In our research on lightweight structural materials, we have focused on the fabrication of Al/Mg bimetallic components using the lost foam casting process. This technique offers near-net-shape capabilities and the advantage of producing complex geometries with a metallurgical bond between aluminum and magnesium. However, as-cast bimetals often require post-processing heat treatment to optimize the interfacial microstructure and mechanical properties. The challenge lies in the fact that conventional heat treatments designed for monolithic alloys may cause defects such as cracking at the interface due to differences in thermal expansion coefficients between the intermetallic compounds and the metal matrix. In this study, we systematically investigated the effects of different heat treatment parameters—cooling method, soaking time, and multi-step homogenization—on the interfacial microstructure of Al/Mg bimetal castings produced by lost foam casting. Our objective was to develop a heat treatment protocol that eliminates cracks and promotes a sound diffusion layer, thereby enhancing the performance of lost foam castings.
The bimetal samples were prepared by the solid-liquid compound process using A356 aluminum alloy and AZ91D magnesium alloy. The lost foam casting setup involved embedding a solid magnesium insert in a foam pattern and pouring molten aluminum around it under vacuum. After solidification, the as-cast samples were subjected to homogenization annealing at 420 °C for various times (6, 14, and 22 h), followed by either air cooling or furnace cooling. Additionally, a multistep homogenization treatment (380 °C for 8 h + 420 °C for 14 h, furnace cooling) and subsequent aging at 200 °C for 20 h were performed. The interfacial microstructures were examined using scanning electron microscopy (SEM) equipped with energy dispersive spectroscopy (EDS). We paid special attention to the formation of cracks, the evolution of intermetallic layers, and the distribution of phases.
Our results revealed that the cooling method after homogenization is critical for the integrity of the interface in lost foam castings. When the samples were air-cooled from 420 °C, we observed severe cracking at two distinct locations: between the Al matrix and the Al₃Mg₂ + Mg₂Si reaction layer, and between the Al₁₂Mg₁₇ + Mg₂Si layer and the Al₁₂Mg₁₇ + δ-Mg eutectic layer. These cracks are attributed to the large thermal mismatch stresses that develop during rapid cooling. The thermal expansion coefficients of the intermetallic phases are significantly lower than those of the Al matrix, leading to tensile stresses at the interface. In contrast, furnace cooling from the same homogenization temperature completely eliminated these cracks. Moreover, we detected a new diffusion layer formed between the Al matrix and the Al₃Mg₂ + Mg₂Si layer. This layer had a thickness of approximately 60 μm after 6 h of homogenization and consisted of Al(Mg) solid solution with finely dispersed Mg₂Si particles. The EDS analysis confirmed that the matrix phase in this diffusion layer is Al-rich with dissolved Mg, and the second phase is Mg₂Si containing some Al. This layer is believed to form by the inward diffusion of Mg into the Al matrix and the precipitation of Mg₂Si from the supersaturated solid solution during slow cooling.
We then examined the effect of homogenization time under the furnace cooling condition. Samples were held at 420 °C for 6, 14, and 22 h. The thickness of the new diffusion layer increased with time, but the growth rate slowed after 14 h. The measured thicknesses are summarized in Table 1. The growth of the diffusion layer can be described by a parabolic relationship, $$ x = k \sqrt{t} $$, where x is the layer thickness, t is the homogenization time, and k is a rate constant. Fitting our data yields k ≈ 24.5 μm·h⁻¹/². This is consistent with diffusion-controlled growth in which Mg atoms migrate into the Al side, forming a solid solution and reacting with Si to form Mg₂Si precipitates.
