A Hypothesis for Cast Iron Microstructures

JOHN CAMPBELL The various microstructures of cast irons are reviewed, including carbidic and graphite forms (ake, compacted, spheroidal, and undercooled, etc.), exploring whether the presence of externally introduced defects in the form of oxide double lms (bilms) in suspension in melts seem to provide, for the rst time, a uniform explanation for all the structures and their properties. Silica-rich oxide bilms provide the substrates on which oxysulde particles form, nucleating graphite. The presence of the lm provides the favored substrate over which graphite grows, which leads to the development of ake graphite. The addition of limited Mg to form compacted graphite destroys all but a remnant of the silica-rich bilms. The oxide lm remnant is stabilized by the presence of the graphite nucleus, which causes the graphite to grow unidirectionally in a lamentary form. The addition of excess Mg destroys all traces of the oxide bilms, leaving only the original nuclei, around which graphite is now free to entirely enclose, initiating the spherical growth mode. Undercooled graphite is the true coupled growth form, nucleated at even lower temperatures in the absence of favorable lm substrates in suspension; the graphite adopts a continuous growth mode in a matrix of austenite. Carbides in mottled and white irons form on the oxide bilms that often lie along grain and interdendritic boundaries, which explains the apparent brittleness of these strong, hard phases. In most cases of nonspheroidal growth modes (ake and misshaped spheroids), it is proposed that the impairment of the mechanical properties of irons is not strongly determined by graphite morphology but by the presence of oxide bilms. Spheroidal graphite iron has the potential for high properties because of the absence of bilms. DOI: 10.1007/s11663-009-9289-0 The Minerals, Metals & Materials Society and ASM International 2009

I.

INTRODUCTION

THE forms of graphite in cast irons have been the subject of intense interest and huge research eorts mainly since the 1950s, but a full understanding has been elusive. Readers are referred to the review by Loper[1] for a wideranging synopsis covering many details not included in this study. Here, a dierent review is made of the literature, exploring the possibility of a unifying approach based on the hypothesis that oxide lms (as doubled-over bilms) are present in liquid irons. Recently, a comprehensive understanding of the microstructure of Al-Si alloys has been proposed in terms of bilms, explaining both the mechanism of modication and the structures of hypoeutectic and hypereutectic alloys.[2] Bilms (double lms), usually oxides, are introduced into molten metals at every pour or stirring event. These surface lms, which are doubled up during the process of entrainment into the bulk, seem to be of major signicance for the development of solidication structure.[3,4] Their outer surfaces seem to
JOHN CAMPBELL, Emeritus Professor, is with the Department of Materials and Metallurgy, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK. Contact e-mail: jc@campbelltech.co.uk. This article is based on a presentation given in the 3rd Shape Casting Symposium, which occurred during the TMS Spring Meeting in San Francisco, CA, February 1519, 2009, under the auspices of TMS, the TMS Light Metals Division, the TMS Solidication Committee, and the TMS Aluminum Processing Committee. Article published online September 1, 2009.
786VOLUME 40B, DECEMBER 2009

be favored substrates for the precipitation of many, if not all, second phases during solidication, whereas the inner unbonded interface acts as a crack. The lms are generally so thin they are eectively invisible to casual observation, which leads to a liquid that is invisibly cracked and, on solidication, to a solid whose cracks may or may not be visible. Nakae and Shin,[5] among many others, have drawn attention to the analogous features of Al-Si and Fe-C alloys. This article is an extension of the bilm hypothesis, which is apparently valuable to an understanding of the Al-Si system as well as to a possible understanding of the various morphologies of carbon in the form of graphite and carbides in the Fe-C alloy system. Naturally, the presence of invisible defects in opaque liquids is not easily conrmed directly, although it will not be impossible in the longer term. In the meantime, this article surveys the experimental evidence from the literature to ascertain whether indirect evidence is supportive of this hypothesis.

II.

THE EVIDENCE FOR OXIDE BIFILMS IN CAST IRON

De Sy[6] has shown that liquid cast iron generally contains signicant quantities of oxygen in solution in excess of its solubility. He concluded, on the basis of careful and rigorous experiments, that the undissolved fraction of oxygen was present as SiO2 particles. Interestingly, by heating to 1550 C, he conrmed the
METALLURGICAL AND MATERIALS TRANSACTIONS B

expectation that the SiO2 solids dissolved because they became less stable than CO but reappeared on cooling once again. Hartman and Stets[7] report not only the presence of SiO2 in suspension but also olivine, 2(Mg,Fe)OSiO2 is found in those irons that contain Mg. Homan and Wolf[8] nd a variety of oxides, including SiO2, FeO, and MnO among others, but superheating and holding at high temperature eliminated many of these. In an elegant study of the thermodynamics, Mampaey and Beghyn[9] show how mainly SiO2, together with some FeO, forms in a typical melt when cooling from 1480 C to 1350 C. It seems reasonable to speculate that these oxides almost certainly would not be compact spheres, cubes, rods, etc. but would most likely be in the form of lms. Only lms would have a suciently low Stokes velocity (one or two orders of magnitude lower than particles as a result of their greatly increased drag) to remain in suspension for long periods of time associated with these experiments, as well as the long periods during which irons are held molten in holding furnaces. In any case, of course, the lm morphology is to be expected. During melting in the cupola as droplets of iron rained down, the natural enfolding of the surface lm of SiO2 of each droplet would ensure a natural population of SiO2-rich bilms. Additional treatments or handling, such as pouring actions, stirring in induction furnaces, and the oxide introduced from the surface of the charge (whether steel, pig, or foundry returns), would increase this already large, natural population. Even so, it is well known that iron from electric furnaces is more liable to chill formation problems in thin sections than cupola iron, and this eect has been widely accepted as the loss of nuclei (in agreement with proposals made in this article), especially during an extended time in holding furnaces or pouring systems. Silica-rich oxides in irons are not stable in some conditions. For instance, they will be expected to go into solution if the melt is held above approximately 1450 C for any length of time.[9] Also, they are rapidly reduced to Si metal, and their oxygen is removed as MgO when Mg is added to the melt. A dierent population of oxide bilms denitely exists in ductile irons that have been cast turbulently. In this case, the presence of Mg stabilizes the magnesia, MgO, in the surface lm, although Si might also contribute, thus forming a magnesium silicate MgOSiO2, which is equivalent to MgSiO3. Both magnesium oxide and magnesium silicate are extremely stable and represent permanent damage folded into the liquid metal and subsequently transferred to the casting. Although the bilms are known to have an initially compact morphology as a result of the turbulence during their formation and are relatively harmless as cracks, the subsequent straightening of these bilms by various natural processes, such as the growth of dendrites, which creates extensive planar cracks, is common; this process results in extraordinary structures (Figures 1 and 2)[10] and the phenomenon of brittleness in so-called ductile iron in the form of plate fracture, as discussed earlier by the author.[3] Thus, at least two dierent populations of oxide bilms seem to exist in suspension in liquid iron, which
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 1A so-called ductile iron casting illustrating brittle fracture as a result of dendrite-straightened bilms.[10]

Fig. 2Optical micrograph through the fracture surface of platefractured ductile iron shows misshapen graphite nodules growing on bilms straightened by dendrite growth.[10]

include the following: (1) the hypothesized silica-rich bilms as a natural population in equilibrium with the melt, the amount of silica-rich phase being predictable by thermodynamics, and (2) the known magnesia-rich bilms in ductile irons are the result of mechanical accidents that involve the turbulent entrainment of the
VOLUME 40B, DECEMBER 2009787

surface lm into the bulk liquid, which results in nonreversible damage. In passing, it is worth drawing attention to the possibility that other nonoxide bilms are to be expected in cast irons. For instance, nitride bilms probably form in cast irons giving rise to the nitrogen ssure defects associated in the past with high nitrogen binders.[3] Also, of course, carbon bilms exist in irons in those cases in which hydrocarbon gases from the mold disassociate on the hot surface of the melt, so that if surface turbulence occurs, the resulting carbon lm can be folded in. Such lustrous carbon defects are not necessarily permanent because the carbon can slowly dissolve, which explains why such features are observed in only relatively thin section castings but not in thicker sections.[3] In this review of the microstructures of cast irons, only oxide bilms seem to have signicant roles.

nucleate graphite. Only those shells with low Mn/S ratio seem to act as nuclei. The mechanisms proposed by Riposan[16,17] are highly convincing and are in agreement with the general consensus that few nuclei exist prior to inoculation but are enhanced in eectiveness and number by preconditioning and inoculation, and their nal composition consists of mixed oxides and suldes. Many authors have reported the benecial eects of S, for instance Chisamera et al.[18] In this article, the mechanisms proposed to explain the various morphologies of graphite are based on the possibility of nucleating particles (probably based on oxysuldes as described by Riposan[16,17]) eective for all types of graphite, but the subsequent growth of the graphite is aected by the presence or absence of oxide bilms of dierent types. This approach is the rst to suggest separate functions of the graphite nucleating particles and graphite growth substrates.

