Abstract: Ceramic core is a critical component in the super-alloy turbine blade casting. In our previous work, a novel multi-phase MgAl2O4/MgO porous ceramic was prepared for this purpose. The most important property was that it crumbled completely after hydrothermal treatment in just pure water, due to the hydration of MgO. In this work, the hydration process of the MgO embedded in the inert matrix was investigated in detail. The collapse behaved as an interior destruction without any bulk expansion of the sample. The hydration percentage was the only factor related to the water-collapsibility. The morphology of hydration product indicated that the reaction advanced in particular direction. Based on the finite element analysis for the expansion effect on the porous structure, the interior-collapsing mechanism was proposed. During the hydration process, the MgO grains exerted pressure to the surrounding matrix and induced the collapse in the adjacent structure. This process took place throughout the matrix. Finally, the sample crumbled completely to the powders. No bulk dilatation was detected before the powdering, indicating that the collapse process would not exert pressure outward. Thus the alloy blade would not be damaged during the removal of the ceramic core. It was also predicted that the decrease in the MgO grain size was beneficial to the water-collapsibility.
Keywords: ceramic cores; porous structure; MgO; hydration; interior-collapsing mechanism
Ceramic core is a critical component in the super-alloy turbine blade casting process to form the complex cooling channel . Due to the harsh service environment, ceramic cores require specific properties, such as good high temperature mechanical strength, creep resistance, leachability, and easy approach to remove after the alloy casting. It is important to select the proper method to remove the ceramic core without damaging the blade. Considering the differences between the chemical properties of the alloy and ceramic, dissolution in concentrated alkaline solution is an applicable method. For these reasons, only silica-based and alumina-based ceramic cores are widely used up to now. Both of them still have many flaws. For silica-based cores, the main problem is the poor performances at elevated temperatures. Alumina-based cores overcome this shortcoming, but the removal process is much more difficult. Many attempts have been conducted to improve the comprehensive performances by adjusting the composition or the micro-structure [2–6]. And the application of additive manufacturing is one of the hot researched areas recently [7–11], but the essential problem remains. On the other hand, with the increase of alkali concentration, the damage of alloy blade cannot be ignored . Thus, reducing the alkalinity is beneficial to minimize the damage to the alloy blade. In the ideal situation, the removal process takes place in pure water. In order to achieve this goal, a new ceramic core system needs to be established.
In our previous work , a multi-phase ceramic core based on MgAl2O4 and MgO was developed. It exhibited excellent high temperature performance. At the same time, the removal process was fulfilled by hydrothermal treatment in just pure water. The main component of the multi-phase ceramic is MgAl2O4. As is well known, it is inert no matter in acidic, alkaline, or neutral environment. Thus, the collapsing mechanism was completely different from the corrosion mechanism of the silica and alumina cores. It was found that the MgO was hydrated into Mg(OH)2 during the watercollapsing process, but the detail of the collapsing mechanism was not clear.
The hydration mechanism of magnesia has been studied for decades. Fruhwirth et al.  investigated the dissolution and hydration kinetics of magnesia at different pH values. They proposed that hydroxide was produced by the precipitation after the solution was supersaturated. This result was supported by Refs. [15–19]. However, this mechanism was not sufficient to describe the reaction process. Typically, some researchers indicated that the hydroxide generated prior to the dissolution [20,21]. And, empirically, dissolution–precipitation mechanism is not able to explain the structural destruction induced by the hydration. Since the existing studies mainly focus on the kinetics of the reaction, there is little research on how the MgO grain expands. A new perspective is required to illustrate the expansion behavior of MgO hydration in a confined structure and its interaction with the inert matrix
In this work, the hydration of MgO embedded in the porous MgAl2O4 structure was investigated systematically. Phase composition and micro-structure were detected in order to illuminate the reaction process and the expansion effect. The expansion behavior of MgO and its effect to the porous matrix were computationally studied by finite element analysis. Finally, an interiorcollapsing mechanism was proposed.
