Available online at www.sciencedirect.com
Acta Materialia 60 (2012) 6573–6580 www.elsevier.com/locate/actamat
Double-atomic-wall-based dynamic precipitates of the early-stage S-phase in AlCuMg alloys S.B. Wang, J.H. Chen ⇑, M.J. Yin, Z.R. Liu, D.W. Yuan, J.Z. Liu, C.H. Liu, C.L. Wu ⇑ Center for High-Resolution Electron Microscopy, College of Materials Science and Engineering, Hunan University, Changsha, Hunan 410082, People’s Republic of China Received 5 June 2012; received in revised form 24 July 2012; accepted 12 August 2012 Available online 17 September 2012
Abstract Developments of high-strength aluminum alloys have always faced a difficult problem: owing to their small size, the early-stage strengthening precipitates are difficult to characterize in terms of composition, structure and evolution. Even for the widely used AlCuMg alloys, in which the phenomenon of precipitation hardening in metals was first discovered by Wilm more than a century ago, the essential questions remain: how many different precursors exist for the most effective strengthening precipitates (referred to S-phase), and how do they transform to the S-phase? Here we employ atomic-resolution electron microscopy imaging and first-principles energy calculations to address these problems. Our study demonstrates that the early-stage S-phase precipitates are highly dynamic in both composition and structure. Having their own genetic double Cu–Mg atomic walls to guide their evolution, these dynamic precipitates initiate, mature and grow with thermal aging following three evolution paths, leading to the S-phase precipitates formed, without exception, with even numbers of Cu–Mg atomic layers. Ó 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Aluminum alloy; Microstructure; Phase transformation; Electron microscopy
1. Introduction AlCuMg alloys are among the most important lightweight structural materials, exhibiting excellent toughness and strength. In particular, their heat resistance is superior to that of other aluminum (Al) alloys. The wide range of applications for which these materials are used include the aerospace industry (e.g. for fuselage structures). The mechanical properties of these alloys depend largely on the fine precipitates formed upon thermal aging [1,2]. These precipitates act as obstacles to dislocation movement in the Al grains, strengthening the alloys [3–5]. In high-strength AlCuMg alloys with relatively high Mg/Cu ratios, such as the Al–4.57Cu–1.34Mg (wt.%) alloy investigated in the ⇑ Corresponding authors. Tel.: +86 731 88664009; fax: +86 731 88664010 (J.H. Chen), tel.: +86 731 88664010 (C.L. Wu). E-mail addresses:
[email protected] (J.H. Chen),
[email protected] (C.L. Wu).
present study, three types of fine precipitates may form. The first type of precipitates that always form upon aging are solute clusters. Being fully coherent with the matrix, the Cu/Mg clusters contain a few to tens of solute atoms and are typically less than 4 nm in diameter [6,7]. They are less ordered in structure and therefore less effective at hardening the alloys. They can be visualized by atom-probe ion-field microscopy [6,8], but normally cannot be seen in scanning transmission electron microscopy (STEM) and in high-resolution transmission electron microscopy (HRTEM). The second type of precipitates that may form in the alloys are the 1-dimensional (1-D) crystals known as Guinier–Preston–Bagaryatsky (GPB) zones [6,9,10] (Fig. 1a). The GPB zones are effective hardening particles and can form rapidly if the aging temperature is higher than 190 °C [10–13]. Their formation can be depressed if the alloys are aged at 180 °C or below [2,6,9,14,15]. Significant progress has recently been achieved in characterizing their structures with STEM [15,16] after pioneering efforts had
1359-6454/$36.00 Ó 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.actamat.2012.08.023
6574
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
Fig. 1. Illustrative overviews of the precipitation microstructures in the Al–4.57Cu–1.34Mg (wt.%) alloy upon thermal aging. (a) The GPB zones formed after aging at 190 °C for 2 h. The scale bar is 20 nm. (b) The S precipitates formed after aging at 180 °C for 18 h. The scale bar is 200 nm. The insets in (a) and (b) are the HRTEM images of a few GPB zones and a S precipitate, respectively. The scale bar is 2 nm. All images are taken along the [1 0 0]Al//[1 0 0]S direction. (c) A 3-D drawing of bunches of the S precipitates that line up in a zigzag manner in the Al matrix. (d) The shape, orientation and Perlitz-Westgren (PW) (crystallographic structure) model of a lath-like S-phase crystal.