III.

GRAPHITE NUCLEI IV. FLAKE GRAPHITE IRON (FGI) AND INOCULATION

Mizoguchi et al.[11] have demonstrated that austenite is ineective in nucleating graphite. In fact, they nd that undercoolings below the liquidus of between 200 C and 400 C are required to trigger nucleation by austenite. A more unfavorable nucleus would be dicult to imagine. The question therefore arises, What does nucleate graphite? This question is all the more intriguing following the work by Mampaey and Xu,[12] in which they found that a single population of nuclei could explain both gray and ductile irons. There is a growing consensus that both ake and spheroidal graphite nucleate on similar, if not identical, nuclei (for instance, Warrick[13]) composed of particles of complex oxides and suldes. This was the conclusion reached in the rst study after the development in the U.K. of the microprobe analyzer in 1974.[14] Many conrmations of this nding have since been made (for instance Skaland[15]), which suggests that the oxysulde mix of the various elements will have a spectrum of lattice spacings ensuring that at least part of the compound will match graphite, and therefore possibly constitute a favored substrate. As an example of an excellent recent study, while working on preconditioning treatments for gray irons (treatments involving small additions of elements such as Al prior to inoculation possibly to enhance the population of naturally occurring nuclei in uninoculated irons), Riposan[16,17] denes a three-stage model for the nucleation of graphite: (a) Small oxides (<2 lm diameter) are formed in the melt (from the preconditioner, usually containing Al and/or Zr, leading to the oxides consisting mainly of alumina or zirconia). (b) Complex suldes (<5 lm diameter) nucleate on the oxides (mainly based on MnS with low Mn/S ratio but complicated by additions, particularly Ca, Sr, and Ba). The complex suldes form a shell around the central oxide. (c) Graphite nucleates on parts of the sulde shell. Riposan[16,17] nds that MnS shells with Mn/S high ratios are numerous in the matrix but do not seem to
788VOLUME 40B, DECEMBER 2009

Uninoculated iron is characterized by poor control of the graphite morphology. Flakes occur but are relatively few in number and uncontrolled in size. The relatively few opportunities for the carbon to precipitate lead to relatively large regions of the iron elsewhere being supersaturated with carbon. Thus, carbide precipitation is likely in places. The mechanical properties of the iron are generally poor. In general, as will become clear during the progress of this account, it seems that some nuclei exist prior to inoculation, but their number and eectiveness cannot be relied on. The gradual introduction of the inoculation process occurred from about 1920 onward, and its development continues to the present day (for instance Skaland[15] and Hartung et al.[19]). The inoculation process was found to increase greatly the number of nuclei available, giving a copious crop of graphite akes of good uniformity of size, with a reduced tendency to carbide formation and a consequent benet to the mechanical properties and machinability of the iron. The inoculation process is often carried out by addition to the melt of granules of a graphite promoter such as a ferrosilicon, which contains perhaps 50 to 75 wt pct Si. Interestingly, the melting point of such an inoculant is close to 1210 C, which is below liquid iron casting temperatures, usually in the range 1350 C to 1400 C. Thus, the added particles melt. However, the particles take time to melt and time to disperse, forming temporary supersaturated regions of liquid rich in silicon surrounding the melting and dissolving inoculant particles. Hurum[20,21] was the rst to draw attention to this phenomenon, but it has been studied by several others since (for instance Fredriksson[22]). Harding et al.[23] show how this region is eectively undercooled by several hundred degrees Celsius as a result of its constitution of nearly undiluted graphite promoter, thus providing a region in which a high driving force exists for the nucleation of graphite. Inoculation with graphite or SiC, and so on, acts similarly to promote regions of
METALLURGICAL AND MATERIALS TRANSACTIONS B

high-constitutional undercooling, which encourages graphite nucleation and growth on any suitable available substrates that happen to be in this region. The overwhelming driving force explains the wide variation of successful nuclei, which, in other circumstances, would be expected to be of only mediocre, if any, eectiveness. Because of the additional trace alloying elements in the inoculant (particularly group IIA elements of the periodic table, including Mg, Ca, Sr, and Ba), the MnS shells around the oxide centers are modied in these regions in which the additions are concentrated (prior to their dissipation and dilution in the melt). All this is taking place in the region highly constitutionally undercooled with respect to graphite, thus naturally providing the driving force for nucleation of graphite in precisely the location needed for maximum eectiveness. Bearing in mind that many (if not all) second phases precipitate on bilms as preferred substrates,[4] it seems reasonable to assume that these new graphite nuclei would also preferentially nucleate on bilm substrates. Thus, graphite would, in turn, nucleate from those nuclei already sitting on the oxide bilms. Figure 3 schematically shows a graphite nucleus formed on an oxide bilm approaching an undercooled region provided by a dissolving inoculant particle. On entering this region, the nucleus experiences a massive driving force as a result of hundreds of degrees of eective undercooling, which forces graphite to form around the nucleus. Harding et al.[23] point out that once nucleated in the regions of high driving force for initiation, the graphite particles attached to their nuclei now will emerge from these regions into the general melt where they will become unstable and start to redissolve. The observations by Loper and Heine[24] conrm that graphite can nucleate in both hypoeutectic and hypereutectic irons at 1400 C, well into the liquid range, high above the expected liquidus temperatures. Mampaey[25] conrms that graphite forms in the melt prior to the appearance of austenite. (These observations are contrary to expectations based on the equilibrium diagram based, of course, not only on equilibrium behavior but also on the behavior of alloys of perfectly uniform composition, neither of which applies during the mechanism of inoculation of cast irons.)

Given sucient time, all the inoculant particles will have melted and dispersed, leaving no pockets of undercooling, and all the embryonic particles of graphite nucleated in the undercooled regions will have re-dissolved. This is almost certainly the phenomenon known to all foundry personnel as fade of the inoculation eect. However, graphite embryos on their nucleating particles will take time to go into solution, so that if the melt is cast without undue delay, many will survive to reach the regime close to the freezing temperature where they will now start to regrow, as observed experimentally by Loper and Heine.[24] Feest et al.[26] nd that although the Si-rich inoculant particles disperse relatively rapidly, the graphite formed rapidly in these regions is slow to redissolve. This is reasonable because the graphite initially grows quickly in the highly constitutionally undercooled region, but it will subsequently dissolve in the open melt rather slowly because the graphite will nd itself only modestly above its equilibrium eutectic temperature, so that the matrix would be eectively nearly saturated in carbon. The newly forming graphite cannot grow completely around the nucleating particle because the particle itself has itself grown on the planar bilm substrate so that at least one of its faces is inaccessible (Figure 4(c)) (work by Riposan[16,17] indicates that only parts of the nucleant particles are active nucleation sites, so that initial growth is rather irregular). The silica-rich bilm will form a next best substrate for graphite, so although insuciently favored to cause nucleation, it is suciently favored to support the subsequent growth of the graphite. Thus, in gray irons, the graphite extends across the bilm, leading to the fairly at morphology of akes in gray iron. The akes grow in regions ahead of the solidication front (i.e., above the general eutectic freezing temperatures) because of the energetically favored growth of graphite on the oxide substrates in suspension (Figure 5).[27] The growth morphology of graphite, extending in the directions in its basal plane, would favor the straightening of the bilm (Figure 3(d)). The bilm would be expected to be extremely thin, possibly measure in nanometers, its minimal rigidity exerting negligible constraint of the advancing graphite crystal. The freedom from restraint would explain the development of relatively perfect crystals of graphite as observed growing ahead of the coupled eutectic graphite (Figure 5).

