Salar Niknafs, Faculty of Engineering, University of Wollongong, Wollongong NSW 2522 Australia. +61-0401-492-732; e-mail: firstname.lastname@example.org
Solidification microstructure is a defining link between production techniques and the mechanical properties of metals and in particular steel. Due to the difficulty of conducting solidification studies at high temperature, knowledge of the development of solidification microstructure in steel is scarce. In this study, a laser-scanning confocal microscopy (LSCM) has been used to observe in situ and in real-time the planar to cellular to dendritic transition of the progressing solid/liquid interface in low carbon steel. Because the in situ observations in the laser-scanning confocal microscopy are restricted to the surface, the effect of sample thickness on surface observations was determined. Moreover, the effect of cooling rate and alloy composition on the planar to cellular interface transition was investigated. In the low-alloyed, low-carbon steel studied, the cooling rate does not seem to have an effect on the spacing of the cellular microstructure. However, in the presence of copper and manganese, the cell spacing decreased at higher cooling rates. Higher concentrations of copper in steel resulted on an increased cell spacing at the same cooling rates.
There is a strong relationship between the pertaining solidification conditions and material properties. However, solidification studies in iron-based alloys are particularly difficult due to the high melting point. Real-time observations of solid/liquid interface transitions from planar to more complex morphologies have in the past been restricted to the study of transparent organic materials. Jackson & Hunt (1965) introduced real-time solidification studies in a Bridgman-type directional solidification apparatus of transparent organic materials (such as succinonitrile), which have been shown to display similar solidification behaviour as metals. Succinonitrile solidifies at about 329K, which simplifies the study of solidification phenomena, but the question remains as to what extend such solidification studies replicate the solidification behaviour of steel. Sen et al. (1997) and more recently Yonemura et al. (2006) suggested the use of X-ray transmission microscopy to study the solid/liquid interface during the solidification of metals, but they were unable to observe the evolution of the interface.
The coming of age of high-temperature laser-scanning confocal microscopy (LSCM) has made it possible to study in situ the solidification of metals and such a microscope has been used to make observations of the influence of solidification parameters and alloy additions on the solid/liquid interface instability in low carbon steel. Chikama et al. (1996) were the first to directly observe the planar to cellular and cellular to dendritic solid/liquid interface transition in Fe-C alloys using the LSCM. They conducted solidification experiments on steel specimens by melting the whole sample in an alumina crucible and observing the progressing solid/liquid front during solidification. As a result of melting the whole sample, nucleation of heterogeneous nuclei occurred upon solidification and resulted on the growth of numerous individual grains. Phelan (2002) subsequently proposed a different technique to make possible real-time observations of interface instability in the LSCM by using rectangular samples, in which he melted half of the sample to avoid heterogeneous nucleation of grains following solidification. After establishing the solid/liquid interface in the middle of the rectangle, he cooled the samples at different rates and observed the interface transition from planar into cells. However, due to the presence of a meniscus, high-quality observations could not be made of the progressing solid/liquid interface in this rectangular geometry. In this study, a concentric solidification technique was used to overcome these difficulties and the influence of cooling rate, sample thickness and copper additions on the solid/liquid interface morphology during non-steady state growth could be determined. The effect of copper additions on interface instability of low carbon steel is specifically important because copper is inevitably present in scarp-based electric-arc furnace steelmaking operations, of which the twin-roll strip casting process is an example.
Assessment of high-temperature laser-scanning confocal microscopy as a tool to study instability mechanisms
The study of solid/liquid interface instability has traditionally been conducted in a Bridgman-type furnace using directional solidification experiments. Esaka & Kurz (1984) suggested that the critical solid/liquid interface velocity (Vc) for the planar to cellular transition can be calculated by , where GL is the temperature gradient in the liquid, DL is the diffusion coefficient in the liquid and ΔT0 is the equilibrium melting range at C0. The Bridgman furnace is designed to produce controlled temperature gradients within an electrically resistance heated furnace in which the sample is placed. Desired temperature gradients in the liquid can be produced by a number of tapered heating coils. The apparatus is also equipped with a synchronized motor, moving the stage at the desired rate, while keeping the solidification front in a stationary position. The Bridgman apparatus design allows precise control on the velocity of the solid/liquid interface and provides relatively good control of the temperature gradient in the liquid ahead of the interface. However, solidification events at elevated temperatures cannot be observed in real-time and hence, in situ observations of the solidification of steel cannot be carried out in this system.
