Atomic Pillar-Based Nanoprecipitates Strengthen AlMgSi Alloys

See allHide authors and affiliations

Science  21 Apr 2006:
Vol. 312, Issue 5772, pp. 416-419
DOI: 10.1126/science.1124199


Atomic-resolution electron microscopy reveals that pillarlike silicon double columns exist in the hardening nanoprecipitates of AlMgSi alloys, which vary in structure and composition. Upon annealing, the Si2 pillars provide the skeleton for the nanoparticles to evolve in composition, structure, and morphology. We show that they begin as tiny nuclei with a composition close to Mg2Si2Al7 and a minimal mismatch with the aluminum matrix. They subsequently undergo a one-dimensional growth in association with compositional change, becoming elongated particles. During the evolution toward the final Mg5Si6 particles, the compositional change is accompanied by a characteristic structural change. Our study explains the nanoscopic reasons that the alloys make excellent automotive materials.

Aluminum is essential to modern civilizations because of its light weight, strength, and workability. Its many applications include fuel-efficient transportation vehicles (e.g., it comprises about 80% of a commercial aircraft's unloaded weight), building construction, and food packaging. Pure aluminum is soft and has little strength or resistance to plastic deformation. However, alloyed with small amounts of other elements, it can provide the strength of steel at only half the weight. With thermal treatments, the added alloying elements can form nanometer-sized precipitates, which act as obstacles to dislocation movement in the crystal (atomic matrix), strengthening the aluminum. This phenomenon is known as precipitation hardening, and the hardening nanoprecipitates are named GP zones after the pioneer work by Guinier and Preston on AlCu alloys (1, 2).

AlMgSi accounts for a large percentage of the total aluminum production in the world. With appropriate pre-aging treatments, AlMgSi alloys can be pressed easily into a given form and then strengthened rapidly by annealing for a very short duration (<30 min) at about 180°C [i.e., by a characteristic two-step age-hardening process (3)]. This important property of AlMgSi alloys, called the quick-bake hardening response, has led to increased applications in the automotive industry (37), such as outer panel materials that have a strength/weight ratio optimal for fuel efficiency and environmental protection. It has long been understood that AlMgSi alloys are strengthened when needle-like monoclinic precipitates form (514), but to date little is known about the structures of the particles responsible for the quick-bake hardening response (57). Whereas the final structure of the needlelike particles has been determined (8, 9), their early-stage development is difficult to characterize.

Because the nanoprecipitate structures are not well understood, the hardening nanoparticles in AlMgSi alloys are named ambiguously: They are referred to as GP(I) and GP(II) zones, pre-β″ and β″ phases, or Si/Mg co-clusters and GP zones in the early stages (411). The essential questions remain: How many different hardening particles exist, and how do they transform from one to another? To answer these questions, we used high-resolution transmission electron microscopy (HRTEM) and computational analysis to assess their initial structures.

We studied aluminum that was alloyed with 0.43% Mg and 1.2% Si (15). The homogenized alloy was heated at 560°C and then water quenched to 20°C. The best quick-bake hardening response can be achieved if the hardening annealing is performed immediately after water quenching. However, this is not practical for automotive body sheet applications, because the sheets have to be stored (up to months) and shipped at room temperature. Any storage (natural aging) will quickly degrade the alloy's quick-bake hardening response because of the formation of natural-aging clusters, which delay the formation of the hardening particles upon annealing (47). Hence, for advanced applications, pre-aging has to be applied in aluminum factories to make the quick-bake hardening response largely independent of the storage time. Therefore, we studied samples with and without pre-aging (15). The specimens were prepared by electropolishing for HRTEM investigations (15).

For straight annealing at 180°C, the hardening particles start appearing in ∼1 to 5 min and then grow rapidly along the Al 〈100 〉 directions into needles. Slightly elongated particles ∼2 to 6 nm in length and ∼2 nm in width lead to a hardness increase of the alloy from 75 to 85 Vickers hardness (HV) (Fig. 1A). In 30 min, they become needles of 18 nm long on average without coarsening in width (Fig. 1B), and the hardness increases to ∼120 HV. As annealing continues, the needles grow rather slowly and still do not change notably in width. A peak hardness of ∼130 HV can be reached after annealing for ∼3 hours. With continued annealing, coarsened needles ∼4 nm in width can appear (Fig. 1C), and the hardness decreases slowly. The final-stage hardening particles have been identified as the Mg5Si6 (β″) phase (8, 9).