2. 1 Sample preparation
Multi-phase ceramic samples were prepared by reaction sintering. Alumina (5 μm, 99.9% purity; Pantian nano Materials Co., Ltd., Shanghai, China) and magnesia (1 μm, 99.9% purity; Yijin new Materials Co., Ltd., Beijing, China) were employed as raw materials. The molar ratio of the reactants, n(Al2O3):n(MgO), was fixed to 1:1.5. After ball milling with ethanol, the mixture was dried, crashed in mortar, and sieved by 100-mesh sieve. Then the powders were pressed into pellet at 50 MPa. The green body was sintered at 1600 ℃ for 2 h with a heating rate of 5 ℃/min. After that, a thermal treatment at 1550 ℃ for 0.5 h was conducted to simulate the casting process.
2. 2 Hydrothermal treatment
Hydrothermal treatment was taken place in a hydrothermal reactor. 0.5 g sample was cut from the sintered body and soaked in 50 mL reaction media. The reactor was placed in an oven at 110, 120, and 130 ℃ for different duration time. For most samples, the reaction medium was deionized water. An acid-treated sample was prepared in comparison. In this case, 0.1 mol/L acetic acid was selected as the reaction media, and the hydrothermal treatment was conducted at 110 ℃ for 12 h.
The hydration percentage of MgO (α) was calculated by Eq. (1):
where nMg(OH)2 and nMgO are the molar fractions of the Mg(OH)2 and MgO in the samples after hydrothermal treatment, respectively. The samples were characterized by the X-ray diffractometer (Ultima IV, Rigaku, Japan) with Cu Kα radiation. The data were collected in the range from 10° to 90°, with a step size of 0.02°. The scan rate was 1 (°)/min. Then the phase composition was calculated by the Rietveld refinement method using the FullProf program.
2. 3 Micro-structure
The micro-structure was investigated by the scanning electron microscope (JSM-7500, JEOL, Japan).
2. 4 Mechanical properties
Compression test and three-point bending test were conducted in order to obtain the compressive strength and the bending strength of the samples. In the threepoint bending test, the sample dimension was 3 mm × 4 mm × 36 mm, and the span distance was 30 mm. In the compressive test, the sample dimension was ϕ20 mm × 10 mm. The loading rate was 0.5 mm/min for both tests. According to the test principle, the bending strength was regarded as the tensile strength of brittle material.
The samples used in the tests were porous material. Since it is difficult to determine the exact relation between the porosity and the sample strength, an approximate correction was conducted to estimate the failure stress (σfailure) of the matrix, according to Eq. (2) :
where σmeasured is the sample strength measured in compression test and three-point bending test and ρ is the porosity of the sample, which is measured by the Archimedes principle in kerosene.
3 Results and discussion
The Gibbs free energy of the reaction between MgO and liquid water is presented in Fig. 1. The Gibbs energy of this reaction is negative in the whole temperature interval (298–500 K), indicating that the reaction is thermodynamically favored. ΔG was also related to the force of crystallization (σcrystal) which was defined as a pressure needed to stop the reaction by pushing it toward equilibrium. It could be calculated by Eq. (3) :
At 110 ℃, the force of crystallization was 1.85 GPa, a much higher value compared to the failure stress of the sample (Table 1). As a result, the hydration process could provide sufficient driving force to destroy the structure of the sample.
Table 1 Mechanical properties of the sample
Fig. 1 Gibbs free energy of the reaction: MgO(s) + H2O(1) = Mg(OH)2(s). The thermodynamic parameters were cited from Ref. .
The X-ray diffraction (XRD) patterns of the samples treated in different conditions are presented in Fig. 2. The observed peaks were matched with the standard PDF card. It was found that there were spinel (MgAl2O4), periclase (MgO), and brucite (Mg(OH)2) in all the samples, as expected. With the hydration time increasing, the peaks belonging to the periclase decreased, while the peaks belonging to the brucite increased.
