Abstract
Strips of magnesium alloy AZ31-B were cast on a simulator of a horizontal single belt caster incorporating a moving mold system. Mixtures of CO2 and sulfur hexafluoride (SF6) gases were used as protective atmosphere during melting and casting. The castability of the AZ31-B strips was investigated for a smooth, low carbon steel substrate, and six copper substrates with various textures and roughnesses. Graphite powder was used to coat the substrates. The correlation between strip thickness and heat flux was investigated. It was found that the heat flux from the forming strip to the copper substrate was higher than that to the steel substrate, while coated substrates registered lower heat fluxes than uncoated substrates. The highest heat flux from the strip was recorded for casting on macrotextured copper substrates with 0.15 mm grooves. As the thickness of the strip decreased, the net heat flux decreased. As the heat flux increased, the grain sizes of the strips were reduced, and the SDAS decreased. The mechanical properties were improved when the heat flux increased. The black layers which formed on the strips’ surfaces were analyzed and identified as nanoscale MgO particles. Nano-Scale particles act as light traps and appeared black.
1. Introduction
Magnesium is the lightest structural metal in common use [1]. Similarly, supplies of magnesium ores are virtually inexhaustible. Magnesium alloys normally have very good castability and machinability, as well as excellent specific strength and stiffness [2]. However magnesium alloys have some difficulty during rolling due to hexagonal close packed (hcp) lattice structure [3]. Meanwhile, a fine grain structure increases strength and ductility by promoting the operation of nonbasal slip systems and limiting twinning in magnesium alloys [4]. Strip casting of magnesium has become important in recent years. For reducing the cost of thin sheets of magnesium alloys, strip casting technologies such as horizontal single belt casting (HSBC), twin roll casting (TRC), and twin belt casting (TBC) have been developed [3]. With a strip casting process, magnesium alloy strips can typically be produced in thicknesses of 1–10 mm [1]. Direct strip casting, or HSBC, as a near-net-shape casting process, has potential use in the processing of aluminum, copper, zinc, and lead alloys, directly into sheet products.
Generally, most metals and alloys are amenable to direct casting into plates, strips, or ribbons. However, a metallurgical understanding of these materials is needed to determine their suitability for casting into thin-gauge strips. Technically, the evaluation of a process must take into account the melting point of the alloy, the freezing range, the oxidation resistance in both the liquid and solid states, the heat transfer behavior, the fluidity of the melt, and the number and type of the liquid-to-solid, and solid-state, transformations that may occur. There is a particular emphasis on magnesium alloys as these are the major candidates for large-scale production by this processing route, given the poor hot rolling capabilities caused by its hexagonal close packed structure [5, 6]. The present study was carried out to investigate the possibility of directly casting magnesium strip products on a horizontal single belt caster (HSBC). The alloy studied was the AZ31-B magnesium alloy.
2. Experimental Procedure
2.1. Raw Materials and Melting Unit
Commercial magnesium alloy AZ31 grade B bar ingots obtained from Magnesium Electron Co. were used as raw material in the present experiments. The raw bar ingot materials were cut into smaller pieces and prepared for melting. Graphite powder, comprising particles of 0.5–0.6 μm, obtained from Asbury Carbons Co., was used to coat the casting substrates. The graphite was of a synthetic variety, and the particles in general were flake shaped.
2.2. Strip Casting Simulator
A schematic overview of the strip casting simulator is shown in Figure 1. The equipment includes the following: a containment mould, a substrate onto which the melt can be poured, a tundish, a motor that drags the substrate at preselected casting speeds, and a data acquisition system. The simulator can be set to produce strips of mm dimensions. The casting substrates can be coated with different materials such as the graphite used in the present research. In order to measure local heat fluxes, two K type thermocouples were placed in each segment of the substrate; one was set near the surface and the other was placed slightly below the first thermocouple.
2.3. Substrate Specification
Two types of material substrates were used in the present experiments, steel with a polished surface and pure copper with six differently textured inserts/segments. Figure 2 illustrates the schematics along with photos of the six copper substrates ( inch segments). Table 1 provides dimensional specifications of different areas of copper chill substrate. Each segment had a different macrotexture.
