Acta Materialia 51 (2003) 4251–4266 www.actamat-journals.com
Transmission electron microscopy study of the evolution of precipitates in aged Al–Li–Cu alloys: the q⬘ and T1 phases Ritsuko Yoshimura a,1, Toyohiko J. Konno a,∗, Eiji Abe b, Kenji Hiraga a a
Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan National Institute for Materials Science, Sengen, Tsukuba 305-0047, Japan
b
Received 24 June 2002; received in revised form 13 December 2002; accepted 6 February 2003
Abstract We have investigated the structures of the q⬘ and T1 precipitates in Al–1.6wt%Li–3.2wt%Cu and Al–2.4wt%Li– 3.2wt%Cu alloys aged at 220 °C. The q⬘ precipitates in the 1.6 wt%-Li alloy are those known for the Al–Cu binary system (a = 0.40 and c = 0.58 nm); whereas those in the 2.4 wt%-Li alloy exhibited two atypical structures. One, named a type I TB⬘ plate in this study, is isostructural to the known q phase with a large c value of about 0.64 nm, having a habit plane parallel to the matrix {1 0 0}a; the other, type II TB⬘, is characterized by a = 0.41 and c = 0.61 nm, having a habit plane inclined at about 20° with {1 0 0}a, while maintaining a coherent interface. Also images of {1 1 1} precipitates in the 1.6 wt%-Li alloy revealed a continuous change from the T1 phase (c = 0.935 nm), to a structure with c = 0.87–0.90 nm. The image and small lattice parameter suggest that this {1 1 1} precipitate is likely to be the ⍀ phase. 2003 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved. Keywords: Al–Li–Cu alloys; Aging; Transmission electron microscopy; Structure
1. Introduction In age-hardened ternary Al–Li–Cu alloys, the peak strength is known to arise from a complex microstructure, in which the d⬘, q⬘ and T1 precipi-
∗ Corresponding author. Present address: Osaka Prefecture University, Department of Metallurgy and Materials Science, 1-1 Gakuencho, Sakai, Osaka 599-8531, Japan. Tel.: +81-72254-9318; fax: +81-72-254-9912. E-mail address:
[email protected] (T.J. Konno). 1 The Furukawa Electric Co. Ltd., R&D Laboratories, Ichikawa, Chiba, 290-8555, Japan.
tate phases are finely dispersed [1–3]. The addition of other minor alloying elements, such as Mg, Zr, Ag, etc., further complicates the precipitation behavior, but results in desired mechanical properties [4–7]. For example, Gayle et al. [8] reported that T1 particles are likely to be the dominant strengthening phase among the S⬘, q⬘, and T1 phases in the optimally aged commercial Al–Cu– Li–Ag–Mg alloys; Huang and Ardell [9], on the other hand, established the addition rules and the contribution of d⬘ precipitates to the strength of aged Al–Li–Cu alloys. Recent studies on Al–Cu– Li–Mg-based alloys have endeavored to distinguish the role of the T1 and ⍀ phases, the latter
1359-6454/03/$30.00 2003 Acta Materialia Inc. Published by Elsevier Science Ltd. All rights reserved. doi:10.1016/S1359-6454(03)00253-2
4252
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
being the principal strengthening phase in the Al– Cu–Mg–Ag alloys [5,10–13]. While there is general agreement on the fact that these precipitate phases are crucial to the strength of the family of Al–Cu–Li-X (X = Ag, Mg, Zr) alloys, various questions still remain as to the evolution and structure of the precipitates [6,14]. The evolution of the q⬘ phase, which is essential in giving rise to the peak strength in the Al–Cu binary alloys [15], has not been investigated extensively in the aged Al–Li–Cu ternary alloys so far. While the q⬘ phase is believed not to play a dominant role in the strength of the Al–Li–Cu alloy because of its relatively small amount in the ternary system [8], it is nonetheless interesting from a scientific point of view. The q⬘ phase, Al2Cu, is a semicoherent metastable precipitate with a tetragonal structure of a = 0.404 and c = 0.580 nm [15– 17]. It grows as thin plates on {0 0 1}a planes ({0 0 1} planes of the Al matrix). There were some conflicting early reports as to the presence of strain fields near the precipitates [18–20], but subsequent works clearly established that the stress fields originating from elastic misfit of the coherent broad faces could influence the growing behavior of the q⬘ phase itself [21,22]. The effect of artificial stress during the precipitation of the q⬘ phase was demonstrated by Hosford and Agrawal [23], while Perovic et al. [24] showed that the q⬘ phase autocatalytically grows by successive nucleation of the q⬘ platelets, leading to linear arrays of the plateshaped precipitates. In addition, the ledge mechanism for the growth of the q⬘ phase has been described in detail by Dahmen and Westmacott [25]. In the early stage of aging of Al–Li–Cu alloys, there is ubiquitous precipitation of the metastable d⬘ phase (Al3Li, L12 by Structurbericht) [8,26]. The lattice constant of this phase, 0.4308 nm [1,27], is only slightly smaller than that of the Al matrix, 0.404 nm, so that the d⬘ phase precipitates coherently in a spherical shape [28,29]. When Cu monolayer precipitates, i.e., the GP-I zones are present, however, the d⬘ phase nucleates and grows heterogeneously around the zone [8,26]. In fact, we have shown that the presence of lenticular d⬘ particles greatly stabilizes the GP-I zone: the hybrid structures made up of the GP-I zone sandwiched
by the d⬘ particles are stable even at around 200 °C [30], well above the reported GP-I solvus in the binary Al–Cu system [31]. In view of the strain fields created around the q⬘ phase, and the presence of the d⬘ phase, the interaction between the two phases in the Al–Li–Cu ternary alloys constitutes an interesting area of research. In the present paper, we chose to use the term q⬘ precipitates to describe the planer precipitates exhibiting a similar transmission electron microscopy (TEM) contrast with that of the q⬘ phase in the Al–Cu binary system; however, in the end, it will be shown that this precipitate in the Al–Li–Cu system deviates from that which is known for the Al–Cu system. The T1 phase, which was initially reported by Hardy and Silcock [32], is based on the hexagonal structure. They suggested that it belongs to the space group P6/mmm or one of its maximal nonisomorphic subgroups with lattice constants of a = 0.4964 and c = 0.9345 nm. To our knowledge, the first TEM study on Al–Li–Cu aged alloys was performed by Noble and Thompson [33], who observed that hexagonal T1 platelets appear on {1 1 1}a planes with an orientation relationship of [0 0 0 1]T1 / / [1 1 1]a and [1 0 1¯ 0]T1 / / [1 1 0]a. Subsequent electron diffraction studies by Huang and Ardell [34] and convergent electron diffraction studies by Vecchio and Williams [35] were in agreement with the earlier suggestion of the space group P6/mmm. In contrast, high resolution electron microscopy (HREM) studies by Howe et al. [36] and by Cassada et al. [37] suggested, respectively, a structure based on trigonal P3¯ m1 space group with A1BA2C... layer sequence and hexagonal P6¯ m2 with a similar layer sequence. A slightly modified structure was proposed by Radmilovic and Thomas [38] in order to account for the observed contrast variation in their HREM study of the planar precipitates. Inconsistencies on the proposed structure for T1 led Van Smaalen et al. [39] to examine the structure using single crystal X-ray diffraction technique. Although their results agreed with a structure based on P6/mmm, the rather high consistency index led them to suggest the possibility of other symmetries, including orthorhombic Cmmm. These investigations indicate that there may be some underlying reasons
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
accounting for the difficulty in the unequivocal determination of the structure of the T1 phase. In view of these questions, we carried out an investigation on the precipitation behaviors of the q⬘ phase and T1 phase in the aged Al–Li–Cu ternary alloys. The choice of the ternary system was made to avoid possible complications on the precipitation behavior arising from segregating elements such as Ag or Mg. The study presented here is also a natural continuation of the joint paper [30], and seeks to provide an overall view of the precipitation process.
