coexistence of clusters, gpb zones, s″-, s′- and s-phases in an al–0.9% cu–1.4% mg alloy
TRANSCRIPT
COEXISTENCE OF CLUSTERS, GPB ZONES, S0-, S '- AND
S-PHASES IN AN Al±0.9% Cu±1.4% Mg ALLOY
A. CHARAI 1{, T. WALTHER{ 1, C. ALFONSO 1, A.-M. ZAHRA 2 and C. Y. ZAHRA 2
1Laboratoire de Me tallurgie, EDIFIS, UMR 6518 du C.N.R.S., Faculte des Sciences et Techniques deSt Je roà me, F-13397, Marseille, cedex 20, France and 2Centre de Thermodynamique et de
Microcalorime trie, C.N.R.S., 26, rue du 141eÁ me R.I.A., F-13331, Marseille, cedex 3, France
(Received 11 February 1999; received in revised form 7 November 1999; accepted 7 November 1999)
AbstractÐThe precipitation phenomena in an Al±0.9 at.% Cu±1.4 at.% Mg alloy were studied by high res-olution electron microscopy and calorimetry in order to clarify the transformation sequence. This papershows for the ®rst time that, at around 2008C, there is coexistence of four di�erent phases: (i) partiallyordered nanometer-sized Mg-rich clusters of ellipsoidal shape, (ii) ordered GPB zones of one monolayerthickness, (iii) monoclinic semi-coherent S0 phase, and (iv) orthorhombic semi-coherent S ' which evolvesinto the incoherent equilibrium S phase of Al2CuMg. At room temperature, Mg-rich clusters precipitatebefore Cu±rich GPB zones. Clusters are considered to be precursors of the S ' phase, whereas GPB zonesmay transform into S0. New lattice parameters for S0 and S ' are proposed. 7 2000 Acta Metallurgica Inc.Published by Elsevier Science Ltd. All rights reserved.
Re sumeÂÐLes phe nomeÁ nes de pre cipitation dans un alliage Al±0.9 at.% Cu±1.4 at.% Mg ont e te e tudie spar microscopie e lectronique aÁ haute re solution et par calorime trie. Nous montrons, pour la premieÁ re fois,qu'aux environs de 2008C, il y a coexistence de quatre phases: (i) des amas de taille nanome trique, partiel-lement ordonne s, riches en magne sium et de forme ellipsoõÈ dale, (ii) des zones GPB ordonne es sur un planatomique, (iii) la phase S0 semi-cohe rente de structure monoclinique, et (iv) la phase S ' semi-cohe rente destructure orthorhombique. Cette dernieÁ re phase e volue vers la phase d'e quilibre S (Al2CuMg) incohe rente.A tempe rature ambiante, les amas (riches en magne sium) pre cipitent avant les zones GPB (riches encuivre). Nous montrons que les amas peuvent eà tre conside re s comme les pre curseurs de la phase S ' alorsque les GPB se transforment en S0. En®n, nous proposons de nouveaux parameÁ tres de maille pour S0 etS '. 7 2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved.
ZusammenfassungÐAusscheidungsphaÈ nomene in einer Al±0.9 at.% Cu±1.4 at.% Mg Legierung wurdenmittels hochau¯oÈ sender Transmissions-Elektronenmikroskopie und Kalorimetrie untersucht, um die Trans-formationssequenz der verschiedenen Phasen zu klaÈ ren. In dieser Arbeit wird zum ersten Mal gezeigt, dabbei einer Temperatur von ca. 2008C vier verschiedene Phasen gleichzeitig existieren koÈ nnen, naÈ mlich: (i)teilgeordnete Mg-reiche Cluster von ellipsoider Gestalt, (ii) geordnete GPB-Zonen von einer MonolageDicke, (iii) monokline, semi-kohaÈ rente S0-Phase, und (iv) orthorhombische semikohaÈ rente S '-Phase, welchein die inkohaÈ rente Gleichgewichtsphase S (Al2CuMg) uÈ bergeht. Bei Raumtemperatur werden Mg-reicheCluster vor den Cu±reichen GPB-Zonen ausgeschieden. Die Cluster werden als VorlaÈ ufer der S '-Phaseangesehen, wohingegen die GPB-Zonen S0 bilden koÈ nnen. FuÈ r S0 und S ' werden neue Gitterparametervorgeschlagen. 7 2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved.
