysatthemostcatalyticinterfaces.Finally,Zhangetal.[21]
proposedthatmetaldustingoccurredbythegraphitenucleation andgrowthalongthecatalyticplansintersectingtheinitial sur-face.Althoughthereisnogeneralagreementonthemetaldusting corrosionmechanisms,theywereextendedtoCr2O3-andAl2O3
-formingalloysbutlocally,atdefectsoftheprotectiveoxidescale
[4,22,23].However,internaloxidationcouldalsoplayakeyrolein thedegradation[24–26].Thestudyofthecorrosionmechanismsof Cr2O3-orAl2O3-formingalloysweremadeeitheratatmospheric
pressureattemperatureshigherthan600◦C[5,13,14,24–26]orat highpressureat540◦Ctoreproduceindustrialconditions[26,27]. Unfortunately,gasflowrateishardlyspecifiedinthestudies detail-ingcorrosionmechanisms,anditseffectisnotdeveloped,despite thefactthatthegasmixturesareoutofequilibrium.Combiningthe fewstudieswhichobservedtheinfluenceofgasflowrateonthe kineticsofcarbondepositionormetaldusting[28–30],itcan any-waybeconcludedthatgasvelocityinfluencesdegradationkinetics uptoamaximumflux.Indeed,whengasvelocityishighenough, thefrequentrenewalofthegasmixtureatthesurfaceofthe sam-plekeepsthereactivegasmixtureconstantlyequaltotheinputone atthereactionfront.However,thepossibleinfluenceofgasflow rateoncorrosionmechanismwasnotreported,particularlyat rel-ativelylowtemperature.Thepresentworktendstodeterminethe
Fig.1.Opticalmicrographsof(a)theexternaland(b)theinternalfaceofa800HTsampleafter4000hofexposure.
Fig.2.Opticalmicrographofthecrosssectionofa800HTsampleafterMurakami’setchingand4000hofexposure.
pittingcorrosionmechanismsoccurringatatmosphericpressure below600◦Cunderalowgasvelocity.
2. Experimentalprocedure
A metal dusting experiment was carried out in a 47.25CO-47.25H2-5.5H2Ogasmixtureat570◦Candatmosphericpressure
on800HTsamples.Thealloyisaustenitic.Itscompositionis 29Ni-45Fe-21Cr-1.3Si-0.9Al-0.7Ti-0.3C-0.2Cu-0.2Mo-0.2Pforthemain elements,inat.%.Watervapourpressurewasestablishedby sat-uratinga 50CO-50H2 mixturewithwaterat35◦C.Thegasflow
ratewasabout0.2L/s/cm2 ofsamplesurfaceareaand thegas
velocitywasabout18m/s.Theexperimentwasconductedina heatedhorizontalquartztube.Samplesrestedonanalumina sam-pleholder.Twoidenticalsampleswerepositionedupstreamand downstreamtoconfirmthatthegascompositiondidnotevolve duringitsflow.Consideringonlyequilibrium ofsyngasreaction (CO+H2=C+H2O)and water decomposition(H2O=H2+½O2), a
carbonactivity(ac)of33andanoxygenpartialpressure(PO2)of
2.1×10−27bar,weredetermined,respectively.Thesyngas reac-tionwasusedtodetermineac,asusualinmetaldustingstudies,
becauseofitsfasterkineticscomparedtoBoudouardreaction[29]. WaterdecompositionwasusedtodeterminePO2 insteadof
car-bonmonoxidedecompositionforthesamereason [31].Sample surfacewasgroundusingP600SiCgritpaperandtheedgeswere chamfered.Thesampleswereremovedevery500h,cleaned ultra-sonicallyinethanolandweighted.Picturesofbothsidesofeach sampleweretakenateverywithdrawalbyaKeyenceVHX-1000E digitalmicroscope.The imageresolutionwasabout4m/pixel. Ramanspectroscopywascarriedouteverytwowithdrawalsusinga LabramHR800YvonJobinspectrometerequippedwithaconfocal microscope(magnificationsare10×,50×and100×).Asthe dif-ferent800HTsamplesshowedsimilarbehaviour,onesamplewas removedafteratotalexposureof4000h.Itsnetmasschangewas −4.49mg/cm2.Topographicalmappingswerecarriedoutfor
sev-eralpitsonanextendedfieldconfocalmicroscopeAltiSurf520with a2-mstepusingachromaticsensorwitha10nmaxial resolu-tion.ThesamplewasalsoexaminedwithaSEM/FIBFEIHELIOS 600icoupledwithanEDXanalyserAztecAdvancedequippedwith aSDDdetector.Analyseswereperformedusinga5keVaccelerating voltage.