been made with different tools [17,18]. The third type of precipitates in the alloys are the most effective strengthening particles known as S-phase precipitates. They are 3D-nanocrystals of an equilibrium phase with the composition of Al2CuMg, and are coherently embedded in the Al matrix upon forming (Fig. 1b). Being typically lath-like, a S-phase precipitate may have a large length up to a few hundred nanometers, a width up to a few tens nanometers and a thickness up to a few nanometers (Fig. 1c and d). If less developed in width, such a particle may appear rod-like. The S-phase precipitates have a principle orientation relationship (OR) with the matrix [14,19], but can tolerate a few degrees of deviation from their principle OR [20,21]. Although they may precipitate individually, the majority of S-phase precipitates prefer to line up in a zigzag manner, forming bunches of long S-phase “fibers” and thereby strengthening the Al matrix (Fig. 1c). The S-phase has a centered-orthorhombic crystal structure (Fig. 1d) with lattice parameters aS = 0.400 nm, bS = 0.923 nm and cS = 0.714 nm, as originally proposed by Perlitz and Westgren [22,23]. Detailed calculations have demonstrated that the S-phase has a strong Cu–Mg bonding with ionic features in its Cu–Mg atomic layers [22]. Without knowing the precise precipitation sequence of the alloys upon thermal aging, significant controversies have long been existed about the development of S-phase precipitates [6,9,11,14,19,20,24,25]. To date the following precipitation sequence has generally been believed: supersaturated solid solution (SSSS) ! solute clusters ! S00 /GPB ! S, still ambiguously implying that the GPB
zones, or a somewhat different phase named S00 [6,9,13,14,19,25], are the precursors of the S-phase. To thoroughly understand the scenario of S-phase formation, atomic-resolution imaging, quantitative image simulations and energy calculations are required to determine the early structures of the S-phase particles upon their initiation. In the present study, using a combination of two atomicresolution imaging techniques, i.e. high-angle-annulardark-field (HAADF) imaging in STEM and through-focus exit-wavefunction reconstruction (TF-EWR) imaging in HRTEM, the initiation, maturation and growth of the Sphase precipitates in an Al–4.57Cu–1.34Mg alloy were followed in detail at the atomic scale. The early-stage precipitates of S-phase were first monitored by visualizing the heaviest atoms in the structures, i.e. Cu, with HAADF imaging, and then all the atoms in these structures were imaged with TF-EWR imaging. Three different evolution paths were revealed for the S-phase formation without the presence of any GPB zones. It is shown that specifically well-defined dynamic precipitates do exist as precursors of the S-phase precipitates. 2. Materials and methods 2.1. Alloy samples and thermal aging The materials used in the present work were supplied in the form of homogenized rolled sheets with the composition Al–4.57Cu–1.34Mg–0.81Mn (wt.%). Prior to the hardening stage, the alloy was solution heat-treated at 495 °C for 1 h and then water quenched to room temperature (20 °C). Subsequent thermal aging was carried out in an oil bath at 150, 180, 190 and 220 °C. It was observed that at an aging temperature lower than 180 °C, the S-phase precipitates formed more rapidly than the GPB zones. A series of the alloy specimens aged at 180 °C for periods of 0–180 h was prepared for investigation, in which the GPB zones were much depressed and formed after the Sphase formation. Specimens for the STEM and HRTEM observations were prepared first by mechanical polishing and then by electropolishing until perforation. 2.2. Atomic-resolution electron microscopy imaging 2.2.1. The HRTEM instrument A FEI Tecnai F20 HRTEM instrument, operating at 200 kV, was employed in the present study. Two atomicresolution imaging techniques are available with this microscope: one is to perform TF-EWR over a focus-varying series of about 20 images per series that are automatically recorded in HRTEM imaging mode [26–29]; the other one involves performing HAADF imaging in STEM imaging mode. The highest point resolution of the microscope achievable by performing TF-EWR is 0.14 nm (the socalled information limit) [30], while the best point resolution achievable in HAADF STEM is 0.16 nm. The key issue with performing atomic-resolution imaging was to
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