Table 1. Thickness of the new diffusion layer after homogenization at 420 °C with furnace cooling.
| Homogenization time (h) | Diffusion layer thickness (μm) |
|---|---|
| 6 | 60 |
| 14 | 95 |
| 22 | 118 |
Beyond the new diffusion layer, we also observed changes in the existing interfacial layers. The Al₁₂Mg₁₇ + δ-Mg eutectic layer (reaction layer III) coarsened with longer homogenization time. The δ-Mg grains within this eutectic increased in size, as shown by comparing samples treated for 14 h and 22 h. This coarsening is attributed to the dissolution of Al₁₂Mg₁₇ into the δ-Mg solid solution during high-temperature holding, followed by the reprecipitation of finer particles upon slow cooling. Meanwhile, the Al₁₂Mg₁₇ phase distributed in the Mg matrix gradually dissolved into the primary Mg grains with increasing time. After 22 h, only a few isolated Al₁₂Mg₁₇ particles remained. This homogenization of the magnesium side improves the compositional uniformity and may reduce the brittleness of the bimetal.
To further optimize the interfacial structure, we implemented a multistep homogenization treatment: first at 380 °C for 8 h, then at 420 °C for 14 h, followed by furnace cooling and a subsequent aging treatment at 200 °C for 20 h. This procedure is designed to promote more complete dissolution of coarse intermetallics and to achieve a fine, uniform precipitation during aging. The resulting microstructure exhibited no cracks and the diffusion layer between Al and Al₃Mg₂ + Mg₂Si remained intact with a thickness comparable to that after single-step homogenization at 420 °C for 14 h. However, the distribution of phases in all reaction layers was more homogeneous. The Mg₂Si particles in the Al₃Mg₂ + Mg₂Si layer and in the Al₁₂Mg₁₇ + Mg₂Si layer were finer and more uniformly dispersed. Notably, the Al₁₂Mg₁₇ phase in the Mg matrix precipitated as fine lamellae after aging, rather than the coarse network observed in as-cast samples. This lamellar structure is beneficial for mechanical properties because it provides a combination of strength and ductility through a dispersion-strengthening effect.
The formation of the new diffusion layer and the elimination of cracks in furnace-cooled lost foam castings can be understood by considering the thermal stress evolution during cooling. We developed a simple analytical model based on the mismatch in coefficients of thermal expansion (CTE). The CTE of pure Al is about 23×10⁻⁶ K⁻¹, while that of Mg is about 26×10⁻⁶ K⁻¹. The intermetallic compounds Al₃Mg₂ and Al₁₂Mg₁₇ have CTEs approximately 18×10⁻⁶ K⁻¹ and 20×10⁻⁶ K⁻¹, respectively. During cooling from 420 °C to room temperature, the thermal strain mismatch between Al and Al₃Mg₂ is $$ \Delta \varepsilon = (\alpha_{\text{Al}} – \alpha_{\text{IMC}}) \Delta T $$, where $\Delta T \approx 400\ \text{K}$. This gives a strain of about 0.2%. If the cooling rate is high, the stress cannot be relieved by creep or diffusion, leading to cracking. In furnace cooling (rate ~0.1 K/s), the slow cooling allows stress relaxation through atomic diffusion and plastic flow, thus avoiding cracks. Additionally, the formation of the Al(Mg) + Mg₂Si diffusion layer may act as a compliant buffer that accommodates some of the mismatch, further reducing the risk of interfacial failure.
The growth kinetics of the diffusion layer can be described using Fick’s second law. Assuming one-dimensional diffusion of Mg into Al with a constant surface concentration, the concentration profile is given by $$ C(x,t) = C_0 \,\text{erfc}\!\left(\frac{x}{2\sqrt{D t}}\right) $$. The thickness of the layer is often defined as the distance at which the concentration reaches a threshold value. If we approximate that the layer thickness scales with $\sqrt{D t}$, the parabolic growth observed is consistent with diffusion control. The effective diffusivity D can be estimated from our data: from 6 h to 22 h, the thickness increased from 60 to 118 μm. Using $x_2^2 – x_1^2 = 4D(t_2 – t_1)$, we obtain $D \approx 1.2 \times 10^{-14}\ \text{m}^2/\text{s}$ at 420 °C. This value is reasonable for grain boundary diffusion in Al at elevated temperatures. The presence of Si in the alloy likely accelerates the diffusion by forming Mg₂Si precipitates that consume Mg and maintain a concentration gradient.
