Structures and formation mechanisms of dislocation-induced precipitates in relation to the age-hardening responses of Al-Mg-Si alloys

1. Introduction

Al-Mg-Si alloys are widely used in transportation, such as automobiles, high-speed trains, and airplanes owing to their high strength-to-weight ratio, good formability, excellent weldability and corrosion resistance [1]. The superior properties of Al-Mg-Si alloys are mainly derived from a large amount of dispersed nano-sized precipitates formed during aging treatment, which can effectively impede the moving dislocations [[2], [3], [4]]. The mostly believed precipitation sequence of Al-Mg-Si alloys is the following [[5], [6], [7], [8], [9]].

Supersaturated solid solution → clusters/GP zones → β"→ β' (/B'/U1/U2)→ β, in which all the metastable precipitates in the alloys have needle/rod/lath shapes with their longest dimension parallel to <100>_Al directions. The needle-like monoclinic β"-precipitates (Mg₅Si₆) and their early-precursors mostly form at the peak-aging stage, and are the most effective strengthening precipitates in Al-Mg-Si alloys [5,6]. The β'-phase (hexagonal Mg₉Si₅) [10] usually appears upon over-aging prior to the equilibrium β-phase, and sometimes may be accompanied by the B'-phase (hexagonal Al₃Mg₉Si₇) [11], the U1-phase (trigonal MgAl₂Si₂) [12] and the U2-phase (orthorhombic Mg₂Al₂Si₂) [13]. Recently, a new phase called β'-2 was found to coexist with one β'-phase precipitate in an over-aged Al-Mg-Si alloy with an impurity level of Cu addition, which is isostructural with β'-Cu (Ag), but having Cu replaced by Si, and Al by Mg [14].

Nevertheless, our recent study has demonstrated that the phases formed in Al-Mg-Si alloys are closely related to the composition of the alloys and the heat treatment applied [15,16]. In most cases, the phase formed in the early-stage of aging is the β"-phase. While in some special cases, for example, in high-temperature-aged (above 250 °C) Mg-excess alloys with ordinary solute content (Mg + Si) > 1 wt.% [15] and in common-temperature-aged (∼180 °C) non-Mg-excess but lean alloys [16], reversal phenomena of precipitation sequence will occur, where the β'-phase can form directly without the formation of the β”-phase. These phenomena indicate that nucleation of the β'-phase and nucleation of the β"-phase can be parallel or independent of each other, largely depending on the Si-content in the matrix [16]. Hence, by their major hardening phase to be formed upon artificial aging rather than by their composition, Al-Mg-Si alloys can be more specifically classified into “normally-β”-hardened” alloys and “normally-β'-hardened” alloys. The reason for adding the word “normally” to the two phrases is that under circumstances the “normally-β”-hardened” alloys may change into β'-hardened alloys abnormally [15,16]. The same may also happen for the “normally-β'-hardened” alloys, as will be shown in the present study.

Another important issue concerning the precipitation behavior of Al-Mg-Si alloys is the following. In industrial practice, products made of age-hardening Al alloys often have to experience a certain degree of small-deformation (with strain < 10%), such as straightening or forming, before the final aging treatment takes place. In this case, the precipitation occurs in a material containing dislocations introduced by small-deformation. Therefore, it is important to know the effect of small-deformation on the precipitation behavior and correspondingly on the mechanical properties of Al-Mg-Si alloys. It has been reported that in the presence of dislocations introduced by 10% deformation prior to 200 °C-aging in a Si-excess Al-Mg-Si alloy, the precipitate type is changed from the β”-precipitates in the non-deformed condition to the "string-like" precipitates, the "elongated type" precipitates, associated with the formation of the B'-precipitates at dislocations [17]. It has also been shown for an AA6060 alloy with 10% deformation prior to 190 °C-aging that nearly all precipitates are formed at dislocations. They are mostly disordered precipitates, but partially are the β' (/B'/U2)-precipitates [18,19]. In addition, although the total yield strength of the pre-deformed alloy is slightly improved due to the dislocation strengthening, the precipitation strengthening effect is reduced actually because of the change in precipitate microstructure [19].