4253
STEM) imaging, we used a 200 kV field-emission transmission electron microscope (Jeol JEM2010F) equipped with a scanning unit. An objective lens was designed as an ultrahigh resolution type pole-piece (spherical aberration Cs = 0.5 mm), providing a minimum probe of approximately ~0.15 nm with a convergence angle of ~10 mrad [43,44]. The annular detector was set to collect the electrons scattered at angles between 120 and 600 mrad, which are considered high enough to achieve the incoherent imaging condition.
3. Results 2. Experimental A preliminary Al–Cu alloy ingot was prepared by arc-melting, which was then melted together with Li metal by induction melting using a boronnitride crucible under an Ar atmosphere. It was subsequently furnace-cooled under a high vacuum. The composition of the ingots was examined by inductively coupled plasma spectroscopy, which yielded the following values: Al: 95.2, Li: 1.6, Cu: 3.2 wt%, and Al: 94.4, Li: 2.4, Cu: 3.2 wt%. These alloys will hereafter be called 1.6 wt%-Li alloy and 2.4 wt%-Li alloy, respectively. The alloy ingots were then rolled to sheets of 0.15 mm thickness. The sheets were solution-treated at 550 °C for 30 min in an evacuated Pyrex tube, and then quenched in ice water. Immediately after quenching, the specimens were aged at several different conditions. TEM samples were cut from the sheet and electropolished in a 75% methanol and 25% nitric acid solution cooled down to ⫺35 °C using a Struers Tenupol-3 twin jet electropolishing machine with an applied potential of 13 V and current of 200 mA. We used a JEOL-3010 electron microscope operating at 300 kV for conventional TEM observations, and a JEOL 4000EX (resolution: 0.17 nm) operating at 400 kV for high resolution transmission electron microscopy (HREM). Computer simulations of HREM images were performed using MacHREM software, which is based on a multislice algorithm proposed by Ishizuka and Uyeda [40] and Ishizuka [41,42]. For high angle annular detector dark-field–scanning transmission electron microscopy (HAADF–
Fig. 1(a),(b) is that of bright field (BF) and dark field (DF) images, taken along the [1 0 0] zone axis of the 2.4 wt%-Li alloy annealed at 220 °C for 11 h, together with the corresponding selected area diffraction (SAD) pattern (inset). This SAD pattern shows streaks arising from the GP zones, superlattice spots from the d⬘ phase, spots from the q⬘ phase, and spots from the T1 phase. The DF image, obtained using 0 1 1 reflection of the d⬘ phase, reveals a number of lenticular contrasts of approximately 50 nm in length as indicated by black arrows, as well as spherical contrasts of the same size, which can be identified as the edge-on and face-on images of d⬘ particles flanking the GP-I zones, respectively. A detailed description on the composite structure was provided in a separate paper [30]. It can also be noticed that there are two other types of precipitates in this specimen. First, the DF image (Fig. 1(b)) shows contrasts of long precipitates of more than 200 nm along the {1 0 0}a, as indicated by white arrows. These contrasts are images of the d⬘ phase precipitated around the q⬘ platelets, as will be shown in Fig. 3. On the other hand, the BF image shown in Fig. 1(a) reveals layer-like images making approximately 35° with {1 0 0}a. These images represent the platelets precipitated on the {1 1 1}a, and hereafter designated as {1 1 1} precipitates. The absence of the contrasts of the {1 1 1} precipitates in the DF image (Fig. 1(b)) demonstrates that they are not surrounded by the d⬘ phase. Fig. 2(a),(b) shows a portion of SAD patterns obtained from the 1.6 wt%-Li alloy annealed at
4254
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
Fig. 2. SAD patterns of (a) 1.6 wt%-Li alloy annealed at 220 °C for 2 days, and (b) 2.4 wt%-Li alloy annealed at 220 °C for 11 h. The positions of spots arising from the q⬘ phase are indicated by arrows. The 0 0 2q⬘ and 0 1 1q⬘ spots in (a) yield the lattice parameter c = 0.58 nm; whereas those in (b) yield c = 0.61 nm.
Fig. 1. TEM images and SAD pattern (inset) of the 2.4 wt%Li alloy annealed at 220 °C for 11 h, taken along the [1 0 0]a zone axis. (a) BF image. Large {1 1 1} precipitates (arrowed) are present among smaller precipitates. (b) DF image obtained with an aperture at one of the 0 1 1 superlattice spots of the d⬘ phase. Long precipitates indicated by white arrows are the d⬘ phase precipitated around the q⬘ particles, whereas smaller lenticular precipitates, some of which are indicated by black arrows, are the d⬘ phase flanking the GP zones.
220 °C for 2 days and the 2.4 wt%-Li alloy annealed at 220 °C for 11 h, respectively. In Fig. 2(a), spots and streaks due to the d⬘ phase disappeared; instead spots arising from the T1 phase are seen at 1/3 0 2 2Al and its equivalent positions. Also, indicated by arrows are weak spots arising from the q⬘ phase. The lattice parameters obtained from these spots are a = 0.40 and c = 0.58 nm, values known for the Al–Cu binary system [15– 17]. On the other hand, Fig. 2(b) shows that the spots arising from the q⬘ phase, e.g., 0 0 2q⬘ and 0 1 1q⬘, are slightly elongated along the [0 0 1]a
direction, with the center of the spots slightly shifted toward the [0 0 1¯ ]a direction. The lattice parameters obtained from these spots are a = 0.40 and c = 0.61 nm. Fig. 3(a) is an HREM image of one of the q⬘ particles found in the 2.4 wt%-Li alloy annealed at 220 °C for 11 h. The image of the q⬘ phase shown here is similar to the one reported previously [45]. An inset on the left is the simulated image of the q⬘ phase, based on the accepted structure (space group I4/mmm; Cu and Al at 2a and 4d sites, respectively). The contrast of the surrounding thin regions are not those expected from the face centered cubic (fcc), but those of the L12 type, showing that most of the q⬘ platelets are sandwiched by the layered d⬘ phase. This observation is in general agreement with a report by Howe et al. [46]. On the other hand, the lattice parameters of the q⬘ phase deduced from this HREM image are a = 0.40 and c = 0.64 nm, deviating signifi-
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
Fig. 3. HREM images of atypical q⬘ precipitates in the 2.4 wt%-Li alloy annealed at 220 °C for 11 h. (a) q⬘ precipitate surrounded by the thin d⬘ phase. The observed c value is 0.64 nm, considerably larger than 0.58 nm, commonly known for the q⬘ phase in the Al–Cu binary system. Inset on the left is a simulated image based on the q⬘ phase. (b) q⬘ plates precipitated vertically or horizontally are flanked by the d⬘ phase, whereas the inclined q⬘ precipitate seen on the right is free from the d⬘ phase. (c) An enlarged image of the tilted “q⬘ phase” in (b), showing coherency between the precipitate and the matrix.
