Oxidation induced void formation in TEM specimens of Al–Cu–Li alloy

Identification of voids using defocus contrast technique

It is interesting to consider why the existence of such small voids has not been observed until now in this particular alloy system. One possible answer is that these voids were too small to be seen by conventional in focus imaging, and thus, only the defocus contrast technique10^–12 can reveal their presence. Small (0·5–5 nm) strain free objects, such as voids, exhibit very little amplitude contrast and thus are not easily recognisable under the in focus condition. Upon defocusing, however, the appearance of Fresnel fringes around the edge of these voids will make the voids visible. Fresnel fringes originate from interference between electron waves going inside and outside a void. Since these small voids can become visible only by a change in phase, for example, by defocusing, they are called phase objects. 11 11,12 The voids seen in Fig. 1 are thus phase contrast images decorated with Fresnel fringes. The use of the defocus contrast technique also confirmed that the as quenched (and naturally aged) alloy did not contain any voids. Thus, the voids observed in this study did not originate from a quenching process, whereby voids could nucleate by agglomerations of quenched in excess vacancies.

It should be remembered that voids are not the only phase object in this material. Small particulate matter, such as precipitates, however, can also act as a phase object. To establish the identity of phase objects unambiguously, a through focusing exercise is required. As seen in Fig. 2, the image of a phase object undergoes contrast reversal upon through focusing. In the case of voids, a spherical or a penny shaped void appears roughly as a white circle surrounded by a dark ring in underfocus conditions, whereas it shows up as a black circle surrounded by a white ring in overfocus conditions (see voids marked with symbols X and Y in Fig. 2). Thus, it is concluded that white circular features appearing in the underfocus condition in Fig. 1 are voids. If these objects show up as a dark circle with a white ring in underfocus conditions, they are not voids but could be particles with a higher mean inner potential than that of the alloy matrix or could be oxide/foreign particles lying on top of the specimen surface. For detecting these phase objects, therefore, it is essential to image them in the underfocus condition. The use of the defocus contrast technique has thus established that the objects attached to the T₁ precipitates are voids.

OPEN IN VIEWER

Figure 2

Images of voids taken in a overfocus (Δf = +580 nm) and b underfocus (Δf = −580 nm) conditions. Δf denotes amount and sign of defocus. Two corresponding voids are marked with symbols X and Y.

Origin of voids

The next question is how such small voids are formed in this alloy. It was established that the observed voids did not originate from the quenching process, whereby voids could nucleate by agglomerations of quenched in excess vacancies. A simple calculation involving the number of vacancies required to form a 3 nm diameter void (∼900) and the measured void density (∼1·3×10¹⁶ cm⁻³) indicates that even if the equilibrium concentration of vacancies at the solution heat treatment temperature (∼10⁻⁴) was maintained at room temperature, there would be insufficient vacancies to create the voids observed. Another consideration was that the voids could have been introduced during TEM specimen preparation. To test this idea, two other preparation methods (FIB lift-out and conventional mechanical polishing/ion milling) besides the electropolishing technique were used to see how the type of thinning method might affect the void structure seen in Figs. 1 and 2. It was found, however, that the void structure was essentially the same regardless of the method of specimen preparation used. Thus, it was concluded that voids were not introduced by the type of TEM specimen preparation method.

The only other possible source of void formation is due to oxidation. In fact, there is one previous paper that reported8 observations of large (>1 μm) voids in an Al–Li related alloy (Al–2·58Li–1·06Cu–0·56Mg–0·07Fe–0·02Si–0·021Ti–0·08Zr, in wt-%) after high temperature (∼500°C) oxidation in air. Although the alloy used in the present study was not deliberately oxidised at such a high temperature, a lower temperature oxidation process could occur during and after the 150 or 160°C aging of bulk samples. If such an oxidation was possible in this alloy, voids could have originated by the same mechanism as that proposed for the high temperature oxidation.8 However, aging under vacuum did not affect void formation. According to observations by Papazian et al.,8 fast moving Li atoms are preferentially oxidised, creating a Li denuded zone near the alloy surface. The Li denuded zone is not replenished fast enough by slow moving Al atoms, thus leaving excess vacancies inside the zone. These excess vacancies will agglomerate and finally form voids at some distance below the alloy surface. As will be discussed further in the next section, an underlying mechanism for this void formation is due to Kirkendall effect, and it can certainly explain why small voids were formed by oxidation in the alloy. It should be noted that oxidation in the alloy may have occurred either during the 150/160°C aging or during TEM specimen preparation performed at ambient temperatures. Spontaneous oxide formation is also expected to continue upon room temperature storage even after the TEM specimen preparation. It will be shown that oxidation occurring during the post-TEM preparation stage is most crucial in determining the possibility of void formation.