(a)
(b)
2.4. Melting and Pouring of Molten Metal
The melt was prepared in a closed steel crucible, in which the atmosphere of the melt was protected from air ingress at all steps in the casting process, so as to avoid any oxidation or possible burning. The protective atmosphere comprised a mixture of sulfur hexafluoride, 0.5% SF6, in a carrier gas of carbon dioxide, 99.5% CO2.
The molten metal was heated using an induction furnace, to a temperature of 710°C, which is above the pouring temperature, 700°C, in order to provide sufficient superheat for skimming and metal transferring purposes. The molten metal was poured directly from the crucible into the tundish to avoid any large temperature drops. Before pouring, the tundish was preheated up to 150°C for drying the refractory walls inside the tundish. At the same time, the substrate was preheated to 30°C for avoiding any moisture. Transferring and pouring the molten metal were done manually. The substrate was propelled by a loaded spring at a constant speed of 0.5 m/sec.
2.5. Calculating Heat Flux and Applying IHCP to a Horizontal Single Belt Casting Simulator
Beck put forward a nonlinear estimation method to deal with phase changes and temperature dependent thermal properties of the solidification process used for solving the IHCP [9]. To treat experimental data, statistical principles and the concept of amplitudes temperatures are applied to the thermal capacity and heat conduction of the substrate during subsurface temperature measurements. This application is performed using a nonlinear estimation method to solve the delayed and diminished thermal response problems. The heat flux is taken to be a constant or a linear function of time within a given time interval. This is the principle of Beck’s nonlinear estimation technique, according to which the heat flux is then determined for that period according to the following function: where is the number of internal points in the temperature measurement excluding those used for boundary condition. is the number of temperature measurements per time interval. is the future number of time intervals considered for the heat flux calculation at each time interval. and are the calculated and measured temperatures of location and the time instant , respectively [10].
The IHCP (Inverse Heat Conduction Problem) applied for the interface between the melt and substrate is shown schematically in Figure 3; the bounding conditions for the governing differential equation (2) are expressed in (3), (4), and (5) [11] as follows:
The corresponding differential equation and boundary conditions for sensitivity coefficients are [2] where is a sensitivity coefficient and subscript denotes the time when it was applied. In this investigation, temperatures are recorded using two thermocouples for each segment of the substrate connected to an Omega Data Acquisition System. In order to convert the time versus temperature data to time versus heat flux, the necessary IHCP software was developed from first principles by Isac et al. [7].
3. Results and Discussion
3.1. Effect of Substrate Material on Heat Flux
Materials of the substrate in the strip casting process are one of the important issues which affect casting parameters. Table 2 shows the values of the maximum heat fluxes recorded for different substrate materials. The casting conditions (speed, superheat, and thickness) were the same.
The thermophysical properties of different substrate materials used in the present experiments are summarized in Table 3.
The measured interfacial heat flux is related to the thermal conductivity and thermal diffusivity of the substrate. Hence, the large difference between the thermal conductivity, , of steel and copper is reflected in the value of the heat flux at the mould/melt interface. The bare copper substrate with no graphite coating had the highest cooling capacity and produced the highest heat flux at the mould/melt interface. Because of the poor thermal conductivity of graphite and of steel, the steel substrate coated with a graphite layer had the smallest heat flux at the mould/melt interface.
3.2. Effect of the Surface Topography of the Substrate on the Interfacial Heat Flux
In the HSBC casting procedure, the metal/mould interface is not in perfect thermal contact. Localized heat flows through the actual contact points between the metal/mould interfaces are significantly less than in the case of perfect contact.
Consider the melt hanging between two parallel running peaks at a distance of apart; then the melt sag () depends on the melt surface tension () and the metallostatic pressure () for a nonwetting substrate. The radius of the metal curvature, , is [12] where is the melt surface tension, is the melt density, is the gravitational constant, and is the melt height. The melt sag can be calculated as [5, 13]
However significant thermal resistance exists at the substrate/melt interface because of trapped air, oxide layers, gaps made by shrinkage of the solidifying shell from the interface, thermal expansion of the mold, and so forth.