cantly from a = 0.404 and c = 0.58 nm of the q⬘ phase in the Al–Cu binary system [15–17]. Fig. 3(b) is another HREM image found in the same alloy. On the left, the q⬘ platelets are seen to be flanked by the d⬘ phase. Note that a pair of facing d⬘ phase across the q⬘ platelets are in anti-phase relationship, as reported previously [30,46]. At the center of the figure, it is seen that the q⬘ precipitate makes an angle of about 18° to (0 0 1)a. When this
4255
picture is viewed from a glancing angle, it reveals that the c plane of the q⬘ platelet is gradually making this angle. Fig. 3(c) is a magnified image of this tilted q⬘ particle, showing complete coherency between the tilted q⬘ platelet and the matrix. It can also be pointed out that the contrast due to d⬘ phase is very weak around this tilted q⬘ platelet. Similar disappearance of a segregating element, e.g., Ag, is also known to occur around the ⍀ phase in the Al–Cu–Mg–Ag system, as reported recently by Moore et al. [47]. Fig. 4(a),(b) is that of an SAD pattern and a BF image of the 1.6 wt%-Li alloy annealed at 220 °C for 2 days, viewed along the [1 1¯ 0]a. The SAD pattern shows, in addition to spots arising from the a phase, streaks in the ⬍1 1 1⬎a direction with strong diffuse intensities at about every 1/4 1 1 1a reciprocal distance, and sharp spots at 1/3 2 2 0a and 1/3 1 1 3a (and their integral multiple) positions. The streaks arise from the shape factor of the {1 1 1} precipitates, seen edge-on in Fig. 4(b). The fourfold variation in the streak further suggests that this phase possesses a layered structure with a period roughly equal to four times the interplanar spacing of {1 1 1}a. Assuming that the lattice constant c of the {1 1 1} precipitates corresponds to exactly four times the spacing of {1 1 1}a, one obtains c = 0.935 nm, in agreement with the reported c value for the T1 phase [32]. The spot at 1/3 2 2 0a can be identified as 1 0 1¯ 0T1, as arising from the T1 phase precipitated on (1 1¯ 1)a or (1¯ 1 1)a planes. The position of these spots gives the other lattice parameter, a = 0.496 nm. The spot at 1/3 1 1 3a, which appears elliptic when closely examined, can be identified as a cross-section of the streak from the 0 1 1¯ 0T1 reciprocal lattice point with the Ewald sphere [26,34,37]. On the basis of these observations, the term T1 phase will be used in a limited manner for the {1 1 1} precipitates with c = 0.935 nm. Fig. 5 is an HREM image of the same specimen, showing edge-on images of the {1 1 1} precipitates. Close examination of the micrograph not only reveals that the contrast of the precipitates varies from place to place, but also that the lattice parameter c of this {1 1 1} precipitate abruptly decreases in the area indicated by letters c and e. It should be noted that we found such local
4256
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
Fig. 5. HREM image of the {1 1 1} precipitates found in the 1.6 wt%-Li alloy annealed at 220 °C for 2 days. Note that the contrast of the precipitates varies from place to place. Also, the lattice parameter c of the precipitate shown at the center (arrowed with letter d) decreases in the upper and lower regions, as indicated by letters b and f.
Fig. 4. (a) SAD pattern of the 1.6 wt%-Li alloy annealed at 220 °C for 2 days, viewed along the [1 1¯ 0] axis. Weak spots, such as those at 1/3 2 2 0a and its multiple positions, as well as streaks in the ⬍1 1 1⬎a directions, are due to the {1 1 1} precipitates. The fourfold intensity variation in the streaks suggests that the c value is four times that of {1 1 1}a spacing. (b) BF-TEM image of the same specimen, showing {1 1 1} precipitates viewed either edge-on or inclined with respect to the [1 1¯ 0] axis of the matrix.
changes in the lattice parameter only in the 1.6 wt%-Li alloy. Fig. 6(a)–(f) is that of magnified HREM images of each area labeled in Fig. 5, showing variations in the layered structure of the {1 1 1} precipitate. The image shown in Fig. 6(a) has a lattice parameter c = 0.94 nm, as obtained with the {1 1 1}a spacing as a reference. This image resembles the HREM images of the T1 phase reported in the literature [36–38], which some authors refer to as “classic T1” [10]. On the other hand, the HREM
image shown in Fig. 6(b), which was obtained from the left part of the large precipitate shown in Fig. 4, exhibits an apparent layer sequence described by ABAB... type with lattice parameter c = 0.87 nm, which is significantly smaller than 0.935 nm, the value of the T1 phase [32]. Fig. 6(c) is an image of the transition region from (b) to (d). At the center of the image, a continuous change in the layer structure, accompanied with a change in c value can clearly be seen. Fig. 6(d) was taken approximately from the center of the large {1 1 1} precipitate shown in Fig. 5. This precipitate is characterized by the lattice parameter c = 0.94 nm, approximately the value for the T1 phase, although the image differs from that of the classic T1. Fig. 6(e) shows another transition region from (d) to (f), in which a continuous change in the structure is seen in the area indicated by an arrow. Finally, in Fig. 6(f), we can see that the apparent layer sequence is more adequately described by AABB... type, and the lattice constant obtained from the layered structure is c = 0.90 nm. The insets in Fig. 6(a),(d) are, respectively, a simulated image of the T1 phase based on the model proposed by Huang and Ardell [34] and that obtained with a slightly modified model, which will be discussed
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
4257
Fig. 6. Enlarged HREM images of the ⬍1 1 1⬎ precipitates shown in Fig. 5. ((a)–(f) correspond to the regions arrowed in Fig. 5 with respective letters.) (a) image of the “classic” T1 phase, with c = 0.94 nm; (b) a region with c = 0.87 nm; (c) an abrupt transition in the structure is indicated by the arrow; (d) atypical image of the T1 phase with c = 0.94 nm; (e) another abrupt transition is indicated by the arrow; (f) a region with c = 0.90 nm. Insets in (a) and (d) are simulated images of the T1 phase along the ⬍ 1 0 1¯ 0 ⬎ axis, obtained with the model based on space group P6/mmm [34], and with a slightly modified model with P6mm, respectively; while those in (b) and (f) are simulated images of the ⍀ phase along [0 1 0]⍀ and [3 1 0]⍀ direction, respectively (see text).