Although the effect of small-deformation on the precipitation behavior in Al-Mg-Si alloys has been extensively studied and eff ;orts have been made to understand the structural characteristics of dislocation-induced precipitates [[17], [18], [19], [20], [21]], using mostly resolution-limited conventional high-resolution transmission electron microscopy (HRTEM), the following three issues still remain to be investigated or answered:

(1)What are the precise atomic-structures of all those disordered precipitate complexes and how do they form upon aging?

(2)Why all the dislocation-induced precipitates grow parallel to each other along the same needle-direction?

(3)What kinds of links exist between the matrix precipitates and the dislocation-induced precipitate complexes?

In the present study, atomic-resolution electron microscopy was used to precisely characterize the precipitate microstructures in slightly pre-deformed Al-Mg-Si alloys. Our results reveal that the majority of dislocation-induced precipitates are short-range ordered while long-range disordered polycrystalline precipitates and multiphase composite precipitates. The formation mechanisms beneath their complex structures are revealed. Moreover, the dislocation-induced precipitation shall result in a coarser precipitate microstructure in the matrix of the “normally-β”-hardened” alloy, while result in a reversal of the precipitation pathway in the matrix of the “normally-β'-hardened” alloy, leading to a reduced age-hardening potential in the former alloy and an improved age-hardening potential in the latter alloy.

2. Experimental procedure

Two alloys with chemical compositions of Al-0.75 Mg-0.75 Si-0.14 Fe-0.24 Cr-0.02 Ti (wt.%) and Al-1.00 Mg-0.50 Si-0.14 Fe-0.24 Cr-0.02 Ti (wt.%), referred respectively as Alloy A (Mg-Si-balanced) and Alloy B (Mg-rich) hereafter, were investigated in the present study. Two thermal aging processes were performed for comparison: (i) Direct artificial aging (abbreviated as AA treatment hereafter). Sliced specimens were firstly solution-treated at 560 °C for 0.5 h and then quenched in cold water to room temperature. Without apparent natural aging they were finally aged at 250 °C for various times in an oil bath. As will be shown later, upon the AA treatment the alloy A will form the β"-precipitates, whereas the alloy B will form the β'-precipitates in their Al-matrix. (ii) Pre-deformation + artificial aging (abbreviated as PDA treatment hereafter). Immediately after the same solution and water-quenching treatment as that for the AA-treated samples, samples were plastically deformed by 5%, followed by a final 250 °C-aging. The time interval between the pre-deformation and the artificial aging is less than 10 min to avoid apparent natural aging effect [15,22].

The Vickers hardness was measured with a load of 4.9 N and a dwell time of 10 s. Each hardness value was the average value of at least 5 indentations. TEM and HRTEM observations were performed with a FEI Tecnai F20 TEM operated at 200 KV. HAADF-STEM (high-angle annular dark-field in scanning TEM imaging mode) characterization was carried out with a FEI Titan Cubed Themis G2 operated at 300 kV. All microstructure images were taken from a <100>_Al zone axis. The TEM/STEM specimens were prepared firstly by mechanical polishing and then by electro-polishing until perforation.

3. Results and discussion

3.1. Age-hardening responses

Fig. 1 shows the age-hardening curves of the AA-treated samples and the PDA-treated samples plotted against the aging time. It can be seen that the effects of pre-deformation on the age-hardening behavior of the two alloys are significantly different. The main points shown in Fig. 1 are as follows. (i) Before aging treatment, the PDA-treated samples possess higher hardness values than the AA-treated samples do, owing to the strain hardening caused by the dislocations introduced by pre-deformation [23,24]. (ii) For Alloy A, the peak hardness of the PDA-treated sample is lower than that of the AA-treated sample. Whereas for Alloy B, the situation is just the opposite. More details of the hardness results will be discussed in association with the TEM results in Section 3.5.