Fig. 2 XRD patterns of the samples treated in different conditions. The samples after hydrothermal treatment are in deionized water at (a) 110 ℃, (b) 120 ℃, and (c) 130 ℃. (d) The samples after hydrothermal treatment are in 0.1 mol/L acetic acid at 110 ℃ for 12 h. ▲spinel (MgAl2O4); ● periclase (MgO); ♦ brucite (Mg(OH)2).
The Rietveld refinement analysis was performed on all these samples. The detail of the refinement is presented in Figs. S1–S4 in the Electronic Supplementary Material (ESM). Based on the phase composition calculated from the refinement, the rate curves at different reaction temperatures are shown in Fig. 3. Increasing in the reaction temperature leads to a significant acceleration in the reaction. And the reaction rate decreased continuously as the reaction progress, due to the product formed on the surface, which hindered the further reaction. The reaction kinetics was consistent with the modified shrinking core model . However, the reaction rate was lower than that of the other studies, indicating the tightly confined MgO grain in the matrix and the small reaction interface.
Fig. 3 Hydration rate curves at different reaction temperatures.
The typical appearance of the samples after hydrothermal treatment is displayed in Fig. 4. If the hydration percentage was lower than 60%, little change was observed. No dimensional change was detected at this stage. When the hydration percentage reached about 60%, cracks appeared in the samples. Further hydration generated more cracks in the sample. Finally, the sample crumbled completely to the powders. This trend was almost the same when treated at different temperatures, indicating that the hydration percentage was the only factor, having an effect on the structure of the sample.
Fig. 4 Typical appearance of the samples after hydrothermal treatment with (a) α < 60%, (b) α ≈ 60%, and (c) α ≈ 80%.
The phase compositions of the two samples after hydrothermal treatment at 110 ℃ for 12 h are presented in Table 2. The reaction media are deionized water and acetic acid. It was found that the content of MgAl2O4 in the acid-treated sample was much higher than that in the water-treated sample, indicating the dissolution of MgO. Since the acetic acid was in excess of the MgO in the sample, the supersaturation of the Mg2+ ion could never be reached. According to the dissolution–precipitation mechanism, there should be no brucite phase in the sample, but it is not what happened here. Although less brucite generates in the acidic environment, the hydration percentage is comparable to the watertreated sample. This implied that the dissolution process and hydration process were independent to each other. That is to say, the hydration process could not be explained by the dissolution– precipitation mechanism. It is worth noting that the shape of the acid-treated sample is intact, as shown in Fig. 4(a). This indicates that the dissolution of MgO is detrimental to the water-collapsibility.
Table 2 Phase compositions of the samples hydrothermal treated at 110 ℃ for 12 h with different reaction media
Figure 5(a) exhibits the surface morphology of the sample before hydrothermal treatment. The energy dispersive spectroscopy (EDS) mappings of this region are presented in Fig. 5(b). Since Mg existed in both MgO and MgAl2O4 phase, it was difficult to distinguish the two phases from the original data. Thus, a processed figure was provided by subtracting the brightness in the Al mapping from the Mg mapping. Then the points with low intensities were eliminated. Thus, the region with denser dots was regarded as MgO. According to the EDS analysis, the possible MgO grains are labeled in Fig. 5(a). Compared to the theoretical volume fraction (~10%), the proportion of MgO detected was much lower. Since the resolution of EDS is about 1 μm, it is difficult to distinguish the grains in smaller sizes. Considering the grain size of the raw MgO powder, it was reasonable that most MgO grains remained were smaller than 1 μm after the solid state reaction. Thus, there were more MgO grains dispersed in the structure which could not be observed.
Fig. 5 (a) Surface morphology of the sample before hydrothermal treatment. The area with higher Mg content is labeled by the yellow circle. They were regarded as MgO. (b) EDS mapping of (a).