later. On the other hand, the insets in Fig. 6(b),(f) are simulated images of the ⍀ phase along the [0 1 0]⍀ and [3 1 0]⍀ axes, respectively. Fig. 7 is a HAADF–STEM image of the {1 1 1} precipitate viewed edge-on in the same specimen as shown in Figs. 5 and 6. The somewhat wavy lattice image is due to the specimen drift during the imaging, which took 30 s. Layers containing Cu appear brighter here because of the large atomic number (Z = 29, vs. Z = 13 for Al). The spacing between strong white contrasts is equal to twice that of the {1 1 1}a spacing, which gives c = 0.935 nm, if one assumes a four layered structure. This value suggests that the image shown here is that of the T1 phase. We cannot, from this image alone, determine the precise amounts of Cu or Li in each layer, but the variation in contrast suggests
Fig. 7. HAADF–STEM image of a {1 1 1} precipitate viewed edge-on, found in the same specimen as that shown in Figs. 3– 5. Note the variation in constrast, which can be characterized by ABCB... layer sequence, indicating the presence of a mirror plane perpendicular to the c axis.
4258
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
that this particular {1 1 1} precipitate has a chemical sequence of ABCB..., with the average mass of each layer being B ⬎ C ⬎ A. This variation can well be described by each of the three models proposed so far [34,36,37]. For example, in the model by Huang and Ardell [34], A is assigned to a layer with Li / Al = 2 / 1, B to a layer with Al / Cu = 1 / 1 and C to a layer with Li / Al = 1 / 2. Note the presence of a mirror symmetry in this chemical sequence. Further annealing resulted in the total depletion of phases other than the {1 1 1} precipitates. For example, a BF image of the 2.4 wt%-Li alloy annealed for 17 h at 350 °C (not shown) was similar to the one shown in Fig. 4. Fig. 8(a),(b) is that of, respectively, BF and DF images of the 2.4 wt%-Li alloy aged at room temperature for 4 months after annealing at 220 °C for 2 days, viewed along [1 0 0]a, together with an inset SAD pattern. Surprisingly, we found that the number of {1 1 1} precipitates was lower than that observed in the as-annealed specimen, whereas GP zones, the d⬘ phase and the q⬘ phase re-emerged. The formation of GP zone and the d⬘ phase is evident from streaks and superlattice spots in the SAD pattern. The DF image, which was taken using 0 1 1 superlattice spot of the d⬘ phase, reveals that lenticular d⬘ particles surround the GP zones as initially observed in the as-annealed specimen, Fig. 1. Moreover, some of the d⬘ platelets are found to make 35–45° to {1 0 0}a planes, as indicated by arrows, suggesting that these d⬘ particles now precipitate in the vicinity of the prior {1 1 1} precipitates. Fig. 9 is an HREM image of one of the q⬘ platelets in the specimen shown in Fig. 8. This image shows that (i) the habit plane is tilted by about 20° from (0 1 0)a, (ii) the a and c axes are both tilted by about 5° from [0 0 1]a and [0 1 0]a, respectively, and (iii) the interfaces nevertheless exhibit well-defined coherency. Also note that there is no d⬘ phase surrounding this q⬘ platelet. The observed lattice parameters are a = 0.42 and c = 0.62 nm, once again, deviating significantly from the values for the q⬘ phase in the Al–Cu binary system [15–17]. We made essentially identical observations for the 1.6 wt%-Li alloy: most of the {1 1 1} precipi-
Fig. 8. TEM images of the 2.4 wt%-Li alloy aged at room temperature for 4 months after annealing at 220 °C for 2 days, viewed along [1 0 0]a, together with an inset SAD pattern. (a) BF image, showing disappearance of the T1 particles and reemergence of the GP zones, the q⬘ plates and the d⬘ particles, (b) DF image obtained using one of the 0 1 1 superlattice spots arising from the d⬘ phase. Pairs of lenticular contrasts indicated by the arrows suggest that these d⬘ phase precipitated in the vicinity of the prior {1 1 1} precipitates.
tates in this sample have also disappeared after the room temperature aging for 6 months. In order to disprove that the apparent reduction in the number of the {1 1 1} precipitates have originated from unlikely compositional changes of the whole sample during the room temperature anneal, we reaged these specimens at 220 °C for 2 days. Fig. 10(a) is a BF image of the 2.4 wt%-Li alloy and an inset SAD pattern, viewed along [1 0 0]a, right after the second aging at 220 °C for 2 days. As seen, the specimen is dominated once again by the {1 1 1} precipitates. The SAD pattern shows spots at 1/3 0 2 2 positions (and their integral multiples), which can be identified as 1 0 1¯ 0T1 and
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
Fig. 9. HREM image of one of the q⬘ platelets found in the specimen shown in Fig. 8. The a and c axes are tilted from ⬍0 0 1⬎a by about 5°, whereas the habit plane makes approximately 20° from (0 1 0)a. The observed lattice parameters of the q⬘ platelets are a = 0.42 and c = 0.62 nm.
0 1 1¯ 2T1 type reflections. Hence, the observed reduction in number density of the {1 1 1} precipitates described in Fig. 8 is not an artifact, but it rather suggests that some of the {1 1 1} precipitates are metastable at room temperature and dissolved during room temperature aging for several months. Additional aging for 6 months at room temperature, however, did not bring about second dissolution of the {1 1 1} precipitates. Fig. 10(b) is a BF image and an inset SAD pattern of the alloy shown in Fig. 10(a) after room temperature
4259
Fig. 10. TEM images of the 2.4 wt%-Li alloy and inset SAD patterns, viewed along [0 0 1]a, after the second aging at 220 °C for 2 days. (a) BF-TEM image obtained immediately after the second aging. The T1 particles precipitated once again (compare this picture with Fig. 8), (b) BF-TEM image obtained 6 months after the second aging. The T1 particles grew slightly, but basically remained unchanged. A few GP zones are indicated by the arrows.
annealing for an additional 6 months. As seen, the specimen is still dominated by {1 1 1} precipitates, which are on average slightly larger than those shown in Fig. 10(a), even though close examination reveals that GP zones are also present in this specimen, as indicated by arrows. Also, the SAD pattern shows weak superlattice spots arising from the d⬘ phase. The emergence of small amounts of GP zones and d⬘ phase, however, might have originated from the difference in solubilities of Cu and Li at 200 °C and at room temperature. It may be concluded from these experiments that at least some of the {1 1 1} precipitates produced right after the solution treatment at 550 °C dissolved after room temperature aging for 4–6
4260
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
months; whereas the {1 1 1} precipitates produced after the second aging at 220 °C (preceded by room temperature annealing for several months) did not exhibit such metastability. The contrasting behaviors described here suggest that the structure and thermal stability of the so-called {1 1 1} precipitates may in fact depend on precipitation kinetics.