3.2. TEM/HRTEM overviews of typical precipitate microstructures

Four peak-aged samples marked with open black circles in Fig. 1 were selected for TEM and HRTEM observations. Fig. 2(a) and (b) shows the overall precipitate microstructures in Alloy A samples subjected to AA treatment and PDA treatment, respectively. With an average length of about 46 nm, the precipitates in the AA-treated sample form homogeneously in the Al-matrix (Fig. 2(a)). While in the PDA-treated sample, apart from the homogeneously distributed precipitates in the matrix (referred to as matrix precipitates), there exist a significant proportion of precipitates that obviously parallel to each other and line up (as marked by dotted-line circles in Fig. 2(b)). These type of precipitates have been believed being those precipitates nucleated at the dislocations introduced by deformation [18] and are referred to as dislocation-induced precipitates. Moreover, comparing the PDA-treated sample with the AA-treated sample, the matrix precipitates formed in the former sample have the average length of 73 nm, which are much larger than those formed in the latter sample.

Conventional HRTEM characterization was further performed on the same samples in Fig. 2(a,b) to identify the crystal types of the precipitates. Typical images are shown in Fig. 3. All the precipitates in the AA-treated sample are the β”-precipitates (Fig. 3(a)) [5]. While with the introduction of pre-deformation, the precipitate structures become much more complicated in the PDA-treated sample. Although most of the matrix precipitates are still the β"-precipitates (Fig. 3(b)) [5], nearly all the dislocation-induced precipitates exhibit varying degrees of disorder and are hardly to be identified, as shown in Fig. 3(c)-(f). Nevertheless, the complexity ofentified, as shown in Fig. 3(c)-(f). Nevertheless, the complexity of these dislocation-induced precipitates is that many of them may show locally recognizable periodic features at some imaging conditions. For instance, unit cells of the β'-phase [10] and unit cells of the U2-phase [13] can be seen locally in the precipitates shown in Fig. 3(c) and (d), respectively. The precipitate in Fig. 3(e) exhibits the periodicity of 0.99 nm along <510>_Al, which is one of the characteristics of the B'-phase [11]. But there also exist some precipitates that seem to be completely disordered without any recognizable features, such as the precipitate shown in Fig. 3(f).

Fig. 4(a) and (b) shows the overall precipitate microstructures in the Alloy B samples subjected to AA treatment and PDA treatment, respectively. It is seen that the AA-treated Alloy B contains very coarse precipitates with an average length of about 147 nm (Fig. 4(a)). Further HRTEM observation demonstrates that they are the β'-precipitates (Fig. 4(c)) [10], which is in agreement with our previous work [15]. For the PDA-treated Alloy B, its precipitate microstructure consists of quite fine matrix precipitates and relatively coarse dislocation-induced precipitates (Fig. 4(b)). The average length of the matrix precipitates is about 70 nm, which is much shorter than that of the matrix precipitates formed in the AA-treated Alloy B (Fig. 4(a)). Further HRTEM characterization shows that most of the matrix precipitates are unexpectedly changed to the β”-precipitates (Fig. 4(d)) [5]. For the dislocation-induced precipitates, as marked by dotted-line circles in Fig. 4(b), their crystal structure features revealed by HRTEM are quite similar to those observed in the PDA-treated Alloy A and therefore are not shown repeatedly here.

In a brief summary, introducing deformation dislocations before aging will significantly change the precipitate microstructures in both Alloy A and Alloy B, but in different manners. For the “normally-β”-hardened” Alloy A, formation of dislocation-induced precipitates results in coarsening of matrix precipitates, but would not change their precipitate structure type. While for the “normally-β'-hardened” Alloy B (when aged at 250 °C), formation of dislocation-induced precipitates changes not only the morphology, but also the structure type of matrix precipitates. The above observed microstructure changes can well explain the property changes demonstrated in Fig. 1(a,b). However, in order to fully understand the mechanism beneath these microstructure and property changes, detailed atomic-structures of the dislocation-induced precipitate complexes have to be revealed.