Table 4 presents the results of measured heat fluxes for different surface topographies. The maximum heat fluxes were measured in segment III, for both coated and uncoated substrates, while minimum heat fluxes were measured in segment VI. We can conclude that the substrate topography of segment number six had an increased thermal resistance compared to that of segment number three.
Consequently, the maximum surface contact of melt-mould occurred with segment III and the highest number of air pockets was in segment VI.
3.3. Effect of Coating of Substrates on Measured Heat Fluxes
Figure 4 illustrates the measured heat fluxes of casting strips of magnesium AZ31-B alloy on substrates with various roughnesses, with and without a graphite coating. The measured heat fluxes on graphite-coated substrates were all lower than those on equivalent (matching) bare substrates.
3.4. Effect of Strip Thickness on Heat Flux
The distance between the nozzle and the substrate was decreased, so as to reduce the thickness of the strips produced from 3 to 1 mm. The substrate was coated with a layer of fine graphite to a thickness of 60 μm. Based on Figure 5 and the following equations, it is evident that if one increases the strip thickness, the amount of heat transferred from the strip to the substrate must increase. Essentially, the total amount of the heat loss per unit area required to complete solidification of the liquid metal is where is the density of liquid metal, is the thickness of strip produced, is the specific heat, is the casting temperature, is the melting point, and is the latent heat of fusion/mass. Note that is the sensible heat of the melt and is the latent heat component.
3.5. Microstructural Analysis
Figure 6 shows a microstructure of AZ31-B magnesium alloy close to the strips top surface when casting 3 mm strip on segment III of the copper substrate without any graphite coating.
It has been claimed by Rappaz and Gandin [13] that the dendrites developed at the interface with the substrate were not inclined, their growth direction being verticality upwards given that the direction of the fluid flow directly influences the growth direction of dendrites at the melt/substrate interface. The dendritic growth mechanism in magnesium AZ31-B alloys has been elaborated by Vander [14]. It is concluded that a similar principle governs the dendritic growth for magnesium AZ31-B alloy investigated in this work. During the solidification of the alloy, a symmetrical solute field at the dendrite tip is established. Accordingly, the solute gradient is uniform on both sides of the dendrite tip during the growth. Thus the main origin of the dendrite growth direction is deemed to be the solute distribution. According to the literature [15, 16], the nonhomogeneous distribution of solute atoms between dendrite arms in a magnesium AZ31-B alloy is mainly due to the fact that the solidification takes place over a range of temperatures. In fact, there is insufficient time for atomic diffusion to redistribute the solute both within the liquid in the vicinity of the solid-liquid interface and within the solid, since cooling occurs so rapidly through the two-phase (L + S) regime. By comparing the microstructures of the bottom surfaces of strips, it can be concluded that the grain size depends significantly on the heat flux. The results of microstructural analyses of all samples are summarized in Table 5.
As shown in Table 5, with decreasing heat flux and increasing thermal resistance between the substrate and the strip, the grain size increases. The grain size of the top surface is larger than that of the bottom surface because of solidification delay due to smaller heat fluxes at the top surface. Secondary dendrite arm spacing (SDAS) is found to be directly related to heat flux and thermal contact resistance. The SDAS decreases when air pockets are generally trapped at the substrate/melt interface, which dramatically reduces the heat flux, as explained earlier. The occurrence of these air pockets can subsequently influence the microstructure of the strips. Consequently, the grains tend to grow up from the bottom surface to the top surface resulting in a columnar pattern of grain growth. The effect of these air pockets on the grain structure is clearly shown in Figure 7. A similar observation pertaining to the effect of air pocket formation on the strip’s microstructure has been reported by Dubé et al. [17].