In addition to the diffusion layer, we also quantified the coarsening of the δ-Mg grains in the Al₁₂Mg₁₇ + δ-Mg eutectic. From SEM images, the average grain size of δ-Mg increased from about 10 μm after 6 h to 18 μm after 22 h. This coarsening follows a $t^{1/3}$ law typical of Ostwald ripening: $$ d^3 – d_0^3 = K t $$, where d is the average grain diameter and K is a rate constant. A linear fit of d³ vs. t yields K ≈ 4.7×10⁻²⁰ m³/s. This coarsening is driven by the reduction of interfacial energy and is promoted by the high solubility of Al in Mg at 420 °C.
The multistep homogenization plus aging treatment produced an even finer and more uniform microstructure. During the initial low-temperature step at 380 °C, the diffusion rates are slower, allowing more nucleation sites for precipitates. The subsequent higher-temperature step dissolves coarse phases, and the slow furnace cooling allows controlled precipitation. Finally, the aging at 200 °C leads to the formation of fine lamellar Al₁₂Mg₁₇ within the Mg grains. The lamellar spacing can be controlled by the aging time and temperature. In our case, the lamellar structure had an interlamellar spacing of approximately 0.5 μm, which is consistent with a discontinuous precipitation mechanism. This structure enhances the strength of the magnesium side via precipitation strengthening and may also improve the toughness of the interface region.
In summary, our investigation of heat treatment effects on Al/Mg bimetal produced by lost foam casting reveals that furnace cooling after homogenization is essential to avoid cracking and to promote the formation of a beneficial diffusion layer. The new diffusion layer, composed of Al(Mg) solid solution and Mg₂Si, grows parabolically with time. Prolonged homogenization coarsens the eutectic structure but homogenizes the Mg matrix. A multistep homogenization plus aging treatment further refines the microstructure, yielding a uniform lamellar precipitation of Al₁₂Mg₁₇ in the magnesium side. These findings provide a practical heat treatment strategy for lost foam castings of Al/Mg bimetal components, enabling the production of sound, high-performance lightweight structures. Future work will focus on correlating these microstructural features with mechanical properties such as shear strength and fracture toughness.
Table 2. EDS analysis results for different phases in the interfacial layers after various heat treatments (representative points).
| Position | Al (wt%) | Mg (wt%) | Si (wt%) | Phase |
|---|---|---|---|---|
| New diffusion layer (matrix) | 88–92 | 8–12 | 0–1 | Al(Mg) solid solution |
| New diffusion layer (particles) | 2–6 | 56–62 | 32–36 | Mg₂Si (with dissolved Al) |
| Reaction layer I (Al₃Mg₂ + Mg₂Si) | 60–64 | 36–40 | 0–1 | Al₃Mg₂ |
| Reaction layer I (Mg₂Si) | 5–8 | 56–60 | 32–38 | Mg₂Si |
| Reaction layer II (Al₁₂Mg₁₇ + Mg₂Si) | 45–50 | 50–55 | 0–2 | Al₁₂Mg₁₇ |
| Reaction layer III (Al₁₂Mg₁₇ + δ-Mg) | 35–40 | 60–65 | 0 | Al₁₂Mg₁₇ (eutectic) |
| Mg matrix after homogenization | 10–18 | 82–90 | 0 | δ-Mg (supersaturated) |
| Precipitate after aging | 35–40 | 60–65 | 0 | Al₁₂Mg₁₇ (lamellar) |
In conclusion, the lost foam casting process combined with a well-designed heat treatment can yield Al/Mg bimetal components with a crack-free interface and a graded diffusion layer. The cooling rate after homogenization must be controlled to avoid thermal stress failure. The diffusion layer thickness can be tailored by adjusting the homogenization time, and the multistep treatment offers additional microstructural refinement. These insights are directly applicable to industrial lost foam castings, where the integrity of the bimetal interface is critical for load-bearing applications. Our ongoing research aims to quantify the mechanical response of these interfaces and establish optimal heat treatment schedules for different alloy combinations.