Oxidation induced excess vacancy formation followed by generation of Kirkendall voids

During oxidation, two types of Kirkendall phenomena can exist in an alloy. This is schematically illustrated in Fig. 3. The type I Kirkendall diffusion occurs inside the oxide M–O between Li⁺/Al³⁺ and O²⁻. A condition for the occurrence of Kirkendall diffusion requires the following diffusivity inequality where , and denote the diffusivities of Li⁺, Al³⁺ and O²⁻ respectively. The formation of hollow core oxide particles is claimed1^–6 to be due to the type I Kirkendall effect. Type II occurs inside the alloy M and involves the following inequality: D_Li>D_Al, where D_Li and D_Al are the diffusivities of Li and Al atoms in the Al–Li alloy. Which Kirkendall diffusion dominates depends on the nature of the rate determining step. The reason why the type I Kirkendall diffusion is not significant in the alloy is that the formation kinetics of the surface oxide is not the rate determining step.

OPEN IN VIEWER

Figure 3

Two types of Kirkendall processes, I and II, occurring during oxidation of Al–Li alloy M: type I occurs in oxide M–O between Li⁺/Al³⁺ and O²⁻, whereas type II involves between Li and Al in Al–Li alloy, M (M = Al–Li alloy; M–O = Al–Li oxide)

The fact that the formation kinetics of the surface oxide in the alloy is not the rate determining step can be explained in terms of the nature of the surface oxide. Al generally forms a very coherent oxide with limited thickness.13 In the case of an Al–Li alloy, Li may break up this coherency by forming more open structure, which allows molecular oxygen to migrate without going through an ionisation process and then to react with the surface of the alloy. If this is the case, an escape of Li atoms to the surface is unlimited. Consequently, the diffusion of Al and Li atoms inside the alloy matrix becomes the rate limiting step. It was also found8 that the concentration profile of Li measured by the nuclear microprobe technique can be best fit to theoretical predictions if the surface Li contents are assumed zero or near zero. This result strongly suggests that surface oxides on the Al–Li alloy are more defective than that on pure Al.

Based on the above discussion, it can be concluded that the void formation observed in the alloy can be attributed to the type II Kirkendall effect, in which Li atoms are faster diffusing species, thus causing an uneven flow of Al and Li atoms. The diffusivity difference between Li and Al atoms can be estimated using published diffusivity data. Based on diffusivity data listed in Table 1, the diffusivities of Al, Li and Cu in an Al matrix at 20, 150 and 160°C are computed, as seen in Table 2. It is clear that the diffusivity of Li at 20°C is seven orders of magnitude larger than that of Al. This large difference in diffusivity is indeed consistent with possible Kirkendall effect between Al and Li atoms in the alloy used in the present study.

OPEN IN VIEWER

Table 1

Published diffusivity values for Al, Li and Cu in Al–Li–Cu alloy*

D	D0, cm2 s−1	Q, kJ mol−1	References
DAl	0.137	123.5	14
DLi in Al	1.09×10−4	66.37	15
DCu in Al	0.654	136.0	14

*Using a formula D = D₀exp (−Q/RT), where D₀ is the pre-exponential factor (cm² s⁻¹), Q is the activation energy (kJ mol⁻¹), R is the gas constant (R = 8.3144 J mol⁻¹ K⁻¹) and T is the temperature (K).

OPEN IN VIEWER

Table 2

Calculated diffusivity values for Al, Li and Cu atoms in Al at 20, 150 and 160°C*

T, °C		20	150	160
D, cm2 s−1	D Al	1·32×10−23	7·70×10−17	1·73×10−16
	D Li in Al	1·60×10−16	6·95×10−13	1·07×10−12
	D Cu in Al	3·72×10−25	1·05×10−17	2·57×10−17

*These values were obtained from diffusion data listed in Table 1.