obtain images that truly reflect the real atomic structures of the observing objects (within the resolution limitation of the microscope). 2.2.2. HAADF imaging in STEM In the HAADF-STEM imaging mode, the obtained image shows a Z-contrast (where Z represents the atomic numbers of the atoms in the specimen) proportional to Z1.72.0 [31–33]. Hence the HAADF image contrast is relatively easy to interpret in terms of the structure, so long as the composition of the observed material is known, though the light atoms are more difficult to visualize than the heavy atoms when they have very different Z values and coexist in the structure. The imaging parameters are as follows: operating power 200 keV, half-angle of 12 mrad for probe convergence and a collection inner semi-angle of 36 mrad. 2.2.3. TF-EWR imaging in HRTEM For atomic-resolution imaging by TF-EWR in HRTEM imaging mode, three steps are required to obtain suitable images. Firstly, each focus-varying series of about 20 images of the observing object is recorded automatically with the Tecnai F20 microscope. Secondly, the image series is processed by the FEI software TrueImage [27,34] to recover the complex wavefunction at the image plane. Thirdly, the aberrations in the obtained wavefunction, such as defocus, coma, twofold astigmatism as well as threefold astigmatism, should be corrected in order to retrieve the final correct wavefunction below the specimen, which has always been a difficult step in the procedure and requires quite a lot of experience in practice [26,27,34–36]. This step becomes much easier if an aberration-corrected HRTEM instrument is used to record the image series [37]. After removing the lens aberrations, the phase-map of the obtained wavefunction shows an atomic-resolution image, the contrast of which reflects the projected atomic potential of the observing object, so long as the specimen thickness is sufficiently thin (to roughly meet the phase-object approximation) [3,38]. 2.2.4. Image simulation analysis Image simulations were performed to verify the atomicresolution images obtained by TF-EWR, and to refine the obtained structure models, using the MacTempas image simulation package [39]. If the structure model proposed by atomic-resolution imaging was correct, the simulated wavefunction would accurately match the one experimentally retrieved by TF-EWR. In image simulations, supercells larger than 4 4 1 Al unit cells were used in order to include both the Al matrix and the precipitates. Simulations of images, exit wavefunctions and electron diffraction patterns have for decades been a standard tool in HRTEM [40–44]; in particular, to determine the defect structure inside a known crystal (i.e. when an internal standard exists), the image simulation analysis can be performed quantitatively [3,22,37,45].
6575
2.2.5. First-principles energy calculations The first-principles calculations were performed using density functional theory (DFT) implemented in the Vienna Ab initio Simulation Package (VASP) with efficient ultrasoft pseudopotentials [46–50]. The generalized gradient approximation (GGA) with the exchange–correlation functional of Perdew and Wang [51,52] was employed. Convergence tests indicated that 260 eV was a sufficient cut-off for the ultrasoft pseudopotential to achieve high precision in the AlCuMg system. For all structures, k-point sampling using Monkhorst–Pack method [53] was used to achieve a precision of 0.1 kJ mol1. The thermal contributions and zero-vibration energy differences or contributions were ignored because of their small effects on this alloy system [22]. Sufficiently large supercells were used in all calculations in order to include the interface energy and the strain energy for the particles [22]. The formation enthalpies were defined and calculated in kJ mol1 per solute atom [54]. 3. Results 3.1. Atomic-resolution HAADF imaging of dynamic S precipitates Fig. 2a–d shows the HAADF images of early-stage precipitates in the sample series, revealing the first type of Sphase formation. Fig. 2a and b demonstrates that even in the very early-stage (1–6 min of aging), {2 1 0}Al-plane-oriented plate-like precipitates with double Cu layers appeared. They exhibit the same OR and line up in the same manner as the well-developed S-phase precipitates do (Fig. 3), though their structures have not yet become the matured S-phase structure. Fig. 2c shows a thin S-phase precipitate with only two Cu layers. The insets in Fig. 2a–c show that three different structures exist among these earlystage precipitates with double Cu layers. They are different not only in Cu–Cu distances but also in Cu–Cu orientations. We use the nomenclature GPS2-I (Fig. 2a), GPS2II (Fig. 2b) and S2 (Fig. 2c) for these particles, respectively, meaning that the thinnest S-phase (S2) precipitates are of double Cu layers and have two of their own characteristic Guinier–Preston (GP) zones or precursors also with double Cu layers. The GPS2-I and the GPS2-II zones appear mostly in the samples aged for 1–6 min, whereas most of the S2-particles form after 1 h of aging. Fig. 2d shows a developed S-phase precipitate in the sample aged for 18 h (the peak-hardness age). It was observed that all S-phase precipitates have even numbers of Cu layers without exception, as indicated by and referred to S2, S4, S6, . . ., S2n in Fig. 2 (where n = 1,2, . . .). The second type of S-phase formation occurs when a particle initiates with two Cu layers with a larger separation (Fig. 2e), and then matures by absorbing solute atoms to replace its internal Al atoms, directly forming a matured S4 precipitate (Fig. 2f). We link such precursors of S4 precipitates to GPS4 zones. The third type of S-phase forma-
