ADVANCEDENGINEERING.MATERIALS

SupportingtheModeIofDuctilelranDendrilicSolidification**

ByHenrique M. C. M. Santos, * A. M. P.Pinto,M. C. P. L.Jacinto, and C. P.M. Só

The literature about graphite formation in ductile iron pro-poses the experimentally supported theory lhat this phaseforms in the melt.[1-4]

This is naturally accepted as a consequence af the euteclicreaction of the iron-carbon equib"briumdiagram. This reaction

rOOJ Dr. H. M. C. M SanfosMetallurgyand M1Jteria1sDepartmentEngineering School

Porto UniversityPorto (Portuglil)Email: hsanfoS@fe.up.ptDT. A. M. P. Pinto

Mechanical Engineering DepartmentMinho UníwrsityMinho (Portugal)

0,. M. C. P. L. Jacinto

Mochtmical Engineering DepartmenfPorto Polytechnic InsHtute lind INEGl

Porto (Portugal)Dr. C. P. M. SdMateriais Centre

Porto UniTJeTSityPorto (Portugal)

rMJ Thisworkispari oftheprojedPRAXlS XXI 3/3.1/CEG/2Jj66/95financed by the Pcn1uguesePRAXlS progmm. 'I'hDnJesare due toFERESPE for having provided some of the material used in theexperiments. TIrePorto Mining and GeolDgylnstitufe is thankedfor having provided facilities for someof fhe microanalysis.

514 1A38-1656/00/0808-0514 $ 17.50+.50/0 ADVANCED ENGlNEERlNG MATERIALS2000. 2, No. B

- _.0______

ADVANCEDENGINEERING.MATEIIALSSantos, Pinto, Jacinto, SójSupportina the Modal of Ductile Iron Dendritic

requires that during cooling, at 1150-1100°C, the liquid with aeutectic composition decomposes into two solid phases, puregraphite aOOaustenite with a earbon content dose to 2 % inweight. li the ductile irons have a generaIly eutectic composi-tion, the presence of neither pro-eutectic graphite nor auste-nite seems necessary for eutectic graphite formation.

According to the proposed solidification path, the graphitenodules become encapsulated by an austenite case, resultingin a 10ss of conmct between liquid and graphite. Carbon willthen have to diffuse from the liquid through the austenite tothe graphite to allow further growth 01both solid phases.

This growth mechanism is responsible for a specwc segre-gation pattern at lhe scale oí the eutectic ce1l.During solidifi-cation. a carbon gradient has to exist against the graphitephase; this composition gradient constitutes the drlving forcefor carbon diffusion. This means that the caJbon content mustincrease with the distance to the nodule, achieving its equili&-rium content in the homogeneous liquid far from the graphitenodule. In fact, the liquid surrounding the austenite shell hasto be carbon enrlehed, otherwise diffusion will nol occur. 50 acarbon-rlch liquid will be the last melt to solidify and will belocated at the eutectic cel1boundaries.

Simultaneously with carbon segregation, alIoying elementsalso play a role in chemical partitioning: some will remain inthe liquid phase and wiIl concentrare in the last regions toso-lidify and olhers will migrate to the solid matrlx surraundingthe graphite nodule, being associated with the early-solidifiedregions. Carbide-forming e1ements like Mn, Cr, and Mo wi1lsegregare (positively) to the eutectlc ceUboundarles. Non-car-bide forming e1ements like Si, Cu, aOONi will segregate (neg-ative1y) to the so-called graphite/matrix interface. An effec-tive partition, or segregation, coefficient " is defined for eachelement as the ratio between the coment af the element in tOOgraphite nadule interface and the content of the sarne elementin the eutectic ceUboundary; k is greater than unity for ele-ments with negative microsegregation and less than unity forelements with positive rnicrosegregation.

Ali these issues were established long aga in the literatureand are supported by numerous analytical techniques. [4.5-31

lf this madel reflects reality, the described made of micro-segregation would be detected in the matrix joining two adja-ceM graphlte nodules, where it intercepts a cel1boundary.

We have been dealing with this microsegregation questionfor a long time, being engaged for more than ten years in aU&-tempered ductile iron (AO!) studies, in whieh microsegrega-tion phenomena are important due to their effect on the ld-netics of austenite decomposition in the bainitic domam.Some results obtained did not appear to be coherent with themadel presented above.