3.3. Atomic-structures of the dislocation-induced precipitates

In order to reveal atomic-structures of the dislocation-induced precipitates, atomic-resolution HAADF-STEM imaging was carried out on the peak-aged Alloy A and Alloy B samples. All the HAADF-STEM images shown in this study have been FFT filtered to remove the high-frequency noises. For the sake of convenience in the following structure analysis, Fig. 5 gives schematic drawing of the unit cells of known precipitates [10,11,13,14] encountered in our HAADF-STEM characterization. Note that the structure portion(s) enclosed by triangles in Fig. 5(a), by hexagons in Fig. 5(b), by regular hexagon in Fig. 5(c) and by triangle-like hexagon in Fig. 5(d), represent the characteristic sub-unit(s) of B', U2, β' and β'-2, respectively. These features are useful in analysis of the dislocation-induced precipitate complexes.

Fig. 6(a) shows a low-magnification HAADF-STEM image of the cross-sections of the typical dislocation-induced precipitates in the PDA-treated Alloy A. The atomic-resolution HAADF-STEM images show that these precipitates can be divided into three groups, as indicated by red, blue and yellow dotted circles respectively in Fig. 6(a). It is interesting that all the precipitates indicated by red dotted circles are polycrystalline U2-phase precipitates, in which the U2-grains have at least two different orientations. Fig. 6(b, c) shows two examples of the atomic-structures of this type of precipitates. Two orientation relationships (ORs) between the U2-phase and the Al-matrix are observed here, i.e., OR1: [100]_U2 // <310>_Al, [010]_U2 // [001]_Al, and OR2: [100]_U2 // <110>_Al, [010]_U2 // [001]_Al. This group of precipitates will be named as polycrystalline U2 precipitate hereafter. Although the grain boundary structures between U2-grains or between U2 and Al-matrix are disordered, the polycrystalline U2 precipitate should have approximately the same composition as that of the U2-phase (Mg₂Al₂Si₂) [13], in which the Mg-to-Si atomic (Mg/Si) ratio is 1.

Two other types of precipitates observed in Fig. 6(a) are shown in Fig. 7. The original images without any marks are given in Fig. 7(a) and (c). From the analysis shown in Fig. 7(b), we can see that apart from the polycrystalline U2 precipitate, this precipitate complex also contains two more other phases, i.e., it is not only a polycrystalline precipitate, but also a multiphase composite precipitate. For the polycrystalline U2 precipitate part, the ORs observed are the same as what have been observed in Fig. 6(b,c). For the lower left portion of the precipitate, unit cells of the B'-phase can be identified, with the OR of [0001]_B' // [001]_Al, < 11$\bar{2}$0 >_B' // <510>_Al (OR3). While for the upper left part of the precipitate, an unknown structure is observed: its unit cell contains sub-units of the B'-phase, but is larger than that of the latter. More evidence for the existence of such a new phase (structure) can be found in Figs. 7(d) and 8 (a), where two precipitates containing large portions of this new structure are again observed. In the following this phase will be referred to as B'-2 phase. Note that when we consider the B'-2 phase as a 3D crystal phase rather than a 2D crystal phase, as observed in the present case, its 3D monoclinic unit cell should be doubled in volume, extending approximately along the [0$\bar{1}$0]_Al direction (Fig. 7(d)).

Before analyzing the structure of the B'-2 phase, one issue concerning the characteristic sub-units (Fig. 5(a)) of the B'-phase and also of the B'-2 phase must be discussed. Based on a previous first-principles calculations investigation [11], the atomic sites at the center of the sub-units of the B'-phase, named as V-sites hereafter (as indicated by the dotted circles in Fig. 7(b)), were considered to be occupied by vacancies, since the formation enthalpy of the B'-phase with Al occupying the V-sites is slightly higher than that with vacancies occupying these sites. However, our atomic-resolution images (Fig. 7, Fig. 8) clearly indicate that the V-sites in both the B' and B'-2 phases can be partially occupied by atoms rather than entirely occupied by vacancies.