4. Discussion The work presented here indicates that the structure of q⬘ platelets in the aged Al–Li–Cu alloy can be different from the well-known tetragonal q⬘ phase in the Al–Cu binary system. It also shows that {1 1 1} precipitates can assume different structures, including those with a smaller c value than that of the T1 phase. We also found that some of these {1 1 1} precipitates dissolved at room temperature. The structure and morphology of the q⬘ precipitate in aged Al–Cu alloys have been well documented [15–17]. For instance, the lattice parameter c has been reported invariant even after re-aging following a mechanical deformation [18]. Also an examination on Moire´ fringes on a 1 µm q⬘ precipitate concluded that the alteration in c, if any, should be less than 0.16%. Thus the broad faces of the q⬘ precipitate are completely coherent, though they may contain widely spaced misfit dislocations at high aging temperatures [48]. Vaughan noted that aging up to 50 days at 240 °C led to the loss of coherency between q⬘ plates and a matrix, and an examination on Moire´ fringes led him to suggest the lattice parameter a = 0.4077 and c = 0.581 nm [17]. He suggested the possible incorporation of vacancies in the q⬘ structure. In the present study on the Al–Li–Cu system, the lattice parameters of the q⬘ plates of 2.4 wt%Li alloy were found to deviate from those known for the Al–Cu binary system; whereas those in the 1.6 wt%-Li alloy did not exhibit such deviation (Fig. 2). This observation indicates that Li atoms play an important role in giving rise to a change in the lattice parameter. Our finding that the tilted platelets did not accompany the flanking d⬘ phase (Fig. 3(b)) also supports this view.
So far, no explicit report on the deviation of the lattice parameters of the q⬘ plates has been given in the literature. However, the DF image of one of the q⬘ precipitates in the Al–2.0wt%Li–3.0wt%Cu alloy aged at 190° for 2.25 h presented by Tosten et al. [49] did show one instance of q⬘ precipitate whose habit plane deviated from {1 0 0}a by about 20° (see Fig. 2(b) of Ref. [49]). A BF image of an Al–1.0wt%Li–4.0wt%Cu alloy annealed at 200 °C for 5 h, in a report by Suzuki et al. [54], also shows tilted q⬘ precipitates. Moreover, a recent HREM observation on the q⬘ plates in an aged Al– 2.5wt%Li–2.0wt%Cu–0.2wt%Sc alloy by Dutkiewicz et al. [50] showed a c value in the range of 0.63–0.64 nm, judging from the micrograph (see Fig. 5 of Ref. [50]), although this rather large value for c was not pointed out by the authors. On the other hand, the HREM image of the q⬘ precipitates surrounded by d⬘ plates presented by Howe et al. [46] did not show any tilting or changes in the lattice parameters. The composition of their alloy was Al–2.1wt%Li–1.3wt%Cu. From these observations and in view of the present results, it may be seen that “atypical q⬘ plates” have been found in alloys that are rich in Cu and/or Li. In the suggested equilibrium Al–Li– Cu phase diagram [32], these compositions fall into regions where, with increasing amounts of Li, the q⬘, TB, T1 and T2 phases are progressively stabilized. However, the T2 phase (Al6CuLi3) is a wellknown quasicrystal [51–53] and kinetically difficult to form during a low temperature annealing. On the other hand, the TB phase (Al7.5LiCu4) has the CaF2 structure, with lattice constants of aTB=0.583 nm, and is closely related to the q⬘ phase [32]. (The q⬘ phase has a tetragonal unit cell with aq⬘ = 0.404 and c = 0.58 nm, and can also be viewed as a distorted CaF2 structure with aCaF2 = 0.57 nm and c = 0.58 nm. These two unit cells are related simply by aq⬘ = aCaF2 / √2 nm.) The structural similarities between the q⬘ and TB phases suggest that the atypical q⬘ plates in Al–Li–Cu alloys may actually be derived from the cubic TB phase. In fact, in the early X-ray study of the precipitates by Silcock, she noted that “It appears that q⬘ can develop continuously into TB” [1]. Under this hypothesis, a tetragonally distorted structure derived from the TB phase is likely to have a lattice
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
parameter a = aTB / √2 = 0.412 nm, which can be compared with 0.41–0.42 nm, a value obtained for tilted q⬘ platelets in this study. These considerations led us to propose that we may term the atypical q⬘ plates in the Al–Li–Cu alloys the TB⬘ plates. In fact, “TB⬘ plates” has been used by several authors as a generic term for {1 0 0}a precipitates in annealed Al–Li–Cu alloys [54,55]. Here, our usage is more specific. The HREM images in Figs. 3 and 9 suggest that TB⬘ plates can be categorized into two groups: type I with a = 0.404 and c⬇0.64 nm; and type II with a⬇0.41–0.42 and c⬇0.61–0.62 nm. (They belong to different crystal systems, as will be shown later.) The type I TB⬘ plate is most probably isostructural to the q⬘ phase, and has a habit plane parallel to {1 0 0}a. On the other hand, the type II TB⬘ plates are those whose habit plane and c plane are inclined from {1 0 0}a by about 20° and 5°, respectively, maintaining complete coherency with the matrix. In Fig. 11, we presented a model based on the HREM image shown in Fig. 9. Here, the c axis of the TB⬘ precipitate is tilted by 5° with [0 0 1]a, while the habit plane makes 20° with (0 0 1)a. In this configuration, four Al unit cells exactly match two TB⬘ unit cells along the c axis of the q⬘ precipitate; 11 Al unit cells match 11 q⬘ unit cells along the a axis. A simple trigonometry allows us to calculate the lattice parameters of the tilted structure in this schematics, yielding a = 0.416 and c = 0.613 nm, which compares favorably with the values obtained experimentally a = 0.42 and c = 0.62 nm (Fig. 9). Note also that the a value is close to 0.412 nm, a value expected from the cubic TB phase. The inclination of the habit plane can also be estimated using the invariant-line theory [56,57]. Namely, when an inclusion undergoing a transformation with principal strains e2 (expansion) and e3 (contraction) along the [0 1 0] and [0 0 1] axes forming a thin plate on the (0 0 1) plane, there exists an unextended line. The angle of this invariant line with the (0 0 1) plane is given by tanq = √⫺e2 / e3. In the present case, the strains are e 2 = 0.0279 and e 3 = ⫺0.2433, yielding q = 18.7°, in fair agreement with the observed value of about 20°. It should be noted, however, that Fig.