In this study, we used a LSCM system and the so-called ‘concentric solidification’ technique, developed and described by Reid et al. (2004) to improve the quality of real-time observations of solidification in a high-temperature laser-scanning confocal microscope at varying cooling rates and thickness of sample. The laser-scanning confocal microscope used in this study was made by Lasertec Corporation, based on Yokohama, Japan. In the concentric solidification technique a thin disk-shaped sample is partially melted in the centre of the disk, producing a liquid pool surrounded by a delta-ferrite rim, as depicted schematically in Figure 1.
The main justification for using the concentric solidification technique is that the surface tension balance between solid, liquid, gas and crucible results in a stretched liquid region with significantly reduced meniscus effect. Consequently, the area at which the solid/liquid interface can be observed is greatly enlarged and the quality of the image significantly enhanced.
Unlike the Bridgman-type furnace, the LSCM apparatus provides the possibility of in situ observations of the progress of solidification as well as the subsequent solid-state phase transformations at temperatures as high as 1873K. A drawback of the conventional LSCM system is that the temperature gradient in the liquid at the solid/liquid interface cannot be controlled. Some modelling attempts have been made by Phelan et al. (2006) to estimate the temperature gradient in the liquid in the concentric solidification technique. Such models revealed that the temperature gradient in the liquid in the concentric solidification technique is continuously increased at decreased pool sizes. As the pool size decreases, the diameter of the surrounding solid rim is necessarily increased and this arrangement results on a greater rate of heat extraction from the liquid into the solid rim. Consequently, in the conventional design of the LSCM it is not possible to apply a constant temperature gradient in the liquid ahead of the solid/liquid interface and unlike the Bridgman-type furnace, the interface instability study in the LSCM reproduces an unsteady-state solidification scenario. As a result the criterion cannot be applied in this study. In addition, the solidification rate in the LSCM is controlled by the rate of cooling of the whole specimen as opposed to controlling the interface velocity in the Bridgman-type furnace. Nevertheless, the in situ observations provided the possibility to study planar to cellular transitions in real-time and revealed some unique phenomena associated with this transition, which will be discussed in more detail below.
A 1.5 kW halogen lamp, located at one focal point of a gold-plated ellipsoidal cavity in the bottom half of the infra-red furnace chamber heats the sample by radiation. The gold-coated ellipsoidal chamber then concentrates the heat on the surface of the specimen, which is located at the other focal point of the furnace chamber. Figure 2 shows a schematic representation of the furnace chamber and the sample holder.
Samples are placed in a cylindrical alumina crucible with an inner diameter of ∼9.2 mm. A platinum specimen holder, to which a B-type thermocouple is attached, holds the crucible in position as shown in Figure 2.
The sample and the holder are then inserted into the top half of the furnace chamber. This section is atmosphere controlled to avoid oxidation of the specimen at elevated temperatures. The top half of the microscope chamber is filled with ultra-high purity argon gas. To ensure a high integrity inert gas, a Super-Clean™ gas filter is placed in the gas train and the purified gas is passed through a stainless steel tube filled with titanium turnings, held at a temperature of 1173K, before in enters the furnace chamber. The purity of the resulting carrier gas before entering the furnace chamber is better than 99.9999%.
Cylindrical LSCM samples of 9 mm diameter were cut into thin disks using an Accutum 3500 diamond cutting machine and ground on 800, 1200, 2400 and 4000 abrasive papers before being polished by standard metallographic techniques, using 6 and 1 μm diamond powder. Two groups of samples were studied as shown in Table 1. Alloy A is low-carbon, low-alloyed steel, where the alloying elements are kept at the lowest possible level to eliminate their effect on the solidification behaviour. Group B are copper and manganese containing steels with approximately the same carbon content of 0.05(wt)%, as Alloy A.