Keywords: Aluminium; Aging; Phase transformations; Transmission electron microscopy (TEM); Di�eren-tial thermal analysis (DTA)
1. INTRODUCTION
The phenomenon of age hardening in an Al±Cu±Mg alloy was discovered early in this century and
led to the development of several types of duralu-min based alloys. It is therefore astonishing that the
precipitation sequence from ternary Al±based solid
solutions containing Cu and Mg in a mass ratio of
about 2.2 (pseudo-binary alloys) is still highly con-
troversial. According to Silcock [1], (probably)
ordered clusters of Cu and Mg, called GPB zones,
appear as ®rst precipitates in the shape of discs on
the {100} planes of the matrix. She could neither
con®rm the zone model published by Gerold and
Haberkorn [2] nor the occurrence of a metastable
coherent S0 phase described by Bagaryatsky [3] and,
Acta mater. 48 (2000) 2751±2764
1359-6454/00/$20.00 7 2000 Acta Metallurgica Inc. Published by Elsevier Science Ltd. All rights reserved.
PII: S1359 -6454 (99 )00422 -X
www.elsevier.com/locate/actamat
{ To whom all correspondence should be addressed.
{ Present address: Institut fuÈ r Anorganische Chemie,
UniversitaÈ t Bonn, D-53117 Bonn, Germany.
later on, by Alekseev et al. [4], but instead proposedthe formation of GPB [2] zones besides metastable
S ' and stable S (Al2CuMg) during ageing above2008C. The semicoherent S ' phase is only a slightlydistorted version of S [1], so that many authors do
not make any distinction between these two orthor-hombic phases. Cusiat [5] detected S0 in an Al±1.2at.% Cu±1.6 at.% Mg alloy, but its structure dif-
fered from that given by Bagaryatsky [3].According to Shih et al. [6], S0 is a partially orderedversion of the GPB zones, and its fast in situ pre-
cipitation from GPB zones sets in during heating ofan Al±1.1 at.% Cu±1.5 at.% Mg alloy.Quite recently, Ringer et al. [7±9] investigated an
Al±1.1 at.%Cu±1.7 at.% Mg alloy and proposed a
new precipitation sequence based on the results oftheir atom probe ®eld ion microscopy (APFIM),high resolution electron microscopy (HREM) and
microbeam electron di�raction studies. The rapidinitial hardening reaction at temperatures >1008Cwas attributed to the formation of Mg±Mg, Cu±Cu
and Cu±Mg co-clusters; GPB zones probablynucleated thereon were detected only late in theageing process (r100 h at 1508C) and were con-
sidered to be responsible for the second hardnessincrease. No S0 phase was observed; S ' was takento be identical to the equilibrium precipitate.In a short note [10], we presented comments on
the articles of Ringer et al. in favour of the for-mation of GPB zones after rapid initial clusteringwhen keeping an alloy of similar composition at
temperatures <2508C, including room temperature(RT); we also con®rmed the existence of the S0phase which, together with the S ' phase, causes the
second hardening stage. In response to these com-ments, Ringer et al. [11] maintained their interpret-ations on the basis of new experiments. The presentpaper intends to examine more closely the con-
ditions of formation, evolution and dissolution ofthese phases, using calorimetric measurements andHREM studies coupled to Fourier analysis. The
concentration of lattice defects is varied by applyingdi�erent quenching rates and by adding a smallamount of Zr to the ternary alloy. Two step ageing
treatments are used to explore the in¯uence onemetastable phase exerts on the other. It is con®rmedthat di�erential scanning calorimetry (DSC) curves
are, in general, excellent indicators of the phasespresent after di�erent ageing treatments [12].