Fig. 3The mechanism of inoculation: (a) a bilm in suspension in the melt together with nuclei already attached from trace contaminants or preconditioners; and (b) the bilm oating into a region of high constitutional undercooling surrounding a dissolving inoculant particle, initiating a graphite ake.
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 4(a) A bilm with precipitated nucleus from natural contaminants or preconditioners; (b) additional nuclei provided by inoculant; (c) graphite nucleation on nuclei in regions of high eective carbon supersaturation as in Figure 3(b); (d) growth of graphite akes along the length of the bilm, straightening the bilm, with a consequent central planar crack in the graphite ake.

VOLUME 40B, DECEMBER 2009789

Fig. 5Extrinsic initiation of straight graphite akes in the liquid, ahead of coupled growth of eutectic graphite and austenite.[27]

Fig. 6Two populations of akes: extrinsically formed on bilms and intrinsically grown as a coupled (undercooled) eutectic.[29]

The mechanism proposed above explains the growth of ake graphite from nucleating particles introduced by inoculation. Particles that seem to be nuclei for the initiation of akes have often been observed, and these particles seem likely to be a universal phenomenon in both ake and nodular irons.[16,17,28] Eventually, the advancing solidication front will overtake those akes growing on bilms oating freely in suspension in the liquid. Thus eventually, these freely oating akes will become incorporated into the solid (Figure 5). Thus, it would be expected to be common to see gray irons with two separate populations of akes: (1) those formed by free growth in the liquid (primary akes) and (2) those formed by coupled growth at lower temperatures (eutectic akes). The coupled growth mode is discussed below. A bimodal distribution of graphite akes is therefore to be expected in most gray iron microstructures, and it is clearly observed in Figure 6.[29] Less obvious but important bimodal distributions are almost certainly common, as may be inferred from the
790VOLUME 40B, DECEMBER 2009

work of Enright et al.[30] On the rst occasion that these authors used the highly sensitive fractal analysis technique to assess microstructures of cast irons, they observed this result. Thus, unless this is an extraordinary coincidence, the populations of primary and eutectic akes should be expected to be mixed in most, if not all, ake graphite irons, even though, of course, such mixtures may not always be obvious to casual observation. In agreement with the mechanism described in this article, Goodrich[31] attributes the Type C iron (ASTM A247), which is characterized by large, straight akes, with some branching, to the result of the growth of the akes in the liquid, unencumbered by the presence of austenite. Primary akes seems to be a good name for these forms. He calls these proeutectic akes. They originate in suspension in the melt and therefore can oat to the upper regions of a casting. The more common Type A graphite akes are similar, displaying only minimal irregularity, suggesting a similar origin and behavior in the melt. Loper and Fang[32] use deep etching to reveal what they call pre-eutectic akes, which exhibit elegant hexagonal symmetry and are apparently largely free from defects. For many other irons, the presence of a dense mesh of austenite dendrites constrains the size and shapes of akes and prevents any signicant buoyancy eects.[1] More usually in castings, the graphite akes are seen to branch relatively frequently. In terms of the bilm substrate, this is straightforwardly understood from the irregular structure of the bilms. During their entrainment from the liquid surface into the bulk melt, they tend not to entrain as nicely parallel double lms, but instead they entrain as randomly folded, messy structures. Thus, folds leading to parts of the double lm at irregular angles to the main bilm fold are to be expected and would account for the branching of growing akes. Experience of variable performance is also to be expected. For instance, on the Monday morning after the melt has been held for the weekend, operators commonly nd the iron has poor graphite structure, particularly if the iron has been held in an acid-lined induction furnace. Riposan[16,17] shows that this is at least partly the result of the gradual loss of the graphite nuclei. Part of the eect would also be expected to be loss of bilm substrates. Similarly, iron heated to high temperature suers a similar degradation of graphite structure, almost certainly as a result of the dissolution of the bilms because of the instability of SiO2 above approximately 1450 C in the presence of carbon. It would be interesting to know whether the melt, after losing its silica-rich bilms at high temperature, would regain its good solidied structure when cooled once again, because although de Sy[6] reports that the silica reappears in the melt on cooling, without some kind of surface turbulence, the form of the silica may not be a bilm, nor even a lm, but it may be a compact particle. As such, it is not likely to be a good substrate for the development of a good ake structure. Even so, the nal pouring of the melt into the mold may provide sucient turbulence to address this problem, which makes the
METALLURGICAL AND MATERIALS TRANSACTIONS B

problem essentially invisible to those attempting to study the eect. In agreement with the prediction that graphite grows on oxide bilms, the eect of oxygen addition to the melt during the pouring of iron into the mold is demonstrated by several authors. For instance, Basdogan et al.[33] and Chisamera et al.[18] found oxygen to be highly eective in converting carbidic irons into beautifully inoculated ake graphite irons. Liu and Loper[34] found that oxygen was necessary to nucleate kish graphite on the surface of gray iron melts. Larger quantities of kish were formed at temperatures below 1400 C, below which SiO2 is stable, and kish was not observed in Si-free melts. Moreover, Johnson and Smart[35] describe a critical experiment in which they use sophisticated Auger analysis to prove that two or three atomic layers of oxygen (and interestingly, sulfur) are present on fracture surfaces of gray irons that were adjacent to a graphite ake (fractured and observed in high vacuum). In contrast, the hollows in the fracture surfaces of spheroidal graphite matrix that had contained spheroids exhibit no oxygen. This behavior is exactly predicted by a bilm hypothesis: If ake graphite formed on one side of the oxide bilms, then the fracture surface would necessarily reveal the oxide, but oxides would be absent in the case of spheroidal graphite iron (as discussed below). For most intermetallics precipitating on one side of a bilm, the halves of the bilm appear on both fracture surfaces. However, gray iron has a curious behavior in that nothing seems to attach to the graphite, so that both halves of the bilm seem to have attached to the matrix. It is right to question, therefore, what role has the bilm played in the fracture because its central unbonded interface has not seemed to act as the decohering surface on this occasion. In fact, it seems likely that the bilm will have nucleated the decoherence, but the continuing decohering action seems to follow the graphite interface. This nonattachment (perhaps we should say active detaching) phenomenon between graphite and other phases is observed elsewhere in bubble trails in gray irons as well as in lustrous carbon lms detached from the matrix and adhered to the sand mold.[36] Although all the above discussion relates to silicabased bilms, evidence indicates that alumina-based, or possibly Al-containing Si-based bilms (for instance, based on mullite or other stable alumino-silicate compound) exist. Carlberg and Fredriksson[37] nd that cast irons based on Fe-C-Si exhibit ne graphite structures, whereas those based mainly on Fe-C-Al display coarse graphite akes. Chisamera et al.[18] conrms that conventional gray irons that contain Al develop coarse graphite akes. It would be reasonable to expect that silica-rich and alumina-rich bilms would have dierent mechanical and chemical characteristics and, therefore, would develop dierent qualities of precipitated graphite. The relatively poor mechanical properties of gray iron seems likely to be more to do with the presence of bilm cracks down the centers of graphite akes (or the sides of graphite akes if graphite grows on only one side of the bilm the impression now is that the ake has
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 7Graphite ake exhibiting a central crack (the solid state precipitation of surrounding temper graphite is also fractured o).[38]

decohered from the matrix) rather than any intrinsic weakness of the graphite itself. It seems possible that graphite is strong in tension perpendicular to its basal plane, because although the interplane bonding is not strong, which leads to the familiar easy shear in its free condition, there are approximately 1015 bonds per mole in the graphite surface. The huge number of bonds will ensure high strength in tension in a situation in which the graphite is encapsulated, preventing easy failure by peeling or shear. A crack down the center of a ake is observed in Figure 7.[38] In their studies of crack initiation and propagation in irons, Voigt and Holmgren[39] report many centerline cracks in graphite akes plus some decoherence from the matrix. The above discussion relates to those graphite akes growing freely in the melt giving rise to large, randomly oriented akes which tend to oat or settle irregularly, creating what has been called in the past an anomalous eutectic.

