Inclusions in Permanent Mold Cast Magnesium
Abdallah Elsayed, Comondore Ravindran, Eli Vandersluis, Sophie Lun Sin
Click here to see this story as it appears in Modern Casting.
Magnesium alloys have been gaining consideration as possible alternatives to aluminum alloys to reduce vehicle weight in aerospace and automotive applications. Magnesium alloys are about 35% lighter than aluminum alloys. However, only 0.3% of the total automotive vehicular weight in North America is composed of magnesium alloys, while 8.3% is composed of aluminum alloys. In terms of total weight, each passenger car contains only 11.02 lbs. (5 kg) of magnesium, yet 264.5-308.6 lbs. (120–140 kg) of aluminum.
The widespread use of magnesium alloys for aerospace and automotive applications is hindered by their high reactivity, which increases the probability of inclusion formation during casting processes. Inclusions in magnesium alloys compromise corrosion resistance, increase porosity, produce unfavorable surface finishes, and reduce mechanical properties, in particular ultimate tensile strength and elongation. Two major types of inclusions in magnesium alloys occur: intermetallic and non-intermetallic. Intermetallic inclusions are almost always iron-rich phases, while non–intermetallic inclusions include sulfides, fluorides, sulfates, chlorides, nitrides, and oxides, with oxides being the most dominant.
Avoiding inclusions in magnesium alloys is difficult due to the many sources from which they arise. Inclusions can arise from reactions with air where magnesium reacts with oxygen to form MgO, reactions with fluxes entrapping flux components (e.g., MgCl2, CaCl2) and flux reactions with oxygen to form MgO. Even with the use of protective atmospheres such as sulfur hexafluoride (SF6), reaction products of MgO and MgF2 can become entrapped in the melt. In addition, melt turbulence during melting, handling, and pouring can be a source of inclusions in magnesium castings.
A wide variety of inclusion assessment techniques are available for magnesium and its alloys. These techniques vary from simple observational methods, such as metallographic and fracture bar examinations, to highly sophisticated online methods, such as liquid metal cleanliness analyzers. Since no industry standard for examining inclusions in magnesium alloys exists, the techniques employed are foundry-dependent, which complicates comparisons between facilities.
This article aimed to characterize and examine the effects in ZE41A and AZ91D magnesium alloys and their influence on microstructure and mechanical properties. By better understanding and documenting metal handling, the resulting scrap reduction, casting quality enhancement, and associated cost reductions will improve foundry competitiveness. This research is part of an ongoing effort to increase the use of magnesium alloys to a significant level in the aerospace and automotive industries and to reduce vehicle weight, fuel consumption, and emission of harmful gases.
The general procedure was the same for both alloys and involved the production of permanent mold tensile castings and fracture bar castings from multiple foundries and characterizing them according to their mechanical properties, grain sizes, microstructures, and inclusion contents. The microstructures, inclusions, and grain sizes were characterized using scanning electron microscopy (SEM) and optical microscopy. The mechanical properties were assessed using uniaxial tensile testing.
For both ZE41A and AZ91D alloy castings, the average yield strength, ultimate tensile strength, and elongation decreased between the start and end of the production run. The results from examination of grain size, microstructure and inclusion analysis indicate that the loss in properties was predominately caused by the accumulation of oxides. For example, the AZ91D castings demonstrated a ~5-10% decrease in UTS and a ~20-30% decrease in elongation, with a smaller change in yield from start to end of production.
For both alloys, an increase in grain size was observed between the start and end of the production run, but the reduction in mechanical properties was mainly attributed to particle-type Mg–Al–O inclusions in the AZ91D alloy and film-type Mg–O inclusions in the ZE41A alloy on the fracture surfaces of the tensile samples and fracture bars. The AZ91D castings had very few inclusions, but they were much larger than those in the ZE41A castings. This research recognized extensive variability of the inclusion levels in the industry and is a precursor to developing industry standards for melt cleanliness in magnesium alloys. This will be a major step in enabling improved quality and enhanced use of magnesium alloys in aerospace and automotive industries.