6576
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
Fig. 2. Atomic-resolution HAADF images revealing the precursors and the paths of S-phase formation. (a) The initial S-phase precursors, named GPS2-I zones, observed in the samples aged at 180 °C for 1–6 min. (b) Another type of S-phase precursor, named GPS2-II zones, observed in the samples aged for 1–6 min. (c) The thinnest (structurally matured) S precipitates, named S2, observed in the samples aged for more than 6 min. (d) A growing S precipitate with several portions of different thicknesses, named S4, S6,. . ., S2n,. . ., respectively, observed in the sample aged for 18 h. (e) A different initial S-phase precursor, named the GPS4 zone, observed in the samples aged for 6 min. (f) An individually formed S4 precipitate observed in the sample aged for 1 h. (g) A S4–GPS4 complex observed in the sample aged for 2 h. (h) A typical intermediate stage, named the S2n–GPS2-II complex, of the S2n precipitate thickening towards S2(n+1), observed in the sample aged for 2 h. The larger scale bars in all images are 2 nm. The smaller scale bars in the insets in (a)-(c) are 1 nm. The brightest dots in all images indicate the positions of Cu-containing atomic columns in the structures studied, viewed along the [1 0 0]Al// [1 0 0]S direction.
tion involves the following process: a S4 precipitate that forms, or more precisely the strain-field in front of it, triggers a GPS4 zone to nucleate, forming a S4–GPS4 complex (Fig. 2g). Such a complex then further develops into a widened S4 particle: the merged particle will be a perfect S4 precipitate if the triggered GPS4 zone closely matches the original S4 particle in atomic planes. Otherwise the merged particle includes a defect. Once a precipitate is matured as a S2n crystal, it may further grow larger in a rather anisotropic and characteristic manner. In its length, a S2n precipitate can grow longer rather normally, since its lattice closely matches the Al lattice. In terms of width, a S precipitate may grow wider only by triggering another precipitate to nucleate first and then merging with it to form a widened particle (see Fig. 2g for a S4 particle). In terms of thickness, a S2n precipitate can grow into a thicker S2(n+1) precipitate only by triggering a GPS2-II zone with a double Cu layer to form on its sides first (Fig. 2h), and then by inducing it to transform to the S-phase structure, leading without exception to a thickened precipitate with even numbers of Cu layers. Fig. 2h indicates that at this intermediate S6–GPS2-II stage of S-phase thickening, one Cu layer of the GPS2-II zone has probably adopted the matured S-phase structure, whereas another Cu layer remains to be transformed, as revealed below in more detail. 3.2. The structures of dynamic S precipitates refined by TFEWR imaging and simulation analysis To understand how early-stage particles evolve, TFEWR imaging was employed to reveal the positions of the lighter Al and Mg atoms in their structures. The obtained
complex exit-wavefunction below the specimen has an amplitude part and a phase part. If the phase part is displayed as a phase-map, it shows an atomic-resolution image, the contrast of which reflects the projected atomic potential of the observing object, so long as the specimen thickness is sufficiently thin (to roughly meet the phase-object approximation) [3,38]. Fig. 4a–c shows such atomic-resolution images of S2, GPS2-II and GPS2-I particles. Since without a known reference structure it is a tedious task to obtain reliable atomic-resolution images of TF-EWR, we use the matured S2 structure as a starting reference (Fig. 4a), which is in good agreement with our previous results obtained using the same TF-EWR technique for larger S-phase particles [22]. It is interesting that beside this S2 precipitate there is a GPS2-I zone with faint contrast, probably still covered by the Al matrix in the viewing direction. Fig. 4b–d demonstrates the atomic-resolution images of three GPS2-I zones and a GPS2-II zone (Fig. 4b), as well as a S8–GPS2-II complex, which is thickening towards a S10 precipitate (Fig. 4d). Since the contrast difference between