The surface morphologies of the samples after hydrothermal treatment at 110 ℃ in deionized water are shown in Fig. 6. After 4 h treatment, no morphology change was observed although ~40% hydration percentage was detected by XRD. This corresponds to the fact that the geometric shape of MgO changes little at a low conversion rate . Significant morphology change took place after 8 h hydrothermal treatment. This is consistent with the shape change displayed in Fig. 4. Flower-like structures appearing on the surface were the accumulation of many thin flakes with random orientation. Combined with the EDS analysis shown in Fig. 6(c) and the XRD pattern mentioned above, these flakes are Mg(OH)2. The distribution of Mg(OH)2 demonstrated that most MgO grains were at the submicron scale and were difficult to distinguish from MgAl2O4.
Fig. 6 Surface morphologies of the sample after hydrothermal treatment at 110 ℃ in deionized water for (a) 4 h, (b) 8 h; (c) EDS mapping of (b).
The fracture morphologies of the samples before and after hydrothermal treatment is exhibited in Fig. 7. The sample after 4 h hydration was also similar to the original morphology, while tiny difference is observed in Fig. 7(b). Some thin flakes appeared on the grain surface, indicating the Mg(OH)2 formation. The presence of Mg(OH)2 became more distinguishable after 6 h hydration. Compared with the flakes on the surface (Fig. 6(b)), a different arrangement of Mg(OH)2 flakes is found in Fig. 7(c). In this case, the flakes were put in order along a specific direction. This indicates that the reaction process for MgO in the confined structure is different to that in the free state. Cracks occured in the adjacent structure of the hydrated MgO grains. After 8 h hydration, more Mg(OH)2 flakes and more cracks appear in the intergranular region (Fig. 7(d)). The geometric shape of Mg(OH)2 flakes became more regular at the same time. This was attributed to the crystal growth.
Fig. 7 Fracture morphologies of the sample (a) before hydrothermal treatment; after (b) 4 h hydration, (c) 6 h hydration, and (d) 8 h hydration.
The fracture morphologies of the sample after hydrothermal treatment in acetic acid are presented in Fig. 8. At low magnification, the morphology was similar to that of the sample before hydrothermal treatment. while the Mg(OH)2 flakes were found at high magnification. Although hydration process continued for 12 h, the morphology of the Mg(OH)2 was similar to the sample treated in deionized water for 6 h (Fig. 7(c)). It was inferred that the crystal growth was attributed to the dissolution–precipitation process. Seen from Figs. 7(d) and 8(b), it is also found that the Mg(OH)2 flakes show highly orientations, no matter what the reaction medium is. This indicated that the hydration of the MgO grains advanced in particular direction.
Fig. 8 Fracture surface morphologies of the sample hydrothermal treated with 0.1 mol/L acetic acid at 110 ℃ for 12 h with different magnification.
In order to illustrate the phenomena mentioned above, a new mechanism was proposed to describe the hydration process. The schematic diagram is presented in Fig. 9. The crystal structure of MgO is displayed in Fig. 9(a). Seen from its (111) plane (Fig. 9(c)), the ion configuration was similar to that of the Mg(OH)2 (Fig. 9(b)). According to the theoretical calculations [27–29], hydration began at the low-coordinated surface site, and the (111) plane was the most stable surface after hydroxylation. Wogelius et al.  proposed a protonation mechanism to explain the proton penetration to a depth at least 500 nm and the formation of brucite-like layer in the acidic environment. Based on this mechanism, hydroxylation process tended to penetrate into the MgO along the (111) plane. In the neutral environment, it was inferred that protons and hydroxyl ions dissociated from water molecules bonded with the oxygen ions and magnesium ions on the surface, respectively. This led to the lattice distortion, increasing the interplanar spacing of the (111) planes (Fig. 9(d)). With water molecules intruding, the reaction continues along the (111) plane and the brucite-like structure generated (Fig. 9(e)). Due to the thermodynamic instability, structural relaxation occured by the distortion of the Mg–O octahedral (Fig. 9(f)). This was the crystal structure of Mg(OH)2 seen from the (0001) plane (Fig. 9(b)). It was found that the volume expansion during the hydration process was anisotropic. Calculated from the lattice parameters, the linear expansions along a-axis, b-axis, and c-axis of Mg(OH)2 were 5.37%, 5.37%, and 95.75%, respectively.
