1. Pre-oxidization of SiC_p

The key to the improvement of composite’s TE rests on the structure, chemistry, and the nature of SiC/Al bonding interfaces,^{Reference Lloyd16} thus the effect of pre-oxidization on the interfacial reaction of the SiC_p/6061 Al composite is first verified. Figure 3(a) shows the SEM morphologies of unoxidized SiC_p, which appears faceted and angular. There are a small number of spot-like particles on their surfaces [Fig. 3(b)], which are identified as SiO₂ particles with reference to the EDS results [Fig. 3(c)]. These small SiO₂ particles should generate during the fabrication process of SiC_p. After being oxidized at 1273 K in air for 8 h, the amount and dimension (∼200 nm) of SiO₂ particles get significantly increased on the SiC_p surface [Figs. 3(d)–3(f)]. Note that the oxygen content on the rest parts except the white particles (indicated by arrows) is also increased, from nearly zero (unoxidized) to 3.5 wt% (oxidized), which suggests the formation of a thin SiO₂ film throughout the surface, in addition to the visible SiO₂ particles. It is reported that the thickness of SiO₂ films on SiC_p oxidized in air at 1373 K for 6 h was approximately 100 nm.^{Reference Lee, Lee, Ahn, Shim and Shi28} So there is a possibility that the thickness of the SiO₂ film by oxidizing SiC_p at 1273 K for 8 h be around 100 nm in this work.

FIG. 3.

(a) SEM morphology of the unoxidized SiC_p, (b) magnified image showing the details of the unoxidized SiC_p surface, (c) EDS spot analysis of the white spot on the surface, (d) SEM morphology of the oxidized SiC_p, (e) magnified image showing the details of the oxidized SiC_p surface, and (f) EDS spot analysis of the white spot on the surface.

The reaction products at the SiC/Al interface in the oxidized and unoxidized SiC_p/Al composites are studied in what follows. As shown in Figs. 4(a)–4(f), the EDS line scanning analysis suggests the enrichment of C, Si (and likely Al) elements at the unoxidized SiC/Al interface, which should form through the chemical reaction^{Reference Lee, Byun, Park and Lee29,Reference Cai, Chen and Zhang30}

(3)$$4{\rm{Al}}\left( {\rm{l}} \right) + 3{\rm{SiC}}\left( {\rm{s}} \right) \to {\rm{A}}{{\rm{l}}_4}{{\rm{C}}_3}\left( {\rm{s}} \right) + 3{\rm{Si}}\left( {\rm{s}} \right)\quad .$$

FIG. 4.

(a) SEM micrograph of the unoxidized SiC_p/6061 Al composite, (b–f) corresponding EDS line scanning analysis of the SiC/Al interface, (g) SEM micrograph of the oxidized SiC_p/6061 Al composite, (h–l) corresponding EDS line scanning analysis of the SiC/Al interface, and (m) XRD patterns of unoxidized and oxidized SiC_p/6061 Al composites.

Since Al₄C₃ obtains a free energy of formation from per mole of carbon as well as a thermal stability comparable to that of SiC.^{Reference Mitra, Rao, Maiti and Chakraborty31} The presence of Al₄C₃ phase in the unoxidized SiC_p/6061 Al composite [marked by arrow in Fig. 4(a)] could be further evidenced by the corresponding XRD pattern, as shown in Fig. 4(m). This result is in good agreement with previous reports on similar unoxidized SiC PRAMCs.^{Reference Cai, Chen and Zhang30,Reference El-Sabbagh, Soliman, Taha and Palkowski32} The Al₄C₃ phase is considerably brittle and unstable,^{Reference Sharifi, Enayati and Karimzadeh18} especially easy to fracture during tensile testing and thus would severely degrade the comprehensive mechanical properties of the composite.^{Reference Sharifi, Enayati and Karimzadeh18,Reference Mitra, Rao, Maiti and Chakraborty31} Besides, the hydrophilic nature of this phase makes the composite very sensitive to corrosive environments.^{Reference Lloyd, Lagace, Mcleod and Morris33} Hence, the formation of the Al₄C₃ phase should be avoided during the fabrication process.