Al and Mg atomic columns are small, image simulations [39–45] and first-principles energy calculations [46,51] were conducted to refine their positions in the GPS2-I, and the GPS2-II structures. Possible atomic compositions and configurations for the structures are optimized to obtain the best match between the simulated images and the experimentally reconstructed ones (the insets in Fig. 4a–d), and at the same time to achieve the minimum formation enthalpies. For image simulation, a starting structure model can directly be read from the atomic-resolution image, and then simulated images can be obtained by adjusting the atomic positions, species, specimen thicknesses, crystal tilts and Debye–Waller factors [3,22,37,45]. When a good
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
6577
Although the observed thickening mechanism of Sphase precipitates has explained well why these particles always have even numbers of Cu–Mg atomic layers in the (2 1 0)-plane without exception (Figs. 2d and 4d), we performed further energy calculations for the particles with both odd and even numbers of Cu–Mg atomic layers. The results show that the S2n precipitates are indeed energetically favored in comparison with the hypothetical S2n+1 precipitates, as shown in Fig. 6. Without having to know their evolution or transformation mechanisms, the presented results, achieved by HAADF imaging and TF-EWR imaging in association with image simulations and energy calculations, suggest three evolution paths of the dynamic S precipitates towards S-phase formation, as summarized by the 3-D illustrations shown in Fig. 7, and as discussed below in more detail. 4. Discussion
Fig. 3. Overview images obtained by HAADF-STEM imaging of the early-stage precipitates and the age-hardening curve of the alloy aged at 180 °C. (a) A low-magnification image of the sample aged for about 1 min. The scale bar is 100 nm. (b) The particles named GPS2-I zones, observed in the samples aged for 1 min. (c) The particles named GPS2-II zones, observed in the samples aged for 6 min. (d) The particles named S2, observed in the samples aged for 2 h. (e) The coarsened S-phase particles observed in samples aged for 18 h. The scale bar in (a) is 100 nm and the scale bars in (b)–(e) are 5 nm. All images are viewed in the [1 0 0]Al/[1 0 0]S direction. (f) The age-hardening curve of the alloy aged at 180 °C for different times, showing two hardening peaks: an early-stage peak (6 min, 120 HV) and a second peak (18 h, 160 HV).
match is achieved between the simulated image and the experimentally obtained one, a refined structure model can be obtained. The refined structure models were then checked for their formation enthalpies using first-principles energy calculations in comparison with other possible hypothetical models, as shown in Fig. 5. The results show that the structure models suggested by the atomic-resolution images and refined by image simulations are indeed energetically favored. Both image simulations and energy calculations suggest that, based on as 2-D crystals embedded in the Al matrix, the compositions of GPS2-I, GPS2-II and (initial) GPS4 zones are Al3MgCu, Al2MgCu and Al6MgCu, respectively. Their structures and formation enthalpies are shown in Fig. 4e. Fig. 4d further confirms that one Cu–Mg atomic layer of the attached GPS2-II zone has adopted the S-phase structure, whereas another Cu–Mg atomic layer still remains in the GPS2-II configuration. These TF-EWR images (verified with image simulations and energy calculations), which allow all the atoms to be visualized, are in good agreement with the HAADF images shown in Fig. 2.
Studying the determined structures and their relative stabilities, the process of initiation, maturation and growth of a dynamic or developing S-phase precipitate becomes understandable in much greater detail. Driven by a reduction in energy, the main precipitate evolution path towards S-phase formation is the following: SSSS ! GPS2-I (Al3MgCu) ! GPS2-II (Al2MgCu) ! S2 (Al2MgCu) . . . ! S2n (Fig. 7a). The transformation from GPS2-I to GPS2-II can be realized through the migration of one of the two Cu–Mg atomic walls towards another by just one atomic distance, leading to the compositional change from Al3MgCu to Al2MgCu (Fig. 8a). The transformation from GPS2-II to S2 involves a structural variation without any compositional change (Fig. 8b). The two Cu–Mg atomic walls of a GPS2-II zone in fact differ in how they transform to the S2 structure. The Cu–Mg atomic wall with inward Cu atoms in the GPS2-II structure can change directly to a Cu–Mg atomic layer of S2, whereas the other one with the outward Cu atoms cannot transform directly (Figs. 2h and 4d support this analysis). This change requires interstitial-vacancy pairs (Frenkel