Al-Li-Cu alloys, as age-hardenable alloys with the advantages of low density and high stiffness, have been widely applied in the automobile and aviation industries [1]. As well known, the Al-Li-Cu ternary system exhibits a complicated precipitation sequence, showing aspects of both Al-Li and Al-Cu binary systems. The binary sequence of Al-Li gives the δ' (Al₃Li) phase [2,3], the formation of which is known to depend strongly on the Li content, whereas the Al-Cu sequence leads to θ′ (Al₂Cu) precipitates [4]. Hono et al. [5] found that the early stage precipitation behavior is dominated by the Li content in matrix, and much fine δ′ particles always can be clearly seen in an as-quenched matrix containing more than 2 % Li. During heating, the ternary Al-Li-Cu system can generate a number of phases, the most important being the T₁ phase (nominally Al₂CuLi) [6]. Over the past decades, great efforts have been made to further improve the mechanical properties of Al-Li-Cu alloys [6,7]. However, for some casting components with complex shapes and sizes, it is hard to improve their mechanical properties by even a proper heat treatment, let alone by a “solution treatment + pre-strain” procedure. This is attributed to the nucleation mechanism of T₁ phases: most of them heterogeneously nucleate on dislocations or on grain boundaries [8], which results in a low density and heterogeneous distribution of T₁ phases in casting Al-Li-Cu alloys. It has been widely accepted that microalloying with appropriate elements can modify the precipitation behavior of nano-sized strengthening phases and hence to improve the mechanical properties of Al-Li-Cu alloys [7,9,10].

Additions of microalloying elements Sn [11], In [12] and Cd [12,13] have been demonstrated to significantly enhance the precipitation hardening effect of Al-Cu alloys. These elements exhibit much higher vacancy binding energies than Cu [14], leading to the formation of X-Cu-vacancies clusters and longer time vacancies available for diffusion. Moreover, the segregation of these microalloying elements at the interface of θ′/matrix reduces the interfacial energy and thereby increasing the nucleation rate of θ′ [15]. However, Bourgeois et al. [16] argued that the major influence of Sn addition in Al-Cu alloys is to promote the nucleation of θ′ with “magic” thicknesses, corresponding to minima in volumetric and shear misfit strain. Hence, although the mechanism of microalloying addition, such as Sn, In in Al-Cu-based alloys, has been reported in numerous publications, some debates still exist and there is no report about the microalloying effect of Cd in Al-Li-Cu alloys.

Based on the inspiring microalloying effects in precipitation-hardenable Al-Cu alloys, it is of great interest to find a similar microalloying method to enhance the precipitation of strengthening phases in casting Al-Li-Cu alloys. In the present article, the age-hardening kinetics of an Al-Li-Cu alloy was substantially enhanced by only a minor addition of Cd (0.06 at.%), which may propose a new mechanism to effectively improve the precipitation hardening effect of this alloy.

Two experimental alloys were prepared by casting in the lab [17]. Table 1 shows the actual compositions of the alloys measured by inductively couple plasma-atomic emission spectroscopy (ICP-AES). Specimens were heat-treated by a solution (500 °C × 8 h +560 °C × 16 h, cold water quenching) + artificial aging (175 °C with various time) process. The hardness was determined on mechanically polished samples by CARAT 930 Automatic Vickers hardness testing machine with a load of 10 kg and a dwell time of 15 s. Transmission electron microscopy (TEM) observations were performed using JEOL 2100 transmission electron microscope and JEM ARM300 F Cs-corrected transmission electron microscope equipped with HAADF detector. Atom probe tomography (APT) analysis was performed in a LEAP-5000XR microscope, which was operated in voltage pulsed mode with a specimen temperature of 30 K, a pulse repetition rate of 200 kHz and a pulse fraction of 20 %. Data reconstruction and visualization were performed with commercial software (Cameca IVAS).

Fig. 1 displays the precipitation behavior of the two alloys during isothermal aging at 175 °C as determined by hardness. Prior to aging, the hardnesses of the two alloys are very similar and both lie around 83 HV. Upon further artificial aging at 175 °C, the hardness of Cd-containing alloy increases much faster than that of the Cd-free alloy. Furthermore, the absolute hardness values of the Cd-containing alloy corresponding to various aging times are all about 10 HV higher than those of the Cd-free alloy. Remarkable improvements in both the precipitation kinetics and the precipitation hardening effects of the studied alloy with only 0.06 at.% Cd were clearly observed, which indicates the significant role of Cd element in the artificial aging process.

In order to further understand the precipitation behavior, the corresponding microstructures of the two studied alloys were characterized using TEM. Before aging, both alloys are characterized by a predominantly homogeneous distribution of fine δ′ precipitates within the matrix accompanied by heterogeneous dispersion of Al₃Zr dispersoids, as evidenced by Fig. 2(a) and (b). Precipitation of Al₃Zr dispersoids occurred during the solution treatment. The L1₂ superlattice spots mainly arise from δ′ (Al₃Li) precipitates and Al₃Zr dispersoids. Fig. 2(c) and (d) shows comparative bright-field (BF) and dark-field (DF) images of the Cd-free and the Cd-containing alloys aged at 175 °C for 1 h. In the conventional TEM images, except for the precipitation and growth of δ′ precipitates, no T₁ precipitates have been observed in the two alloys. However, in the HAADF-STEM images, more ultrafine nanoparticles with an average size of ~2.5 nm (marked by white arrows) have formed in the Cd-containing alloy, which is consistent with the pieces of literature [18,19].