Based on the Z-contrast atomic-images, inter-atomic distances and local similarities with the B'-phase, the atomic structure model of the B'-2 phase can be constructed, given that the V-site columns are partially occupied by Al atoms. Table 1 lists the experimentally suggested lattice parameters and atomic coordinates of the B'-2 phase with a composition of Si₁₃Mg₁₄Al_2(5+x) (0 < x < 1). A further systematic first-principles calculations study is needed to fully understand the coexistence of different phases observed in the present work, in which not only the role of strain energy, but also that of interfacial energy for these phases should be considered.

The a-axis of the B'-phase and the c-axis of the B'-2 phase have the same lattice parameter (1.04 nm) and both are parallel to <510>_Al. Along this direction, the two phases are coherent with the matrix (the misfit value between the two phases and the matrix calculated to be 0.7%). Then it is understandable that in the cross-sections of the B'-phase and the B'-2 phase observed in the investigated condition, they have one (for the B'-phase) or half (for the B'-2 phase) unit cell growing along <510>_Al.

A series of images reveal that pure B'-phase and pure B'-2 phase, as the cases given in Figs. 8(b) and 7 (d), respectively, are rarely encountered. Both the B'-phase and the B'-2 phase tend to coexist with the (polycrystalline) U2 precipitate to form two-phase (as the example shown in Fig. 8(a)) or three-phase (as the example shown in Fig. 7(b)) composite precipitates, generating three precipitate types termed as B'/U2 composite precipitate, B'-2/U2 composite precipitate and B'/B'-2/U2 composite precipitate. It is worth mentioning that since both the B'-phase and the B'-2 phase (with Mg/Si ratios of 1.29 and 1.08, respectively) are Mg-rich [11], the above three composite precipitates are Mg-rich. Nevertheless, their accurate compositions cannot be determined without knowing the specific number of atoms in the atomic columns at boundaries and interfaces.

Interestingly, the fraction of each phase in the multiphase composite precipitates is variable from precipitate to precipitate. For instance, in a B'/U2 composite precipitate, in one case the B'-phase dominates, while in another case the (polycrystalline) U2 precipitate dominates. In the latter case, few unit cells and/or sub-units of the B'-phase are present, see the precipitates shown in Fig. 9. Here new ORs that differ from those mentioned earlier are observed for both the U2-phase and the B'-phase, i.e., OR4: [100]_U2 // <710>_Al, [010]_U2 // [001]_Al, OR5: [100]_U2 // <410>_Al, [010]_U2 // [001]_Al, and OR6: [0001]_B' // [001]_Al, < 11$\bar{2}$0 >_B' // <710>_Al.

In addition to the precipitates demonstrated above, two types of dislocation-induced β'-phase precipitates were observed in the PDA-treated Alloy A (Fig. 10). One is lath-like β'-precipitate with Mg/Si ratio of 1.8 [10]. Detailed image analysis reveals that a small portion of this β'-precipitate contains some β'-2 unit cells (Fig. 10(a,b)). Fig. 10(b) shows an enlarged image of the area marked by the white dashed box in Fig. 10(a), in which the β' unit cell and the β'-2 unit cell are indicated by white lines and blue lines, respectively. The second one is a multiphase composite precipitate consisting of the β'-phase and the (polycrystalline) U2 precipitate, as shown in Fig. 10(c), which is also Mg-rich, designated as β'/U2 composite precipitate in this paper. It should be mentioned that, like the U2-phase and the B'-phase, the β'-phase was also observed having different ORs with the matrix. A recent study has reported similar phenomenon in an over-aged Al-Mg-Si-Cu alloy [25].

For the PDA-treated Alloy B, its dislocation-induced precipitate types are mainly the β'/U2 composite precipitates (Fig. 11(a,b)), the polycrystalline U2 precipitates (Fig. 11(c)) and the lath-like β'-precipitates (Fig. 11(d)). They are somewhat different from those formed in the PDA-treated Alloy A. There are almost no B'-phase, B'-2 phase and the related composite precipitate variants in the PDA-treated Alloy B. Fig. 11(b) shows a special β'/U2 composite precipitate, in which the β'-phase parts repeatedly interrupt the continuous formation of the (polycrystalline) U2 precipitates. Note that such kind of precipitate would be easily recognized as completely disordered precipitate in normal HRTEM imaging due to image delocalization effect [26].