pairs) generated by thermal fluctuations. It is known that a Frenkel pair, or Frenkel defect, forms in a crystal lattice when an atom leaves its lattice site and becomes an interstitial, creating a vacancy at its original site. Frenkel pairs occur due to thermal vibrations, and theoretically there will be no such defects in a crystal at 0 K. In the Al lattice an Al self-interstitial occupies a special position to form (together with its nearest Al-atom) a dumb-bell configuration [55,56]. However, it has been shown by previous work on a dilute Al–Cu alloy that a Cu-interstitial prefers to locate at the octahedral interstitial positions [57], as shown in Fig. 8c and d. Imagine that a thermal Frenkel pair occurs at the outwards Cu atoms on one of the two Cu–Mg atomic walls of a GPS2-II zone (Fig. 8c). The pair will recover most of the time. However, if the vacancy diffuses away (i.e. an Al atom moves in), a Cu atom shifted (by 0.5a) into an octahedral position will be caged and survive for a rela-
6578
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
Fig. 4. Atomic-resolution TF-EWR phase-images matched with image simulations and energy calculations to reveal the structures and their transformations of the S-phase precursors. (a–d) The TF-EWR images of a matured S2 precipitate (a), a GPS2-II zone (b), several GPS2-I zones (a–c) and a S8-GPS2-II complex (d), are well matched with the simulated images (as the insets), obtained based on the inserted structure models, respectively, (a) for a specimen thickness of 6 nm and a crystal tilt of 0.46°; (b) for a specimen thickness of 4 nm and a crystal tilt of 1.15°; (c) for a specimen thickness of 5 nm and a crystal tilt of 0.86°; (d) for a specimen thickness of 5 nm and a crystal tilt of 0.29°. All images were viewed along the [1 0 0]Al//[1 0 0]S direction and all the scale bars are 1 nm. (e) The formation enthalpies DH obtained by first-principles energy calculations for the illustrated GPS2-I, GPS2-II, GPS4, S2 and S4 supercell models.
Fig. 5. A number of hypothetic GPS2-I and GPS2-II structures in comparison with those suggested by atomic-resolution imaging with respect to their formation enthalpies DH, plotted against the number of atomic-layers existing between the two Cu atomic layers in these structures, showing that the experimentally proposed structure models are energetically favored.
tively longer period [57]. Although unlikely, it is still possible in thermodynamic statistics that several Cu-interstitials caged at several connected octahedral positions on this Cu– Mg atomic wall survive together for a moment due to (local) thermal fluctuations (Fig. 8d). Energy calculations
Fig. 6. First-principles calculations for the S2n particles and the hypothetic S2n+1 particles embedded in the Al matrix, showing that the S2n precipitates are indeed energetically favored.
show that if this occurs, the GPS2-II structure will automatically relax to the S2 structure to reach an energy minimum, as shown in Fig. 9. The proposed microscopic mechanisms for the transformations from GPS2-I to GPS2-II and from GPS2-II to S2 do not imply that all the atoms in such zones have to move collectively upon transforming. Our model is the following: once a portion larger than a critical size has transformed to the more stable structure, this segment will serve as a stable nuclei for a rapid transformation of the remaining portion
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
6579
Fig. 9. First-principles energy calculation results, showing that once the outwards Cu atoms are shifted upwards and caged in octahedral positions by moving the vacancies away, the GPS2-II structure will relax to the S2 structure to reach an energy minimum.
Fig. 7. 3-D illustrations of the evolution scenarios of the double-Cu–Mgatomic-wall-based dynamic S-phase precipitates in AlCuMg alloys upon thermal aging. (a) The most frequently observed evolution path of a dynamic precipitate undergoing several distinguished metastable stages towards S-phase formation. (b and c) The other two less frequently observed evolution paths of the dynamic precipitates towards S-phase formation, some of intermediate stages of which are still difficult to characterize.
Fig. 8. Schematic illustrations of proposed transformation mechanisms for the evolving S precipitates. (a) Transformation from GPS2-I to GPS2II. (b) Transformation from GPS2-II to S2. (c) A Cu atom shifts into an octahedral position and leaves a vacancy at its original site, forming a Frenkel pair. (d) A couple of Cu atoms might be caged in several connected octahedral positions, if a few Frenkel pairs occurred due to thermal fluctuations and the vacancies moved away. This configuration may lead the formation of a S2 nuclei in a GPS2-II zone. VC indicates vacancy.