Fig. 1.

HRTEM images of typical hardening precipitates in AlSiMg alloys. (A to D) Overviews of these 〈100 〉-oriented particles after direct annealing at 180°C for 5 min (A), 30 min (B), and 4 hours (C), and after a characteristic two-step age-hardening process (15) (D). (E to H) The lattice images of the particles shown in (A) to (D), showing that these particles have the same monoclinic unit cells. To see the morphology of short particles clearly [(A) and (D)], the high-resolution imaging mode in HRTEM has to be used, whereas the diffraction-contrast imaging mode provides better overviews of the needlelike particles [(B) and (C)]. The arrows in (G) indicate a stacking fault in the particle.

After appropriate pre-aging, most of the particles formed are spherical and ∼2 nm or smaller (small round particles in Fig. 1A). Typical hardening particles formed after two-step age-hardening processes (15) yield quick-bake hardening responses with hardness increases between ∼75 to 85 HV and ∼110 to 120 HV (Fig. 1D). Although shorter than those seen in Fig. 1, B and C, these particles are also elongated along the Al 〈100 〉 directions. If many particles become elongated after pre-aging, the alloy may demonstrate the quick-bake hardening response but lose the good formability necessary for automotive parts.

Analysis of many HRTEM images of these nanoparticles reveals that they all have the same monoclinic lattice (Fig. 1, E to H, and figs. S1 and S2). These lattice images show that (i) most particles have only ∼1 to 2 monoclinic unit cells in width (∼2 nm), whereas the final particles can be up to 4 nm in width (Fig. 1G) and (ii) most particles are coherent with the matrix, but final particles can be semicoherent and some of them may include stacking faults.

It is difficult to determine the monoclinic nature of these particles, especially the short ones, for several reasons. (i) The image delocalization due to microscopic aberrations leads to mixing of the particle image with the matrix image (16). As a result, the particles appear as disordered clusters under most imaging conditions (15). (ii) The differences between Mg, Al, and Si columns in electron-scattering power are small for the specimen thicknesses used (typically ∼5 to 15 nm). (iii) For thick specimens, particles are likely covered by the matrix. Hence, longer particles (>5 nm) and atomic imaging in HRTEM must be used to reveal the structure of early-stage precipitates.

We used through-focus exit-wavefunction (EW) reconstruction (1521) to image the atomic structure of early-stage needles (Fig. 1B). From a recorded through-focus series of 20 HRTEM images, we retrieved the electron wavefunction at the exit plane of the specimen. The EW's phase was then used as an atomic image of the specimen. The obtained image clearly resolves all of the atom columns in the precipitate (Fig. 2).

Fig. 2.

Deduction of the precipitate structure. (A) The phase-image of the reconstructed EW of a precipitate in the matrix. The bright dots represent atom columns. The insets are the calculated images from a structure model (left) with varying specimen parameters: 1, changing the Si-Si distances in the model; 2, adjusting the crystal tilt and thickness; 3, adjusting the positions and compositions of atomic columns; 4, including the DW factors of atoms. (B) Refinement results: The calculated images (insets) match with the reconstructed images for both the precipitate and the matrix [Al(m)].

To deduce the precipitate structure through modeling and image simulation, we considered the following. (i) The early precipitates are the result of the Mg/Si aggregation and substitution of Al in the matrix. (ii) Previous studies indicate that atoms in such precipitates should take the coordinates of either y = 0 or y = ½ (8, 9). (iii) The known atomic sizes of Mg, Si, and Al allow us to exclude unreasonable configurations (e.g., the Mg–Mg bonds should not be the shortest bonds). We first assumed that our specimen was a weak-phase object, for which the EW's phase is proportional to the projected potentials of atom columns (21, 22). Therefore, the brightest spots in Fig. 2 are Si (the heaviest), the least bright spots are Mg, and the others are Al. Using these guidelines, we obtained an initial structure model.

The model was repeatedly adjusted such that the phase image of the EW calculated from the model matched well with the phase image of the EW reconstructed from the experimental image series. The refinement took into account all parameters that may influence the image contrast (Fig. 2A). The EW calculations were performed with the MacTempas image simulation software (23). Furthermore, it was necessary to match the phase images not only for the precipitate but also for the matrix (as an internal standard) under the same conditions (Fig. 2B) (15).