When the SiO₂ film is coated on the surface of the SiC_p which have been oxidized, the content of C element at the SiC/Al interface is greatly decreased, while those of Mg, O (and possibly Al) elements are distinctly increased [Figs. 4(g)–4(l)], indicating that the formation of the Al₄C₃ phase might be suppressed and some new phases containing Mg, O, and/or Al elements may form at the oxidized SiC_p/6061 Al interface. It is proposed that Al could react with SiO₂ to form Al₂O₃ through the following reaction:

(4)$$4{\rm{Al}}\left( {\rm{l}} \right) + 3{\rm{Si}}{{\rm{O}}_2}\left( {\rm{s}} \right) \to 2{\rm{A}}{{\rm{l}}_2}{{\rm{O}}_3}\left( {\rm{s}} \right) + 3{\rm{Si}}\left( {\rm{s}} \right)\quad .$$

Besides, when Mg element exists in the composite, Mg is likely to attack the SiO₂ to form MgO^{Reference Shi, Yang, Lee, Zhang, Lee and Wu34}:

(5)$$4{\rm{Mg}}\left( {\rm{l}} \right) + {\rm{Si}}{{\rm{O}}_2}\left( {\rm{s}} \right) \to 2{\rm{MgO}}\left( {\rm{s}} \right) + {\rm{M}}{{\rm{g}}_2}{\rm{Si}}\left( {\rm{s}} \right)\quad .$$

Another possible reaction suggested is^{Reference Shi, Yang, Lee, Zhang, Lee and Wu34}

(6)$$4{\rm{Al}}\left( {\rm{l}} \right) + 2{\rm{MgO}}\left( {\rm{s}} \right) + 3{\rm{Si}}{{\rm{O}}_2}\left( {\rm{s}} \right) \to 2{\rm{MgA}}{{\rm{l}}_2}{{\rm{O}}_4}\left( {\rm{s}} \right) + 3{\rm{Si}}\left( {\rm{s}} \right)\quad .$$

Since the volume fraction and activity of Mg in touch with SiO₂ films is significantly lower than that of Al with SiO₂, Al first reacts with SiO₂. Hence, it is much more possible that MgAl₂O₄ should form through the combination of reactions (4), (7), and (8) shown as follows:

(7)$$3{\rm{Mg}} + {\rm{A}}{{\rm{l}}_2}{{\rm{O}}_3} \to 3{\rm{MgO}} + 2{\rm{Al}}\quad ,$$

(8)$${\rm{MgO}} + {\rm{A}}{{\rm{l}}_2}{{\rm{O}}_3} \to {\rm{MgA}}{{\rm{l}}_2}{{\rm{O}}_4}\quad .$$

Among the above reaction products, the free energy of formation per mole of MgO is the lowest, followed by that of SiO₂ and Al₂O₃.^{Reference Mitra, Rao, Maiti and Chakraborty31} The change of free energy in reaction (8) is negative as well,^{Reference Mitra, Rao, Maiti and Chakraborty31} indicating that Al₂O₃ and MgO could react to form MgAl₂O₄. This statement can be further verified by the XRD pattern of the oxidized SiC_p/6061 Al composite [Fig. 4(m)]. It was reported that the MgAl₂O₄ phase was prone to form within SiC/Al composites having low Mg contents (≤4.0 wt%) at high temperatures (>873 K), whereas the MgO phase tended to generate in a high Mg contents at low temperatures.^{Reference Lee, Lee, Ahn, Shim and Shi28} This agrees well with the present observations considering 0.8–1.2 wt% Mg content and the processing temperature of 933 K for the oxidized SiC_p/6061 Al composite in this work. The MgAl₂O₄ phase existing on the SiC_p surface has excellent thermal stability, which could inhibit the formation of harmful Al₄C₃ by acting as the protective barrier.^{Reference Mitra, Rao, Maiti and Chakraborty31} More importantly, the interfacial bonding strength of the SiC/Al composite containing MgAl₂O₄ phase is evidenced to be 3 to 4 times higher than those having either no interfacial reaction products or containing only Al₄C₃.^{Reference Lee, Lee, Ahn, Shim and Shi28} Hence, by oxidizing the SiC_p at 1273 K for 8 h, the formation of Al₄C₃ in the SiC_p/6061 Al composite is effectively restrained due to the presence of MgAl₂O₄ at the interface. Herein higher mechanical properties, especially the TE of the oxidized SiC_p/6061 Al composite, can be achieved.