4261
Fig. 11. Schematic model of the habit plane between the type II TB⬘ plate and the a matrix shown in Fig. 9. In this model, the c axis of the TB⬘ plate is tilted by 5° from [0 0 1]a, while the habit plane makes 20° with (0 0 1)a. The lattice constants of the TB⬘ plate are set so that four Al unit cells match two TB⬘ unit cells along the c axis, whereas 11 Al unit cells match 11 TB⬘ unit cells along the a axis, resulting in a = 0.416 and c = 0.613 nm. In this model, the lattice constant b of the TB⬘ plate along the projection line (normal to the drawing) is assumed to be that of the a matrix. This makes the TB⬘ plate orthorhombic.
4262
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
11 is a projected image and no change in the lattice parameter was assumed along the line of the projection (normal to the plane of drawing). In other words, the model in Fig. 11 implies that the type II TB⬘ plane shown in Fig. 9 belongs to the orthorhombic crystal system with a = 0.416, b = 0.404, c = 0.613 nm. The volume of the orthorhombic unit cell is 0.103 nm3, which is only slightly larger than 0.099 nm3, the value expected from the cubic TB phase. The possible alteration of the atomic positions in the q⬘ phase in the Al–Cu system was in fact reported earlier from a slightly different viewpoint by Silcock and Heal [58]. They noted that the q⬘ structure consists of alternate Al and Cu layers, but in order to make all the layers close packed, two positions must be added to the Cu layers. The observed integrated intensity in their X-ray diffraction pattern led them to suggest that approximately one in eight Al atoms from an Al site is removed and occupies an unusual position in the Cu layers. In other words, they suggested that, in the q⬘ phase with space group I4/mmm, in which 2a sites are occupied by Cu atoms and 4d sites by Al, some Al atoms should occupy a portion of 2b site. If the occupancy of this 2b site is large, then the whole structure may experience an increase in its volume. Also, taking into account the aforementioned suggestion by Vaughan [17] on the possible incorporation of vacancies in the q⬘ structure in the Al– Cu binary system, and the affinity of Li atoms with Al and Cu atoms, as exemplified by the formation of several ternary phase, e.g., T1 and T2, phases, the presence of the TB⬘ plates may not be too surprising. It is interesting to point out that, in the study of the precipitation behavior of Al–Cu–Li– Mg based alloys, Gayle et al. [5] noted that the S⬘ phase (Al2CuMg) in the Mg-deficient alloy appeared to have nucleated at the edge of the q⬘ phase, while they nucleate and grow independently when the Mg content in the alloy increases. As for the T1 phase, early confusions on the existence of a possible variant phase referred to as T1⬘ by some researchers [55], derived from a misinterpretation of the spots that appear as crosssections of diffuse streak from reciprocal points of the T1 phase with the Ewald sphere, as explained by Huang and Ardell [34]. The structural variation
of the {1 1 1}a precipitates in the present investigation, on the other hand, are based on HREM observations, and thus the presence of the atypical structure with different kinds of layer sequences and lattice parameters is evident. The HREM images shown in Fig. 6(a),(d) are both characterized by the observed lattice parameter of c = 0.94 nm, indicating that these are the images of the T1 phase. However, it can be noticed that, while the image of the “classic T1” (Fig. 6(a)) indicates the presence of a mirror plane normal to the c axis, the image shown in Fig. 6(d) does not. This mirror plane is essential in the space group P6/mmm, on which a model by Huang and Ardell [34] is based. It should be mentioned here that the presence of the mirror plane in the HREM image does not exclude the other well-known models, which are based on P6¯ m2 [37] or P3¯ m1 [36], since the rotation–inversion center in these space groups still brings about a “chemically symmetrical” layer sequence, giving rise to an HREM image with an apparent mirror plane. The inconsistencies between the simulated images and the observed asymmetrical contrast of some of the T1 precipitates led researchers to break the restriction imposed by the space group. For example, Howe et al. [36] obtained a reasonably good matching for some of their HREM images by changing the Cu/Al ratio in their model. Huang also noted this difficulty and tried, without much success, to reproduce diffraction intensities by manipulating the Cu and Al positions [14]. Also in the HREM study on the commercially important Al–Cu–Li–Ag–Mg alloy (Weldalite 049), Herring et al. [10] noted that, in addition to the classic T1 phase, there are {1 1 1} precipitates that are similar to, but not exactly like, the so-called T1 phase. They obtained lattice images of the precipitates which were not reproduced by simulation of the existing model structures for the T1 phase, in spite of the fact that the lattice parameters of the {1 1 1} precipitates agreed with those reported for T1. In view of these known difficulties, we first simulated the HREM image of the T1 phase based on the model proposed by Huang and Ardell [34], which is characterized by a chemical layer sequence of ABCB, where A is a layer with Li / Al = 2 / 1; B, Al / Cu = 1 / 1; C, Li / Al = 1 / 2. Note
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
that this chemical layer sequence is in accord with the HAADF–STEM image shown in Fig. 7. The result obtained for sample thickness t = 36 nm and defocus value of ⌬f = 24 nm (underfocus being positive) is inset in Fig. 6(a). On the other hand, because the image shown in Fig. 6(d) does not exhibit a mirror symmetry, we changed the layer sequence to AB1CB2, where B1 is a layer with Al only, while B2 is a layer with Cu only, making the space group P6mm. The resultant simulated image, obtained with t = 36 and ⌬f = 60 nm is inset in Fig. 6(d). It must be emphasized that, because of the different defocus values employed and uncertainties involved in the structural models, the results of the simulation here should not be considered conclusive; rather they indicate that some of the T1 precipitates differ from the previous models in that they do not possess a mirror plane perpendicular to the c axis nor a rotation–inversion center. Likewise, the presence of a twofold axis perpendicular to the c axis brings about a chemically symmetrical sequence. Hence, out of the original suggestion made by Hardy and Silcock [32], the present consideration leaves P6mm for a possible space group of the atypical T1 structure shown in Fig. 6(d). The most striking finding was the presence of structures having c ⬍ 0.935 nm, i.e. c = 0.87 and 0.90 nm (Fig. 6(b),(f)), and their continual transformation from the T1-like phase (Fig. 6(d)). Such transition has never been reported in the literature. The substantially small lattice parameter suggests that this structure cannot be characterized as T1 phase. In fact, the image of this precipitate resembles more the ⍀ phase, reported for the Al– Cu–Mg–Ag alloys, than the T1 phase. The ⍀ phase, which precipitates on {1 1 1}a as a platelet with composition of Al2Cu, was initially described by Auld [59,60] as having a monoclinic structure, which is best regarded as a distortion of the structure of tetragonal q precipitates (Al2Cu) found in over-aged Al–Cu alloys. In a detailed TEM investigation of the Al–Cu–Mg–Ag alloy on the ⍀ phase, Knowles and Stobbs [61] suggested that this phase is characterized by the orthorhombic unit cell of a = 0.496, b = 0.859, c = 0.848 nm. They also examined the orientation relationship (OR) between the ⍀ precipitates and the a matrix, and