Fig. 12 demonstrates the formation frequencies of all observed dislocation-induced precipitate complexes in the two PDA-treated alloys, estimated according to the atomic-resolution HAADF-STEM examinations on hundreds of these precipitates. It can be seen that in the PDA-treated Alloy A, the main precipitate complexes are the polycrystalline U2 precipitates and the B' (/B'-2) related composite precipitates, while in the PDA-treated Alloy B, the main precipitate complexes are the polycrystalline U2 precipitates and the β'/U2 composite precipitates (the latter account for a much higher proportion than the former). In both alloys, the proportions of single-phase precipitates (i.e., β' (/B'/B'-2)) are quite low. It should be pointed out that there always exist multiphase composite precipitates in each bunch of the dislocation-induced precipitates. As mentioned earlier, the Mg/Si ratio of the polycrystalline U2 precipitates is close to 1, but all the multiphase composite precipitates are Mg-rich (Mg/Si ratio > 1). Therefore, the overall composition of each bunch of the dislocation-induced precipitates is Mg-rich.

3.4. Formation mechanisms of the dislocation-induced precipitates

The dislocation-induced precipitates tend to form and line up continuously along small-angle boundaries formed by dislocations, as shown in Fig. 13(a). Our results reveal that they usually appear as short-range ordered while long-range disordered polycrystalline precipitates and multiphase composite precipitates, which can hardly be observed in non-deformed Al-Mg-Si alloys. Actually, this phenomenon can be understood as the consequence of a high precipitation kinetics at dislocations (Fig. 13(b)). There are reasons and evidences for such a judgement. Firstly, the diffusion rates of solute atoms at dislocations are significantly larger than that in the matrix [27], leading to a rapid segregation of solutes and vacancies at dislocations [28]. Secondly, there is evidence in our study that the time needed to form precipitates in the PDA-treated sample is much shorter than that in the AA-treated sample (Fig. 14). Furthermore, the early precipitation occurring in the PDA-treated sample mostly appears at the dislocations which form small-angle boundaries (Fig. 14(c)).

The detailed characterization of the dislocation-induced precipitate complexes also provides useful hints to understand the formation mechanism beneath their complex morphology and structures. (1) It has been shown that all the precipitates in one bunch are still needle-like along a <100>_Al direction, but prefer to line up and to grow parallel to each other with a same <100>_Al length direction. This clearly indicates that the parallel dislocation lines at the small-angle boundary have guided the nucleation and growth direction of these well-aligned precipitates (Fig. 13), since otherwise they could grow long in other equivalent <100>_Al directions. (2) We have observed that the polycrystalline U2 precipitates frequently appear, whereas the polycrystalline precipitates of β'-phase and B'/B'-2 phases are hardly seen among the dislocation-induced precipitate complexes, though different possible orientations do exist for all these phases, apart from the fact that their easy-growth direction has been fixed to the associated dislocation line direction (nonetheless they could still rotate around the dislocation line direction to a different orientation). Table 2 lists a few possible orientations for the U2-phase and the B'-phase and their corresponding lattice misfit values. It can been seen that the misfit values of different orientations for the U2-phase are very similar, whereas those for the B'-phase are quite different, indicating that the U2-phase may form energetically-equivalent variants with different orientations, while the B'-phase prefers to form with one energetically-favorable orientation, the same for the β'-phase and other phases. (3) It has been shown that in both alloys a large percentage of the dislocation-induced precipitates are Mg-rich composite precipitates with the Mg/Si ratios larger than one (Fig. 12). This clearly indicates that the dislocations attract more Mg than Si atoms to segregate around them, such that the Mg/Si ratios around dislocations are generally larger than that of the matrix, leading to rapid formation of the Mg-rich late-stage precipitates at dislocations upon aging, prior to the formation of the matrix precipitates. Moreover, frequent formation of composite precipitates also implies that the solute concentration and composition fluctuation near dislocations may vary from one region to another. As such, different phases can nucleate simultaneously in the early-stage of precipitation and finally form composite ones upon further growth and development, as illustrated in Fig. 13.