2. EXPERIMENTAL
The investigations were carried out on two tern-ary alloys having the following compositions and amass ratio Cu/Mg of 1.6:
. alloy A: 0.87 at.% (2.03 mass%) Cu, 1.44 at.%(1.28 mass%) Mg, besides 0.002 at.% Sn, 0.006at.% Si, 0.001 at.% Zn;
. alloy B: 0.90 at.% (2.1 mass%) Cu, 1.46 at.%
(1.3 mass%) Mg, 0.03 at.% (0.09 mass%) Zr,
besides 0.04 at.% Si, 0.01 at.% Fe, 0.01 at.%
Ti.
Some experiments were also performed on an Al±
0.85 at.% (1.97 mass%) Cu binary alloy.
The cast ingots of the alloys were cold rolled,
homogenized and machined into 1 mm thick
discs. Quite generally, solution treatment was per-
formed for at least 0.5 h at 5258C and followed
by air cooling or water quenching to 208C (RT);
direct cooling to the ageing temperature was also
employed. Pre-ageing treatments were accom-
plished in metallic furnaces.
For the microscopic observations, specimens
were mechanically ground down to about 0.2 mm
thickness. Electrolytical polishing was carried out
on foils 3 mm in diameter with the help of a
Struers Tenupol. A double jet of a 3:1 metha-
nol:nitric acid solution performed the thinning at
about ÿ208C until formation of a small hole. A
Jeol 2010 F, URP 22 microscope operated at
200 kV and equipped with a ®eld emission gun
(Cs=0.5 mm and smallest spot size=0.5 nm)
and a detector for X-ray analysis (Energy
Dispersive Spectroscopy, EDS, type Kevex
Quantex Delta Plus) was used.
Two sorts of calorimetric studies were underta-
ken. Isothermal heat ¯ow measurements were car-
ried out with the help of di�erential Tian-Calvet
microcalorimeters which possess excellent base line
stabilities as well as high sensitivities (down to the
microwatt); experiments were stopped after 4±7
days (d). Direct cooling in air to the ageing tem-
perature was realized by ®xing a furnace above a
calorimeter. The heat e�ects vdH/dtv measured over
time represent the kinetics of the process under
study, as no correction for thermal inertia is necess-
ary in the case of long lasting reactions. In each
run, 20 discs, 15 mm in diameter, were examined
using an equal mass of high purity aluminium as
reference material. Initial heat e�ects due to the pla-
cing of the specimens in the calorimeter were esti-
mated with pure aluminium discs and subtracted
from the experimental curves.
A DuPont di�erential scanning calorimeter (DSC
of the heat-¯ux type), model 990, was employed in
order to follow the evolution of a variety of struc-
tural states during heating at di�erent rates ranging
from 0.5 to 508C/min (in general at 208C/min
between ÿ20 and 5208C). The 1 mm thick speci-
mens had a diameter of 5.9 mm. The base lines
obtained by scanning pure aluminium discs on the
reference and the laboratory sides were reproducible
over months within a band of 1 mW. They were
slightly curved and could be represented by poly-
nomial expressions of 2nd and 3rd order. The heat
capacity di�erence between the alloy and pure alu-
minium was well within the band given by the base
line variations.
2752 CHARAI et al.: CLUSTERS, ZONES AND PHASES
Fig. 1. Lattice images (a, c, e) and moduli of their Fourier transforms (b, d, f) of GPB zones in alloyA. (g), (h) are Fourier ®ltered versions of (e).