3. Resultsanddiscussion
Corrosionofthesampleswascharacterisedbypitting(Fig.1). Aspitting ismorepronounced attheedgeofthesample,mass lossdoesnotrepresenttherealmaterialdegradationbutrather thesampledegradationwithasizeandshapeeffect.This hetero-geneouspittingcouldbeexplainedbyanenhancedcarburisation, internaloxidationandCr-depletionattheedgesduetoahigher surface/volumeratio.Itisalsocertainlyduetoanoxidescale con-taining more defects due to higher mechanical stresses at this location.Thefollow-upofpittingisthenamoreaccuratewayto evaluatemetaldustingcorrosionofmaterialscapabletoforma protectiveoxidescale,asperformedin[32,33].
A carburisedzone composedofM7C3 carbideswasrevealed
below thepits byMurakami’setchingoncross section(Fig.2), accordingtoVanderVoort[34].Consideringthealloycomposition, itisassumedthatthiscarbidesareiron-chromiumcarbides.This carburisedzone wasthickeratthecentreofthepits.This mor-phologywasobservedforeveryanalysedpit.M23C6carbideswere
identifiedatgrainboundariesbyMurakami’setching[34].They werealsoassumedasiron-chromiumcarbides.
It wasfoundthat,atthesurfaceof thesample, pitdiameter growsataconstantrate(Fig.3a).Thiskineticscanbecharacterised byalateralpitgrowthrateconstantkdof0.82±0.02m/handof
1.05±0.04m/hontheexternalside(facingthequartztube)and ontheinternalside(facingthecentreofthequartztube), respec-tively.Alinearlateralpitgrowthkineticshasalreadybeenreported for800HTalloyathighpressureinapreviousstudy[33].Asall pitspresentasimilarmorphology,thefollowingcharacterisation ofasinglepitatthecentreoftheinternalfaceextendstoevery pit.Pitgrowthis characterisedbytheappearanceofsuccessive corrosionrings,notedCRAtoCRE(forCorrosionRingAtoE),on
Fig.3btheCRAwasvisibleafter2000hofexposure.TheCRBand theCRC,respectivelyatthecentreandaroundCRA,weredetected after2500hofexposure.StudyofotherpitsshowedthattheCRB appearedwhenthepitdiameterwasabout250m,whiletheCRC appearedwhenthedistancebetweenthepitedgeandtheCRBwas equal.TheCRDwasalsovisibleafter2500hofexposureandwasthe mostexternalring.After4000h,theCREwasdetectedaroundthe CRD.AthingrooveseparatestheCRDandtheCRE.Thisthingroove isalwaysatabout150mfromthepitedgeand100mfromthe
Fig.3.(a)EvolutionofthediameterofthepitatthecentreoftheinternalfaceonFig.1showingexperimentaldata(squares)andlinearregression(dottedline)and(b)
opticalimagesofthepitevery500hbetween2000hand4000hofexposure.
Fig.4.TopographymappingofthepitofFig.3after4000hofexposure.#1,#2and
#3indicatewhereFIBabrasionweremade.
CRC,asobservedonnumerouspits.Carbonand(Fe,Cr)3O4spinel
oxideweredetectedbyRamanspectroscopyovertheentirepit, exceptfortheCRAandtheCRBwhereonlycarbonwasdetected. Thealmostperfectcircularityofthepitsandthecorrosionrings showsthattheirgrowthdoesnotdependonthemicrostructureof thealloy.
Thestudyofthepittopography(Fig.4)revealedthattheCRB andtheCRCarebelowtheinitialsamplesurface,downtoabout 70m.TheCRDandtheCREareseveralmicrometresabovethe initialsurface,whereastheCRAisatabout20mabovetheinitial surface.Thethin groovebetweentheCRDandtheCREisa few micrometresbelowthem.Thesamecharacteristicswereobserved onotherpits,thenumberofcorrosionringsdependingonthetime spentsinceincubationofthepit.Theobservationsandassertions proposedfurtherinthispaperforasinglepitarethenextendable toeveryobservedpit.
FIB analysis carried out at the edge of the pit (see #1 on
Fig.4)showsinternaloxidationoccurringbelowaCr-depletedarea (Fig.5a).GraphitisationoftheCr-depletedalloyintheinternal oxi-dationzonebut alsobeneaththeoxidescaleisalsovisible.The Cr-depletedzonebeneaththeoxidescalehasathicknessofabout 1mandaveryfinerecrystallizedmicrostructure.Itisclearthat internaloxidation pushesthe protectivechromia scale upward. Thisis due to thelargevolume increase when Crcarbides are oxidised.Indeed,thevolumeexpansionofthealloyduetoCr7C3
formationisabout5%consideringallthechromiumreacted.The volumeexpansionofthealloyduetocarbideoxidationintoFeCr2O4
isabout48%,i.e.41%comparedtothecarburisedalloy.The inter-naloxidationthicknessbelowtheCr-depletedalloyincreasesas theresultofsolid-statediffusionofoxygeninthemetallicmatrix.