4263
found it to be virtually identical to that of the ORs presented by Vaughan and Silcock [62] for the q phase and the a matrix, which was designated a Vaughan II orientation by Vaughan and Silcock. In addition, the structure and precipitation kinetics of the ⍀ phase have been studied in detail [63– 65]. For example, Garg and Howe [64] have shown that the ⍀ phase actually belongs to the tetragonal system, and designated qM for the precipitate. In what follows, we use ⍀ for the {1 1 1} precipitates, but this choice does not imply that the tetragonal structure should be excluded. We carried out HREM image simulation for the images shown in Fig. 6(b),(f), based on the orthorhombic structure for the ⍀ phase. Because of the OR: (0 0 1)⍀ / / (1 1 1)a; [0 1 0]⍀ / / [1¯ 1 0]a, the ⍀ precipitates have two variants when viewed along ⬍1 1 0⬎a. That is, [0 1 0]⍀ and [3 1 0]⍀ can be parallel to [1 1 0]a. The results of simulation were in a general agreement with those reported previously [61]. We superimposed simulated images obtained with t = 30 and ⌬f = 60 nm for [0 1 0]⍀ or [3 1 0]⍀ in Fig. 6(b),(f), respectively. Note that the employed defocus value here is the same as that used for the inset image of the atypical T1 phase (Fig. 6(d)). The agreement between the experimentally obtained images and simulated ones is reasonably good, supporting that they can indeed be the ⍀ phase, even though the lattice parameter c in these images is slightly larger than 0.848 nm of the ⍀ phase. It is true that no ⍀ phase has in fact ever been reported for Al–Li–Cu alloys, despite the structural similarity between T1 and ⍀ phases: the latter is known to appear only in Al–Cu–Mg based alloys with trace additions of Ag [66–68]. However, studies on Al–Cu based alloys that contain both Li and a combination of Ag and Mg did indicate complexity of {1 1 1} precipitates [10,13]. Gayle et al. have studied mechanical properties and precipitation behavior of Al–Cu–Li–Ag–Mg alloy with a rather small amount of Li content (0–1.31 wt%) and noted that it is difficult to distinguish ⍀ from T1 other than HREM. Since the structure of the ⍀ phase can be considered as a modified q phase (Al2Cu, tetragonal) [60,61], its nucleation on the {1 1 1}a has stimulated a number of studies. An atom probe study on
4264
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
the evolution of the ⍀ phase by Reich et al. [68] showed that the nucleation of ⍀ on the {1 1 1}a planes is initiated by the co-clustering of Ag and Mg. In the later stage of nucleation, these elements are shown to be strongly segregated to the broad face of the ⍀ plates. Recent HAADF–STEM study by Hutchinson et al. [69] and energy-filtered TEM study by Moore et al. [47] also support this view. Thus, the initial segregation of impurity atoms on the {1 1 1}a can be considered to be a key element to the nucleation of the ⍀ plates, as opposed to the q phase. In the present study, the ⍀-like phase was found to continuously evolve from the region with a lattice parameter of the T1 phase. This continuous change from the T1 (or T1-like) phase to the ⍀like phase did not require a specific nucleation process for the {1 1 1} precipitate. These considerations led us to suggest that, when a T1 precipitate whose suggested composition is Al2LiCu, suffers a shortage of Li supply during the growth, it may make a continuous transition to Al2Cu, leading to the occurrence of the ⍀-like structure with c = 0.87–0.90 nm, rather than to the independent nucleation and growth of the tetragonal q structure. Finally, we also showed that at least a portion of these {1 1 1} precipitates dissolved after 4–6 months of room temperature aging, and were replaced with other precipitates including the d⬘, q⬘, and TB⬘. If this is a universal occurrence, the dissolution of the principal strengthening phase has obvious implications. But as described in Figs. 8 and 10, the dissolution mostly is limited to the {1 1 1} precipitates produced by annealing at 220 °C right after solution treatment. We cannot give a plausible explanation for these events, but suggest that the initially observed dissolution of the precipitates is related to the appearance of a number of structures that are different from the classic T1 phase, as seen in Fig. 6. We can also point out that there are abundant vacancies right after the quenching, which accelerates the whole precipitation process; whereas that, in the second annealing after prolonged aging, such vacancies are not present. This difference may result in a different nucleation and growth kinetics for the {1 1 1} precipitates. It is also of interest to note that, in the so called T6 temper, the peak strength is brought about by an artificial aging carried out
after the T4 temper, which is defined as an aging at ambient temperature for more than 1000 h [8].
5. Conclusions We have investigated the structure of the precipitates in Al–1.6wt%Li–3.2wt%Cu and Al– 2.4wt%Li–3.2wt%Cu alloys, aged at 220 °C. The major results are as follows: 1. The lattice parameters of the q⬘ precipitates in the 1.6 wt%-Li alloy are the same as those known for the Al–Cu binary system (a = 0.40 and c = 0.58 nm); whereas those in the 2.4 wt%Li alloy are a = 0.40 and c = 0.61 nm, as obtained from the SAD pattern. HREM observation further revealed the presence of atypical q⬘ plates with a = 0.40 and c = 0.64 nm, having a habit plane parallel to {1 0 0}a. They are likely to be isostructural to the q⬘ phase, and it is proposed to call them type I TB⬘ plates. 2. Some of the precipitates in the 2.4 wt%-Li alloy exhibited a = 0.41 and c = 0.61 nm, and were found to grow away from {1 0 0}a, while maintaining coherency with the a matrix. These atypical precipitates have been interpreted as a distorted TB phase, and termed type II TB⬘ plates in this study. Assuming coherency along the projection direction, it was suggested that they belong to the orthorhombic class. 3. The SAD pattern of the {1 1 1} precipitates showed diffraction spots expected from the hexagonal T1 phase with a = 0.496 and c = 0.935 nm. The HREM images of the precipitates in the 1.6 wt%-Li alloy, however, exhibited several different contrasts. The image with c = 0.94 nm having a mirror plane perpendicular to the c axis can be interpreted by the existing models; whereas that with c = 0.94 nm but without the mirror plane has been interpreted using a model with a space group P6mm. 4. In the 1.6 wt%-Li alloy, the c value of the {1 1 1} precipitate exhibited a continuous change from 0.94 to 0.87–0.90 nm. The HREM images of the latter regions are similar to those expected from the ⍀ phase (Al2Cu). Although the observed c is larger than 0.848 nm, the value
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
for the ⍀ phase, the present results indicate that, when the supply of Li atoms is limited, the T1 phase can make continuous transformation from T1 to ⍀-like structures. 5. At least a portion of the {1 1 1} precipitates produced right after solution treatment was dissolved at room temperature after 4–6 month aging for both the alloys investigated, whereas the {1 1 1} precipitates produced after a prolonged aging did not show such metastability.
[16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26]
Acknowledgements
[27]
This work was supported by the Grants-in-Aid for Scientific Research (C-2: 08405043), provided by the Ministry of Education, Science, Sports and Culture of Japan. The authors thank Prof. G. Itoh for his helpful comment. T.J.K. also thanks the financial support provided by the Inter-university Cooperative Program of IMR, Tohoku University.