CHARAI et al.: CLUSTERS, ZONES AND PHASES 2753
3. RESULTS
3.1. Electron microscopy investigations
Alloys A and B were examined after di�erentheat treatments performed within the calorimeters
using the techniques of HREM, microdi�ractionand EDS. In the case of microdi�raction and X-ray
analysis, the electron beam is focused down to 0.5±1 nm in size and induces strong damage such as
contamination and defect creation that, in extremesituations, can produce a hole in the sample. Theresults obtained under these conditions might not
be representative of the initial alloy. HREM images,recorded using a slow scan camera with around 0.1 s
Fig. 2. Lattice images (a, c, e) and moduli of their Fourier transforms (b, d, f) of small clusters. Theclusters in (a, c) and the ordered clusters in (e) presumably represent the same particle type, viewed
along two perpendicular h100iAl directions.
2754 CHARAI et al.: CLUSTERS, ZONES AND PHASES
exposure time, and their Fourier transforms are bet-
ter suited for this investigation, since the total elec-
tron dose can be reduced signi®cantly.
We limit ourselves here to the description of a
very interesting microstructural state attained after4 h hold at 2008C following water quenching of
alloy A from 5258C. We distinguish ®ve di�erent
types of precipitates which will be described in the
following in the order of their typical sizes. A more
detailed precipitation sequence is then proposed in
Section 4. Our nomenclature follows that estab-lished for structures of precipitates in this alloy sys-
tem from di�raction techniques.
In Figs 1±5 lattice images and the moduli of their
Fourier transforms (di�ractograms) are shown for a
[100]Al orientation. They show that GPB zones,
Fig. 3. Lattice images (a, c) and moduli of their Fourier transforms (b, d) of coherent precipitates ofthe S0 phase. In (b) the reciprocal vectors of the monoclinic S0 unit cell are marked by arrows, assum-
ing a primitive lattice. (e) is the sum of (b), (d).
CHARAI et al.: CLUSTERS, ZONES AND PHASES 2755
clusters and precipitates of the S0, S ' and S phases
can all co-exist in aged AlCuMg alloys. We point
out that this is the ®rst direct observation by latticeimaging in a transmission electron microscope of
the simultaneous presence of clusters, ordered GPB
zones and precipitates denoted as S0 in this alloysystem.
Figure 1 shows lattice images of GPB zones
which consist of thin, fully coherent platelets on
{001}Al matrix planes. For better visibility their
habit planes are marked by arrows in the images ofFigs 1(a), (c) and (e). Their Fourier transforms
exhibit di�use streaks between the lowest indexed
Bragg spots of the Al matrix, in agreement withelectron di�raction patterns reported in [13, 14].
These platelets are fairly di�cult to detect even at
high magni®cations, due to their small size (widthof 421 nm, thickness of 0.2 nm). A lack of obser-
vation of these only faintly visible platelets in stan-
dard electron microscopic imaging is certainly noproof of their non-existence. Figure 1(e) shows such
a platelet at higher magni®cation. Filtering the
Fourier transform, the modulus of which is dis-
played in Fig. 1(f), may be achieved in two ways.One can either suppress all the Bragg spots of the
Al matrix lattice before the inverse transform is car-
ried out. The real part of the result is shown inFig. 1(g) and shows a line of weak alternating
black-and-white contrast at the location of the GPB
zone. This demonstrates the existence of a (001) fre-quency which is probably due to ordering of Cu
and Mg/Al atoms on alternating (002)Al matrix
planes, similar to recently published results on theAlMgSi system [15]. Applying instead narrow slit
masks to the Fourier transform of the image such
that mainly the intensity in the di�use streaks along
2[010]Al contributes and calculating the modulusof the resulting image yields Fig. 1(h), from which
it can be concluded that the di�use scattering orig-
inates almost completely from a thin platelet asingle monolayer (0.2 nm) in thickness.
In Figs 2(a)±(d) we observe clusters of probably
Fig. 4. Lattice image (a) and modulus of its Fourier transform (b) of a semi-coherent precipitate of theS ' phase. In (b) the reciprocal vectors of the S ' unit cell are marked by arrows.
Fig. 5. Lattice image (a) and modulus of its Fourier transform (b) of a precipitate of the S phase. In(b) the reciprocal lattice vectors of the S unit cell are marked by arrows.