Thus,itfollowsaparaboliclaw[35].Inthesametime,theinternal oxidationthicknessdecreasesbythematerialremovalontopof itresultingfromthevolumeexpansionduetocarbideoxidation. Carbideoxidationoccursattheinternalcarburisation/internal oxi-dationinterface.Ontopofit,theremovaloftheCr-depletedalloy andtheoxidescale,consecutivelytovolumeexpansion,takesplace attheinternaloxidation/atmosphereinterface,wheretheinduced stressismaximised,seeFig.5aPitlateralgrowthoriginatesfrom thisremovalandischaracterisedbythelateralpitgrowthrate con-stantkd,explicatedearlier.Thecombinationofsolidstatediffusion
andmaterialremovalleadstothefollowingrateequation: dxox
dt = kox
xox −kd
(1) withxox theinternaloxidation thicknessbelowtheCr-depleted
zone,koxtheparabolicinternaloxidationconstantandtthe
dura-tion of exposure. This equation shows that the thickness xox
increasesuntilitreachesaconstantvalueatthesteadystate,xSS ox, i.e.whendxox/dt=0: xSS ox=k ox kd (2) Basedonobservationsthisthicknessiscomprisedbetween3and 5m. Combining the xSS
ox and kd values measured in the
cur-rentstudy,theparabolicinternaloxidationconstantiscomprised between 2.8×10−12 and 1.5×10−11cm2/s. This experimental
valueiscomparedwiththetheoreticalvaluecalculatedusingthe internaloxidationkineticmodel[35,36]andconsideringchromium andirondiffusionmuchslowerthanoxygendiffusionandFeCr2O4
astheinternaloxide.Thisleadtothefollowing: kox=ε3 4 Deff OC s O 1 3CFe0 + 2 3CCr0 (3) withalabyrinthfactor(takenequalto1)C0
FeandCCr0 thebulk
ironandchromiumconcentrationsrespectively,andCs
Othe
oxy-gen concentration within the Cr-depleted alloy at the internal oxidation/atmosphereinterface.Cs
OwascalculatedusingSievert’s
lawandafreeenergyofoxygensolubilisationof−129.5kJ/mol− obtainedwithamixturelawbetweenpureNi[37]andpure␥-Fe
[38]andconsideringtheFe/Niratioofthe800HTalloy.The effec-tiveoxygendiffusioncoefficient,Deff
O,wasdeterminedusingHart’s
formula[39]: Deff O =
1−q Lı Db O+ q LıD gb O (4)whereqistheshapefactorofgrains,consideredascubic(i.e.aq valueof3),Listhegrainsizeofthealloy(whichisabout500m),ı
Fig.5.SEMmicrographinBSEmodeofFIBcrosssections(a)#1attheedge(CRE),(b)#2betweenCRAandCRCand(c)#3atthebottomofCRCofthepitonFig.3after
4000hofexposure.
isthewidthofgrainboundaries(takenequalto1nm),Db
oisthe
oxy-gendiffusioncoefficientinthebulk.Itsvalueof1.2×10−10cm2/s
wasdeterminedbyalinearmixturelawbetweentheoxygen diffu-sioncoefficientsinpureNi[37]andpure␥-Fe[38]withtheFe/Ni ratiocorrespondingtotheoneof800HTalloy.Dgb
o wasestimated
consideringitwasabout5ordersofmagnitudehigherthanDb o.The
obtainedvalueofDeffo ,equalto2.0×10−10cm2/s,isabouttwotimes
higherthanDbo.BasedonEq.(3),avalueofkox=2.3×10−11cm2/s
isobtained.Thisvalueishigherbutinreasonableagreementwith theexperimentalone−2.8×10−12to1.5×10−11cm2/s,
consider-ingthefinermicrostructurebeneaththeoxidescaleisnottaken intoaccountin calculation,inadditionwiththelackof dataat thislowtemperature,particularlyfordiffusionatgrainboundaries. TheuseofalinearmixturelawtodetermineCsOandDeffo atthis
lowtemperatureisalsoasourceoferror.Moreovergraphitisation occursinthesametimeasoxidation,increasingpitlateralgrowth (Fig.5a).Finally,thetheoryusedtocalculatetheparabolicconstant kox considersprecipitationofinternaloxideandnotoxidationof
internalcarbides.Theagreementbetweencalculatedand experi-mentalparabolicoxidationconstantsallowstoconcludethatpit lateralgrowthkineticsisconsistentwithacontrolbythekinetics ofinternaloxidationofcarbidesbelowtheCr-depletedzone.The detectionof(Fe,Cr)3O4asinternaloxideinsteadofthemorestable
Cr2O3canbeexplainedbytheoxidationofthe(Fe,Cr)7C3carbides
inaCr-depletedmatrix.Thereleasedcarbonisexpectedtodiffuse inwardtoformnewcarbides,asobservedbyMeijeringetal.[40]
andotherstudies(seeRef.[31]).