Fig. 9 Crystal structures of (a) MgO and (b) Mg(OH)2, (c–f) schematic diagrams of the structure transformation during the MgO hydration process. Green, red, and white balls represent Mg2+, O2−, and H+ ions, respectively.
Finite element analysis was employed to simulate the interaction between the MgO grain and the matrix during the expansion. Since the materials applied in simulation were isotropic in the xy-plane, two-dimensional analysis was used to simplify the calculation. The model is presented in Fig. 10. A 30 μm × 30 μm square was selected as the simulation area. A 1 μm × 1 μm MgO grain was trapped in the structure. Two models were presented to simulate the extreme cases. In the case A (Fig. 10(a)), the c-axis of Mg(OH)2 was perpendicular to the contact surface. In the case B (Fig. 10(b)), the c-axis of Mg(OH)2 was parallel to the contact surface. If the tensile stress or compressive stress exceeds the corresponding failure stress, structural failure is declared to occur in this region.
Fig. 10 Two extreme models to simulate the stress induced by the volume expansion during MgO hydration: (a) The c-axis of Mg(OH)2 is perpendicular to the contact surface, denoted as case A; (b) the c-axis of Mg(OH)2 is parallel to the contact surface, denoted as case B.
There was another typical structure of the MgO grain embedded in the MgAl2O4 matrix. Since the stress state in this model shows little difference, it is presented in Figs. S5 and S6 in the ESM and not discussed in detail.
Thermal expansion was used to represent the volume expansion during the MgO hydration process. For this purpose, the thermal expansion coefficient of MgAl2O4 was set to 0. The thermal expansion coefficients of Mg(OH)2 along the x-axis and y-axis were 0.537× 10-3 and 9.575×10-3 ℃-1, respectively. This represents that the c-axis of Mg(OH)2 crystal is parallel to the y-axis of the coordinate system. A thermal expansion of 0–100 ℃ could simulate the expansion with 0–100% of MgO hydration. The elastic constants were listed in Table 3.
Table 3 Elastic constants of MgAl2O4 and Mg(OH)2[30,31]
The stress state after expansion is presented in Fig. 11. In the case A (Figs. 11(a) and 11(b)), intense stress was induced in the simulating structure. It was several times larger than the failure stress of the sample (Table 1) in a wide range. However, in the case B (Figs. 11(c) and 11(d)), both the tensile stress and compressive stress decreased dramatically.
Fig. 11 (a) Tensile stress and (b) compressive stress of the case A. (c) Tensile stress and (d) compressive stress of the case B.
Generally, the orientation relationship between the c-axis of Mg(OH)2 and the contact surface was random. Thus the reality of the volume expansion was the combination of the cases A and B. From the arrangement of Mg(OH)2 flakes seen in the Fig. 7(d), it was inferred that the case A was the major component during the expansion. This was attributed to the contact with water [32,33]. Considering a model as shown in Fig. 12, MgO single crystal with random orientation was confined by the fixed surface. There were two equivalent (111) planes in the MgO crystal. The edges ab and cd were connected to the inert matrix. They were fixed to a certain extent because of sintering. The edges bc and ad were free and exposed to water. As a result, the hydration process tended to carry on along the plane labeled by the red arrow.
Fig. 12 Model represents the random orientation of the MgO single crystal confined by the fixed surfaces. Two equivalent (111) planes are presented. The red arrow indicates the preferred hydration plane.
