Localized dissolution of grain boundary T₁ precipitates in Al-3Cu-2Li

3.1 Nucleation and growth of T₁ plates within grains and at grain boundaries

We briefly describe the nucleation and growth of T₁ within a grain as the nature of T₁ plates and their interfacial characteristics within grains are related to the grain boundary T₁ precipitates. It has been suggested that the nucleation of T₁ occurs by the dissociation of matrix dislocation into partials (Cassada, Shiflet, & Starke, 1991; Howe, Lee, & Vasudevan, 1988), as the c-axis spacing of T₁ is 4 times the (111) spacing of Al. It was argued that the nucleation is related to the movements of 1/6 <112> partial dislocations, which transform the fcc stacking to hexagonal stacking. Thus, two partial dislocations are needed to form a T₁ unit cell. However, we have observed that the T₁ precipitates within a grain nucleate via the Guinier-Preston (GP) zone formation. Figure 1 is a HRTEM image close to the [011] zone showing the early stage of T₁ plates, indicated by arrows, formed along the {111} planes. The fast Fourier transform (FFT), obtained from the matrix containing a number of such platelets, shows diffuse streaks developed along the 111 directions (see inset of Figure 1). These platelets do not have the lattice spacing of one unit cell of T₁ phase in the 111 direction of the matrix, suggesting that the nucleation of T₁ plate takes place via GP zones. The T₁ phase has a hexagonal lattice with a=0.496 nm and c=0.935 nm. Smaalen, Meetsma, DeBoer, and Bronsveld (1990) proposed using the X-ray structure refinement technique that the structure contains several corrugated layers and does not have stacking of four hexagonal layers with a spacing of 0.233 nm. Recently, a TEM-based HAADF imaging (Donnadieu et al., 2011; Dwyer, Weyland, Chang, & Muddle, 2011) showed that the T₁ structure is more complex, and along the c-axis, nine layers were observed. As the structure contains several corrugated layers and does not have the stacking of four hexagonal layers with a spacing of 0.233 nm, the nucleation of the T₁ phase cannot be explained by the dissociation of matrix dislocations. Based on our TEM observations, we conclude that T₁ forms by thermal fluctuation in the solid state via GP zone formation.

Figure 1:

HRTEM image close to the [011] zone showing the early stage of T₁ plates formed along the {111} planes. Inset shows FFT of the outlined box in the main image.

With continued aging, T₁ plates lengthen on the {111} habit planes to a significant extent, from 100 nm to approximately 1000 nm. As both the broad faces of the T₁ plate are coherent with the matrix, it thickens with ledge growth mechanism at a lower rate as compared to the rate of lengthening, which is consistent with previously reported observations (Dorin, Deschamps, Geuser, & Sigli, 2014; Howe et al., 1988). Figure 2A is a bright field TEM image for sample aged for 40 h showing the considerably long T₁ plates close to the [011] zone. In addition to T₁ plates, we observe θ′ and δ′ plates. The corresponding diffraction pattern obtained from a number of T₁, θ′, and δ′ precipitates in the α-Al matrix is presented in Figure 2B. The δ′ is not readily visible in the bright field because of the high contrast. The diffuse streaks in the diffraction pattern along the [111] direction are due to T₁ plates, and diffraction spots can be seen from δ′. In addition to T₁, θ′, and δ′ precipitates, a lath-like phase could be observed, and the diffraction spots of this phase are indicated by white circles.

Figure 2:

(A) Bright-field TEM image showing the T₁ plates, θ′ plates, and δ′ dots close to the [011] zone. Two variants of T₁ are indicated by arrows. (B) The corresponding diffraction pattern close to the [011] matrix zone obtained from a number of T₁, θ′, and δ′ precipitates in the matrix.

While the T₁ plates grow along the {111} habit planes within a grain with both broad faces parallel to the {111} planes of the matrix, such conditions are not easily satisfied for the growth of T₁ at grain boundaries. Most grain boundaries were observed to be faceted or serrated upon aging even for 2 h as a result of the formation of plate-like T₁ precipitate at the boundary. TEM observations, presented later, showed that such serrated portion consists of {111} segments at one side of the boundary, suggesting that the habit plane criterion is met mostly at one side of the broad face of the T₁ plate. Such grain boundary configuration would decrease over all free energy as the serrated portion consists of {111} segments. As the grain boundary serration is associated with the formation of T₁ plates at grain boundaries, it was observed in aged sample only. Figure 3A shows the faceted appearance of the grain boundary after 4 h of aging. We also confirmed the grain boundary faceting with electron backscatter diffraction (EBSD) imaging as shown in Figure 3B. The area fraction or coverage was estimated by measuring the length of T₁ plates at the boundary over the total length of grain boundaries. The area fraction was obtained from SEM images of approximately 20 grain boundaries containing T₁ plate. The area fraction or coverage of grain boundary T₁ plates as a function of aging was shown in Figure 3C. It was observed that the area fraction saturates after 40 h of aging. At underaged conditions, it is considerably high and increases with subsequent aging. The area fraction, for example, after 4 h of aging is 65%, and at 150 h, the area fraction is more than 80%. Recently, in situ small-angle X-ray scattering was used by Dorin, Deschamps, Geuser, Lefebvre, and Sigli (2014) to measure the evolution of mean thickness of T₁ within grains. They also employed differential scanning calorimetric measurements to study the evolution of volume fraction of T₁ plates within grains. However, this type of study is not suitable to measure the volume or area fraction of grain boundary T₁.