[28] [29] [30]
References
[37]
[1] Silcock JM. J. Inst. Met 1959–60;88:357. [2] Jensrud O. In: Baker C, Gregson PJ, Harris SJ, Peel CJ, editors. Aluminium–lithium alloys III. London: Institute of Metals; 1986. p. 411. [3] Sainfort P, Guyot P. In: Baker C, Gregson PJ, Harris SJ, Peel CJ, editors. Aluminium–lithium alloys III. London: Institute of Metals; 1986. p. 420. [4] Kumar KS, Brown SA, Pickens JR. Scripta Metall. 1990;24:1245. [5] Gayle FW, Tack WT, Swanson G, Heubaum FH, Pickens JR. Scripta Metall. Mater. 1994;30:761. [6] Kumar KS, Brown SA, Pickens JR. Acta mater. 1996;44:1899. [7] Lee KH, Lee YJ, Hiraga K. J. Mater. Res. 1999;14:384. [8] Gayle FW, Heubaum FH, Pickens JR. Scripta Metall. 1990;24:79. [9] Huang B-P, Ardell AJ. Acta Metall. 1988;36:2995. [10] Herring RA, Gayle FW, Pickens JR. J. Mat. Sci. 1993;28:69. [11] Boukos N, Flouda E, Papastaikoudis C. J. Mat. Sci. 1998;33:4015. [12] Huang B-P, Zheng Z-Q. Acta Mater. 1998;46:4381. [13] Boukos N, Flouda E, Papastaikoudis C. Phil. Mag. 2000;80:1055. [14] Huang BP, Ardell AJ. Scripta Metall. 1992;27:755. [15] Silcock JM, Heal TJ, Hardy HK. J. Inst. Met. 1953– 54;82:239.
[38]
[31] [32] [33] [34] [35] [36]
[39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49]
[50] [51] [52] [53] [54]
4265
Hornbogen E. Aluminium 1967;43:115. Vaughan D. Phil. Mag. 1968;18:1305. Laird C, Aaronson HI. Trans. TMS-AIME 1968;24:1393. Weatherly GC. Acta Metall. 1971;19:181. Phillips VA. Acta Metall. 1973;21:219. Headley TJ, Hren JJ. Phil. Mag. 1976;34:101. Stobbs WM, Purdy GR. Acta Metall. 1978;26:1069. Hosford WF, Agrawal SP. Met. Trans. A 1975;6A:487. Perovic V, Purdy GR, Brown LM. Acta Metall. 1981;29:889. Dahmen U, Westmacott KH. Phys. Stat. Sol. (a) 1983;80:249. Huang B-P, Ardell AJ. In: Baker C, Gregson PJ, Harris SJ, Peel CJ, editors. London: Institute of Metals; 1986. p. 445. Axon HJ, Hume-Rothery W. Proc. Roy. Soc. A 1948;193:1. Noble B, Thompson GE. Metal Sci. J. 1971;5:114. Williams DB, Edington JW. Metal Sci. 1975;9:529. Yoshimura R, Konno TJ, Abe E, Hiraga K. Acta Mater 2003;51:2891. Beton RH, Rollason EC. J. Inst. Metals 1957–58;86:77. Hardy HK, Silcock JM. J. Inst. Met. 1955-56;84:423. Noble B, Thompson GE. Metal Sci. J. 1972;6:167. Huang B-P, Ardell AJ. Mat. Sci. Tech. 1987;3:176. Vecchio KS, Williams DB. Met. Trans. A 1988;19A:2885. Howe JM, Lee J, Vasude´ van AK. Met. Trans. A 1988;19A:2911. Cassada WA, Shiflet GB, Starke Jr. EA. J. Physique Colloque C3 1987;48:397. Radmilovic V, Thomas G. J. Physique Colloque C3 1987;48:385. Van Smaalen S, Meetsma A, De Boer JL. J. Solid State Chem. 1990;85:293. Ishizuka K, Uyeda N. Acta Cryst. 1977;A33:740. Ishizuka K. Ultramicroscopy 1980;5:55. Ishizuka K. Acta Cryst. 1982;A38:773. James EM, Browning ND, Nicholls AW, Kawasaki M, Xin Y, Stemmer S. J. Electron Microsc. 1998;47:561. James EM, Browning ND. Ultramicroscopy 1999;78:125. Fujita H, Lu C. Mater. Trans. JIM 1992;33:892. Howe JM, Laughlin DE, Vasude´ van AK. Phil. Mag. A 1988;57:955. Moore KT, Howe JM, Veblen DR. Phil. Mag. B 2002;82:13. Weatherly GC, Nicholson RB. Phil. Mag. 1968;17:801. Tosten MH, Vasude´ van AK, Howell PR. In: Baker C, Gregson PJ, Harris SJ, Peel CJ, editors. Aluminium–lithium alloys III. London: Institute of Metals; 1986. p. 483. Dutkiewicz J, Simmich O, Scholz R, Ciach R. Mater. Sci. Eng. A 1997;234–236:253. Ball MD, Lloyd J. Scripta Metall. 1985;19:1065. Marcus MA, Elser V. Phil. Mag. B 1986;54:L101. Konno TJ, Ohsuna T, Hiraga K. J. Alloys. Comp. 2002;342:120. Suzuki H, Kanno M, Hayashi N. Keikinnzoku 1982;32:88 [in Japanese].
4266
R. Yoshimura et al. / Acta Materialia 51 (2003) 4251–4266
[55] Rioja RJ, Ludwiczak EA. In: Baker C, Gregson PJ, Harris SJ, Peel CJ, editors. Aluminium–lithium alloys III. London: Institute of Metals; 1986. p. 471. [56] Dahmen U. Acta Metall. 1982;30:63. [57] Dahmen U, Westmacott KH. Acta Metall. 1986;34:475. [58] Silcock JM, Heal TJ. Acta Cryst. 1956;9:680. [59] Auld JH. Acta Cryst. 1972;A28:S98. [60] Auld JH. Mater. Sci. Tech. 1986;2:784. [61] Knowles KM, Stobbs WM. Acta Cryst 1988;B44:207. [62] Vaughan D, Silcock JM. Phys. Stat. Sol. 1967;20:725. [63] Garg A, Howe JM. Acta Metall. Mater. 1991;39:1925.
[64] Garg A, Howe JM. Acta Metall. Mater. 1991;39:1939. [65] Garg A, Chang YC, Howe JM. Acta Metall. Mater. 1993;41:235. [66] Muddle BC, Polmear IJ. Acta Metall. 1979;37:777. [67] Ringer SP, Hono K, Polmear IJ, Sakurai T. Acta Mater. 1996;44:1883. [68] Reich L, Murayama M, Hono K. Acta Mater. 1998;46:6053. [69] Hutchinson CR, Fan X, Pennycook SJ, Shiflet GJ. Acta Mater. 2001;49:2827.