In addition, to maximize composite’s TE for a certain volume fraction of reinforcing particles, some researchers have reported that the following features should be achieved for the PRAMCs^{Reference Lloyd16,Reference Tong35} : (i) a fine (<10 μm), uniform reinforcement size distribution; (ii) control of reinforcement shape; (iii) uniform reinforcement distribution; (iv) a ductile matrix; (v) a high interfacial strength. In what follows, all these factors will be examined in sequence so as to reveal their contributions to toughening mechanisms.

2. Reinforcement size distribution

The mean diameter of the angular SiC_p in this work is measured to evaluate its size distribution. The results shown in Fig. 5(a) suggests that the general particle size distribution fits Gaussian distribution. Nearly 90% of the SiC_p size are below 10 μm, which are mainly distributed around the average value of 6.94 μm. This authenticates that the general distribution of SiC_p size is quite uniform, which can reduce the possibility of premature failure at particular (generally at large) particles during tensile testing,^{Reference Lloyd16} and therefore contribute to a high TE.

FIG. 5.

(a) Particle size distribution of SiC_p utilized in this work, (b) schematic illustration showing definition of aspect ratio in irregular SiC_p, (c) corresponding statistical results, (d) application of the quadrat method, only SiC_p inside and in touch with the top and right side of each quadrat are counted, and (e) results of quadrat analysis based on ten micrographs accompanied with theoretical distribution.

3. Reinforcement shape

The aspect ratio of D/d, as schematically shown in Fig. 5(b), is utilized to evaluate the SiC_p shape. Statistical results disclose that the majority of SiC_p have an aspect ratio varying between 1 and 4 with only a small portion of SiC_p aspect ratio being larger than 6 [Fig. 5(c)]. Although the long and angular particles are more efficient strengtheners (like whiskers) than spherical ones in transferring load from the matrix to reinforcements,^{Reference Majumdar and Pandey36} which results in a higher tensile strength, they are extremely harmful to the TE because of the rapid initiation and growth of voids at sharp angles. Geni and Kikuchi^{Reference Geni and Kikuchi37} favored this statement by the findings that increasing the SiC_p aspect ratio (0.25 → 0.5 → 1 → 2 → 4) increased the overall tensile strength, while degrading the TE inversely. A maximum TE and a medium tensile strength were achieved only when the aspect ratio approached 1 (i.e., spherical particles). Therefore, it can be expected that the utilization of SiC_p with an average aspect ratio of 2.43 for the present composite should balance well between tensile strength and TE.

4. Reinforcement distribution

The SiC_p distribution in the synthesized composite was analyzed quantitatively through the quadrat method,^{Reference Tzamtzis, Barekar, Babu, Patel, Dhindaw and Fan22} to evaluate the effectiveness of the powder thixoforming method in uniformizing particle distribution. Figure 5(d) shows a characteristic example of how this method was implemented on the micrograph, and the results as a frequency scatter of the amount of reinforcement particles in each quadrat, N _q, in association with the according theoretical curves are displayed in Fig. 5(e). The shape of N _q distribution is affected by the magnitude of clustering tendency with a higher particle uniformity making the N _q distribution more symmetric.^{Reference Boostani, Mousavian, Tahamtan, Yazdani, Khosroshahi, Wei, Xu, Gong, Zhang and Jiang38} As proven mathematically,^{Reference Rogers39} a clustered, random, and regular spatial distribution follows the negative binomial model, Poisson model, and binomial model, respectively. The observed particle distribution for the SiC_p/6061 Al composite is more of resemblance to the binomial curve and lies away from the negative binomial curve, which indicates a uniform SiC_p distribution in this composite. More quantitatively, the skewness (β) is calculated to be 0.29 for the powder thixoforming composite. By contrast, it is 0.821 for the high-pressure die-casting composite,^{Reference Tzamtzis, Barekar, Babu, Patel, Dhindaw and Fan22} 0.53 for the gravity-casting composite,^{Reference Karnezis, Durrant and Cantor23} 1.04 for the squeeze-cast composite,^{Reference Karnezis, Durrant and Cantor23} and 0.65 for the roll-cast composite,^{Reference Karnezis, Durrant and Cantor23} respectively (composite here all refers to SiC_p/A356 PRAMCs). An increase in β represents an increase in reinforcement clustering. It is therefore concluded that the powder thixoforming method provides a great advantage for the fabrication of PRAMCs with a more uniform distribution of the reinforcements when compared with conventional methods, which leads to a more homogeneous microstructure and thus high mechanical properties, especially the TE.