With the artificial aging progress, the hardness of Cd-containing alloys is significantly improved compared with that of the Cd-free alloy. Fig. 3(a) and (b) shows comparative BF and DF images of the Cd-free and the Cd-containing alloys aged at 175 °C for 8 h. The sizes of spherical δ' precipitates in both the two aged alloys are obviously larger. However, the numbers of δ' precipitates are less than those of the non-aging alloys. Moreover, the pre-existing Al₃Zr dispersoids, which could offer additional nucleation sites for δ′ precipitates, resulting in the formation of core/shell complex dispersoids (Al₃(Li, Zr)). In fact, due to the large size (∼ 40 nm) and very low number density of Al₃Zr and Al₃(Li, Zr) dispersoids, the contribution of these two dispersoids on strengthening effect is subordinate. According to former literature [17], the precipitation of T₁ should be sluggish in cast Al-Li-Cu alloy due to insufficient dislocation density, which is confirmed by the Cd-free alloy. However, a higher number density of fine T₁ precipitates which are newly formed can be observed in the Cd-containing alloy, as shown in Fig. 3(b). In the Cd-containing alloys aged for 32 h, the former observed metastable phases have further evolved. As shown in the inspection of Fig. 3(d), it is evident that the number density and the size of T₁ precipitates increased significantly. However, there are hardly any T₁ precipitates in the grain interior of the Cd-free alloy, as shown in Fig. 3(c). It is worth mentioning that a high density of fine T₁ platelets inhabiting at the grain boundary regions is observed in Fig. 3(c2) (marked by red arrows).

As revealed by conventional TEM (Fig. 2, Fig. 3), the enhanced precipitation of T₁ phases should be attributed to the minor Cd addition. To the author's knowledge, this phenomenon has never been reported before.

The microalloying effect of Cd in Al-Cu alloys has been reported in kinds of literature which proposed that the Cd-rich nanoparticles reduce the interfacial energy and the lattice mismatch along the c direction for the θ′/α-Al interface and hence stimulate θ′ phase precipitation by providing good matching [20,21]. Therefore, we can speculate that the microalloying effects of Cd in Al-Li-Cu alloys may act in a similar manner. On one hand, the rapid aggregation of the Cd-Cu-vacancy clusters enhanced the nucleation of T₁. Wolverton [14] obtained the binding energies by first-principle calculation and proposed that the binding energy between Cd atoms and vacancies was 0.14 eV, which was much higher than 0.02 eV for Cu atoms and vacancies, leading to the formation of clusters of Cd-Cu-vacancies and longer time vacancies available for diffusion. Hu et al. [19] reported that the Cd-Cu-vacancy clusters formed in the incubation stage had a higher diffusion coefficient than Cu atoms, leading to a high growth rate of strengthening phases. Previous studies [22,23] also indicated that when the Cu/Cd atomic ratio was lower than 65, Cd atoms are sufficient for the formation of Cd-Cu-vacancy clusters and the Cu/Cd atomic ratio in this paper is 12.5. Therefore, Cd-Cu-vacancy clusters are formed during the early stage of aging in Cd-containing alloy, and the nucleation and growth of the T₁ precipitate may enhance by agglomeration of these mobile clusters.

On the other hand, the segregation of Cd atoms at the T₁/matrix interface reduces the interfacial energy of T₁ nuclei. According to the EDS elemental maps, no Cd segregation can be experimentally found at the T₁/matrix interface. However, in Fig. 4(a), the bright Z contrast, which is similar to the adjacent Cd-rich region, is also evident near the T₁/matrix interface with one or two atom layers. Hence, the segregation of Cd atoms at the T₁/matrix interface is possible. This segregation was not observed at the interface in energy dispersive spectroscopy (EDS) mapping, probably because the content of the element is lower than the detection limit of EDS.

Furthermore, we can infer that the Cd-rich precipitates formed in the very early stage of artificial aging act as heterogeneous nucleation sites for T₁ precipitates. The mechanism of such kind of nucleation behavior is similar to the way that Cd nucleates α-Al(Mn, Fe)Si dispersoids in Al-Mn-Fe-Si [24].

To elucidate the origin of enhanced precipitation, the atomic-resolution HAADF-STEM images of the typical microstructure of the Cd-containing alloy viewed along the 〈110〉_Al direction are shown in Fig. 4.