Onecannotice that thecalculated parabolicconstant ofthe internaloxidationkineticsisclosetotheonesgenerallyobserved underanoxidescale.Indeed,theoxygenpartialpressuresusually observedinmetaldustingenvironmentsareclosetotheones corre-spondingtothemetal/oxideequilibrium.Forinstance,theoxygen
partialpressureof theinputgasmixtureis PO2=2.1×10−27bar
whichisonlyoneorderofmagnitudelowerthantheoxygenpartial pressureoftheFe/FeOequilibriumdeterminedviaanEllingham magnitudelowerthantheoxygenpartialpressureoftheFe/FeO equilibriumdeterminedviaanEllinghamdiagram:PFe/FeOO
2 ≈10
−26
bar.Theinternaloxidationkineticsinthisstudyisthenequivalent totheoneunderaFeOoxidescale,consideringtheNi/Feratioof the800HTforthematrix.
FIBanalysiscarriedoutattheinterfacebetweentheCRAandthe CRC(Fig.5b,#2onFig.4)showsa8m-thickcontinuousinternal oxidationlayerintheCRA.Althoughtheinnerzoneofthislayeris composedofspinelandCr-depletedalloy,theouterzoneis com-posedofspinelandcarbon.ItisconcludedthattheCr-depleted alloyisgraphitisedbythehighlycarburisingatmosphere.Athick carbon depositisvisible ontopofthealloy.Thisadherent car-bonexplainswhytheCRAishigherthantheothercorrosionrings, asobservedonFig.4.Itisalsoconsistentwiththecarbonsignal obtainedbyRamanspectroscopy.
AtthebottomoftheCRC(Fig.5c,and#3onFig.4)andonthe slopingcorrosionfrontvisibleonFig.5b,itcanbeseenthatthe thicknessoftheinternaloxidationlayeristhinnerthandescribed previouslyforothercorrosionrings.Itisalsodiscontinuous.Locally, carbondepositisindirectcontactwiththecarburisedalloy.This morphologyisverydifferentfromwhatisobservedfortheCRAand theCRB.Suchdifference,associatedwithafasterinwardgrowthof thecorrosionring,showsthatthedegradationmechanisminthe CRCdiffers.Thislocaliseddegradation,occurringonlyinthis“deep” ring,couldbeexplainedbythelackofatmosphererenewalinthis area.Oxygenconsumptionbyinternaloxidationcoupledwiththe lackofgasrenewalleadstoareducedoxygenconcentrationinthe gasphaseandatthemetalsurface.Thiscouldexplainwhythe inter-naloxidationlayeristhinnerthanattheplanarsurfaceofthepit
Fig.6.Schemeoftheproposedfour-stepmechanism:(a)oxidationofcarbides,thearrowsrepresenttheoutwardpushoftheoxidescaleduetointernaloxidation;(b)
graphitisationoftheCr-depletedalloy;(c)enhancedgraphitisationatthebottomofthethermalcrack(themechanismisdetailedintheinset);(d)lateralgrowthbyinternal
oxidationandinwardgrowthbyenhancedgraphitisationatthebottomofthecracks,leadingtonewdeepcorrosionrings.CorrosionringsA–Earerecalledforeasycomparison
withFig.3.
andwhyitisdiscontinuous.Inaddition,thelackofoxygendoesnot allowtheformationofainhomogeneousinternaloxidationlayer. Thisresultsinadirectexposuretothecarburisingatmosphereof theCr-depletedalloyoftheinternaloxidationandofthecarburised alloy.TheCr-depletedalloyisthengraphitisedinagreementwith theFe-Cr-C phaseequilibrium.Moreover,thelocaloxygen con-sumptionleadstoahighersurfacefractionofalloyavailableforCO andH2adsorption,enhancingthegraphitisation.