Transmission electron microscopy specimen as vacancy confining geometry

A nanoparticle has a convenient geometry for confining oxidation induced excess vacancies. If such a geometrical configuration is responsible for oxidation induced void formation, a TEM specimen also has geometry suitable for trapping vacancies between two top and bottom surface oxide layers. For this reason, a surface oxide layer for TEM specimens can be considering as a two-dimensional vacancy mask, as opposed to a three-dimensional one for a nanoparticle (see Fig. 4). In the two-dimensional geometry, oxidation induced vacancies can dissipate laterally over the film plane, although they can be confined between two surface oxide films separated by the specimen thickness (100–150 nm). Note that the degree of vacancy confinement decreases with increasing specimen thickness. Thus, vacancies are less confined in TEM specimens than those in nanoparticles, making the formation of voids less plausible. The average diffusion distance of Li atoms can be estimated using the relationship x = (2Dt)^1/2, where x is the average distance that an atom can travel, D is the diffusivity and t is the time. Li atoms can travel the average of ∼10 nm in 1 h at room temperature. This diffusion distance is sufficient to form a thick oxide, as seen in Fig. 5. Figure 5 is a plan view lattice image taken from an electropolished TEM specimen of the aged alloy. There is a thin edge region containing only the amorphous oxide film M–O and a mixed amorphous oxide/crystalline metal layer M+(M+O). This amorphous oxide layer is indeed thick as predicted. In fact, this oxide film grows thicker upon room temperature storage.

OPEN IN VIEWER

Figure 4

Comparison of two oxidised metal geometries (M = Al–Li alloy; M–O = Al–Li oxide)

OPEN IN VIEWER

Figure 5

Lattice image from electropolished TEM specimen of aged alloy, showing thin edge region containing only amorphous oxide film M–O and mixed amorphous oxide/crystalline metal layer M+(M+O) [M = Al–Li alloy; M+(M–O) = (Al–Li alloy)+(Al–Li oxide)]

In support of an idea that voids might have been generated inside TEM specimens, the authors speculate that the observed voids were formed not during bulk aging at 150 or 160°C but were generated during and/or after TEM specimen preparation. Since the present TEM specimens were prepared at room temperature, the authors conclude that room temperature oxidation is responsible for the observed void formation in the alloys. Nevertheless, it should be remembered that the TEM specimen geometry is not the only requirement for observing oxidation induced voids. In addition, a site for void nucleation is needed after vacancy supersaturation is achieved.

T₁ phase as preferential site for void nucleation

Transmission electron microscopy observations of the aged alloys have revealed that many voids are attached to the T₁ phase. This result strongly suggests that the T₁ phase must contain a preferential site for void nucleation. It was previously shown16 that the hexagonal T₁ phase grows by the lateral motion of the ledge on the (0001) plane. Thus, the ledges are an important element for the growth of the T₁ phase. The single ledge height corresponds to the c value of this hexagonal phase,17 which is 0·940 nm. Many of the voids attached to the T₁ phase suggest that the ledges lying on the broad (0001) face might be preferred sites for void nucleation. These ledges represent the coherent/semicoherent interface between the T₁ phase and the Al alloy matrix. From the orientation relationship of ,18 the amount of mismatch at the ledge between the T₁ phase and Al can be estimated. The c value of 0·940 nm for the T₁ phase and a (111) interplanar spacing of 0·2338 nm for the Al alloy can be used to estimate the amount of misfit present at these ledges. Here, the value of pure Al was used as an approximate value for the lattice constant of the alloy. Multiplication of the (111) interplanar spacing by 4 yields 0·9352 nm, which can be compared with c value of T₁ of 0·940 nm, leading to a modest misfit value of 0·5%. As seen in Fig. 6, atomic imaging by HAADF indeed indicates that the four {111} planes of the alloy align with the unit cell height of the T₁ phase. This mismatch of 0·5% causes the ledge region of the precipitate to exert a local tensile stress in the Al alloy matrix. This tensile stress could in turn help attract oxidation induced excess vacancies towards the ledge region. When a critical vacancy concentration for void nucleation is reached, voids can be generated preferentially at these ledges. Figure 7a is a TEM image of voids attached to the T₁ phase of the alloy, and Fig. 7b is a schematic view showing how voids might form from oxidation induced excess vacancies at the ledges of the T₁ phase under the influence of a tensile stress.

OPEN IN VIEWER

Figure 6

T₁ phases taken by high angle annular dark field imaging: structural model of T₁ phase proposed by Huang and Ardell19 is inserted in: a a single layered T₁ phase showing A/B/C/B/A/ sequence and its structural relationship to 111 planes of alloy; b doubled layered T₁ phase

OPEN IN VIEWER

Figure 7

a a number of voids (marked with black triangle) attached to T₁ precipitate of alloy and b schematic illustration showing how voids could nucleate preferentially at ledges of T₁ precipitate from agglomerations of excess vacancies under influence of tensile stress σ_T, which is developed from misfit between alloy matrix and T₁ phase