A recent publication proposing a new approaeh to ductileiron solidificationl91induced us to recall some data from pre-vious work soo to conduet some new experiments to anaJyzethis particular problem in more detail. These facts, and theirinterpretation in the light of the new ideas about ductile ironsolidification. are presented in this communication.

In this work several ductile iron alloys have been studied.They differed in composition and in the processing routes.Two of the aIloys (AI and A2) were supplied by an iron foun-dry and the olher twa (A3 and A4) were produced under la&-oratory conditions.

The alloys A1 and A2 were electricaDymelted in a furnaceof 150 kg capacity with a base composition suitable for ferriticductile iron production and treated by the "fIow-through"process; Cu and Mn were added to the rnelt after nodularizingit, in the form of electrolytie CUand Fe-Mn (75 wt.-% Mn). Al-Ioys A1 and A2 were cast in resin-bonded sand in the form Dfcylinders. AI cylinders were 25 mm in diameter and 200 mmlong; A2cylinders were 80 mm in diameter and 300 mm longo

Themeltingof alloysA3and A4was coOOuctedin a core-less induction fumace of 100 kg capacity. The melt was trea-ted by a sandwich processoAlloys A3 and A4 were cast in re-sin-bonded sand and gravity die-cast. In gravity die-casting,two inoculation stages were performed. Y-blocks 25 mmthick were cast in open molds made of furanic resin sand. Forgravily die-casting, a permanent die with an area of 365 x500 mm2 and maximum thickness of 30 mm was used; cylin-ders of 200 mm in height and 30 mm diameter were poured.

Table 1 displays the average composition of the aJloys, ob-tained by atómic emíSSÍonspectrometer ana1ysis on cbill castdiscs of several samples.

The graphite nodularity of ali alloys is better than 85%.The nodule density of the sand-aOOdie-cast aJ]oys is about300 nodules/mm2 and 2000 nodu1es/mm2 respectively. Dif-ferent fractions of pearlite and femte were identified as theconstituents af tOOmeta1licmatrix in the as-cast state.

The microsegregation of Si, Cu, and Mn was evaluatedusing two different approaches. A scanning electron micro-scope equipped with an energy dispersive spectrometer wasused to characterize the distribution in alloys AI and A2;a mi-croprobe equipped with a wavelength dispersive spectrometerwas used to characterize the distrlbution in alloys A3 aOOA4.

The microsegregation studies were perforrned aver trans-verse sections.

The regions to be microanalyzed were se1ected using anirnaginary straight line defined between two adjacent no-dules. This line was divided in several equal segments andthemícroanalysiswasperformedat theend of eachsegmento

As the intemodular distance was not constant, it was nor-malized. At least ten internodular distances, from differentfields of view, were studied per samp1e, corresponding tosome hundreds of ana1yses.

Tubl/! 1. Alloymmpasition (,ut.-'1ó).

ADVANCED ENGINEERtNG MATER1ALS2000, 2, No. 8 51S

AUoy C Si MIl Cu

AI 3.82 3.30 0.56 0.96

A2 331 2.50 0.56 1.(19

A3 3A1i 3.06 (lU 0.30A4 3.43 3.04 0.55 0.92

ADVANCEDENGINEERINGMATEIIALS Santos, Pinto, Jacinto, S6/Supporting the Model of Ductile Iron Dendritic

This means that the profile of the Si/Mn and Cu/Mn raliasversus the distance to the graphite matrlx interface should bea U-shaped curve. Our resu1ts c1early contradict this state-ment: in a great number of cases the Si/Mn and Cu/Mn ra-tios versus the graphite matrix interface distance do not givea U-shaped curve. In some cases, there is Si and Cu enrich-ment at the interface wbile in othen; there is a depletion ofthese elements and a Mn enrichment. This behavior is similar

for alI alloys studied in this work.In the face of this peculiar result, several more nodule/

matrix interCacesin alloy A2 were submitted to microanaly-sis; in each interface four measurements were performed at900 to each other, with the aim of confirming the occurrence

On the a110ysAI and Al, Si, Mn, and Cu contents weremeasured using SEM/EDS (Jeo135C/Noran-Voyager), acquir-ing data at each point aver the 2 x 211m2probe. The microa-nalysis was conducted at 15 keV excitation voltage and theacquisition time was 2 min.

Camebax equipment was used for the microanalysis of al-loys A3 and A4 using an accelerating voltage of 15 keV and abeam cunent of 21 nA. Reference standards of pure elementswere used for the quantitative analysis.