Table 1. Chemical composition of specimens.
In Group B samples, copper and manganese were deliberately added to investigate the effect of these alloying elements on the solidification behaviour because these elements change the stability of interface by their segregation and their influence on the solid/liquid interfacial energy. In addition, the solidification rate (cooling rate in the LSCM or the interface velocity in directional solidification setup) and the temperature gradient also affect the interface stability. By contrast to the Bridgman apparatus where the temperature gradient can be precisely controlled, the temperature gradient cannot be controlled in the confocal microscope. The cooling rate can be controlled precisely up to a maximum of 100 K min-1 by using an Omron ES100P digital PID controller, but cooling rates higher than 100 K min-1 and up to a maximum of 2500 K min-1 can be achieved by turning off the heating lamp in the microscope for short periods of time. However, there is a high risk of damaging the specimen holder in the microscope by thermal shock and, therefore, solidification experiments at high rates were only performed on a limited number of experiments.
Because real-time observations in the LSCM are limited to events occurring on the surface of a specimen, it was necessary to ensure that the solidification events observed on the surface are representative of bulk behaviour. For this reason, sample thicknesses of 150, 250 and 450 μm were examined in Alloy A. Additions of copper and manganese to Alloy A reduced the surface tension of the liquid and as a consequence, the liquid pool ruptured repeatedly in 150 μm thick Group B alloys. It was found that a minimum thickness of 250 μm was required to establish a stable liquid pool in Group B samples. Geometrical details of the specimens used in the experiments are shown in Table 2.
Table 2. Sample geometry.
Once a high integrity argon atmosphere was established in the LSCM furnace chamber, the specimens were heated at a rate 45 K min-1 into the δ phase. The heating rate was then reduced to 1 K min-1 until the centre of the specimen begun to melt and a liquid pool was established. When the pre-determined pool size was reached, the temperature was kept constant and sufficient time was allowed to reach a steady state. Samples were then cooled at cooling rates ranging from 5 to 100 K min-1 and the solidification behaviour observed in real-time. A feature of the concentric solidification technique is the reproducibility of experiments. Once the solidification observations are made, the melt pool can be recreated to the original size with the same delta-ferrite grain structure in the rim, using a second heating cycle. Therefore, the effect of cooling rate on solidification from the same initial rim microstructure and pool diameter can be obtained. The in situ observations are recorded in real-time digital movie format at 30 frames per second. The resulting movie is then analysed by the use of SolTrack image-processing software, developed by Griesser (2012). By using this software, it is possible to precisely measure the pool radius at any given time and also track the progression of the solid/liquid interface with time.
Results and discussions
In situ observation of solidification in alloy A
Alloy A samples with a thickness of 150 μm and an initial pool radius 3.00 ± 0.15 mm were cooled at cooling rates that varied between 20 K min-1 and 100 K min-1. In this range of cooling rates the planar interface developed a perturbed wave which eventually became stable and developed into cells, as shown in Figure 3.
When a cooling rate of 150 K min-1 was imposed on 150 μm thick samples, a different solidification microstructure was observed. Following the planar to cellular transition and at the onset of cellular growth, the cellular structure transformed into the microstructure presented in Figure 4. Closer examination of the ‘black spots’ in Figure 4 revealed that they are grooves which form at the liquid bays in the inter-cellular regions.
To further study the morphology of these grooves, the solidified delta-ferrite structure was oxidized immediately upon solidification. The ‘Oxisurface’ technique developed by Niknafs (2007) allows retention of the topographic surface structure and traces of the grain boundaries of the oxidized phase at room temperatures. Following further cooling to room temperature, the sample was polished with 1 μm diamond powder and the surface studied in an SEM. The observed surface structure is shown in Figure 5.