A representative micrograph from Foundry A of the grain structures of the ZE41A castings produced toward the start and end of a production run is shown in Figure 1. The samples were extracted from tensile mold castings. At the start of pouring, the average grain size of the castings from Foundry A was 22 ± 1 µm and increased to 38 ± 7 by the end of the production run. For Foundry C, at the start of pouring, the average grain size of castings was 27 ± 2 µm and increased to 32 ± 3 µm by the end of the production run. The grain structures in Figure 1 were very spherical in shape, especially near the start of the production runs. With the minor grain coarsening toward the end of the production run, the grain structure begins to deviate from its spherical shape.
Such a coarsening and deviation from spherical grain structure during holding were expected due to zirconium losses, which may occur from reactions with iron crucibles or settling of zirconium particles over time. However, this coarsening is not expected to be significant in reducing mechanical properties.
The grain structures of the AZ91D castings from Foundry B produced using the tensile mold are shown in Figure 2 with Foundries A, B, and D all having similar looking microstructures with just a variation in grain sizes. At the start of pouring, the average grain sizes of the castings from Foundries A, B, and D were 77 ± 4, 49 ± 8, and 63 ± 9 µm, respectively. At the end of pouring, the average grain size for Foundries A, B, and D increased to 112 ± 12, 58 ± 15, and 78 ± 20 µm, respectively. The variation in grain sizes is attributed in part to the range of pouring temperatures used at each foundry.
This coarsening of the grain structure can be attributed to the transformation of grain-refining Mn–Al particles to less potent Mn–Al–Fe compositions during holding, as observed in high-purity Mg–Al alloys. The increase in grain size over the course of the production run is more significant for the AZ91D alloys than for the ZE41A castings. In addition, the AZ91D castings from all foundries had grain structures that contained both small spherical grains and larger irregular grains. Both these phenomena are likely due to the relatively weaker and less abundant Mn–Al refining particles in AZ91D as compared to often excess zirconium added to ZE41A.
Inclusions are known to act as stress risers, and their presence on a fracture surface can indicate their role in fracture initiation. Figure 3a shows an inclusion that likely initiated failure in a sample collected at the start of the experimental trials. Analysis of the inclusion using EDX indicated it was rich in magnesium, zinc, and oxygen. The inclusion is likely a Mg–O-based inclusion with zinc contributions from the alloy matrix. The lack of any iron in the analysis eliminates the possibility of the inclusion being an iron-based intermetallic. The inclusion in Figure 3a also has a fold or crack defect at its interface with the magnesium matrix. Similar results were observed with the samples from Foundry C, as shown in Figure 3b where a Mg–O-based inclusion (indicated by the arrow) was observed with poor interfaces with the magnesium matrix. The Mg–O inclusions appeared mainly as films sitting atop the fracture surfaces. These Mg–O films accumulate during the production run, becoming entrapped in the molten metal during sampling, pouring, and holding. This defect indicates that the observed Mg–O inclusion was weakly bonded to the magnesium matrix, making it a likely source of failure during tensile loading.
The corresponding SEM image of the fracture surface depicts dimple-like features and confirms the absence of inclusions on the surface. These dimples usually indicate good casting ductility. Samples from Foundry B were similar to those from Foundry A with no inclusions evident on the fracture surface, and their microstructure does not contain any noticeable cleavage planes. Samples from Foundry D from the start of the experimental trials were also free of inclusions.
Some of the fracture bars from Foundries A and B were virtually inclusion free, while the maximum inclusion areas were under 2%. If tensile samples were prepared, Foundry B likely would produce castings with mechanical properties very similar to those of Foundry A. On the other hand, the castings from Foundry C contained the highest median inclusion area and had a maximum inclusion area of about 9%. This can be attributed to the fact that Foundry C was the only foundry to use 100% remelted metal and had the highest pouring temperature, exacerbating oxidation. It appears that a melt cleaning measure using filtration or argon bubbling is necessary to enable use of 100% recycled metal. All of the foundries produced some castings that appeared completely inclusion free.