For both techniques the weight percentage of each eltmtentwas calculated using a program accounting forZAF corrections.

Once silicon and copper segregate negatively during thesolidification of ducnle iron and manganese segregates posi-tively, one may assume that the high-er concentration of Si and Cu are as-sociated with the lower Mn contentoTherefore, our microanalysis resultswiII be presented in a way that shows

this relationship; the réltios Si/Mn ~ 10and Cu/Mn are plotted versus Mn incontento Figure 1 shows these rela-tionships for alloy AI.

The results show that there is a

trend for high concentrations of Siélnd eu to be associated with low con-centrations of Mn and vice-versa.This behavior is common to all four

aDoys, as shown in FIgure 2. whichpresents ali the col1ecteddata.

According to the classical acceptedsolidification model described above,the solidification of the metallic ma-

trix begins dose to the graphite no-dules, enveloping them and avoidingtheir contact with the liquid phase;tbis implies that the highest levels ofSi and Cu are associated with the re-gions that first solidify. The graphitematrix interfaces are then Si and Cuenriched and Mn depleted, whi1st thelowest contents of Si and Cu are asso- Fig.2. RLlaJilllo!5i/MnarrlCu/Mr1vs. Mnfora/lfouralÚ1!fS.

dated with the last regions to solidifya.e., the eutectic cell boundaries),which are supposed to be Mn en-riched.

Figure3 disp1aystbe Si/Mn and cCu/Mn evolution along several inter- ~ 4nodular lines for al10y A2. In agree- U'Jment with tbe microsegregation mod-el, the higher Si/Mn and Cu/Mnratias must be associated with re-

gions dose to tbe graphite matrix in-terface and the lower figures of thoseratios must be associated with eutec-tic cen boundaries.

20

..15 .

so0.0 0.5 1.0

"eMn1.5

F/g. 7. RLlallon 01Si/M" and Cu/MIl vs. Mil/li lhe CA~ ai allO!jAt.

40

."li'

10

o0.0 0.5 %Mo 1.0 I.S

8

2

o0.0 0.2 0.4 0.6 0.8

NormaIisaIDislancet.o

8

6

2

o0.0 1.0 t.S0.5

12

9

3

5 1.0O. %Mn 1.5

4

3

~2u

0.2 0.4 0.6 0.8Normaliscd Distante

1.0

Fig. J. Evollltüm of SilMn alld ClI/Mrl us. rlOTlllll/iz.d inlenwdu/ar distaru:e in several i"temoduh:r "'giarrs in lhe CtlSeof 11110)1A2 (t/w hOl"izontlllli~ i"dicat. lhe S/IMn and C,,/MII 110'''/''''/ rtttio of the 11111»1).

516ADVANCED ENGINEERING MATERIALS2000, 2, No. 8

ADVANCEDENGINEERINGAlATERIALSSantos, Pinto, Jacinto, Sá/SuPPOrtirl9 tt.e Model of Duclile Iron Dendritic

of Si/Mn and Cu/Mn maximum figures. Figure 4 p]ols theSi/Mn and Cu/Mn ratios in the neighborbood of the nodule/matrix interface.

The results show that the graphite matrix interface is notsystematically Si and Cu enriched and Mn dep]eted: in fad,more then 50 % aí the analyzed points present a Si/Mn ratiolower than the nominal ratio for the alloy (which is 4.45in thecase ai alloy A2). The Cu/Mn nominal ratio for the alloy 151.% and more than 30% of the ana1yzed points give lowerfigures. It was a1so verified in each graphite matrix interlacetbat there was a strong dispersion of Si/Mn and Cu/Mn ra-tio, achieving variations as high as 3:1.

Table 2 supports this verification, showing the maximumand minimum figures for the Si/Mn and Cu/Mn ratios insevera! nodule/matrix interíaces.

Nominal Ratio SiIMn = 4.5

2-3 3-4 4-S S-6 6-7 7-8 1-9 9-10 10-11

SiIMn

NominalRaIio~ =2.0

The results reported above are nol consistent with the melttheory that 5Uggeststhat graphite is fonned diredly flom themelt and soon surrounded by an austenite shell, requiringcarbon diffusion for further growth. Tbis growth model im-poses on the austenite shell a higlt content of graphitizing ele-ments (Si,Cu) and a low content of carbide forming elements(C, Mn).