2756 CHARAI et al.: CLUSTERS, ZONES AND PHASES
atoms and vacancies in the matrix which are very
similar to the ones described by Radmilovic et al.[16]. They yield no clear discrete di�raction spotsbut give rise to di�use spots in the range of 2.4±
2.8 nmÿ1. The diameter of these clusters in the(100)Al matrix plane has been measured as 1.7 20.3 nm. This matches remarkably well the width
along [001]Al of the ordered clusters depicted inFig. 2(e). Therefore, it can be suggested that Figs
2(a), (c) and (e) represent the same clusters viewedalong two perpendicular h100iAl directions. Thisview is corroborated by the strong contrast the clus-
ters exhibit in Figs 2(a) and (c). For small sphericaldisordered clusters within a foil we would expect amuch fainter lattice fringe contrast. If, on the other
hand, these clusters are ellipsoidally shaped, withextension along [001]Al and ordering parallel to
(011)Al planes, then the experimental observationswill be understood easily. Let us point out thatthese clusters were analyzed using EDS and exhibit
an enrichment in Mg and Cu as shown in [10, 17]where they were mistakenly labeled as ternary GPBzones. The Mg/Cu ratio was found to vary from
one cluster to another making the exact chemicalcomposition di�cult to determine. Note that, in
any case, these clusters exhibit Mg enrichment.Given their size, these clusters should not be takenas the larger S0, S ' and S precipitates, also observed
in this specimen.Figure 3 depicts two S0 precipitates oriented at
908 with respect to each other. As both variantsoccur within a single grain, di�raction patterns inX-ray studies always reveal squares of discrete spots
around the {110}Al matrix positions, as in Fig. 3(e)which was obtained by superimposing Figs 3(b) and3(d). If these spots were blurred in di�raction exper-
iments from larger sample volumes, this could bemistaken for the {110} spots of a cubic P or I
(b.c.c.) phase. The measured spacings in the Fouriertransforms correspond to lattice vector lengths of0.25 and 0.32 nm which are similar to the (112)S
and (111)S re¯ections of the well-known orthor-hombic S phase. An interpretation in terms of var-iants of the S phase is, however, not possible, as
Gupta et al. [18] have shown that for all orientationvariants of the S phase, the only Bragg re¯ections
close to h100iAl zone axes are of {112}S and{131}S type. We can index the di�ractograms ifinstead we de®ne a monoclinic unit cell with an
angle b=91.720.58. The images and di�ractogramsthen indicate a periodic twinning of S0 on planesparallel to (020)Al, i.e. perpendicular to the habit
planes of the precipitate. The monoclinicity agreeswell with a value of 91.38 suggested in one model
for S0 [19] but contradicts the value of 988 predictedin another [20]. For a primitive (P) lattice type andsemi-coherence, the unit cell parameters measured
would be aS0=0.320 2 0.008 nm, bS0=aAl andcS0=0.254 2 0.003 nm. For a C lattice type, thespots marked in Fig. 3(b) by arrows would have
Miller indices of (200)S0 and (001)S0 instead, andthe value of aS0 would be only 0.16 nm. In any
case, this measured unit cell would be much smallerthan earlier reports suggested [19]. Our result de®-nitely excludes the possibility that the S0 phase
could be one variant of the S phase, as suggestedby Ringer et al. [11] in their response to our com-ments [10].
Figure 4 shows a precipitate which is semi-coher-ent with the matrix as can be seen from the sur-rounding dark contrast due to strain. The
di�ractogram is in accordance with an orthorhom-bic C lattice (angle measured between the axesmarked: 90.4 2 0.88) with unit cell dimensions ofbS '=0.89 nm and cS '=0.76 nm [the visible spots are
of (020)S ' and (001)S ' type]. The semi-coherence isre¯ected in the assigned unit cell parameters di�er-ing from those of the stable (and also orthorhom-
bic) S phase measured below, while the unit cellvolume is identical to that of the S phase within theexperimental error bars. The unit cell parameters
determined by us di�er signi®cantly from thosereported by other groups (see, e.g. [21]). The S ' andAl lattices are rotated by about 268 such that the
(023)S ' and (041-)S ' spots almost coincide with the
(002)Al and (020)Al spots, respectively, the latterappearing extended as expected for a strainedmatrix lattice.