V. COUPLED EUTECTIC GROWTH OF GRAPHITE AND AUSTENITE In this section, we move the focus from anomalous to the truly regular, classic eutectic form, in which austenite and graphite grow in a coupled mode. In the absence of suitable nuclei that have formed on oxide substrates in suspension in the melt, the carbon in solution will be unable to precipitate. Thus, the melt will continue to undercool until the undercooling nally becomes sucient to provoke precipitation on some other (less favorable) substrate. Only relatively few such nuclei will operate, activating in those parts of the melt that are especially cool, such as those regions close to
VOLUME 40B, DECEMBER 2009791

the mold walls. The subsequent evolution of heat will inhibit other nuclei from becoming active. All the coupled eutectic forms described below are therefore not especially inuenced by the (rather rare) nucleation events. They are mainly inuenced by the continuous growth process. At modest undercoolings, the coupled growth takes the form of rosettes, which are often called cells (Figure 8).[40,41] Thus, it seems likely that a single initiation event on a nucleus, which is often sited on the mold wall, expands the coupled growth front as a hemisphere to form the rosettes, or cells (Figure 8). The cells are beautifully regular structures with interake distances now strictly controlled by diusion in the boundary layer immediately ahead of the advancing front. Whether the rosette form is a strictly coupled growth, not requiring the presence of bilms, is not clear. For instance, it may be some kind of aggregate of bilms that, the graphite having nucleated, subsequently develops along radially oriented bilms to generate the appearance of a single growth phenomenon. Those bilms present between akes, at less than a diusion distance from neighbors, will lose any graphite to their neighbors by a natural coarsening process, and so these bilms become invisible. Thus, a clearly formed rosette structure may be capable of generating from a random morass of bilms. A careful examination of graphite akes to determine whether they contain bilms, and an examination of interake regions to check for bilms, will help to settle such questions. At lower undercoolings, ner, more highly undercooled eutectic, which is sometimes designated Type D or E according to ASTM specication A247, seems in general to have been avoided for general engineering castings. This is possibly because the interake diusion distance is now so small that only ferrite can be formed, limiting the strength of such irons. During coupled growth, akes have to realign their growth direction continually because of the intrusion of their neighbors into their growth space. Because the

growth direction of graphite is mainly parallel to the basal (0001) plane, this means that the crystal has to develop faults to allow it to change direction. Probably all coupled eutectic graphites experience this eect. However, in particular, this explains the coral type of graphite morphology (to be discussed later), which is highly faulted.[42] We would expect, therefore, that ne graphite morphologies, types D and E graphite, would be highly faulted, containing high defect densities, whereas rosette (or cell) graphite would represent an intermediate case as a result of its larger spacing. Primary ake graphite, as mentioned above, would be expected to contain the least faults.

VI.

SPHEROIDAL GRAPHITE IRON (SGI) (DUCTILE IRON)

When sucient magnesium is added to the melt, the oxide bilms are completely eliminated. In the case of silica-rich bilms, the silica will be reduced by magnesium to (1) silicon metal, which will dissipate into solution in the matrix, and (2) solid magnesium oxide that will precipitate probably on the pre-existing nuclei that originally sat on the lms, augmenting these original particles. The reaction is simply SiO2 2Mg Si 2MgO The total loss of bilms means that only solids remaining in suspension in the melt are the original particulate nuclei, possibly augmented by additional MgO. If sulfur is also present in the melt, then the MgO is likely to contain a component of MgS. These compact nuclei are now the only nucleation sites available for the precipitation of graphite. The precipitating graphite grows over the compact nucleus, wrapping completely around it so as to form a compact initiating morphology. The disappearance of the bilms and the initiation of spheroids are shown schematically in Figure 9. The wrapping around process (Figure 9(d)) may consist of renucleation of many separate microscopic grains of graphite on favorable fragments of the oxysulde surface. The growth mode is probably some kind of addition of carbon atoms to spiral growth steps generated by h0001i oriented screw dislocations (Figure 10).[43] In this way, the radial structure of graphite nodules develops from the graphite grains growing radially out from the compact nucleus to form the familiar approximately spherical nodule (Figure 11).[44] The careful study by Riposan[16,17] on ake graphite irons shows that the particulate nuclei do not seem, in general, to be completely wrapped around by graphite, but graphite forms preferentially on isolated regions of the particle. On the one hand, if such a behavior also exists for ductile iron, as seems possible, then this unpromising start to nodularity will emphasize the possible importance of plastic constraint of the surrounding austenite matrix as will be discussed below. On the other hand, the addition of Mg, in addition to eliminating the silica-rich bilms, might also somehow
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 8A scanning electronic microscope (SEM) image of a deeply etched rosette of ake graphite, expanding to form a cell (courtesy of Fras et al.[40,41]).
792VOLUME 40B, DECEMBER 2009

Fig. 9(a) The melt with bilms and sundry contaminant or preconditioned nuclei; (b) the elimination of the silica-rich bilms by addition of Mg; (c) the survival of existing nuclei and additional nuclei from inoculation; (d) the nucleation of graphite, wrapping completely around existing nuclei (particularly if they happen to pass through constitutionally supercooled regions); (e) growth of spheroids.

Fig. 10The probable structure of a graphite nodule.[43]

aect the ability of the nuclei to work more eciently, so that a complete wrapping around eect might be promoted. Subsequent research will clarify this point. Later, as the graphite grows, Stefanescu[45] concludes in his review that all the evidence points to nodules initially growing freely in the liquid, subsequently developing a shell of austenite, and nally contacting and becoming incorporated into an austenite dendrite. A minor modication of this development may be envisaged, in which the graphite nodule does not grow a shell of austenite until it contacts an austenite dendrite. At that moment, a shell of austenite would be expected to wrap itself rapidly around the nodule. Painstaking metallography would be required to clarify this detail. Anyway, whatever the ner details of the encapsulation process, the shell of austenite seems a key feature associated with the growth of spheroids. It seems possible that the spherical morphology of the graphite nodules may be encouraged by the mechanical constraint provided by the nodule having to force its growth against the resistance provided by its surrounding shell of austenite.[46] Many studies have clearly revealed the deformation of austenite dendrites by the growth of internal nodules (Figure 12).[47] This lumpy morphology has been attributed to various mechanisms, all of which are likely to contribute to some degree, as follows: (a) Ruxanda et al.[48] and Stefanescu[45] assume the protrusions to be the natural growth shapes arising from cooperative growth of austenite and graphite by diusion from the liquid. (b) Buhrig-Polackzed and Santos[49] indicate in a schematic illustration that the contact between nodules
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 11Graphite nodules in an austempered iron indicating nucleation on a small central inclusion.[44]

with their austenite shells and the austenite dendrites results in their mutual assimilation, to create a local bump on the dendrite. (Some subsequent surface smoothing driven by surface energy would be expected to occur rapidly.) (c) Deformation of the dendrite by plastic ow, locally expanding the surrounding solid to accommodate the increasing volume occupied by the graphite has to be important. This eect seems to have been generally overlooked, but it is important and worthy of examination, as discussed below. The pressure developed in a thick spherical shell (Figure 13) expanding plastically because of internal pressure is quantitatively expressed by[50] P 2Y ln b=a 1 where P is the internal pressure, Y is the yield stress, and b and a are the external and internal radii of the shell,
VOLUME 40B, DECEMBER 2009793

Fig. 12The distortion of dendrites as a result of the internal growth of nodules (after Hillert[47]).

Fig. 13A thick shell expanding plastically because of internal pressure. A perturbation radius r is not favored because a higher local pressure is required. Thus, sphericity is encouraged.

respectively. The logic is as follows: If a perturbation to the spherical graphite shape were to occur, having a necessarily smaller radius r, the pressure to extend plastically the growth at this location would (according to Eq. [1]) be increased (Figure 12(b)). Thus, growth of the extension of smaller radius would be discouraged because additional pressure would be required to stabilize the perturbation. The easier spherical growth mode, simply expanding the uniform radius a, would therefore be encouraged. It seems, therefore. that there is some qualitative justication for believing that mechanical forces stabilize the spherical growth mode of the nodule. However, it is useful to ascertain whether there is quantitative justication for this mechanism. If we take Y to be approximately 6 MPa[51] for austenite at the melting point of iron, and a = 2 nm and b = 20 nm, then we nd P = 30 MPa approximately. Even at values of a = 20 lm and b = 200 lm, P is of course unchanged, which indicates that there is a substantial restraining pressure on the growth of the nodule during most of its life. With regard to the possible asymmetric eect of a perturbation of radius r, taking r = a/2 to a/10 locally increases P to approximately 35 to 60 MPa, respectively. Thus, a rounding eect caused by mechanical smoothing