5. Matrix ductility

The aforementioned factors all play a crucial role in improving the TE of the composite but are independent of solution treatment. But the matrix ductility should be affected by the solution treatment. Figures 6(a)–6(h) show the fracture surfaces of the SiC_p/6061 Al composite solutionized at 808 K for various durations. It is evident that the fracture surface of the as-fabricated composite is generally flat besides the existence of a few traces of large caves [Figs. 6(a) and 6(e)], which should result from the decohesion of primary α-Al particles considering that the mean size and morphology of these pulling-out caves generally agree with those of primary α-Al particles (16.51 μm). The presence of eutectic phases in the grain boundaries [Fig. 1(b)], which is prone to fracture under a low tensile load, accounts for the brittle fracture of the as-fabricated composite. An increase in solution time decreases the amounts of eutectic phases [Fig. 2(d)], leading to the formation of plentiful dimples on the fracture surface of the solution-treated composite for 2 h [Figs. 6(b) and 6(f)]. The dimples indicate ductile fracture,^{Reference Pineau, Benzerga and Pardoen40} whose amount becomes much larger with increasing solution time to 6 h [Figs. 6(c) and 6(g)]. Further increasing solution time to 14 h coarsens the primary particles to a large extent^{Reference Zhang, Chen and Qin7} and numerous flat facets appear on the fracture surface [Figs. 6(d) and 6(h)]. The flat facets are thought to result from the cracks developing through primary α-Al particles^{Reference Zhang, Chen, Cheng and Li41} and are indicative of brittle fracture, thus the ductility of the solution-treated composite for 14 h are degraded to some extent. Overall, the ductility of the SiC_p/6061 Al composite matrix first increases (0–6 h) and then decreases (after 6 h) in the process of solution treatment, which is in good agreement the variation of TE with solution time (Table I).

FIG. 6.

Fracture surfaces of the SiC_p/6061 Al composites solutionzied at 808 K for various durations (a and e) 0 h (as-fabricated state), (b and f) 2 h, (c and g) 6 h, and (d and h) 14 h; (e–h) are the enlarged views of the squares in (a–d), (i) typical fracture surface of the SiC_p/6061 Al composite containing SiC/Al interface debonding and SiC_p fracture with corresponding results of EDS analysis, and (j) statistical result of fractured SiC_p to the whole failed ones as a function of solution time.