Fig. 4(a) and (c) shows the typical HAADF-STEM images of T₁ precipitate which is characterized by a dark line squeezed in between two bright lines in the Cd-containing alloy aged at 175 °C for 8 h. Small bright nanoparticles (at the arrowhead) can also be observed at the broad face and/or terminal of many platelets. The EDS analysis in Fig. 4(b) and (d) show that the nanoparticles are enriched with Cd, which is consistent with its relatively high Z contrast (Z_Al = 13, Z_Cu = 29, Z_Cd = 48). Besides, individual Cd-rich nanoparticles, which were not attached to the precipitates, were also observed in the matrix as shown in Fig. S2 in supporting information.

According to the phase diagram calculation [24], Cd has a maximum solubility of 0.52 wt% at 635 °C and decreases sharply with the decrease in temperature. At temperatures below 300 °C, the solubility of Cd is very limited (≤ 0.01 wt%). Due to rapid cooling during the quenching process, the matrix is supersaturated with Cd atoms. Therefore, during the following artificial aging, decomposition of the supersaturated solid solution has a strong driving force, leading to the precipitation of Cd-rich nanoparticles. It is worth pointing out that although Cu is also supersaturated in the matrix during the aging process, the diffusivity rate of Cu in Al is 1.1 × 10^-16 cm²/s at 175 °C, which is much lower than that of Cd (3.6 × 10^-15 cm²/s) [24,25]. Thus, the precipitation of Cd-rich nanoparticles is prior to T₁ precipitates. Therefore, it is highly likely that the Cd-rich nanoparticles formed in the initial stage of artificial aging act as heterogeneous nucleation sites for T₁ precipitates, which consequently leads to a higher number density of T₁ precipitates in the Cd-containing alloy. Most importantly, the minor addition of Cd not only plays an important role in the precipitation of T₁ precipitates but also has an obvious interaction with δ′ phase.

Due to the deficiency of HAADF and EDS in detecting the Li element, we used atom probe tomography (APT) to characterize atoms partitioning and interaction of elements. The distribution of Cu, Li and Cd atoms solutes and NND analysis of Cd-containing alloy aged at 175 °C for 8 h are shown in Fig. 5(a). Most Cu atoms were randomly distributed, while the segregation of Li and Cd atoms occurred obviously. Moreover, the Li-rich precipitates, namely δ′, were apparently enriched with Cd atoms. The δ′ precipitates displayed by 16 at.% Li iso-concentration surface were observed in Fig. 5(c). The composition profile of δ′ precipitates is also shown in Fig. 5(c). The core of δ′ precipitates mainly consisted of Li (23.68 ± 1.16 at.%) and Cd (0.31 ± 0.11 at.%), suggesting that Cu and Al atoms were transferred from the core to the matrix where their concentrations significantly decreased to 0.35 ± 0.15 at.% and 74.25 ± 1.13 at.%, respectively. However, we are not able to infer which sub-lattice sites were occupied by Cd atoms in the Al₃M structure from the composition profile. Furthermore, the radial distribution function (RDF) is a useful method to describe the interaction between elements. This pair-correlation function is equal to unity if there is no correlation between X and Y atoms and is superior to unity in the case of a positive correlation [26]. For the Cd-containing alloy, the one autocorrelation Li-Li and the three cross-correlations Li-Cd, Li-Cu and Li-Al are shown in Fig. 5b. The Li-Li autocorrelation and Li-Cd cross-correlations are both positive, with values estimated to be above unity, which indicates that Li and Cd atoms have a strong trend for agglomeration, while Cu seems to be excluded from these precipitates. To the best of the authors’ knowledge, this is the incorporation of Cd in δ′ precipitates being reported for the first time. However, first-principles calculations are necessary to identify the site occupation of Cd atoms and judge its effects on the lattice parameter of the δ′ precipitates.

It is noted that it is not likely to use Cd in commercial alloys due to its poisonousness. However, the highlight of this work is to propose an effective method to promote the heterogeneous nucleation of strengthening precipitates in Al-Li-Cu alloys. Thus, other non-toxic elements with similar effects on precipitation hardening aluminum alloys (e.g. Sn or In) are worth examining in terms of enhancing the precipitation.

To summarize, a trace (0.06 at.%) addition of Cd significantly increased both aging kinetics and precipitation hardening of an Al-Li-Cu alloy. A detailed HAADF-STEM analyses revealed that Cd-rich nanoparticles are associated with the rim-facets and/or broad-facets of the T₁ platelets. It is highly likely that these nanoparticles formed at the initial stage of artificial aging act as heterogeneous nucleation sites for T₁ precipitates, which consequently leads to a higher number density of T₁ precipitates in the Cd-containing alloy. APT confirmed that Cd atoms are preferentially segregated within δ′ precipitates and this microstructural change is closely correlated to the strong interaction between Li atoms and Cd atoms. However, the sub-lattice sites occupied by Cd atoms in the Al₃M structure has yet to be further identified. ‘

HAADF-STEM and 3DAP works were carried out at State Key Laboratory of Functional Materials for Informatics of Shanghai Institute of Microsystem and Information Technology. The authors would like to thank Mr. Yangming Cheng for providing technical support and fruitful discussions. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jmst.2019.11.032.