of the forces to expand the austenite shell seems to be important. Although a creep model rather than the above plastic model might give a somewhat more accurate result, the above result can be relied on to give us an order-of-magnitude estimate of the eect. Clearly, more work is required to conrm this preliminary indication. Johnson and Smart[35] use the sophisticated and respected perturbation analysis by Mullins and Sekerka to suggest that interfacial energies are of importance in spherodizing graphite nodules up to a diameter of perhaps 50 nm, after which the spherical form can no longer be stabilized. Thus, much speculation by earlier authors that interfacial energies may be important in dening the shape of spheroidal graphite seems irrelevant. This leaves mechanical stabilization as a potential critical mechanism. This conclusion is reinforced by the recent evidence[16,17] that the early phases of growth of the graphite around the nucleus are anything but regular. Clearly, a strong spherodizing eect is needed that neither the nucleation mechanism nor surface energy can provide. In general agreement with this conclusion, Jiyang et al.[52] used color etching to reveal the austenite shells around graphite. They found that if the shell formed quickly and completely, then the nodule developed as a sphere, whereas slow-developing or nonenveloping shells led to misshapen nodules. In passing, it may be signicant that on the addition of Mg causing dissolution of the silica-rich oxide bilms, any residual air trapped between the lms is expected to be released. The air trapped in a bilm will be expected to lose its oxygen by continued oxidation of the matrix, followed by nitrogen that would react to form nitride,[53] leaving mainly argon. In this way, it seems possible that clouds of ne argon bubbles will be released into the melt. It seems likely that some Mg vapor will also diuse into the bubbles. It is not easy to dene the sizes of such bubbles with any accuracy. For instance, a bilm of 100 lm square and an average gas gap of 1 lm would yield a pore of approximately 20 lm diameter. A similar area bilm of an average 10-nm gas gap would form a pore approximately 5 lm in diameter. A small bilm of

794VOLUME 40B, DECEMBER 2009

METALLURGICAL AND MATERIALS TRANSACTIONS B

10 lm square and 10-nm average spacing would create a 1-lm-diameter pore. Thus, it seems that a fog of bubbles in the range of approximately 1 to 20 lm is to be expected. It is intriguing, but not perhaps relevant, that a theory proposes that nodules nucleate from Mg bubbles in suspension in the melt.[54,55] However, attractive this hypothesis might be to explain graphite coatings inside pores in solidied castings, as a result of the reduction in strain energy involved, any strain energy relief in the liquid state is zero, and the reduction of surface energy to encourage such precipitation in the liquid state seems negligible. Furthermore, the theory requires the incorporation of solid particulate oxide nuclei into the bubble, but a successful penetration of a particle into the bubble would depend critically on a reduction in interface energies. This is unlikely for particles formed by precipitation in situ in the liquid, which will be in perfect atomic contact with the melt (i.e., will be well wetted, which is a necessary condition for the particle to be a nucleating agent). Such particles will be energetically rejected by bubbles. Thus, although a mechanism for the presence of extremely ne bubbles may be provided by the current analysis, it seems to be irrelevant to the formation of graphite nodules. It does not oer support to the gas bubble nucleation hypothesis. Another interesting aside can be noted. The transformation of the planar cracks sandwiched inside the bilms into clouds of ne bubbles that may oat and escape from the alloy is the essence of the process by which apparently brittle gray iron becomes ductile. (Ductile iron only becomes embrittled once again, as noted below, if oxide bilms are reintroduced by turbulence by handling of the melt or poor lling system design of the casting.) Finally, curious observations, such as that reported by Yamamoto et al.,[56] in which ake is converted into nodular iron by simply purging the melt with ne bubbles of nitrogen, argon or carbon dioxide become explicable. From experience in the light metals industries, it is known that purging with gases can eliminate

bilms from melts. Thus spherodization seems to be achievable via a purely mechanical route, replicating the condition achieved chemically by the addition of Mg.

VII.

COMPACTED GRAPHITE IRON (CGI)

If the addition of magnesium is more carefully controlled to some level intermediate between spheroidal and ake iron, compacted (vermicular based on the Italian for worm-like) graphite is the result (Figure 14).[57] In our bilm model, it is clear that most of the oxide bilms will be dissolved by the addition of Mg. However, small patches may remain if the Mg addition is not too high; the tiny patches on which the original nuclei sat will be resistant to dissolution because they will be stabilized by their attachment to the nuclei. (Naturally, it will have been energetically favorable for the nuclei to attach to the bilm, so that the combination of nucleus and lm will enjoy a reduction in overall energy, stabilizing the combination.) Only half of the bilm will be retained in this way, its distant twin half not enjoying the protective inuence of the nucleus will dissolve and disappear. Only the small part of 1/2 of the bilm together with its unbonded interface, the remnants of the layer of air, will remain (Figure 15(b)). The subsequent nucleation of graphite on the nucleus will result in rapid spreading of growth around the nucleus. On arrival at the nonwetted interface of the residual patch of bilm, this spreading will be arrested (Figure 15(c)). The subsequent growth of graphite is forced to occur not radially but in general unidirectionally away from the bilm residue (Figure 15(d)). Clearly, the growth cannot now be a ake because no bilm is present. Originally, it was thought that it could not be spherical because it could not encapsulate the complete nucleating particle, but the observations by Riposan[16,17] and others indicates that the nucleating particle is often not entirely surrounded in the case of

Fig. 14CGI viewed by (a) SEM deep etching and (b) optical metallography.[57]
METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 40B, DECEMBER 2009795

Fig. 15The formation of CGI by addition of just enough Mg to (a) melts with existing bilms with attached nuclei, to (b) eliminate most of the silica-rich bilms, but not the remnant of lm attached to the nuclei. (c) Inoculation promotes graphite growth on the nuclei; (d) growth continues unidirectionally.

ake iron, and this may also be true of nodular iron. Naturally, the subsequent growth form is not possible to predict and has to depend on completely dierent factors. With regard to the microstructure of growth, Cole[58] had observed a ne, unidirectional spiral structure similar to the worm-like growth mode clearly observed in Figure 14. Liu[59] nds that the growth direction is along the C axis (0001 direction perpendicular to the basal planes) and seems to develop by a spiral dislocation mechanism as witnessed by the coarse and irregular spirals that they observe. The great sensitivity of the compacted graphite morphology to magnesium concentration is corroborated by the proposed bilm mechanism. If the Mg level is too low, then residual bilms will encourage ake graphite, whereas if the Mg level is too high, then the limited stability enjoyed by the residual bilm patches will be overcome, and the last remaining patches of bilm will be dissolved, encouraging the growth of totally spherical grains. Clearly, the nal form of CGI is not strongly inuenced by the inoculation and nucleation events. It develops its continuous worm-like growth because of a continuous growth process. As an important feature of the continued growth of CGI, several workers nd that the graphite seems to stay in contact with the liquid (for instance, References 60 and 61), thus transferring the expansion of the graphite to the liquid, reducing feeding requirements,[62] in contrast to ductile iron in which the graphite transfers its expansion to its surrounding solidied shell, expanding the casting and increasing feeding requirements.

Fig. 16An SEM image of deeply etched CGI closely resembling coupled eutectic coral structure, illustrating the overall similarity of these structures (Courtesy of Thomas Prucha, AFS, 2009).

VIII.

CORAL GRAPHITE

Coral graphite is, as its name suggests, a brous morphology of graphite, formed in rather pure, noninoculated irons. Because of the purity of the alloys, and the fact that it can be grown in alloys containing Ce, no silicarich oxide bilms can be present. The purity and absence of inoculation also means that no oxysulde nuclei are present. Thus, neither akes nor spheroids can form. Thus, coral graphite occurs at high undercoolings (clearly at least one nucleation event is required, but the
796VOLUME 40B, DECEMBER 2009

phenomenon is dominated by growth). After initiation in undercooled regions, the growth of the eutectic is likely to be so fast that it will cover large undercooled regions at the mold wall, thus advancing subsequently on a substantially planar growth front away from the wall. This seems to be a coupled growth mode. For interested readers, even though nowhere in their report do Nakae and Shin[5] mention coral graphite, they present micrographs that clearly illustrate coral growth when the iron is suciently pure, or Ce is added, and when the growth rate is suciently high. They show the close similarity between the coral eutectic structures of Fe-C and Al-Si alloys. CGI and coral forms of graphite (Figure 16) are both ne, brous lamentary morphologies. It is tempting to consider that the resemblance reects an underlying similarity.

IX.