The SEM images reveal shrinkage cavities, which are associated with the planar to cellular interface transition and appear as regular grooves on the surface of the solidified samples.
In specimens 250 μm thick, the perturbation wave initiated at delta-ferrite grain boundaries/liquid metal contact in similar vein to 150 μm thick samples. However, Figure 6 shows that arrays of solid islands formed at regular intervals ahead of the solid/liquid interface. This observed behaviour is possibly due to the fact that the temperature at the centre-line of the disk-shaped specimen is lower than the temperature at the free surface and hence a temperature gradient is introduced at the thickness direction. The imposed temperature gradient in the thickness direction could initiate the planar to cellular transition in the thickness direction. When these cells grow and reach the free surface, they appear as islands as shown in Figure 6.
450 μm thick samples, displayed different solidification behaviour. At cooling rates less than 150 K min-1 an array of dendrites formed ahead of the progressing solid/liquid interface and Figure 7 depicts the typical solidification microstructures of a 450 μm thick sample of Alloy A that was cooled at rates of 20 and 75 K min-1, respectively.
These dendrites, which will be referred to as ‘bulk dendrites’ are orientated such that their primary arms are underneath the surface with the secondary and ternary arms hitting the free surface in the liquid pool. Figure 7(a) also shows that ‘bulk dendrites’ are formed ahead of the solid/liquid interface at the surface whereas the solid/liquid interface on the surface is still planar. Therefore, the observed dendrites are not formed at the surface as a result of the temperature gradient/cooling rate conditions at the interface on the free surface, but their formation is driven by thermal and diffusional conditions in the bulk of the specimen.
Figure 8 schematically illustrates the cross section of Alloy A samples with different thickness.
In the LSCM setup, the top and bottom of the specimen are the hottest parts, where the incident laser beam is focused. Consequently the centre point between these two surfaces is the coldest. In thin (150 μm thick) samples, the solid/liquid interface in the bulk is approximately linear as shown in Figure 8(a). By contrast and due to the higher rate of heat extraction into the solid in thicker samples (450 μm thick), the interface in these samples forms a convex shape as illustrated in Figure 8(c). Hence, as the sample thickness increases, a component of the heat is extracted at an angle to the planar solid/liquid interface and a temperature gradient is established in the thickness direction. In addition, the lower temperature in the sample bulk with respect to the surface favours destabilization of the solid/liquid interface, before the interface on the free surface is perturbed. The degree to which these events occur in the bulk is schematically illustrated in
Figure 8 by the values of X1 and X2, where X1 and X2 are the maximum distance between the tip of a dendrite in the bulk interface and the surface. Accordingly, formation of ‘bulk dendrites’ in 450 μm thick samples as shown in Figure 7 appeared at a greater distance from the interface on the surface than the ‘solidification islands’ in 250 μm thick samples as shown in Figure 6.
To further investigate the effect of cooling rate on the solid/liquid interface transition observed on the surface, Alloy A samples were cooled at varying rates of up to 500 K min-1 and the initial pool radius (R0) and the pool radius at the moment of planar to cellular transition (Rp) were measured using SolTrack software. Because the bulk interface became unstable before the surface interface in 450 μm samples, the effect of cooling rate was investigated in 150 and 250 μm thick samples only. Figure 9 shows the movement of the initial planar interface before it transforms into a cellular structure in 150 and 250 μm thick samples respectively at varying cooling rates, showing the difference between the initial pool radius (R0) and the pool radius at the moment of planar to cellular transition (Rp). It appears that the interface has to move a minimum distance before it transforms into a cellular structure. This distance, which is attained at a cooling rate higher than about 100 K min-1 is approximately 131 μm.
An important observation was that in Alloy A samples, where solute segregation is minimal, the cooling rate had little effect on the spacing of the resulting cellular structures which was about 57 μm. To provide the extreme conditions for the cellular to dendritic transition (CDT) to occur, the microscope power was temporarily turned off, by which a cooling rate of about 2500 K min-1 was achieved. Upon turning off the power, the planar interface instantly transforms into a cellular and then dendritic structure as depicted in Figure 10.