The precipitate of the stable S phase shown inFig. 5 lacks this strain contrast in the images: thereare no dark rims around the precipitates as in Fig. 4
and in the Fourier spectrum the spots due to thematrix are clearly separated from those due to theprecipitate. The lattice parameters measured for this
orthorhombic C lattice are bS=0.921 2 0.006 nmand cS=0.719 2 0.012 nm (aS 1 aAl may beassumed) and thus are in good agreement with var-ious literature data [22].
The present HREM investigation demonstratesfor the ®rst time the simultaneous existence of fourdi�erent phases in an aged AlCuMg alloy: ordered
GPB zones on {001}Al, clusters ordered parallel toh011iAl, semi-coherent monoclinic S0, semi-coherentorthorhombic S ' and incoherent orthorhombic S
precipitates, the latter two phases di�ering only inthe degree of coherency and thus the unit cell par-ameters. It follows from structural considerations
that clusters may be precursors of the S ' phasewhich transforms into S, whereas GPB zones maytransform into S0 (such as in binary Al±Cu alloysGP zones transform into y0). Heterogeneous pre-
cipitation of S ' and S on dislocations was of no im-portance in this well annealed alloy.
3.2. Isothermal calorimetric studies
The decomposition of solid solutions A and Bquenched into water or cooled in air for 5 minwas examined in a calorimeter at 308C in order
CHARAI et al.: CLUSTERS, ZONES AND PHASES 2757
to study the transformation near room tempera-ture (RT). Figure 6 shows the heat ¯ow evol-
ution during the ®rst day of ageing. Each curvemay be divided into two parts, a rapid initialreaction and a second one which proceeds more
or less slowly through a minimum; the last beha-viour is characteristic of a nucleation and growthprocess [23]. In the case of the Zr-bearing alloy,
the minimum is much smaller than for sampleA. This minimum appears sooner for the aircooled sample compared to the water quenched
one. The heat output accompanying the ageingof a water quenched Al±0.9 at.% Cu solid sol-ution is indicated for comparison (curve 5).Following Ringer et al. [7±9], precipitation starts
with the formation of clusters. As the parasiticheat e�ect due to the introduction of the speci-mens into the calorimeter lasts for about 30 min,
we are unable to con®rm this unequivocally. Thesecond process, however, points to GPB zoneformation, although no pre-precipitates could be
detected by HREM after ageing at RT for 7months, which suggests that their sizes are smal-ler than a few atomic columns. We were also
unable to observe GP zones in the RT aged Al±0.9 at.% Cu alloy, whereas we had no problemidentifying them in an Al±1.7 at.% Cu alloy[17].
The enthalpy production during the ®rst halfhour cannot be evaluated correctly, as the parasiticheat e�ects due to the introduction of the specimens
are not reproducible. It is, however, certain that itis greatly reduced in the presence of Zr. The follow-
ing enthalpy values (DH) are derived between 0.5and 96 h of ageing:
alloy A: water quenched ÿ320 J/mol; air cooled
ÿ170 J/molalloy B: water quenched ÿ350 J/mol; air cooledÿ225 J/mol.