MISSHAPEN SPHEROIDS

The presence of ill-formed spheroids, particularly if present in large numbers, is widely known to be associated with the reduction in mechanical properties

METALLURGICAL AND MATERIALS TRANSACTIONS B

of nodular irons. Hughes[44] describes how a good ductile iron can achieve at least 90 pct nodularity but less good irons can fall to as low as 50 pct or less and suer reduced properties. The 50 pct or so component of akes or other non-nodular shapes does not particularly aect properties such as proof strength but greatly reduces those properties sensitive to failure, such as tensile strength and ductility. Hughes,[44] in common with widely held belief, attributes this loss of fracture resistance to the sharp notches at the root of akes. However, this belief presupposes that the akes act as cracks. This would not be true if graphite had high tensile strength perpendicular to its basal plane but is probably only true if the akes are formed on bilms; the bilm provides the unbonded interface, as a crack. It is the presence of the crack provided by the bilm that has to be viewed as the principal cause of failure. In terms of the bilm hypothesis, graphite would be expected to grow on oxide bilms. During the Mg treatment, the silica-rich bilm content of the melt would be eectively eliminated. Thus, in a liquid now cleaned from transient silica-rich bilms, spheroids would be created in suspension. However, many Mg addition techniques are extremely turbulent, so that large quantities of Mg oxide and/or Mg silicates are expected to be created by the turbulent jumping and splashing of the liquid surface, thereby folding in the oxide surface of the liquid to create entrained bilms. These new bilms will be permanent defects formed from highly stable magnesia or magnesium silicate. Prior to pouring, some of these will fortunately oat out, adding to the Mg-rich slag. Thus, given a reasonable time between Mg treatment and the pouring of the castings for separation of Mg-rich bilms (an interesting and clearly important process variable that seems not well researched), not all of these defects will nd their way into the castings to impair the structure and properties. Unfortunately, on pouring into the mold, additional large quantities of Mg-rich oxide bilms are likely to be reintroduced, particularly if the mold lling system is a rather poor design. On contact with an oxide bilm, the subsequent growth of the existing suspended spheroids will be redirected along the plane of the lm. The symmetrical spherical constraint previously provided by the surrounding austenite is also destroyed, aiding the nonspherical development. Thus, the spheroid will grow to become signicantly misshapen. This eect can be observed in Figure 2. In this gure, several oxide bilms have been straightened by the growth of dendrites so as to lie along 100 planes. These planes, now containing a planar bilm crack, lead to disastrous brittleness known as plate fracture (Figure 1) described elsewhere by the author.[3] Meanwhile, the nodules attached to the bilms are clearly poorly shaped. Additional misshapen nodules are evident elsewhere in the structure. These are expected to be lying on random areas of bilm not straightened by dendrite growth. The appearance of poorly shaped spheroids is therefore predicted to be associated with the growth of nodules on
METALLURGICAL AND MATERIALS TRANSACTIONS B

bilms as a result of either (1) insucient dwell time for the damage introduced during Mg addition to oat out, or (2) poor casting practice, in which an otherwise nicely inoculated and spherodised melt is recontaminated with oxides. It is interesting to predict that perhaps more time after the spherodizing treatment to allow the melt to clear, together with a properly designed lling system, or counter-gravity lling system, should completely eliminate poor nodularity. A step in the right direction is presented in the work of Takita et al.,[63] who observe that nodules are converted from misshapes to spherical by the use of a lter to take out the inclusions produced by inoculation. This positive step contrasts with that taken by Liu et al.[64] who, after the Mg addition, added post inoculants. These were highly successful to increase the nodule count but led to a disastrous fall in nodularity. This was almost certainly a result of adding the inoculants through the melt surface, which would have also entrained the surface oxide to create Mg-rich oxide bilms that, on contact, would generate nonspheroids. Furthermore, of course, the signicant reduction of properties associate with malformed spheroids cannot be the direct result of the shape of the spheroids, because they occupy such a small volume fraction of the alloy. The loss of properties is predicted to be the result of the presence of the bilms in the melt, occupying a vastly greater cross sectional area than the spheroids. These extensive bilms simply act as cracks in the casting, signicantly reducing properties.

X.

EXPLODED NODULAR GRAPHITE

Exploded spheroids (Figure 17)[65] are commonly observed in irons subject to graphite otation and especially if the composition of the iron is suciently hypereutectic.[66,67] This undesirable morphology is not easily explained at this stage as a result of relatively little experimental work to clarify the problem. Cole[58] suggested they had suered remelting as a result of being carried by convection in and out of hot zones of the liquid. This seems most unlikely, however, because if the nodule had grown uniformly in a compact morphology, then the uniform graphite would be expected to have a substantially uniform rate of dissolution; solidication and remelting would be expected to be reversible. Because exploded nodules appear exclusively in the otation region of hypereutectic irons two far more likely factors are as follows: (a) Nodules growing in a suciently hypereutectic melt[66] will experience an enhanced driving force for growth because of the carbon supersaturation that develops as the melt cools. This will encourage growth instabilities leading to dendritic rather than planar growth, leading to exploded rather than smooth spheroid surfaces. (b) Austenite will be less likely to form in hypereutectic irons, especially in conditions of carbon enhancement by segregation during cooling in the liquid
VOLUME 40B, DECEMBER 2009797

Fig. 17(a) Spheroid and (b) malformed spheroid; (c) chunky graphite (after Liu et al.[65]); (d) SEM iron image of an exploded spheroid; (e) electron image.[58]

phase. The nodules may have nucleated early in the liquid phase and grown without the benet of the mechanical constraint of the austenite. When not pressurized to remain spherical, the nodule will be free to grow more like a dendrite, developing instabilities that grow into projections to its growth front, nally developing the characteristic exploded forms. Evidence for mechanical restraint as a powerful eect is presented in the section on nodular graphite above.

XI.

CHUNKY GRAPHITE

Chunky graphite is often observed concentrated in the centers of heavy sections of nodular iron castings. Chunky is not a particularly helpful descriptive adjective for this variety of graphite. Its chunkiness is only apparent under the microscope at high magnication; otherwise, it simply appears to be ne, irregular, branched, and interconnected fragments (Figures 17(c) and 18).[68] Once again, the properties of nodular iron are reduced. However, it seems the loss of properties is possibly more associated with the short diusion distances between branches of the graphite laments, promoting the development of ferrite.[65] Liu et al.[59,65] nd evidence that chunky graphite grows along the C-axis direction, as does nodular graphite. Furthermore, they report observations on spheroids that exhibit gradual degeneration, which slowly take on the growth forms of chunky graphite.
798VOLUME 40B, DECEMBER 2009

Fig. 18Graphite nodules and areas of ne, chunky graphite in the thermal center of a 200-mm cube.[68]

Thus, they conclude that chunky graphite is a degenerate form of spheroidal graphite, and their work implies that chunky graphite grows out from spheroids. Itofugi and Uchikawa[69] conrm the identical growth orientations of spheroidal, compacted, and chunky graphites. All these workers observe the characteristic form of chunky graphite, as an apparently stop/start growth in the C-direction consisting of nearly separate pyramidal chunks linked by a narrow neck, like beads on a branching string. The individual chunk sections are composed of layers parallel to the basal plane, but only
METALLURGICAL AND MATERIALS TRANSACTIONS B