TheFIBstudyofthethingroovebetweentheCRDandtheCRE showsamorphologysimilartotheoneobservedatthebottom oftheCRC.Asnoticedpreviously,this thin grooveis alwaysat about150mfromtheedgeofthepit,andatabout100mfrom theCRC.Asalreadyexplained,thealmostperfectcircularityofthe grooveanditssizelargerthantheoneofthealloygrainsreveal thegrooveshapeisnotcontrolledbythealloymicrostructure.One couldthinkthatthisgroovefindsitsorigininthethermalcycling every500h.Toconfirmorinvalidatethis,themechanical proper-tiesandtheevolutionofthecoefficientsofthermalexpansion(CTE) withtemperatureof800HTalloy[41]andof(Fe,Cr)3O4[42]haveto
beconsidered.Atthebeginningofthecooling,theCTEofthe inter-naloxidationzone−determinedusingamixturelawbetweenthe CTEsof800HTand(Fe,Cr)3O4–isinitiallyhigherthantheCTEofthe
800HTalloy.Thiscanbesurprising,butitisduetothefactthat theCTEofthespinelishigherthantheoneofthemetallicalloyat 570◦C.Samplecoolingthengeneratestensilestressintheinternal oxidationzone.Thistensilestressisexpectedtoreachits high-estvalueatatemperaturearound370◦C,whentheCTEsof800HT
andoftheinternaloxidationzonebecomeequal.Thisleadstothe localisedrupturewherethestressismaximised.Thetensilestress necessarytoinitiatethecrackwithintheinternaloxidationlayer happenswhenthewidthoftheCRD–i.e.thedistancebetweenthe edgesofthepitandtheCRC–islongenough.ForaCRDwidthof 250m,thestressismaximisedatabout150mfromtheedge ofthepit.ItsplitstheCRDintwonewcorrosionrings:aninner ring(theCRD)andanouterring(theCRE)—seeFig.3bafter3500h. Below370◦C,theCTEof800HTishigherthantheCTEofthe inter-naloxidationzone.Thestressthendiminishesduringtheresidual cooling.Thecrackclosesandresultsinthethingrooveobserved atroomtemperature.Attheendofthecooling,thereisalmost noresidualstress.Whenreheated,thedefectopensandexposes thecarburised alloybeneaththeinternal oxidelayerdirectlyto theatmosphere.Consequently,theenhancedgraphitisation phe-nomenondevelopsinthisgroove,creatinganewdeepcorrosion ring.Thiswasconfirmedforpitsexperiencingfurtherexposure.
Theproposedexplanationisvalidonlywhenthereisaperfect adhesionbetweentheoxideandthematrix.Yet,asinternal oxi-dationleadstocompressivestresses,whichcouldbemuchhigher thanthetensilestressobtainedbysamplecooling,andCTE mis-match,furtherstudyisnecessary.Nevertheless,itisreasonableto thinkthatthecompressivestressesduetointernaloxidationcan berelaxedbycreepduringthelongdwellsathightemperature, leadingtothethickeningoftheinternaloxidationzone perpendic-ularlytothesurface,whereastensilestressduetoCTEmismatch havenotimetorelaxduringcooling.
Asnoticedpreviously,theformationofthedeepCRBatthe cen-treoftheCRAonFig.3bissupposedtohappenwhenthepit−i.e. theCRAatthistime−reachesadiameterofabout250m.The mechanismgivenpreviouslyforthethingroovecouldalsoapply atthecentreofthepit,i.e.theCRA.TheCRCisalsoexpectedto originatefromthesamephenomenonoccurringtotheCRAonce itreachesagainawidthofabout250m.Theenhanced graphiti-sationmechanismexplainsthedepthoftheCRBandtheCRC.The CRB,theCRCandthethingroovearethenthreestatesofthesame degradationphenomenon.
4. Pittingcorrosionmechanism
Anoverall corrosionmechanisminfourmainstepscannow beproposedtoexplainourobservationsofthepitcorrosion fea-tures.First,carbondiffusesthroughthechromiascaledespitethe protectivepropertiesofthisoxide.Indeed,WolfandGrabke[43]
proposedcarbondiffusedthroughprotectiveoxidescalesinpores andcracks.ZengandNatesan[44]proposedthatdiffusionpaths forcarbonweremicro-channelsofmetallicorcementiteparticles withintheoxide,whileRöhnertetal.[23]proposedcarboncould diffusethrough graphite encapsulatedin the oxidescale. More recently,Youngetal.[45]showedcarbondiffusingthroughaCr2O3
scaleusingatomprobemicroscopy.Afteritsdiffusionthroughthe oxidescale,carboncontinuesitswayinthemetalandcarburises thealloybeneaththeCr-depletedzone–stemmedfromthe forma-tionoftheprotectiveoxidelayer–,forming(Cr,Fe)7C3carbides.As
suggestedbyRöhnertetal.[23],thevolumeexpansionofthealloy duetocarburisation(+5%)couldalsoinducecracksintheoxide scale,allowingthedirectaccessoftheatmospheretothemetallic surface.
Secondly,oxygendiffusesinthealloythroughthedefectsin the oxide scale and oxidises internal (Cr, Fe)7C3 carbides into
FeCr2O4spineloxide(Fig.6a).Thisstepisexpectedtohappenwhen
chromiumconcentrationinthematrixisloweredduetochromium trappingbycarburisation.Oncetheconcentrationistoolowto sup-portexternaloxidation,internaloxidationofcarbidestakesplace
[46].Theresultinglocalvolumeexpansionoftheinternal oxida-tionzone(+41%)disintegratestheoxidescaleandtheCr-depleted alloy.Thisstepcanbeconsideredasthenucleationofthepit.This removalofmaterialexposestheinternaloxidationzone,andthe Cr-depletedzonebeneaththeprotectiveoxidescale,attheedgeof thepitinitiationtotheexternalatmosphere.Thelateralgrowthof theinternaloxidationzonebelowtheprotectiveoxideinducespit lateralgrowth,pushingupwardtheexternaloxidescale.This phe-nomenonisaccentuatedbythegraphitisationoftheCr-depleted alloybeneaththeoxidescalebecauseitcausesanadditionalvolume expansion.ItalsoreducesthethicknessoftheCr-depletedalloy intowhichtheoxygenhastodiffusetoreactwithnon-oxidised carbides.