Although we have confirmed that Si and Cu segregare in-versely with Mn, a clear contradiction was evident when theSi/Mn and Cu/Mn ratios were evaluated along the internod-ular path. The high Si/Mn and Cu/Mn figures were not gen-erally associated with the graphite matrix interfaces, but werepresent elsewhere. fi fad, one should find that somewherealong this path the Si/Mn and Cu/Mn ratios have mínima;quite different profiles were obtained as shown in Figure 3. Ithas been possible to verify that the minima in Si/Mn andCu/Mn ratias may occur even in the nodule/matrix interface,5Uggesting that the metaOic matrix enveloping the nodulesresults from the la5t solidified liquid, which is C and Mn en-ricbed and Si and Cu depleted.

Several microanalyses around each graphite/matrix inter-face, performed for one oí the aUoys, showed that the distri-bution of the chemical elements did not foUow the pattemforecasted by the melt theory for the graphite formation. In-tense chemical dispersion of Si/Mn and Cu/Mn was ob-served and both positive and nega tive microsegregation wasevaluated, as may be demonstrated by the data presented inTable 2. This behavior requires the solid sWTOundingthe nlrdule to be formed from liquid fractions with different campo-sitions.

In view oí these facts we suppose that the ductile iron soli.dification has to occur by another model, diHerent from themelt theory. In a recent paper by Van de Velde,I9Ja dendriticmode! for this process is proposed; the author claims that en-ouglt evidence i5 avaiJable to justify the rejection ai the tradi-tional idea af graphire formation direct1y from the melt. Inlhe past, several researchers advanced theories about soUdifi-cation oí ductile iron based on austenite dendrite furmation.namely the cementite theory and the austenite theory (pro-posed by Dunphy and Pellini and by De Syrv~OOtthey wereignored due the popuIarity af the melt theory, More recent1yStefanescu[21claimed that austenite dendrites pIay an impor-tant role in the solidification of the ductile irons, proposingthat they would form simultaneously with graphite and em-bracc the nodules, producing what the author calls "the eu-tectic grain".

According to Van de Velde,(9)ducti1eiron solidification is aprocess in which the eutectic liquid transforms into austenitedendrites am! enriches in carbon. in fact a mechanism identicalto the one universally accepted for the bainitic reaction. Withthe progress of the reaction, the liquid will finally segregatecarbon to the interdendritic interstices, amongst the dendritebranches. Due to the low specific mass of carbon, a consider-able expansíon i5 associated with its precipitation, the spherl-cal morphology being accepted as natural. This hypothesis isquite compatible with the results presented in this work.

ADVANCED ENGINEERING MATERJAlS 2000. 2, No. 8 517

1-2 2-3 3-4 4-5 S-6 6-7 7-8

CuIMn

Fig. 4. Di.tribJdionfmtuency ofSi/Mn Imd CulMn Mlioin /lU'CIlSI!oflll10yAl.

TRb1e2. SilMn rmdCulMn 1Illios(IIOminoSmtiw, mi,rimum. aM marimllm MIres)01:1m di/ferent norbde/mlllriz jnterfaces. in lhe caSl'Df tlle d110ilAZ.

Nodute/matrix interface

D 4 5 6 7 8 9 10Si/ 4.5

:;i/Mn 2.4 4.5 4.6 33 3.6 4.4 5.0 4.5 4J) 43(uUNSiIMn 7J) 8.6 9.4 7.0 83 8.6 9.1 10.6 8.3 7.46Dax)

Mn 2.0

1.6 2.1 3.8 1.6 1.9 1.9 2.5 3.1 2.1 2.3

CulMIt 5..5 53 7.1 5.6 5.9 7.7 7.4 7.1 6.8 39<-x>

ADVANCEDENGINEERINGMATERIALS

Our results require that, during soUdification, a growingnodule must be in contact simultaneously with early and lateregions to solidify. This is a reasonable way to explain thehigh dispersion of the Si/Mn and Cu/Mn ratios in thegraphite matrix interfuces and the irregular profiles for Si/Mn and Cu/Mn ratios along the intemodular regions. Whena growing nodule has in its neighborhood an early-fonneddendrite branch, its intenace wi11 be Si and Cu enriched.When the nodule touches a residual solidifying melt or alate-formed dendrite branch, its interface will be ~n en-riched. H one accepts that graphite nodules nudeate from theinterdendritic liquid instead of nudeating from the initialmelt, this explanation becomes plausible. In this case, thegrowth of graphite wi1l occur in contact with austeniteformed sooner or later in the solidification process, whichmeans that the austenite will be chemically heterogeneous,making possib1e high compositional variations of the graph-ite/matrix interface.