The obtained structural data (including the specimen thickness and tilt) are shown in table S1. Comparing the deduced Mg2Si2.6Al6.4 structure with the β″ structure (8, 9) and with an intermediate structure proposed for the so-called pre-β″ phase (Fig. 3A) (11), we made several observations. (i) Mg2Si2.6Al6.4 has a high Al content (58%), and its Al atoms are ordered nearly in the same way as they are in the matrix. (ii) The common components of these structures are the Si double atoms. They remain unchanged not only in their location and orientation, but also in their interatomic distance (0.256 nm). This implies that a transformation from Mg2Si2.6Al6.4 to Mg5Si6 may occur without having to break down the monoclinic frame kept by these Si2 components. Two changes are required for such a transformation: replacing all Al by Si/Mg and rearranging atom positions, including a bGP/2 shift of 2 (out of 22) atoms (where bGP is the b lattice parameter of the precipitates, i.e., GP zones) (Fig. 3A). (iii) Notably, all the Si2 atoms have a small Debye-Waller (DW) factor, whereas all the atoms that are expected to change have large DW factors (which describe the displacements of atoms from their average positions resulting from thermal vibration and other local random movements) (table S1). The large DW factors imply that atoms are less evenly aligned along in these columns because of their compositional variations.

Fig. 3.

The structures of different hardening precipitates. (A) The structures projected along cAl (=bGP). (aAl bAl cAl) and (aGP bGP cGP) denote the Al and precipitate lattice vectors, respectively. The Al atoms in (Mg4Al1)Si6 are outlined with red circles. The Mg atoms outlined with red circles indicate the special positions at which the Mg atoms have to shift bGP/2 to become Mg5Si6 (11). (B) The 3D view of Mg2Si2.6Al6.4 and Mg5Si6 particles surrounded by Al, showing that the Si double columns are the common structural components of these precipitates and may serve as the stable pillars in the structure evolution. More Si2 components (yellow ones) can be found in Mg5Si6, but they no longer act as pillars for Mg5Si6 to evolve.

In its first stage, a hardening precipitate can be considered a one-dimensional (1D) crystal containing MxNxQ unit cells (where the cell number in one width direction M = 1 and that in another width direction Q = ∼1 to 2, whereas that in length N equals the needle length divided by 0.405 nm) (Fig. 3B). A 20-nm-long needle contains ∼50 to 100 cells. The main feature of such 1D crystals is the Si double columns. They are not only the common components of different precipitates, but also the stable components of an evolving precipitate (i.e., they act as atomic pillars of the precipitate). All columns between the Si2 pillars can change.

We propose that the hardening precipitates have a dynamic structure, which changes upon annealing but does not break down its skeleton of Si2 pillars. Its initial format can be assumed as Mg2Si2Al7 rather than as the deduced Mg2Si2Si0.6Al6.4, where Si2 indicates the Si2 pillar. Its dynamic composition can be described as [Mg1-u-v(SiuAlv)]2Si2[Al1-x-y(SixMgy)]7 [0 < u(t) < ½, ν(t) ≈ 0, 0 < x(t) + y(t) < 1, where t is annealing time]. Compositionally, the particle starts with [Mg]2Si2[Al]7 and then undergoes a continuous substitution of the Al and (partially) Mg atoms by other atoms, until it reaches Embedded Image. The energy gradients for a compositional change and the local availability of Si/Mg atoms drive the evolution. This model implies that each nanoprecipitate has its own composition and its own stages in evolution.

To qualitatively understand some aspects of the structure dynamics of such nanoparticles, we performed first-principles calculations without directly including the particle-matrix interactions (24, 25). Figure 4A plots the bulk formation enthalpies (Eb) and the volume expansion ratios (VERs) for different structures against their Al content. The larger the VERs, the higher the particle-matrix interaction energies (Ei) (including the interface-energy increase due to the mismatch of atom bonds). VERs, therefore, qualitatively indicate the relative levels of Ei for these structures (if embedded in the matrix). At least for a sufficiently long precipitate, ΣEb (total Eb) is dominant over ΣEi (total Ei), because such particles would not have existed otherwise. Figure 4A shows that (i) the evolution is indeed energetically favored. (ii) Mg2Si2Al7 has the highest Eb but a minimal mismatch with the matrix; i.e., it is least effective for strengthening the alloys [supporting online material (SOM) text]. (iii) Eb markedly decreases, first upon the change from Mg2Si2Al7 to Mg2Si3Al6, and then upon the appearance of β″-type structures (the P1, P2, and P3 structures shown in Fig. 4), in which two specific (AlMg) atoms in the unit cell are shifted by bGP/2 (Fig. 3A).