Thirdly,the Cr-depletedalloyin theinternal oxidation zone isalsographitised(Fig.6b).Moreover,acrackformsthroughthe internaloxidationzoneduetotensilestressduringsamplecooling, onlyifthepitwidthisgreaterthanacriticalsize.Whenreheated, thecrackopens,exposingthecarburisedzonebeneathit.The Cr-depletedalloyisthendirectlyexposedtotheatmosphere.Atthe surfaceofthenewlyexposedCr-depletedalloy,theatmosphere isnotrenewed,duetoamicroclimateinthecrevice.Thelackof gasrenewalresultsina loweroxygenpartialpressure, hencea lowerinwardoxygenflux.Thisgeneratesathinand inhomoge-neousinternaloxidationlayerexposingtheCr-depletedalloyof theinternaloxidationzoneandofthecarburisedalloy,tothelocal atmosphere.Inthesametime,oxygenconsumptionincreases par-tialpressuresofCOandH2.Theadditionofthetwophenomena
leadstoanenhanced graphitisation,hencea fasterlocalinward
degradation.Thisleadstotheformationofacentraldeep corro-siondiskatthecentre ofthepit(Fig.6c),correspondingtothe appearanceoftheCRBatthecentreoftheCRA(Fig.4).
Finally,ifthediameteroftheexternalringisgreaterthanthe criticalvalueduringafollowingcooling,anewcrackwilloccur duringcoolingatthetemperaturewhenthestressismaximised. Whenreheated,thecrackopensandthesameenhanced graphitisa-tionmechanismoccursatthebottomofthecrack,asexplainedfor theformationofthecentraldeepcorrosiondisk.Thisformsadeep corrosionring(Fig.6c).ThiscorrespondstotheCRConthestudied pit,theCRAbeingsplitinaninternal(CRAonFig.4)andanexternal corrosionring,attheedgeofthepit(CRDonFig.4).Theenhanced graphitisationmechanismfollowingcrackformationleadstothe formationofnewdeepcorrosionringsduringthecoolingifthe externalcorrosionringhasreachedacriticalwidth,asobservedon olderpits.Ithastobenoticedthatthe250mwidthmeasuredin thisstudyisnotthiscriticalsizebutcorrespondstothepitgrowth after500hofexposure.Thebeginningofathirddeepcorrosionring isthethingroovebetweentheCRDandtheCRE(thelatterbeingthe externalpartofthecrackedCRD)(Fig.4).Thefinalpitmorphology after4000hisschematisedonFig.6d,basedonFig.4.Fortheoldest pits,i.e.forthelongestdurationsofexposure,thedeepcorrosion ringsmergetoformamassivecentraldeepcorrosiondisk.
Theproposedmechanismcanbeviewedeitherasanentirely novelmechanismspecifictoourexperimentalconditions,oras theelementaryreactionstepsoftheoverallobserveddegradation mechanismsinvolvinginternaloxidation, withmuch highergas flowrate[25,27].Ilhasalsotobenoticedthatasimilar morphol-ogycanbeobservedinpreviousworks,althoughthishasnotbeen mentionednorstudiedintheconsideredpapers [30,47,48].The numberofringsthatcanbeobservedinthesearticlesare differ-entthaninthecurrentstudy.Thenumberofringsinthisliterature datacannotbequantitativelycomparedwithourmodellingsince pitincubationtimeandpitlateralgrowtharenotgiven.
5. Conclusion
Apittingmechanismisdescribedforsamplesof800HTtested undermetaldustingconditionsat1barinCO-H2-H2Omixture,and
withdrawnevery500hfor characterisation.The pitting mecha-nismisbasedoninternalcarbideoxidationandlocalisedenhanced graphitisationoftheCr-depletedalloy.Itcanbedividedin4steps: 1)Pitnucleationandgrowthduetothebreakdownoftheoxide scaleinducedbythelargevolumeexpansionresultingfromthe oxidationofpreviouslyformedcarbides.Thelocalisationofthe attackisduetothepresenceofdefectsintheprotectiveoxide scale.
2)GraphitisationoftheCr-depletedmatrixoftheinternal oxida-tionzoneandtheonethatstemmedfromtheexternaloxide scaleformation.
3)Enhancedgraphitisationduetoamicroclimateatmospherewith loweroxygenandhighercarbonactivitiesatthebottomofa crackinducedbytensilestressintheinternaloxidationzone duringcooling.
4)Pit lateral growth controlledby the kinetics of oxidation of thecarbideswhereasthepitinwardgrowthis controlledby theenhancedgraphitisationatthebottomofthecrack(whose mergingresultsinaninnerdeepcorrosiondisk).
