1. Introduction
In light alloys, particularly aluminium alloys, tensile properties are dramatically affected by the way in which the metal is cast [
1]. In a normal pouring operation, the oxide film on the surface can become entrained by either the folding over of the surface, or by the impingement of drops and splashes. Both mechanisms bring together regions of the dry upper surface of the film, as a dry-surface to dry-surface impingement. Because the surfaces are composed of highly stable ceramics (typically Al
2O
3) with melting points over 2000 °C and are microscopically rough like sand-paper, there is little or no bonding between the two films. This double film, called a ‘bifilm’ for convenience, now acts like a crack in suspension in the liquid. It has been demonstrated that turbulent pouring can introduce dense populations, a snowstorm of cracks into the liquid,
Figure 1. The bifilms can be extremely thin, merely nanometers thick, because the oxide films have limited time to thicken when forming on a surface in the process of expanding and submerging, which will take only milliseconds. However, occasionally, chunks of thickened surface oxide can be involved creating a highly asymmetrical bifilm, with one film mm thick and the other only nm. However, its crack-like behaviour is not expected to be impaired.
There is much evidence that bifilms form even in vacuum melting and pouring conditions, because there always seems to be sufficient oxygen in normal industrial vacuum systems. Naturally, the bifilms are thinner, making them more difficult to detect, and giving the impression that the metal is cleaner, which in a way it is. Nevertheless, the area of the bifilms is probably unchanged, because the accidental geometry of the pour will not have been affected. Clearly, the way to eliminate bifilms is not to attempt to eliminate oxygen in the environment, but simply to avoid the folding over or splashing of the liquid surface. Pouring of metals is to be avoided if possible, and casting systems are now available to achieve this [
2]. In the meantime, while we continue to use our current casting technology, there are other ways forward which will be mentioned below.
A population of bifilms in an aluminium alloy is shown in
Figure 2.
Figure 2a is the radiograph of the small sample of metal poured turbulently so that the bifilms are scrambled and ravelled into compact shapes, visible as faint shadows.
Figure 2b shows the same metal poured at the same time into a mould over which a bell jar was placed and evacuated to reduce the pressure during cooling. The air trapped in the ‘air-gap’ of the bifilms is thereby expanded to unfurl the defects, showing their true sizes, in the region of 10–15 mm, in the radiograph.
In steels containing aluminium or chromium as a major alloying element, alumina-rich or chromia-rich oxide films can occur [
2] which can become entrained as double films (bifilms) in the liquid alloys. Their presence has been confirmed in ultrasonic observations which have been used to monitor their number and size. At lower temperatures in nickel-based alloys containing aluminium, their presence is commonly seen on fracture surfaces, but in steels at higher temperatures and generally lower Al levels, the films have been thinner and often fragmented into particles by Oswald coarsening. In this disguise they have tended to escape identification as originally films [
2]. (This point is taken up again later in the discussion of Figure 5).
This paper explores the possibility that the hot ductility of TWIP (twinning induced plasticity) and TRIP (transformation induced plasticity) steels may be explainable, at least in part, by the presence of casting defects although more experimental evidence is required to confirm such a theory.
The hot behaviour of these steels is dauntingly complex, so this author has relied extensively on the recent review by Mintz and Qaban [
3]. These authors draw attention to the importance of the poor hot ductility of these steels which is seen in cracking behaviour during straightening subsequent to continuous casting, particularly when the Al is high (1.0–1.5%). This failure mode can be a major limitation to the use of these otherwise highly attractive engineering materials.
3. Hot Ductility in TWIP and TRIP Steels
It is significant that Al is often a necessary alloying element to achieve the desired properties of TWIP and TRIP steels, but, perversely, has the worst influence on hot ductility [
3]. Traditionally, this behaviour has been attributed to the presence of AlN on grain boundaries. Naturally, the AlN has been assumed to be brittle, since it leads to the familiar ‘rock candy’ fractures, in which the prior austenite grains, apparently largely un-deformed, are clearly revealed on the fracture surface (
Figure 3).
Interpreting these well-known observations from a bifilm perspective, it is worth drawing attention to the fact that aluminium oxide bifilms are to be expected in liquid steels as a result of the turbulent handling of the liquid metal. If analogous observations from Al alloys apply to steels, the Al
2O
3 bifilms would form first, followed by AlN which would precipitate later, forming on the alumina substrates provided by the bifilms [
2].
On a fracture surface, the precipitated AlN film would be sited on the underside of the thin oxide of the bifilm and would be viewed through the very thin oxide and would normally be overlooked. However, the presence of the oxide could be identified by careful observation in a scanning electron microscope, identifying the oxide by its folds and creases showing that it had been formed on a mobile liquid surface. Such oxide films appear to have been observed on fracture surfaces of TRIP 980 steel [
5].
In most metal and alloy systems researched so far, it is the alumina (occasionally chromia) bifilms which form first in the liquid metal [
6]. All precipitates of second phases, including carbides and nitrides such as AlN, appear to form subsequently and do so, on the bifilms as a favoured substrate. This is the reason for many carbides and nitrides in steels being seen to be associated with cracks. It has always been assumed that the cracks are due to cooling stresses. However, in view of the extreme strength of most carbides and nitrides, and their atom by atom deposition from solution, one would expect such structures to result in a solid, free from stress-raising defects, and therefore having no sites to initiate failure when stressed. There is little doubt (again with Al alloys and other alloys behaving similarly) that the cracks associated with carbides and nitrides appear because they have formed on a crack, a bifilm, as a favoured substrate. These cracks have not been formed by stress; they have formed by surface entrainment. There is now a large literature confirming this mechanism [
2,
7].
Hence, the turbulence of casting in air would be expected to produce double oxide films (not nitride films) based on Al2O3, and will take the form of cracks which have an oxidised internal surface.