Based on this mechanism, the minimum effect would be induced when the case B occupies the maximum proportion, that was, the angle between the c-axis of Mg(OH)2 and the contact surface was 45°. The model and computational results are presented in Fig. 13. It was found that the tensile stress extended along the contact surface while the compressive stress extended perpendicular to the contact surface. Thus, different failure modes would occur in different directions.
Fig. 13 (a) Simulation model of the situation where the angle between the c-axis of Mg(OH)2 and the contact surface is 45°; (b)
tensile stress and (c) compressive stress of the model in (a).
However, the MgO hydration was relatively slow in the real process. The stress in the MgAl2O4 matrix increased gradually. Due to the porous structure of the MgAl2O4 matrix, the genuine stress state could never reach so far away like the computational simulation. A porous structure was modeled to depict the interaction between the local stress and the pores (Fig. 14). The detail simulation process is presented in Figs. S7 and S8 in the ESM. Cracks were generated under the action of tensile stress. Due to the stress concentration effect, the crack would propagate forward until it reached the pore wall. Then the tensile stress in the matrix was released and the crack propagation terminated. The main cause of the local collapse was the compressive stress generated perpendicular to the contact surface of the MgO grain and MgAl2O4 matrix. In the region with compressive stress higher than the failure stress, the intactness of the structure was destroyed. If the failure region was tightly surrounded by the matrix, the compressive stress was able to pass through the failure region. When the failure region reached the pore, the compressive stress was released by breaking the pore wall. Since the connection between the MgO grain and the matrix was broken at this stage, further expansion of the MgO grain would not cause stress in the matrix. The same process occured on all of the MgO grains throughout the matrix, finally leading to the complete collapse. Since the collapse caused by a single MgO grain was limited by the porous structure in the matrix rather than the size of MgO grains, it was inferred that the decrease in the MgO grain size did not affect the local failure process. On the other hand, the decrease in the grain size led to the increase of the grain number, namely, the number of the failure region. Thus, a prediction could be made that the decrease in MgO grain size was beneficial to the water-collapsibility.
Fig. 14 Schematic diagram of the collapsing behavior with the existence of pores. Red line represents the crack induced by the tensile stress. Red area with black squares represents the failure region due to the compressive stress. Detail simulation process is presented in Figs. S7 and S8 in the ESM.
In this work, the hydration process and the collapse behavior of the MgAl2O4/MgO porous ceramic were investigated. It was found that the collapsibility was only related to the hydration percentage of MgO in the sample. The expansion behavior was the critical factor to the structural collapse. Acidic solution treated sample demonstrates that the dissolution process was independent to the MgO hydration. Dissolution of MgO was even detrimental to the water-collapsibility. A novel interior-collapsing mechanism was proposed to describe the structural transformation. Based on this mechanism, the collapse process was described as follows.
At the beginning, water diffused into the matrix through the pores. Mg(OH)2 formed on the MgO surface which was in contact with water. Oriented expansion occured if the MgO was confined by the MgAl2O4 matrix. Pressure was applied to the matrix during this process. Interior-collapsing occurred when the local stress exceeded the failure stress. As the hydration proceeding, the area of the collapsed region increased but was limited by the pores in the matrix. The same process took place throughout the matrix. Finally, the sample crumbled completely to the powders.
It was also predicted that the decrease in the MgO grain size was beneficial to the water-collapsibility. No bulk dilatation was detected in the hydration process, indicating the collapse would not exert pressure outward. This was conducive to the ceramic core application. The multi-phase MgAl2O4/MgO porous ceramic had enormous potential in the ceramic core field.
Declaration: This article is provided by CERADIR™ users or obtained from Internet, the content does not represent the position of CERADIR™. We are not responsible for the authenticity/accuracy of the article, especially the effects of the products concerned. This article is for study only, it does not constitute any investment or application advice. For reprinting, please contact the original author. If it involves the copyright and/or other issues, please contact us and we will deal with it asap! CERADIR™ has the interpretation of this declaration.