of the entire zone (just like crystal growth) (Fig. 8a and b). It is understandable that forming a S2 nuclei in a GPS2-II zone will take a long incubation time, since it requires a couple of its outwards Cu atoms to become the interstitials caged in the octahedral sites of the same {2 1 0} plane and therefore needs a long incubation time to occur (through generation of the Frenkel pairs by locally collective thermal fluctuations of small probability in thermodynamic statistics). This in turn allows GPS2-II zones and S2n–GPS2-II complexes to be frequently observed before they have transformed. The less frequently observed GPS4 zones and S4–GPS4 complexes (Figs. 2e, g and 10) follow the same evolution path towards S-phase formation: SSSS ! GPS4 (Al6MgCu) ! S4 (Al2MgCu) . . . ! S2n (Fig. 7b and c). The transformation from GPS4 to S4 is again a dynamic process. A GPS4 zone has a dynamic composition of Al12-x(t)-y(t)Mg2+x(t)Cu2+y(t) with x(t = 0) 0 and y(t = 0) 0. It may initiate (t = 0) with two distant Cu–Mg atomic walls and a composition of Al12Mg2Cu2 (=Al6MgCu) (Fig. 4e), and evolve both compositionally and structurally towards a S4 precipitate (Al8Mg2+2Cu2+2 = Al2MgCu), as shown in Fig. 10. Fig. 10a clearly shows that outside of the evolving GPS4 precipitate all the atomic columns are well resolved, whereas inside the double Cu–Mg wall the atomic columns cannot be resolved, indicating that the internal structure is transforming and therefore the atoms in the columns are not well aligned or ordered. Fig. 10c demonstrates that even for a
Fig. 10. HAADF images showing more experimental evidences that a GPS4 precipitate may have complex intermediate stages: (a) a GPS4 precipitate with a less-ordered internal structure, (b) an extended S4–GPS4 complex, and (c) a S4–GPS4 complex with a maturing GPS4 precipitate.
6580
S.B. Wang et al. / Acta Materialia 60 (2012) 6573–6580
more matured GPS4 precipitate with more atomic ordering, its structure seems still different from that of the S4 precipitate. Hence due to drastic changes inside the dynamic GPS4 zones, their intermediate structures are difficult to characterize precisely. At this moment it is not clear whether or not a GPS4 zone has more than one structurally distinct stage, e.g. GPS4-I and GPS4-II stages. 5. Conclusions In summary, a detailed scenario of S-phase formation in widely used high-strength AlCuMg alloys has been revealed (Fig. 7). It is demonstrated that the S precipitates are highly dynamic in their early stages of development and have their own characteristic precursors. These particles were analyzed from two different viewpoints to understand their complex evolution in composition, structure and morphology. From one view, a S2 (or S4) precipitate and its precursors, the GPS2-I and GPS2-II (or GPS4) particles, should be regarded as the same object in a dynamic process (like a living object), because they have genetic double Cu– Mg atomic walls in the (2 1 0)Al-plane to guide their evolution towards a structurally matured S2 (or S4) precipitate (Al2MgCu). From another view, these particles can be treated as different physical phases, in order to elucidate the driving forces and the transformation mechanisms from one structure to another. Nonetheless, there are more intermediate stages that are still difficult to characterize. With ultra-high resolving power in electron microscopy and additional computing power in calculations, a better understanding of the atomic-scale metallurgy of this alloy system can be achieved. Acknowledgements This work is supported by the National Basic Research (973) Program of China (No. 2009CB623704); the National Natural Science Foundation of China (Nos. 51171063, 10904034, 51071064); Instrumental Innovation Foundation of Hunan Province (No. 2011TT1003); The Aid Program for Science and Technology Innovative Research Team in Higher Educational Institutions of Hunan Province. We thank Prof. S. Van der Zwaag for providing some of the Al samples. References [1] Fine M. Metall Mater Trans A 1975;6:625. [2] Ringer SP, Sakurai T, Polmear IJ. Acta Mater 1997;45:3731. [3] Chen JH, Costan E, van Huis MA, Xu Q, Zandbergen HW. Science 2006;312:416. [4] Erhard HJ. Light Met 2001;1:127. [5] Liu JZ, Chen JH, Yang XB, Ren S, Wu CL, Xu HY, et al. Scripta Mater 2010;63:1061. [6] Sha G, Marceau RKW, Gao X, Muddle BC, Ringer SP. Acta Mater 2011;59:1659. [7] Marceau RKW, Sha G, Ferragut R, Dupasquier A, Ringer SP. Acta Mater 2010;58:4923. [8] Kazuhiro H. Prog Mater Sci 2002;47:621.