nanometers thick. This characteristically lumpy growth may be the result of a pulsating or irregular advance of the growth front, with the austenite advancing to grow nearly over the top of the graphite, forming the nearly pinched-o neck of the graphite, only to be overtaken again because the carbon in solution will now build up in the liquid ahead of the front, accelerating the next phase of growth of the graphite until the local carbon concentration is depleted once again, and so on. Observations by Kallbom et al.[68] seem consistent with an origin associated with bilms. They observe the chunky graphite to be concentrated in the center of heavy sections, which is explained by the growth of the freezing front pushing bilms ahead by their observations of stringers of graphite nodules. These features are almost certainly sheets of oxides decorated with graphite nodules that have been nucleated on the oxide (analogously to those observed in Figure 2). They conclude that a collaboration among Mg, S, and O is important for chunky graphite formation. Although the above authors provide evidence that suggests the presence of bilms, it seems possible that at the same time there is likely to be an absence of nuclei. This is because so much time is available for particles in suspension to oat out from the centers of heavy sections where chunky graphite is commonly found. In the absence of nuclei, spheroids cannot form. Thus, if a graphite nodule can be initiated somewhere in or near this region, perhaps happening to oat into this region, it will continue to grow. However, it will not enjoy the benet of the presence of an austenite shell in a region of higher temperature and enhanced segregation, so that its growth mechanism seems likely to be asymmetrical. Because the growth will constitute an extension of the spheroid, the spheroidal mechanism of growth along the C-axis is likely to be continued.[59] In the absence of signicant numbers of nuclei, the whole region will ll with a variety of continuous, branching growth, extending along the C-axis direction. The extended size of chunky graphite regions, which are much larger than cells of other types of graphite,[69] corroborate the absence of nuclei in these regions. If the above mechanism is correct, then exploded spheroids can be viewed to be a similar growth mode to chunky graphite, both originating from spheroids. However, why continued growth should not occur in all directions as in the exploded variety, but only in a few select directions (possibly one direction) in chunky graphite, is not clear at this time. It is hoped that in the near future, the correct explanation for the origin of chunky graphite might be elucidated by subsequent careful experiments. The key word here is careful. For instance, the experiment by Asenjo et al.[70] that involves the placement of inoculants in dierent branches of a runner system to dierent sized cavities in a mold to compare the eects of mold inoculation in dierent heavy castings was a clever concept but regrettably awed in execution. This was because, in common with most iron casting, the runners were not designed to be pressurized and ll on a single pass. Thus, a reverse ow is likely to have contaminated the mold cavities, and all the cast
METALLURGICAL AND MATERIALS TRANSACTIONS B

material would have suered from turbulence and air entrainment, therefore containing unknown quantities of oxide bilms. Clearly, in the future, much greater sophistication of melting and casting will be required for experiments designed to clarify the solidication mechanisms for cast irons.

XII.

IRON CARBIDE (CEMENTITE)

Work by Rashid and Campbell[71] has demonstrated the nucleation and growth of carbides on oxide bilms in vacuum-cast Ni-base super-alloys. It would be expected, therefore, that an analogous reaction would occur in Fe-C alloys, because the austenite forming during solidication also possesses a closely similar facecentered-cubic structure, and other conditions, such as the temperature and solidication rates, are all similar. Carbides in irons seem to form preferentially at grain boundaries and often seem to be associated both with residual graphite (sometimes as nodules, malformed nodules, or akes aligned with the boundary) and pores all forming on the same boundaries. This is a clue to their bilm associations. Faubert et al.[72] have studied carbides in heavysection austenitic ductile iron (ADI). Toward the top of their castings, they nd degradation of properties more serious than they would have expected from the carbides themselves. They suspected that the real impairment was caused by the presence of lms that had oated into this region. Almost certainly, the lms would have been bilms (it seems impossible to devise a mechanism by which a single thickness of lm can be introduced into a matrix). Bilms would segregate to grain boundaries and possibly actually constitute the boundary. The presence of the bilm is not only inferred from (1) the cracked carbides but also (2) from the linear rows of nodules viewed in micrographs from this work, (3) from the pores as the residues of air bubbles trapped between the lms, and (4) from the graphite akes sitting in the boundary (called by the authors, unatteringly, degenerate graphite). Both graphite and carbides are expected to form on the wetted, outer surfaces of the bilms. The presence of the central unbonded region (including the pores) between the lms, constituting the crack through the interiors of the carbides, explains the apparent brittleness of the carbides. These intermetallic compounds would otherwise be expected to be strong and resistant to failure by cracking at the modest stresses that can be induced by solidication and cooling. Stefanescu[43] quotes the work of Hillert and Steinhauser,[73] in which the growth of iron carbide eutectic (ledeburite) occurs by the spreading of carbide (cementite) across a plane, followed by the development of a rod type of eutectic at right angles. It is tempting to consider that the original planar expansion would have been facilitated by growth across the surface of a bilm. The bilm would originally have been randomly crumpled but would have been straightened by the progress of the carbide across its face, thus creating an essentially planar crack that would constitute a serious defect in the
VOLUME 40B, DECEMBER 2009799

carbide. Associated branching cracks would have arisen from irregular folds in the bilm. 6.

XIII.

GENERAL

Overall, both nucleation and growth mechanisms inuence graphite morphology, although these mechanisms dominate to dierent extents in dierent circumstances. For instance, nucleation dominates the formation of spheroids and deformed spheroids, whereas for ake graphite, both nucleation and growth are inuential. For compacted, undercooled coral and chunky graphites, nucleation occurs once to initiate each cell, after which continuous growth leads to developing continuous branching morphologies. Indeed, the ner, fairly continuous forms of coral, chunky, and compacted graphites are so similar that they often seem to be confused in the casting literature. They do seem similar in the sense that they all appear to be more-or-less coupled growth forms, advancing together with the austenite. If, as proposed here, the various forms of graphite are signicantly inuenced by the presence or absence of bilms, then it would explain the historical resistance of the phenomenon to explanation so far. This seems typical of bilm phenomena. Furthermore, it seems likely that the principle cause of reduced mechanical properties in all cases of nonspheroidal forms of cast irons is the presence of various kinds of oxide bilms that act as cracks. Ductile iron is ductile because of the absence of oxide bilms, not because of its spherical graphite morphology; but when oxide bilms (mainly magnesia-rich) are entrained by poor casting technology, ductile iron can be seriously reduced in ductility, with castings at times even failing disastrously by the brittle plate fracture mechanism.[3] Naturally, this short account presents only an outline of a new approach to the structures of cast irons. Much more research is needed to prove, enhance, or disprove these proposals. It is hoped that improved control and improved castings will result from these eorts.

7.

8.

9. 10.

growth morphology may be natural but is additionally encouraged by the mechanical constraint of the austenite matrix; the absence of bilms explains the high mechanical properties. Coral morphology nucleates on unknown nuclei at low temperatures, expanding to form cells of coupled growth with austenite, consisting of highly faulted continuous branching laments of graphite in the austenite matrix; bilms play no part in this growth mode. Misshaped spheroids seem to be spheroids that have encountered a Mg-rich bilm, subsequently growing along the bilm and losing sphericity (Figure 2). In addition, the presence of the bilm destroys the symmetrical mechanical constraint of the austenite that favors sphericity. Chunky graphite occurs in heavy-section ductile iron regions. The formation mechanism is not clear at this time. The central regions comprise either (1) regions containing Mg-rich bilms as a result of poor casting techniques or (2) regions from which nuclei have oated out, leaving regions nearly devoid of nuclei and creating graphite akin to coral morphology. As such, it may be a coupled eutectic form, and its beads on a branching string morphology may result from an unstably advancing growth front. Exploded spheroids may be the result of growth in the liquid, without the benet of the mechanical constraint of an austenite shell. Carbides form at low temperatures on oxide bilms. The presence of bilms in the carbides explains the brittle behavior of these strong intermetallics and their common association with both pores and residual graphite fragments.

ACKNOWLEDGMENTS The author is grateful to those who have assisted with the micrographs; to Carl Loper and Riposan and colleagues for inspiring research; and, last but not least, to the painstaking, assiduous, and doubtful reviewer of this article from whom I learned much.

XIV.

CONCLUSIONS

A hypothesis is proposed as follows: 1. Cast iron melts normally contain double lms (bilms) in suspension. 2. Inoculation produces oxysulde particles that nucleate on silica-rich oxide bilms. 3. Graphite nucleates on the oxysulde particles and grows, spreading over the bilms, straightening the bilms, and forming akes of crystallographically near-perfect graphite. The presence of the bilms, as cracks, trapped inside or alongside graphite akes accounts for the poor tensile properties of ake irons. 4. Compacted graphite forms on oxysulde nuclei that occupy bilm residues. 5. Spheroids form in the absence of silica-rich bilms, nucleating on oxysulde particles. The spherical
800VOLUME 40B, DECEMBER 2009

REFERENCES
1. C.R. Loper: AFS Trans., 1999, vol. 107, pp. 52328. 2. J. Campbell: Mater. Sci. Technol., 2009, vol. 25, pp. 12526. 3. J. Campbell: Castings, Elsevier, Oxford, UK, 2003, pp. 17881; 16162; 15860. 4. J. Campbell: Mater. Sci. Technol., 2006, vol. 22 (2), pp. 12745; (8), 9991008. 5. H. Nakae and H. Shin: Int. J. Cast Met. Res., 1999, vol. 11 (5), pp. 34549. 6. A. de Sy: AFS Trans., 1967, vol. 75, pp. 16172. 7. D. Hartmann and W. Stets: AFS Trans., 2006, vol. 114, pp. 1055 58. 8. E. Homan and G. Wolf: Giessereiforschung, 2001, vol. 53 (4), pp. 13151. 9. F. Mampaey and K. Beghyn: AFS Trans., 2006, vol. 114, pp. 637 56. 10. R. Barton: Foundry Trade J., 1985, pp. 11726.
METALLURGICAL AND MATERIALS TRANSACTIONS B