It is worth to mention that deformation dislocations may annihilate upon aging, especially when the samples are aged at high temperatures [29]. In our experiment, we have hardly observed dislocations associated with the well-grown precipitates in the peak-aging condition, indicating a large degree of dislocation annihilation. This implies that the dislocation lines play crucial roles mostly in the early-stage of precipitation. Once their well-aligned nuclei are formed, these precipitates have to grow long parallel to each other and no longer need the guidance of dislocation lines.

3.5. Impacts of dislocation-induced precipitation on matrix precipitation and age-hardening responses

Having revealed and understood the structures and the formation mechanisms of the dislocation-induced precipitate complexes, we may now better understand their impacts on the formation of matrix precipitates and on the age-hardening responses of the alloys. Firstly, the TEM results have shown that for the “normally-β”-hardened” Alloy A, upon the PDA-treatment at 250 °C, the dislocation-induced precipitation does not change the type of matrix precipitates but makes them coarser (Fig. 2). This is understandable quite easily, since the rapid precipitation at dislocations has quickly lowered the super-saturation of solute atoms in the matrix, leading to a lower nucleation rate, i.e., less nuclei of β”-phase, and therefore a larger average grown size of β”-precipitates in the less densely-precipitated matrix. Although the matrix precipitates are still the most effective hardening β”-precipitates in the PDA-treated Alloy A, their total volume fraction becomes significantly smaller due to formation of the dislocation-induced precipitates, which are much less effective in hardening the alloy. As a consequence, the PDA-treated Alloy A is softened upon the peak-aging in comparison with the AA-treated Alloy A, as shown in Fig. 1(a).

Secondly, we have seen that for the “normally-β'-hardened” Alloy B (i.e., the alloy is hardened mostly by the β'-precipitates upon the AA-treatment at 250 °C), upon the PDA-treatment at 250 °C, the dislocation-induced precipitation changes not only the morphology but also the phase-type of matrix precipitates, i.e., the matrix precipitates in the PDA-treated Alloy B become very fine β”-precipitates rather than the very coarse β'-precipitates in the AA-treated sample (Fig. 4). This can now be explained by several concrete facts, though some comprehensive analyses are needed, as briefly discussed below.

(1) It has been reported by our previous studies [16,30] that in Al-Mg-Si alloys the formation of β"-phase and the formation of β'-phase are actually two independent precipitation sequences and therefore the β'-phase can form directly without the β"-phase as its precursor. Upon aging, which of the two phases precipitate first in the matrix can be determined by the following equation [31,32]:

ΔG=$\frac{αΩ}{(k_BT)^2}[ln(C^Si/C_{eq}^{Si})]^{-2}$ (1)

Where α is a newly added constant here, which is defined to be associated with the Mg-concentration. Ω is assumed being constant for a concerned phase (β' or β”) to nucleate during aging, k_B is the Boltzmann constant, T is the temperature, C^Si is the Si-concentration and $ C_{eq}^{Si}$ is a characteristic Si-concentration value, below which a concerned phase cannot nucleate and form. For a concerned phase, when the Mg-concentration is sufficiently high for its nucleation, then α = 1, otherwise, α > 1. And the lower the Mg-concentration, the higher the α value. Moreover, the α value for β'-nucleation (α_β') is generally higher than that for β”-nucleation (α_β") under a same insufficient Mg-concentration value, due to the fact that the formation of β'-phase (Mg₉Si₅) requires a higher Mg-concentration than that of β”-phase (Mg₅Si₆) does. The detailed description of other parameters in Eq. (1) has been given in our previous work [16]. Note that the Ω for β'-nucleation (Ω_β') is much larger than that for β”-nucleation (Ω_β"), because Ω is related with the interfacial coherency energy between the nuclei and the matrix [33]. Fig. 15 illustrates qualitatively the dependence of the nucleation energy barriers ΔG on C^Si for β" and β' under high (or sufficient) Mg-concentration (black lines) and low Mg-concentration (red lines). It is seen that in all cases the ΔG decreases with increasing C^Si. When Mg-concentration is sufficient, ΔG_β" equals to ΔG_β' at the Si-concentration value $C_0^{Si}$1 (Fig. 15). For C^Si > $C_0^{Si}$1, ΔG_β" is lower, whereas for C^Si < $C_0^{Si}$1, ΔG_β' is lower. Nevertheless, when Mg-concentration is low, Mg becomes insufficient for the formation of β', while is still sufficient for that of β". Such that ΔG_β' (Line ①) drops down rather slowly with increasing CSi, as compared with all other lines, due to a larger value of αβ' (> 1). As a result, the minimum Si-concentration value for the nucleation of β"-phase ($C_0^{Si}$2) is lowered as compared with $C_0^{Si}$1 for the case when Mg-concentration is high (Fig. 15).