Asitisexplainedintheproposedmechanism,themorphology ofthepitsismodifiedbythermalcycling.Indeed,thecrackfrom whichinwardgrowthoccursisformedduringsamplecooling car-riedoutforsamplecharacterisation.Thishighlightstheimportance oftheexperimentalprocedure.
Acknowledgements
ThisworkhasbeensupportedbytheFrenchnationalresearch agencythroughtheprojectANRSCAPAC11-RNMP-0016in part-nershipbetweenAirLiquide, Veolia,Sedis,University ofNancy, CIRIMATlaboratory.TheauthorsacknowledgeSabineLerouxofthe ICAlaboratoryfordailymonitoringoftheexperiment.
References
[1]R.F.Hochman,Metaldeteriorationincarbonmonoxideandhydrocarbonsat elevatedtemperatures,in:ThirdInternationalCongressonMetallic Corrosion,USSR,Mir,Mosow,1966.
[2]R.F.Hochman,Catastrophicdeteriorationofhigh-temperaturealloysin carbonaceousatmospheres,in:SymposiumonPropertiesofHigh TemperatureAlloyswithEmphasisonEnvironmentalEffects,LasVegas, Nevada,1976,pp.715–732.
[3]H.J.Grabke,Metaldusting,Mater.Corros.54(2003)746.
[4]D.J.Young,J.Zhang,C.Geers,M.Schütze,Recentadvancesinunderstanding metaldusting:areview,Mater.Corros.62(2011)7–28.
[5]H.J.Grabke,R.Krajak,E.M.Müller-Lorenz,S.StrauB,Metaldustingof nickel-basealloys,Mater.Corros.47(1996)495–504.
[6]S.Strauß,H.J.Grabke,Roleofalloyingelementsinsteelsonmetaldusting, Mater.Corros.49(1998)321–327.
[7]H.J.Grabke,E.M.Müller-Lorenz,B.Eltester,M.Lucas,D.Monceau,Resistance of9–20%Cr-steelsagainstmetaldusting,SteelRes.68(1997)179–185.
[8]H.J.Grabke,E.M.Müller-Lorenz,S.Strauss,E.Pippel,J.Woltersdorf,Effectsof grainsize,coldworking,andsurfacefinishonthemetal-dustingresistanceof steels,Oxid.Met.50(1998)241–254.
[9]R.A.Perkins,W.C.Coons,F.J.Radd,Metal-Dustingcorrosionin
coal-Gasificationenvironments,J.Electrochem.Soc.123(1976)733–749.
[10]H.J.Grabke,C.B.Bracho-Troconis,E.M.Müller-Lorenz,Metaldustingoflow alloysteels,Mater.Corros.45(1994)215–221.
[11]A.Schneider,H.Viefhaus,G.Inden,H.J.Grabke,E.M.Müller-Lorenz,Influence ofH2Sonmetaldusting,Mater.Corros.49(1998)339.
[12]E.M.Müller-Lorenz,H.J.Grabke,Cokingbymetaldustingofsteels,Mater. Corros.50(1999)621.
[13]P.Szakálos,Mechanismsanddrivingforcesofmetaldusting,Mater.Corros. 54(2003)762.
[14]H.J.Grabke,R.Krajak,J.C.NavaPaz,Metaldustingofhigh-temperaturealloys, Mater.Corros.44(1993)89–97.
[15]Z.Zeng,K.Natesan,Relationshipofcarboncrystallizationtothe metal-dustingmechanismofnickel,Chem.Mat.15(2003)872–878.
[16]H.J.Grabke,R.Krajak,J.C.N.Paz,Onthemechanismofcatastrophic carburization—metaldusting,Corros.Sci.35(1993)1141–1150.
[17]E.Pippel,J.Woltersdorf,R.Schneider,Micromechanismsofmetaldustingon Fe-baseandNi-basealloys,Mater.Corros.49(1998)309–316.
[18]C.M.Chun,J.D.Mumford,T.A.Ramanarayanan,Carbon-inducedcorrosionof nickelanode,J.Electrochem.Soc.147(2000)3680–3686.
[19]Q.Wei,E.Pippel,J.Woltersdorf,S.Strauss,H.J.Grabke,Orientation dependenceofmetaldustingnanoprocessesonnickelsinglecrystals,Mater. Corros.WerkstoffeKorros.51(2000)652–656.
[20]R.Schneider,E.Pippel,J.Woltersdorf,S.Strauss,H.J.Grabke,Microprocesses ofmetaldustingonnickelandNi-basealloys,SteelRes.68(1997)326–332.
[21]J.Q.Zhang,P.Munroe,D.J.Young,Microprocessesinnickelaccompanying metaldusting,ActaMater.56(2008)68–77.
[22]H.J.Grabke,Thermodynamics,mechanismsandkineticsofmetaldusting, Mater.Corros.49(1998)303–308.