11. T. Mizoguchi, J.H. Perepezko, and C.R. Loper: AFS Trans., 1997, vol. 105, pp. 8994. 12. F. Mampaey and Z.A. Xu: AFS Trans., 1997, vol. 105, pp. 95103. 13. R.J. Warrick: AFS Trans., 1966, vol. 74, pp. 72233. 14. M.H. Jacobs, T.J. Law, D.A. Melford, and M.J. Stowell: Met. Technol., 1974, vol. 1 (11), pp. 490500. 15. T. Skaland: AFS Trans., 2001, vol. 105, pp. 7788. 16. I. Riposan, M. Chisamera, S. Stan, P. Toboc, E. Ecob, and D. White: Mater. Sci. Technol., 2008, vol. 24 (5), pp. 57984. 17. I. Riposan, M. Chisamera, S. Stan, C. Hartung, and D. White: Mater. Sci. Technol., in press. 18. M. Chisamera, I. Riposan, and M. Barstow: AFS Trans., 1996, vol. 104, pp. 58188. 19. C. Hartung, C. Ecob, and D. Wilkinson: Casting Plant Technol., 2008, vol. 2, pp. 1821. 20. F. Hurum: AFS Trans., 1952, vol. 60, pp. 83448. 21. F. Hurum: AFS Trans., 1965, vol. 73, pp. 5364. 22. H. Fredriksson: Mater. Sci. Eng., 1984, vol. 65, pp. 13744. 23. R.A. Harding, J. Campbell, and N.J. Saunders: Solidication Processing 97, Sheeld, UK, 1997, pp. 48993. 24. C.R. Loper and R.W. Heine: AFS Trans., 1961, vol. 69, pp. 583 600. 25. F. Mampaey: AFS Trans., 1999, vol. 107, pp. 42532. 26. E.A. Feest, G. McHugh, D.O Morton, L.S. Welch, and I.A. Cook: Proc. Warwick University Conf., 1980, pp. 23239. 27. Y.X. Li, B.C. Liu, and C.R. Loper: AFS Trans., 1990, vol. 98, pp. 48388. 28. L. De Rong and Y.J. Xiang: AFS Trans., 1991, vol. 99, pp. 70712. 29. M. Hillert and S. Rao: Proc. Brighton Conf., 1967, pp. 20412. 30. N. Enright, S.Z. Lu, A. Hellawell, and J. Pilling: AFS Trans., 2000, vol. 108, pp. 15762. 31. G.M. Goodrich: Metals Handbook, vol. 15, Oxford, OH, 2008, pp. 785811. 32. C.R. Loper and K. Fang: AFS Trans., 2008, paper 08-065(05), p. 8. 33. M.F. Basdogan, G.H.J. Bennett, and V. Kondic: Proc. Warwick University Conf., 1980, pp. 24047. 34. S. Liu and C.R. Loper: AFS Trans., 1990, vol. 98, pp. 38594. 35. W.C. Johnson and H.B. Smart: Proc. Sheeld Conf., 1977, pp. 12530. 36. J. Campbell: Int. J. Metalcast., 2007, vol. 1 (1), pp. 720. 37. T. Carlberg and H. Fredriksson: Proc. Solidication and Casting of Metals Conf., 1977, pp. 11524. 38. S.I. Karsay: Ductile Iron II, Engineering Design Properties Applications, Quebec Iron & Titanium Corporation, Canada, 1971. 39. R.C. Voigt and S.D. Holmgren: AFS Trans., 1990, vol. 98, pp. 21325. 40. E. Fras, M. Gorny, and H.F. Lopez: AFS Trans., 2007, vol. 115, pp. 117. 41. E. Fras, M. Gorny, and H.F. Lopez: Metall. Mater. Trans. A, 2007, vol. 38A, pp. 38595. 42. P. Zhu, R. Sha, and Y. Li: Proc. Mater. Res. Soc., 1985, vol. 34, p. 3.

43. D.M. Stefanescu: Metals Handbook, ASM, Oxford, OH, 1988, vol. 15, pp. 16881. 44. I.C.H. Hughes: Metals Handbook, ASM, Oxford, OH, 1988, vol. 15, pp. 64766. 45. D.M. Stefanescu: Metall. Mater. Trans. A, 2007, vol. 38A (7), pp. 143347. 46. A.N. Roviglione and J.D. Hermida: Metall. Mater. Trans. B, 2004, vol. 35B, pp. 31330. 47. M. Hillert: in Recent Research on Cast Iron, H.D. Merchant, ed., Gordon and Breach, New York, NY, 1968, pp. 10127. 48. R. Ruxanda, L. Beltran-Sanchez, J. Massone, and D.M. Stefanescu: AFS Trans., 2001, vol. 109, pp. 3748. 49. A. Buhrig-Polackzek and A. de Santos: Metals Handbook, ASM, Oxford, OH, 2008, vol. 15, pp. 31729. 50. P. Chadwick: Int. J. Mech. Sci., 1963, vol. 5, pp. 16582. 51. J. Campbell: Trans. TMS-AIME, 1967, vol. 239, pp. 13842. 52. Z. Jiyang, W. Schmidt, and S. Engler: AFS Trans., 1990, vol. 98, pp. 78386. 53. R. Raiszadeh and W.D. Griths: Metall. Mater. Trans. B, 2008, vol. 39B, pp. 298303. 54. A.A. Gorshov: Russ. Cast. Prod., 1964, pp. 33840. 55. H. Itofugi: AFS Trans., 1996, vol. 104, pp. 7987. 56. S. Yamamoto, Y. Kawano, Y. Murakami, B. Chang, and R. Ozaki: AFS Trans., 1975, vol. 83, pp. 21726. 57. D.M. Stefanescu, R. Hummer, and E. Nechtelberger: Metals Handbook, ASM, Oxford, OH, 1988, vol. 15, pp. 66777. 58. G.S. Cole: AFS Trans., 1972, vol. 80, pp. 33548. 59. P.C. Liu, C.R. Loper, T. Kimura, and H.K. Park: AFS Trans., 1980, vol. 88, pp. 97118. 60. J.Y. Su, C.T. Chow, and J.F. Wallace: AFS Trans., 1982, vol. 90, pp. 56574. 61. F. Mampaey: AFS Trans., 2000, vol. 108, pp. 1117. 62. J.D. Altstetter and R.M. Nowicki: AFS Trans., 1982, vol. 90, pp. 95970. 63. M. Takita, Y. Yang, and H. Nomura: Int. J. Cast Met. Res., 1999, vol. 11 (5), pp. 35762. 64. S.L. Liu, C.R. Loper, and T.H. Witter: AFS Trans., 1992, vol. 100, pp. 899906. 65. P.C. Liu, C.L. Li, D.H. Wu, and C.R. Loper: AFS Trans., 1983, vol. 91, pp. 11926. 66. G.X. Sun and C.R. Loper: AFS Trans., 1983, vol. 91, pp. 84154. 67. A.P. Druschitz and W.W. Chaput: AFS Trans., 1993, vol. 101, pp. 44758. 68. R. Kallbom, K. Hamberg, and L.-E. Bjorkegren: Proc. World Foundry Congr., Harrogate, UK, 2006, paper 184/1-10. 69. H. Itofugi and H. Uchikawa: AFS Trans., 1990, vol. 98, pp. 42948. 70. I. Asenjo, P. Larranaga, J. Sertucha, R. Suarez, J.-M. Gomez, I. Ferrer, and J. Lacaze: Int. J. Cast Met. Res., 2007, vol. 20 (6), pp. 31924. 71. A.K.M.B. Rashid and J. Campbell: Metall. Mater. Trans. A, 2004, vol. 35A (7), pp. 206371. 72. G.P. Faubert, D.J. Moore, and K.B. Rundman: AFS Trans., 1990, vol. 98, pp. 83145. 73. M. Hillert and H. Steinhauser: Jernkontorets Ann., 1960, vol. 144, pp. 52022.

METALLURGICAL AND MATERIALS TRANSACTIONS B

VOLUME 40B, DECEMBER 2009801