6. Interfacial bonding strength

In addition, another phenomenon that varies during solution treatment is the number and mode of failed SiC_p (including SiC/Al interface debonding and SiC_p fracture) on the fracture surfaces of the SiC_p/6061 Al composites [Figs. 6(e)–6(h)]. Note that it is hard to distinguish SiC/Al interface debonding from SiC_p fracture in certain specific particles [Fig. 6(i)], and EDS analysis was utilized to differentiate these two modes since Al and a small amount of Mg could react with SiO₂ through reactions (4) and (5) (stated in section “C”) to enrich Al (and occasionally tiny Mg) element at the oxidized SiC/Al interface (Fig. 4). As shown by the EDS results in Fig. 6(i), it should be SiC/Al interface debonding when there are no Al and Mg elements on the SiC_p surface; otherwise, it belongs to SiC_p fracture. Clearly, the SiC/Al interface debonding [marked by arrows labeled “A” in Fig. 6(e)] dominates the as-fabricated composite by forming pulling-out caves on the fracture surface. When the composite is subjected to solution treatment, SiC_p fracture is frequently observed on the fracture surface [marked by arrows labeled “B” in Figs. 6(f)–6(h)], and the ratio of fractured SiC_p number to the whole failed particle number continuously increases with prolonging the solution time to 6 h and then decreases slightly [Fig. 6(j)]. Both the debonding and fracture of SiC_p are actually the results of stress concentration at the SiC/Al interface when the external stress is up to a certain level.^{Reference Chawla, Jones, Andres and Allison42} When the interface bonding strength is relatively weak, the SiC_p would debond from the Al matrix under uniaxial tensile load, disabling the load transfer from the matrix to SiC_p.^{Reference Chen, Ling, Xia, Huang and Fang43} But when the interface bonding strength is strong enough, external load can be effectively transferred to the SiC_p, which then fracture under a higher tensile stress.^{Reference Chen, Ling, Xia, Huang and Fang43} Therefore, a gradual increase in fractured SiC_p ratio within 6 h, in fact, indicates an enhanced interfacial bonding strength in the SiC_p/6061 Al composite during solution treatment.

In conclusion, the toughening mechanisms of the solution-treated SiC_p/6061 Al composite, as schematically shown in Fig. 7, can be summarized from the perspective of material design, powder thixoforming, and solution treatment as follows (panels labeled “i–vi” in the figure correspond to “i–vi” explanations in what follows): (i) utilization of SiC_p having a mean size of 6.94 μm with nearly 90% of the particle size being below 10 μm reduces premature failure of the composite at large particles; (ii) selection of SiC_p with an average aspect ratio of 2.43 achieves a good balance between tensile strength and elongation; (iii) oxidization of the SiC_p provides a good interfacial bonding strength by forming the MgAl₂O₄ phase at the SiC/Al interface; (iv) the powder thixoforming method makes possible a considerably uniform distribution of SiC_p in the Al matrix eliminating stress concentration at certain particular areas. These four factors lay foundations for the subsequent solution treatment, which dissolves eutectic phases into the matrix alloy leading to (v) a ductile matrix and (vi) a high interfacial strength in the composite.

FIG. 7.

Schematic illustration of the toughening mechanisms of the SiC_p/6061 Al composite during the material design, powder thixoforming, and solution treatment processes.

7. Elongation prediction

In this part, the SF of the SiC_p/6061 Al composite is calculated. In fact, the SF prediction of PRAMCs is a rather complicated issue requiring a detailed damage analysis to have an estimation of the onset of specimen instability. Available models on predicting SF, such as those proposed by Tanaka et al.^{Reference Tanaka, Mori and Nakamura44} or Evensen–Verk,^{Reference Evensen and Verk45} were found to be inadequate for the SF prediction due primarily to their partial neglect of the reinforcement premature failure including reinforcement crack and interface debonding, followed by matrix fracture in the ligament between the reinforcements during tensile testing (predictions from some of the mentioned models will be given in the following for comparison). In this work, the damage model of Majumdar and Pandey^{Reference Majumdar and Pandey36} based on the fracture mechanics of reinforcing particles was adopted to predict the SF of the synthesized composite. This model is found to perform well in SF prediction of PRAMCs, especially when the failed SiC_p are uniformly distributed in the composite and their volume fraction is small. Examination of the failed SiC_p on the fracture surface, as typically exhibited in Ref. Reference Zhang, Chen and Qin7, confirms the applicability in the current composite by showing a relatively homogeneous distribution and a relatively small volume fraction of failed particles. In accordance with the Majumdar and Pandey model, the total SF for the bulk composite (ε_f) can be expressed as^{Reference Majumdar and Pandey36}

(9)$${\varepsilon _{\rm{f}}} = \left( {{1 \over {\lambda \sqrt 2 }}} \right) \cdot \left( {1 - {{\left( {{{6{V_{\rm{p}}}} \over \pi }} \right)}^{{1 \over 3}}}} \right) \cdot \left( {1.285 - 4.403{V_{\rm{p}}} + 4.122{{\left( {{V_{\rm{p}}}} \right)}^2}} \right)\quad ,$$

where V _p is the volume fraction of SiC_p and λ is a parameter corresponding to the local SF, which can be obtained via the following equation in accordance with McMeeking’s results^{Reference Majumdar and Pandey36,Reference Mcmeeking46} :