Figure 3:

(A) SEM image showing faceted grain boundary containing T₁ plates. (B) An EBSD image showing the faceted grain boundaries. The orientation triangle is shown as an inset. (C) Grain boundary coverage of T₁ as a function of aging.

TEM analysis of precipitates at grain boundaries showed that the GBPs at underaged, peak-aged, and overaged conditions are mostly hexagonal T₁ plates. The grain boundary T₁ plates are considerably thicker than the T₁ plates within the grain (Figure 4A). As these grain boundary T₁ plates grow along {111} habit planes at one side of the grain boundary, it shows orientation relation with one side of the matrix grain. Figure 4B is the HRTEM image showing the orientation relation that the (1-11)α-Al is parallel to (0001)T1 and [022]α-Al is parallel to [10-10]T1, which is similar to that of the T₁ plates formed within the grain. A probable formation sequence of T₁ plates at grain boundaries with aging, as shown schematically in Figure 4C, involves nucleation and growth of T₁ precipitates along the grain boundary with the {111} habit planes by continuously readjusting the grain boundary location. The continuous adjustment of grain boundary during the growth of T₁ requires grain boundary migration. Similar grain boundary precipitation behavior of γ′ precipitate has been reported in Al-Ag system (Clark, 1967).

Figure 4:

(A) A bright field TEM image showing the grain boundary T₁. (B) HRTEM of grain boundary T₁ plate showing the orientation relation with the matrix. (C) A schematic diagram showing the mechanism of faceting at grain boundaries as a result of grain boundary precipitation.

3.2 Localized dissolution of grain boundary T₁ in moist air

As these Al₂CuLi precipitates are anodic relative to the surrounding Al solid solution matrix, they get preferentially dissolved during localized corrosion. Here, we are interested to know whether the localized reaction occurs at the T₁/matrix interface or within the T₁ phase in the presence of moisture. The localized dissolution behavior of micron- and nano-sized S phase within grains in Al-Cu-Mg alloys has been studied in water+NaCl environment (Birbilis et al., 2008; Buchheit et al., 1997, 1999). It was observed that the dissolution of S phase is associated with dealloying of Mg and Al, and such a dissolution process leaves behind Cu as remnant (Birbilis et al., 2008). In order to investigate the dissolution or reaction behavior at the nanoscale level, TEM samples were exposed to air for a couple of hours. The grain boundary T₁ phase of the TEM foil will have greater chance to come in contact with moisture, and a number of such grain boundary T₁ were observed to study the behavior. Figure 5 is the HRTEM image of grain boundary T₁ showing the localized reacted regions, indicated by arrows. As a result of this localized reaction, the (0001) lattice planes of T₁ have been destroyed in these regions.

Figure 5:

HRTEM image showing the localized reaction at different regions of grain boundary T₁.

To further investigate the localized dissolution process, we used HAADF imaging. In this imaging mode, the scattering cross-section is approximately proportional to the square of atomic number (Z). The Cu-rich regions will appear bright as compared to the Al-matrix. Figure 6 is the HAADF image of a grain boundary T₁ phase, which appears bright due to its higher Cu content as compared to the matrix. The localized dissolution of the T₁ at grain boundary can be observed mostly in the middle of the T₁. An elongated hole in the middle of grain boundary T₁ could be observed, suggesting that materials from that region have been completely dissolved. As a result of the dissolution, the Cu-rich clusters in form of platelets, which appear brighter as compared to T₁, form onto T₁. This is consistent with the localized dissolution behavior of nanocrystalline S phase in Al-Cu-Mg (Birbilis, 2008). We also observed considerable reaction of T₁ with moisture (Figure 7) after 40 h of aging. Note the area fraction of grain boundary T₁; at 40 h, the area fraction is approximately 80% (see Figure 3C).

Figure 6:

HAADF image showing the localized dissolution of grain boundary T₁. It leaves behind Cu/Cu-rich nanoparticles.

Figure 7:

An optical image showing the extent of reaction at grain boundaries for the sample aged for 40 h. The sample was polished and kept in the laboratory air for more than 12 h.