[23]D.Röhnert,F.Phillipp,H.Reuther,T.Weber,E.Wessel,M.Schütze,Initial stagesinthemetal-dustingprocessonalloy800,Oxid.Met.68(2007) 271–293.
[24]P.Szakálos,R.Pettersson,S.Hertzman,Anactivecorrosionmechanismfor metaldustingon304Lstainlesssteel,Corros.Sci.44(2002)2270.
[25]M.Hansel,C.A.Boddington,D.J.Young,Internaloxidationandcarburisationof heat-resistantalloys,Corros.Sci.45(2003)967–981.
[26]P.Szakálos,M.Lundberg,R.Pettersson,Metaldustingonanaluminaforming Ni-basealloy,Corros.Sci.48(2006)1695.
[27]J.Z.Albertsen,Ã.Grong,J.C.Walmsley,R.H.Mathiesen,W.Beek,Amodelfor high-Temperaturepittingcorrosioninnickel-Basedalloysinvolvinginternal precipitationofcarbides,oxides,andgraphite,Metall.Mater.Trans.A39 (2008)1258–1276.
[28]P.L.Walker,J.F.Rakszawski,G.R.Imperial,Carbonformationfromcarbon monoxide-hydrogenmixturesoverironcatalysts.II—ratesofcarbon formation,J.Phys.Chem.63(1959)140–149.
[29]E.T.Turkdogan,J.V.Vinters,Catalyticeffectofironondecompositionof carbonmonoxide:i.carbondepositioninH2-COmixtures,Metall.Trans.5 (1974)11–19.
[30]J.C.Nava-Paz,H.J.Grabke,MetalDusting,Oxid.Met.39(1993)437–456.
[31]D.J.Young,HighTemperatureOxidationandCorrosionofMetals,Elsevier, Amsterdam,2008.
[32]S.K.Varma,A.Putrevu,M.Pasala,Z.Zeng,K.Natesan,Metaldustingand oxidationat593and704◦C,THERMEC2006(2007)4226–4231,Pts1–5.
[33]A.Fabas,D.Monceau,S.Doublet,A.Rouaix-VandePut,Modellingofthe kineticsofpittingcorrosionbymetaldusting,Corros.Sci.98(2015)592–604.
[34]G.VanderVoort,E.Manilova,J.R.Michael,Astudyofselectiveetchingof carbidesinsteel,Microsc.Microanal.10(2004)76–77.
[35]R.A.Rapp,Kineticsmicrostructuresandmechanismofinternaloxidation—its effectandpreventioninhightemperatureoxidation,corrosion,NACE21 (1965)382–401.
[36]C.Wagner,Reaktionstypenbeideroxydationvonlegierungen,Zeitschrift Elektrochem.63(1959)772–790.
[37]J.-W.Park,C.J.Altstetter,Thediffusionandsolubilityofoxygeninsolidnickel, Metall.Trans.A18A(1987)43–50.
[38]J.Swisher,E.T.Turdogan,Solubility,permeability,anddiffusivityofoxygenin solidiron,Trans.Metall.Soc.AIME239(1967)426–431.
[39]E.W.Hart,Ontheroleofdislocationsinbulkdiffusion,ActaMet.5(1957), 597–597.
[40]J.L.Meijering,Internaloxidationinalloys,in:H.Herman(Ed.),Advancesin MaterialsResearch,5thed.,Wiley-Interscience,NewYork,1971.
[41]INCOLOY®alloy800H&800HT®in,SpecialMetalsCorporationAlloy
specifications,(2004).
[42]J.Robertson,M.I.Manning,Criteriafortheformationofsingle,duplexand breakawayscalesonsteels,Mat.Sci.Technol.4(1988)1064–1071.
[43]I.Wolf,H.J.Grabke,Astudyonthesolubilityanddistributionofcarbonin oxides,SolidStateCommun.54(1985)5–10.
[44]Z.Zeng,K.Natesan,Initiationofmetal-dustingpitsandamethodtomitigate metal-dustingcorrosion,Oxid.Met.66(2006)1–20.
[45]D.J.Young,T.D.Nguyen,P.Felfer,J.Zhang,J.M.Cairney,Penetrationof protectivechromiascalesbycarbon,Scr.Mater.77(2014)29–32.
[46]C.S.Giggins,F.S.Pettit,Corrosionofmetalsandalloysinmixedgas environmentsatelevated-temperatures,Oxid.Met.14(1980)363–413.
[47]Y.Nishiyama,N.Otsuka,T.Kudo,MetaldustingbehaviourofCr-Nisteelsand Ni-basealloysinasimulatedsyngasmixture,Corros.Sci.48(2006) 2064–2083.
[48]C.Rosado,M.Schütze,Protectivebehaviourofnewlydevelopedcoatings againstmetaldusting,Mater.Corros.54(2003)831–853.