(2) Keeping above points in mind, we can now understand the observed precipitation reversal from the β'-formation to the β"-formation in the matrix, which occurs in Alloy B when small-deformation is applied before aging (Fig. 4). Firstly, Fig. 12 has clearly shown that the rapid formation of the dislocation-induced precipitates in the PDA-treated Alloy B consumes both Mg and Si prior to the matrix precipitation and reduces the solute concentration in the matrix. However, by mostly forming the β'-phase (Mg₉Si₅) precipitates they consume much more Mg than Si, leading to the insufficient Mg-concentration case for the nucleation of β'-phase (the red line ① in Fig. 15) in the matrix. As such, in the region of $C_0^{Si}$2 < C^Si < $C_0^{Si}$1 in Fig. 15, a precipitation reversal from the β'-nucleation to the β"-nucleation can occur for the latter matrix precipitates, as what has been observed in the “normally-β'-hardened” Alloy B (Fig. 4c and d).

Hence, having changed the matrix precipitates from the β'-phase to the β"-phase, the PDA-treated Alloy B has an improved age-hardening response than that of the AA-treated Alloy B upon peak-aging at 250 °C.

4. Conclusions

We have systematically studied the microstructures of the dislocation-induced precipitates in two representative Al-Mg-Si alloys and their impacts on the matrix precipitation in relation to the age-hardening responses of the alloys. From the results obtained, the following can be concluded.

(1) Except for a small portion of β' (/B'/B'-2)-precipitates, most of the dislocation-induced precipitates in slightly pre-deformed Al-Mg-Si alloys are short-range ordered while long-range disordered polycrystalline precipitates and multiphase composite precipitates, including polycrystalline U2 precipitates, B'/U2, B'-2/U2, B'/B'-2/U2 and β'/U2 composite precipitates. The polycrystalline U2 precipitates and the B'(/B'-2) related composite precipitates account for a higher proportion in the alloy A with a lower Mg/Si ratio, while the β'/U2 composite precipitates account for a higher proportion in the alloy B with a high Mg/Si ratio.

(2) The formation of dislocation-induced complex precipitates is due to a high nucleation rate and the rapid growth of different precipitate phases. Furthermore, it is revealed that their parallel growth is owing to that all the nuclei orient their easy-growing directions parallel to the associated dislocation lines, though they may rotate around the growth directions.

(3) The formation of dislocation-induced precipitates results in a coarser precipitate microstructure in the matrix of the PDA-treated alloy A, while results in a reversal of the precipitation pathway in the matrix of the PDA-treated alloy B, leading to a reduced age-hardening potential of the former alloy and an improved age-hardening potential of the latter alloy. This can well be explained by the structures and the formation mechanisms of the dislocation-induced precipitate complexes revealed in the present study.

Acknowledgements

This work is supported by the National Key Research and Development Program of China (No. 2016YFB0300801) and the National Natural Science Foundation of China (Nos. 51831004, 11427806, 51671082, 51471067).

---