3.6. Phase Analysis
Scanning Electron Microscopy was used to analyze the chemistry and morphology of the secondary phases present in the strip microstructure, as shown in Figure 8.
The microanalyses using EDX technique, were performed on the intermetallics marked in Figure 8, in order to identify their composition. Figure 9 presents one X-Ray spectrum of the analyzed intermetallics.
EDX analysis, demonstrated in the associated spectra, indicates that the phases contain Mg, Zn, Al, and Mn. The presence of magnesium in the spectra could originate from either the matrix or the intermetallics. Based on the information from the phase diagram, the matrix is α-Mg, and the particles are Mg-Al-Zn phase (most likely (Al,Zn)49Mg32) and Al-Mn (it could be a mixture of Al11Mn4, Al8Mn5, Al9Mn11, and β-Mn(Al)) intermetallics. However, according to the investigation by Cao et al. [18], the α-Mg is a solid solution of Mg-Al-Zn-Mn.
3.7. Analysis of Mechanical Properties of AZ31-B Strips
Figure 10 and Table 6 summarize the mechanical properties of the strip cast AZ31-B strips. The tensile and yield strength are calculated using equations TS (MPa) = 3.4 × BHN and . Local Vickers hardness was taken on the cross section of the strips. There was no significant difference between hardness values taken at the top and the bottom surfaces of the strips.
Based on the experimental results presented above, as the heat flux increases, the hardness and the other mechanical properties increased. In the case of 1 mm thick strip, because of lower weight pressure of the strip on the substrate, the thickness of the air gap between the strip and substrate was increased, and the heat transfer rate from strip to substrate decreased. Based on the lower global heat fluxes, represented as the area under the heat flux-time curve, in strips with 1 mm thickness, grain sizes were large and mechanical properties were lower. This issue was presented in detail in Section 3.5.
3.8. Evaluation of the Black Layer on AZ31-B Strip Surface When Cast in Air
To protect magnesium alloys from oxidation during strip casting, a mixture of CO2 and SF6 is usually used. If not, a black layer film on the surface of the strip forms as shown in Figure 11. This layer is strongly adherent, and severely compromises the surface quality of the strip for commercial proposes.
XRD tests were performed in order to analyze the nature of the black layer, as shown in Figure 12.
Figures 13(a) and 13(b), respectively, present an SEM image and an EDX spectrum of the black layer formed on the strip surface when cast in air.
(a)
(b)
It is well known that, under normal conditions, MgO is white and MgAl2O4 is colorless. According to XRD and EDX patterns, the major component of the layer on the strip is definitely MgO. As shown in the SEM image, the particle sizes of MgO are around 0.1 μm. According to the literature [19–22], if the particle size of the oxide is so small that it acts as a light trap, then it will appear black. It should be noted that Mg peaks in the XRD and EDX patterns emanated from the base metal, not from the thin black layer of MgO particles forming on top of the cast strip. Clearly, it will be necessary to use a protective atmosphere in the commercial strip casting of magnesium alloys.
4. Conclusions
In the present research, the effects of casting parameters on the properties of strips of AZ31-B alloy have been investigated. The effect of heat flux on the microstructure and mechanical properties were studied as well. The following conclusions can be drawn.(1)As substrate roughness increased beyond 0.15 mm for macroscopically grooved substrates, the thermal resistance increased, while heat fluxes decreased.(2)As the thickness of strip increased, the heat flux into the substrates increased.(3)As the interfacial heat flux increased, the grain size and the SDAS across the strip were decreased.(4)No significant differences were recorded between hardness values taken at the top and the bottom surfaces of the strips.(5)As heat fluxes increased and the grain sizes decreased, mechanical properties, TS, YS, and HV, all increased.(6)Microstructural analyses of AZ31-B strips revealed that the finest grain sizes and lowest SDAS were obtained using a copper substrate versus a steel substrate.(7)Coated substrates reduced the capability of heat extraction but stabilized the dimensions of strip and gave a good surface quality.(8)The black layer on the strips cast in air is composed of small particles of MgO (in the 100 nm range).
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.