Fig. 4.

The evolution of a Si2 pillar–based nanoprecipitates upon annealing. (A) VERs with respect to the Al lattice and the bulk formation enthalpies (Eb) with respect to the solid solution, calculated for a few ordered structural configurations plotted against the Al content in the precipitate (atom %). The total energy calculations were performed with the use of the first-principles Vienna ab initio simulation package (VASP) code (24, 25). (B) Schematic illustration of four featured stages of an evolving nanoprecipitate with respect to the hardness profile of the alloy being annealed at 180°C.

The reality is more complex, in that the precipitates start with a small size and the compositional evolution is accompanied by a morphology (length) change (Fig. 1, A and B). This is understandable. For any particle with a certain structure there is a critical size (length) below which ΣEi is dominant over ΣEb and the particle is unstable. Because it is mostly coherent with the matrix, Mg2Si2Al7 has the smallest critical length of all. Once stabilized, a Mg2Si2Al7 particle can grow in one dimension (because along bGP the particles are most coherent with the matrix), but its ΣEb gain will be small (Fig. 4A). However, if the 1D growth is combined with the compositional change from Mg2Si2Al7 to Mg2+xSi2+yAl7-x-y (1 < x + y < 3), the substantial lowering in ΣEb will allow the particle to grow rapidly. It is not surprising that a short Mg2Si27 particle will not evolve to β″ without a substantial growth in advance, given that Mg5Si6 (β″) has the largest critical length of all.

Further compositional changes in a needle-like particle result in the development of β″-type structures. In this stage, because all changes from P1 toward β″ markedly lower Eb (Fig. 4A), one segment of the needle would proceed with further changes rather than wait for the rest of the needle to develop the same composition. At this point, four featured stages of the evolving nanoprecipitate can be identified (Fig. 4B), which is in agreement with the alloy's hardness profile (SOM text).

The dynamic behavior of the nanoparticles changes with temperature. At lower temperatures (e.g., ∼70° to 80°C), the particle's critical lengths at different compositions become larger, such that the nucleation at Mg2Si2Al7 and particularly the further evolution to Mg2+xSi2+yAl7-x-y become slower, or even impossible. This prevents the rapid 1D growth and therefore the further evolution toward β″. At room temperature, even the particle's nucleation at Mg2Si2Al7 becomes impossible. Hence, for automotive application, our study suggests that the particles at Mg2Si2Al7 and at Mg2+xSi2+yAl7-x-y are key particles and are directly responsible for the quick-bake hardening response of AlMgSi alloys. Appropriate pre-aging should nucleate a large number of Mg2Si2Al7 particles but prevent their further evolution. Two mechanisms coexist for quickly strengthening the alloys upon a second heating: (i) the quick evolution of the nuclei from Mg2Si2Al7 to Mg2+xSi2+yAl7-x-y (leading to effective obstacles for dislocation movement), and (ii) the 1D growth of the particles (leading to a rapid increase of the total precipitate volume fraction) (SOM text).

We show that upon annealing, the hardening nanoprecipitates in AlMgSi alloys undergo a rather complex evolution, involving changes in composition, structure, and morphology. Yet, they do not break down their Si2 pillar skeleton. We analyze two complementary points of view to provide a complete explanation of these nanoparticles and their evolution. On the one hand, we consider them as dynamic objects in a nonequilibrium evolution process, analogous to a living process, involving nucleation, growth, and maturation. A hardening nanoprecipitate undergoes various physical changes with age (annealing time), but it remains fundamentally the same object because of its stable identity—the Si2 pillar skeleton. On the other hand, we treat the nanoparticles as distinct phases in a series of transformations; different particles are formed in different stages, and one may transform into another (if energetically favored). This approximation allows an analysis on the driving forces for the evolution. Additional computer power will allow direct incorporation of the particle-matrix interactions in energy calculations (requiring the inclusion of ∼104 atoms in the particle-matrix system), which will lead to an even better understanding of the nanoparticle evolution.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 and S2

Table S1


References and Notes

View Abstract

Navigate This Article