The dendritic microstructure depicted in Figure 10 seems to have formed as a result of the temperature gradient on the specimen surface (G1) and, therefore, is different in origin than the ‘bulk dendrites’ which formed due to the temperature gradient in the thickness direction. In Figure 10 the primary and secondary arm spacings are 128 and 50 μm, respectively.
In situ observation of solidification in group B alloys
Unlike Alloy A that contained minimal solute elements, Group B samples contained various amounts of substitutional alloying elements, most notably manganese and copper. These samples were produced in a steel mould in an induction furnace with argon gas shrouding. After experimenting with different sample thicknesses, 250 μm was chosen as the minimum thickness for carrying concentric solidification experiments on Group B alloys. The planar to cellular transition of the solid/liquid interface in Group B alloys always started at the free surface at a cooling rate of 5–100 K min-1. After the planar interface on the surface developed into a cellular structure, bulk events occasionally started to appear as a planar/cellular transition ahead of the interface on the free surface.
Figure 11 shows an array of cellular grains, which appears to have originated in the bulk before reaching the free surface.
In Group B alloys, the distance the planar interface progresses before it becomes unstable is reduced at increased cooling rates as shown in Figure 12. This behaviour is similar to that observed in Alloy A samples. There appears to be a minimum distance the planar interface has to move before it transforms into a cellular pattern. This minimum length, which is found at higher cooling rates, is about 110 μms.
Unlike Alloy A, cooling rate directly affected the spacing of the cellular structures in Group B alloys. Starting with a liquid pool of 3.80 ± 0.15 mm in radius, alloys of Group B1 were cooled at the rates of 10, 50 and 100 K min-1 and the resulting cellular interfaces are illustrated in Figure 13.
Figure 13 shows that at increased cooling rates the cell spacings of Group B1 samples are reduced. The added copper and manganese in Group B alloy alters both the solute segregation and the capillary effect (solid/liquid interfacial energy) of the low carbon steel. The difference in the cooling rate dependence of the cellular spacing of the Alloy A and Group B alloys may therefore be attributed to the compositional differences between Alloy A and Group B alloys. To further investigate the effect of compositional differences on cellular arm spacings, alloys of Group B with copper additions were cooled at varying rates and the spacing of the resulting cellular microstructures are compared in Figure 14. Cooling rates of 10 and 100 K min-1, respectively were applied and the initial pool radii were in the range of 3.80 ± 0.15 mm.
Figure 14 shows firstly that there is a significant decrease in cellular spacing at an increased cooling rate. Secondly, the cellular spacing is increased by an increase in copper concentration.
The concentric solidification technique in a high-temperature LSCM can be used to study bulk solidification events by observations on the surface as long as the specimen is thin enough. In this study, bulk effects began to influence surface observations when the specimen thickness exceeded 150 μm.
Cooling rate had little influence on the cellular spacing of an Fe-0.05(wt)%C alloy, which was about 57 μm. Increasing the sample thickness resulted in cellular structures in 250 μm thick specimens and dendritic structures in samples 450 μm thick.
The planar to cellular transition results in the formation of shrinkage cavities in the intercellular spaces, which initiates at the onset of the planar to cellular transition.
At increased cooling rates in experiments conducted by the concentric solidification technique, the distance the planar interface moves before it develops a cellular structure is reduced. There appears to be a minimum length that the planar interface has to travel before it is transformed into a cellular structure.
Additions of copper and manganese to the Fe-C alloy reduced the cell spacing and the cooling rate had a significant effect on cell spacing. With 0.06(wt)% and 1.1(wt)% manganese added to the base-alloy, the cellular spacing at a cooling rate of 100 K min-1 decreased to 24 μm.
An increase in copper content from 0.06(wt)% to 0.3(wt)% increased the cellular spacing from 24 to 39 μm, in specimens cooled at a rate of 100 K min-1.