Figure 5 sltows the as-cast mic10stmcture of thc alloy A3die-cast, resulting f10m a special etching procedure, f10mBoutobari et aJ.}1°1used to show the primary soUdificationstructure: the whiter regions indicate high Mn contents_ Jt isobvious that the graplúte nodules are in contact with regionswith quite different Mn contento Rundman[11Jhas al$o Te-ported data showing high Mn content dose to the graplúte/matrix interfaces. AlI these arguments are coherent with Vande Velde ideas and in contradiction with the dassical graph-ite formation theory.

Micro segregation studies over four different ductile ironsshowed that Si and Cu segregate inversely with Mn.

Si/Mn and Cu/Mn determinations showed considerablevariation and the expected decrease with increasing distanceto a nadule was not evident.

A great dispersion of the Si/Mn and CuIMn ratios wasdetermined after local microsegregation measures in the me-taUicmatrix interface dose to the graphite nodule.

A great number of microanalysis measurements in the me-tallic matrix did show that the interfaces dose to the graphitenodules may present Si/Mn and Cu/Mn ratios )ower thanthe nominal composition of thc aDoy.

. . - 'I. - .. ..,.. A' .. .. .~"" f ~. . ~ .· , .IO. 4.. ... .. ..., . . ..' .. . . ,

-~ '. ~ , ... . . ~. .. ... ~ -:.,,~".. .' .' ,. .. --~... !~.. -:.- .~... '" ... # ..~~ ~.~.-. .. . .. . .. 110.. . , ... . .. I .. .lO . . .'y.'O 'ti,. . t lO' ~ .. .... '.. ~........'. . ~.".' ..-AI!'.r. · \ ." ",,'..'. ,_..",,- ... .. . .

:' .. .. ~. .. ". ' e.. ., · . .f · ,... .. . ~ -~ ~r.. .,..'t. '~-:;:_..;,."...",.f ~. _ ~ ·'". .. . .. .. ... ..- ' -- t ... \ .... -" ..~ ... ,. · .

:r8í8 : -- '.. ... lO'I. . ~".. *. '. , ......". ..: ". ... I'. ' e ". , ~....I. ~.. '. . 41"'~... ."Fig. 5. OptÍi:llI as-€llSf IIfÍCTO$'nrdllre 01nlloy A3 .1i~ 'Qsl sholl'inf( the oorÍllti01lf onMn contmt (Il!/rife, n,ens nfll MtJ mric/red reJ{Íons)(x 1001.

The results obtained in this work are in contradiction with

the microsegregation pattern imposed by the mechanism ofgraphite growing from the melt and become plausibl.e if themechanism of dendritic ductile iron solidification is assumed.

R.eceived:DeJ:ember 9,1999FifUll version: Mnrch 27, 2000

[11 c. R. Loper, R.W. Heine, AFS Trans. 1962,583.(2) D. Stefanescu, Proc. 61st World Foundry Congress, Beijing

Giesserei Verlag, Diisseldorf, 1995, 8.[31 S. E. Wetterfall, H. Fredriksson, M. Hillert, I. Iron Steel

ln.<;t.1972, May, 162.

(4) E. Dorazil, High Strength Austempered Ductile lron, EllisHorwood Series in MetaIs and Associated MateriaIs,

Hemel Hempstead. UK 1991,p. 13.(5) J. M. Schissler, J.Saverna, I. Heat Treat.1985,4, 167.(6) N. K. Data, N. N. Engel, AFS Trans.1976,431.(7) J.Janowak,R.Gundlach,AFSTrans.1983,377.(8) R.c. Voight,C.R. Loper,Proc.1st lnt. Conf.ADI, Rosemont,

Clriargo,lUinois,2-4 Apr 1984, ASM, MetaIs Park (OH),1984,p.83.

191 C. A. Van de Velde, The Solidification of Ductile Cast lron

- a Nerv Approach, AFS Special Repert, American Foun-drymen's Society, Des Plaines, IL, 1997.

[10] S. M. A. Boutorabi, J. Campbell, Mater. ChaTact.1993,31. 127.

111] K. B.Rundman, Prac.Int. ADI 5imulationConf.,HelsinkiUniversity of Technology, Helsinki Finland, 1997.

518ADVANCED ENGlNEERING MATERIALS2000, 2, No. 8143S-1656/00/080S0518 $ 17.50+.50/0