[9] Ratchev P, Verlinden B, De Smet P, Van Houtte P. Acta Mater 1998;46:3523. [10] Ralston KD, Birbilis N, Weyland M, Hutchinson CR. Acta Mater 2010;58:5941. [11] Zahra AM, Zahra CY, Alfonso C, Charai A. Scripta Mater 1998;39:1553. [12] Silcock JM. J Inst Metals 1960–61;89:203. [13] Charai A, Walther T, Alfonso C, Zahra AM, Zahra CY. Acta Mater 2000;48:2751. [14] Ringer SP, Hono K, Polmear IJ, Sakurai T. Appl Surf Sci 1996;94– 95:253. [15] Kovarik L, Court SA, Fraser HL, Mills MJ. Acta Mater 2008;56:4804. [16] Kovarik L, Mills MJ. Scripta Mater 2011;64:999. [17] Pennycook SJ, Jesson DE. Phys Rev Lett 1990;64:938. [18] Batson PE. Nature 2002;418:617. [19] Bagaryatsky YA. Dokl Akad Nauk SSSR 1952;87:559. [20] Radmilovic V, Kilaas R, Dahmen U, Shiflet GJ. Acta Mater 1999;47:3987. [21] Winkelman GB, Raviprasad K, Muddle BC. Acta Mater 2007;55:3213. [22] Liu ZR, Chen JH, Wang SB, Yuan DW, Yin MJ, Wu CL. Acta Mater 2011;59:7396. [23] Perlitz H, Westgren A. Ark Kemi Mineral och Geol B 1943;16:13. [24] Wang SC, Starink MJ. Int Mater Rev 2005;50:193. [25] Wang SC, Starink MJ. Acta Mater 2007;55:933. [26] Coene WMJ, Janssen AJEM, Op de Beeck M, Van Dyck D. Philips Electron Opt Bull 1992;132:15. [27] Thust A, Coene WMJ, Op de Beeck M, Van Dyck D. Ultramicroscopy 1996;64:211. [28] Van Dyck D, Op de Beeck M, Coene WMJ. Optik 1993;93: 103. [29] Coene W, Janssen G, Op de Beeck M, Van Dyck D. Phys Rev Lett 1992;69:3743. [30] De Jong AF, Van Dyck D. Ultramicroscopy 1993;49:66. [31] Nellist PD, Pennycook SJ. Ultramicroscopy 1999;78:111. [32] Hillyard S, Silcox J. Ultramicroscopy 1995;58:6. [33] Kirkland EJ, Loane RF, Silcox J. Ultramicroscopy 1987;23:77. [34] Coene WMJ, Thust A, Op de Beeck M, Van Dyck D. Ultramicroscopy 1996;64:109. [35] Coene WMJ, Denteneer TJJ. Ultramicroscopy 1991;38:225. [36] Chen JH, Van Dyck D, Op de Beeck M, Van Landuyt J. Ultramicroscopy 1997;69:219. [37] Chen JH, Zandbergen HW, Dyck DV. Ultramicroscopy 2004;98:81. [38] Cowley JM. Ultramicroscopy 1992;41:335. [39] O’Keefe MA, Kilaas R. Advances in high-resolution image simulation. In: Pfefferkorn conference, Niagara Falls, NY; 1988. [40] Hy¨tch MJ, Stobbs WM. Ultramicroscopy 1994;53:191. [41] Boothroyd CB. J Microsc 1998;190:99. [42] O’Keefe MA, Buseck PR, Iijima S. Nature 1978;274:322. [43] Chen JH, Van Dyck D. Ultramicroscopy 1997;70:29. [44] Spence JCH, Zuo JM. Electron microdiffraction. New York and London: Plenum Press; 1992. [45] Jia CL, Urban K. Science 2004;303:2001. [46] Kresse G, Furthmu¨ller J. Comput Mater Sci 1996;6:15. [47] Kresse G, Furthmu¨ller J. Phys Rev B 1996;54:11169. [48] Kresse G, Hafner J. Phys Rev B 1993;47:558. [49] Vanderbilt D. Phys Rev B 1990;41:7892. [50] Kresse G, Hafner J. J Phys: Condens Matter 1994;6:8245. [51] Perdew JP, Chevary JA, Vosko SH, Jackson KA, Pederson MR, Singh DJ, et al. Phys Rev B 1992;46:6671. [52] Perdew JP, Wang Y. Phys Rev B 1992;45:13244. [53] Monkhorst HJ, Pack JD. Phys Rev B 1976;13:5188. [54] Van Huis MA, Chen JH, Zandbergen HW, Sluiter MHF. Acta Mater 2006;54:2945. [55] Ehrhart P, Schilling W. Phys Rev B 1973;8:2604. [56] Jesson BJ, Foley M, Madden PA. Phys Rev B 1997;55:4941. [57] Klaver TPC, Chen JH. J Comput Aid Mater Des 2005;10:155.