For Foundries A and B, the median inclusion areas were 2–4 times higher for AZ91D than their ZE41A counterparts. Also, for Foundries A and B, the maximum areas of the inclusions in the AZ91D castings were 10 and 25%, whereas in the ZE41A castings they were only 1 and 1.5%. The inclusion assessment for Foundries A and B reveals that the AZ91D alloy tends toward a lesser quantity of inclusions, albeit of much larger sizes, than ZE41A alloy.
A similar inclusion assessment for the fracture surfaces of tensile samples was also conducted. They show similar trends, with Foundry C having the largest inclusions for ZE41A and Foundry B for AZ91D. Foundry B did not provide any ZE41A tensile samples. It is interesting to note that the inclusion areas observed in the tensile samples are much lower than those of the fracture bars. Therefore, the tensile sample fracture surfaces also can be used as a representative means to determine the relative amounts of inclusions in samples but likely underestimates their maximum size. The particle-type Mg–Al–O inclusions in the AZ91D alloy also resulted in a higher measured inclusion area because they tend to be equiaxed in shape and cover a larger surface area than the film-shaped Mg–O inclusions in the ZE41A alloy.
Whereas the ZE41A alloy is much more susceptible to the accumulation of many film-type oxide inclusions throughout the production run, the AZ91D alloy tends to collect a few large particle-type inclusions. This difference in the accumulation of inclusions between the two alloys is likely a contribution of many factors, including oxidation tendencies of each alloy, melt density and viscosity which would influence how inclusions would agglomerate throughout the melts and alloy addition sources.
The film-type inclusions in the ZE41A were more distributed in the samples, while the particle-type inclusions in the AZ91D appeared as agglomerates with a large surface area. It is not possible to relate the decrease in the mechanical properties according to inclusion type, whether it be film or particle type, because of the difference in alloy system (AZ91D or ZE41A) where each inclusion type was dominant. The authors reason the large particle-type Mg–Al–O inclusions are more detrimental because of their faceted nature, larger surface area, and appearance as agglomerates on fracture surfaces. Possible future avenues for research would be to induce inclusions of various sizes and shapes into magnesium alloy melts and measure changes in microstructure and mechanical properties.
The types of inclusions observed in ZE41A and AZ91D magnesium alloy castings from multiple foundries were investigated. The following are the major results:
Mechanical properties decreased in both alloys from the start to the end of the production run for all foundries. This principally depended on the increase in number and size of entrapped inclusions.
Grain size increased in both alloys from the start to the end of the production run for all foundries, especially AZ91D. For ZE41A, a loss of grain-refining zirconium with holding time was the attributing cause, whereas for AZ91D a transformation of grain-refining manganese–aluminum particles to less potent compositions during holding was the reason for the increase in grain size.
The fracture surfaces of tensile samples can be used as a representative means to determine the relative amounts of inclusions in samples but underestimates their potential maximum size. Fracture bars provide a better representation for the range of inclusion sizes in castings as compared to tensile samples because of the much larger sample size and increased number of sampling locations.
The fracture surfaces of the ZE41A alloy contained film-type magnesium–oxygen-based inclusions, whose poor interface with the matrix was likely the source of fracture. The AZ91D alloy fracture surfaces contained mostly particle-type magnesium–aluminum–oxygen spinel inclusions, as well as few smaller iron- based intermetallic inclusions. Whereas ZE41A alloy was susceptible to many small inclusions, AZ91D alloy was more susceptible to few large inclusions.
The film-type inclusions for ZE41A would tend not to agglomerate and are reasoned to be not as harmful as the agglomerated and faceted particle-type inclusions with large surface area observed in the AZ91D alloy.
This article is a summary of a manuscript published in the International Journal of Metalcasting, Elsayed, A., Vandersluis, E., Lun Sin, S. et al. Inter Metalcast (2016). For more information on the manuscript, contact the AFS technical department at 800-537-4237