The extraordinary and unique features of the bifilm consist in its exterior faces consisting of those interfaces which were the original underside of the surface oxide film on the liquid metal, which are therefore in perfect atomic contact with the liquid matrix, and perfectly wetted, in contrast to the interior interface which is perfectly non-wetted (
Figure 1). This duality of ‘interior interfaces un-bonded but exterior interfaces perfectly bonded (to the matrix)’ appears to be central in the mechanical and chemical behaviour of the current quality problems with our metals.
The exterior wetted interface seems to be a favoured substrate for the nucleation and growth of intermetallic and second phases, so if AlN is formed, it is most likely to precipitate on the outer surfaces of oxide bifilms (
Figure 4) [
2].
AlN might be formed even in the presence of low N in solution in the melt, because trapped air in the bifilm (in the ‘air gap’ between the microscopically rough dry oxide surfaces) might provide its 4/5ths of the entrained air (78%N, 21% oxygen) which would continue to react with the metal, thickening the existing oxide film with a layer of nitride [
7]. The aluminium oxide/nitride interchanges have been intensively studied in Al alloys where it is now known that the oxide always forms first, and continues to grow while oxygen in available, but as soon as the oxygen is consumed, then the nitride forms [
6]. Because air is mostly nitrogen, the nitride layer can be significantly thicker than the oxide and may obscure the oxide layer. On a fracture surface the nitride from entrained gaseous nitrogen in the air will appear uppermost, on top of the oxide. In contrast, if N in solution is sufficiently high to precipitate as a nitride, the nitride is almost certain to precipitate on the wetter outer surface of the oxide bifilm, so that on the fracture surface the nitride will appear below the oxide. In either case, the oxide is usually the thinner film so it appears that the AlN is the phase causing the fracture. However, once again, the oxide bifilm is the pre-existing crack, and the AlN would be expected to make a negligible contribution to the fracture.
On occasions, probably when the Al content is high, the thickness and stability of the alumina bifilms prevents this breakup under the driving force of reduction of surface energy, and simply remains as slabs of alumina and/or AlN (
Figure 4).
At lower temperatures in the light alloys, the oxide films can be clearly seen and identified by their generally smooth surface, crossed with fine folds and ripples derived from its origin as a vanishingly thin film on the surface of a liquid.
In contrast, at steelmaking temperatures the morphology of very thin films is not stable. Because of the higher rate of diffusion in high temperature metals the film tends to break up into particles, coarsening to reduce its surface energy [
2,
7]. On a polished micrograph the continuous line of a sectioned film becomes separated as beads on a string.
Figure 5 gives examples. In these cases, the separate inclusions may be separately nucleated inclusion particles on a fragmented bifilm. Alternatively, the bifilm remains inert and intact, but the inclusions have separated by the coarsening reaction. Just possibly, both the bifilm and inclusion films have together balled up to reduce their combined energies. Only careful research will elucidate the actual mechanism for the formation of this string of particles decorating the bifilms occupying the prior austenite grain boundaries.
The fractures along prior austenite grain boundaries are understandable, because during freezing, the growing austenite grains will push bifilms ahead (the grains cannot, of course, grow through the ‘air gap’ of the bifilms). On impinging against other grains, the bifilms will naturally be trapped at the boundary. In a real sense, the bifilms are the grain boundaries (
Figure 1).
An analogous situation occurs in the solid state when recrystallisation occurs: a migrating boundary incurring a bifilm cannot cross its ‘air layer’. For this reason, the majority of bifilms exist in grain boundaries, leading to intergranular failures. Hence, TWIP steels with 0.75%Al and 1.5%Al as shown in Figure 14 of reference [
3], numerous alumina bifilms are present causing grain boundary embrittlement and loss of ductility. The TWIP alloy without aluminium will have fewer and/or weaker bifilms, so boundaries will be enabled to migrate more easily, permitting some recrystallization, and so benefiting ductility.
Where recrystallization (or grain growth) occurs by the motion of boundaries at high temperatures, the eventual trapping of all the grain boundaries would bring recrystallization to a halt. At that stage, the continuing cooling strains will activate strain concentration in the grain boundaries, particularly in those boundaries which are composed of large areas of ‘air gap’. These weakened boundaries will shear first. The strain transfer to surrounding boundaries will subsequently tend to shear less weakened boundaries. The grain boundary sliding action will lead to failure, limiting ductility, as shown in
Figure 6.
It is interesting that the excessive grain boundary sliding of the un-recrystallised austenite appears to reduce the hot ductility of TWIP steels. It seems probable that this can be understood in terms of the higher bifilm content of un-recrystallised boundaries which will have gathered their bifilms by impingement of the growing austenite grains in the solidifying liquid. Such boundaries will be incapable of migrating as in recrystallization, but will naturally slide easily if parts of the boundary are not in contact, but are separated by a microscopically thin ‘air layer’.
The interesting current confusion in the literature is illustrated by the fracture surface of the tensile test piece shown in
Figure 7, which the authors interpret as AlN on the surface of a film. The film is most probably the half of an alumina bifilm, but probably not easily identifiable as alumina because of its extreme thinness. The globular features appear to be a liquid phase, which may have originated as a favoured precipitate on the bifilm (the wetted underside of the film as viewed in the image) but has exuded through holes and cracks in the film. Although the exudate is labelled as AlN, the AlN signal is probably confused by AlN which often precipitates favourably on alumina bifilms (once again on its wetted underside as seen in the image), and in such a thickness as to be easily identified by EDAX through the overlying thin alumina. In any case of course, AlN has a melting point over 2000 C whereas the exudate appears to be a phase with a melting point below that of the matrix steel. A careful re-examination would be worthwhile.