(10)$$\lambda = {{9.4 + 35{\varepsilon _{\rm{B}}}} \over {0.3 + 100{\varepsilon _{\rm{B}}}}}\quad .$$

Among them, ε_B is the plastic strain when there is distinct void initiation and propagation, which can be given by^{Reference Majumdar and Pandey36}

(11)$${\varepsilon _{\rm{B}}} = \left( {{{\varepsilon _{\rm{m}}^{\rm{f}}} \over {\left( {2.2 + 7N)(N + 1} \right)}}} \right) \cdot {\left( {{{\varepsilon _{\rm{m}}^{\rm{f}}} \over {0.002}}} \right)^N}\quad ,$$

where $\varepsilon _{\rm{m}}^{\rm{f}}$ and N are the respective tensile SF of the matrix and work-hardening exponent, which are thought to be identical in value.

As mentioned above, numerous SiC_p in the work would inevitably fail through SiC/Al interface debonding and/or SiC_p fracture during the tensile testing process [Figs. 6(e)–6(h)], resulting in ductility reduction of the composite. Herein, the fraction of failed SiC_p (f) is introduced to modify Eq. (9):

(12)$$f = \left( {{n \over N}} \right) \times 100\% \quad ,$$

where N and n are the number of visible SiC_p on the SEM micrograph and fractographs under the same magnification, respectively. The variations of f with solution time have been given in Ref. Reference Zhang, Chen and Qin7. Finally, a modified model for predicting the SF of the SiC_p/6061 Al composite that considers the effective volume fraction of SiC_p (V _p(1−f)) can be developed in this work:

(13)$${\varepsilon _{\rm{f}}} = \left( {{1 \over {\lambda \sqrt 2 }}} \right) \cdot \left( {1 - {{\left( {{{6{V_{\rm{p}}}\left( {1 - f} \right)} \over \pi }} \right)}^{{1 \over 3}}}} \right) \cdot \left( {1.285 - 4.403{V_{\rm{p}}}\left( {1 - f} \right) + 4.122{{\left( {{V_{\rm{p}}}\left( {1 - f} \right)} \right)}^2}} \right)\quad .$$

The SF of the SiC_p/6061 Al composite given by Eq. (13) along with the experimental values is plotted in Fig. 8 as a function of solution time. The calculation details are listed in Table II. Predictions from this model agree generally well with the experimental values considering the measuring error, especially when compared with those from the Evensen–Verk model^{Reference Evensen and Verk45} and Brown–Embury model^{Reference Brown and Embury47} (taking C = 1.5, ε_n = 0 and V = V _p(1−f) in these models). Note that SF predictions within the solutionization range of 1 h deviate obviously from the experimental values. This might be ascribed to the incomplete dissolution of brittle eutectic phases in this range [Fig. 2(d)], which is detrimental to the composite ductility. For instance, some eutectic phases tend to distribute around SiC_p [Fig. 1(c)], which could serve as crack initiators since these eutectic phases are hard and brittle in nature. In this model, the effects of eutectic phases are only considered partially via the work hardening exponent of the matrix (N). However, the overall harmful effects of the eutectic phases cannot be fully quantified in this calculation due to complexity in its morphology and distribution, consequently leading to the overestimation of SF predictions in this range. When the number of eutectic phases decreases after 1 h, the SF predictions tend to agree well with the experimental values. This calculation again demonstrates the harmful effects of the eutectic phases on composite’s ductility and thus the effectiveness of solution treatment in improving the composite’s ductility. Furthermore, the modified model might provide references for SF prediction of similar PRAMCs that are subjected to solution treatment.

FIG. 8.

SF comparisons of the experimental values and theoretical ones calculated by the modified model in this work as well as by the reported Evensen–Verk model^{Reference Evensen and Verk45} and Brown–Embury model^{Reference Brown and Embury47} in the literature studies as a function of solution time.

TABLE II.

Calculation details for the strain to failure prediction of SiC_p/6061 Al composites by different models.