Furthermore, the precipitation kinetics of meta-stable phases were examined after water quenchingin steps of 108C between 150 and 2508C, and the in-
¯uence of pre-ageing treatments on subsequent age-ing at higher temperatures was assessed. Thecalorimetric curves start with an exo- or an
endothermal e�ect (the latter for ageing tempera-tures between 190 and 2308C), then pass more orless rapidly through a minimum the position of
which depends on several parameters. This mini-mum corresponds to homogeneous S ' formation(see Section 4.3). As examples, some curvesobtained at 1908C on alloy A (Fig. 7) are given for
the ®rst 4 d.Curve 1 in this ®gure pertains to quenched solid
solutions immediately transferred into the calori-
meter at 1908C; the other curves are relative tosamples maintained beforehand at 20 (2), 100 (3) or1508C (4) for di�erent periods. All pre-ageing con-
ditions below 1508C increase the absolute value ofthe minimum, but neither modify its positionstrongly (at around 7 h) nor its ®nal enthalpy value
Fig. 6. Heat ¯ows accompanying precipitation at 308C of alloy A water quenched (1) or air cooled (2),of alloy B water quenched (3) or air cooled (4) and of binary Al±Cu alloy water quenched (5).
2758 CHARAI et al.: CLUSTERS, ZONES AND PHASES
after 4 d ageing amounting to ÿ(3452 20) J/mol.Long pre-ageing periods (>2 d) at 1508C shift this
minimum to shorter times and diminish DH; pre-ageing at 1708C has an even stronger e�ect. Heat¯ow minima still occur at around 1 h up to 2408C,whereas at and above 2508C, the experimentallyexploitable curves are regularly falling with time. At2208C, DH for alloy A is about ÿ600 J/mol
between 0.5 and 96 h, and ÿ550 J/mol for B.About the same values are obtained after coolingthe solid solution directly to 2208C, but the precipi-
tation kinetics is quite di�erent.
3.3. Non-isothermal calorimetric studies
Figure 8(a) was obtained by heating alloy A in
the as-quenched state at 208C/min. The heat ¯owcurve for alloy B (curve 2) shows the same generalpattern as for A, but is shifted to higher tempera-tures [24]. To give an idea, the heats accompanying
the low (mainly GPB zone formation) and hightemperature exothermal (mainly S ' and S for-mation) e�ects for A are equal to ÿ(255220) and
ÿ(360 2 20) J/mol, respectively. In the case of B,the corresponding values are ÿ(250220) and ÿ(340230) J/mol. A discontinuity appears around 2308Cin both alloys; it is attributed mainly to GPB zone4 S0 transformation during heating in the DSC ap-paratus. In fact, HREM observations show the pre-sence of clusters and S0 when heating the alloy at
20 K/min up to 2358C. The last exo- and endother-
mal e�ects are highly asymmetric, especially in the
case of B, and seem to result from an overlap of
two peaks (S ' and S precipitation and dissolution,
respectively). An additional small endothermal
e�ect characterizes all curves for alloy B before the
attainment of the solid solution; it is presumably
due to dissolution of some y (Al2Cu) phase
nucleated on b ' (Al3Zr) particles [25].
The heat evolution in alloy A was studied more
closely up to 2008C at di�erent heating rates (0.5±
508C/min) with the view of determining the acti-
vation energy for GPB zone formation. Each curve
starts with a small heat output [see arrow in
Fig. 8(a)] whose importance increases with falling
scan rate; it may correspond to the end of rapid
cluster formation. The heat value of the main pre-
cipitation e�ect decreases at 508C/min, but remains
approximately constant at the lower rates. A
Kissinger analysis [26] yields an activation energy of
67 kJ/mol.
After 4 d ageing of alloy A at 20, 100 or 1508C[curves 1, 2 and 3 of Fig. 8(b)], the ®rst exothermal
e�ects disappear completely; the strong exothermal
e�ects above 2708C vary with heat treatment. The
minima of the latter decrease in the order 20, 100,
1508C; the corresponding heats are ÿ360, ÿ360 and
ÿ310 J/mol. Also shown are the DSC scans on
samples aged for 4 d at 1708C (curve 4) and 1908C(curve 5); dissolution of the S0, S ' and S phases
occurs. Discontinuities in the ®rst endothermal
Fig. 7. Calorimetric curves obtained at 1908C on alloys A after water quenching (1) followed by ageingat 208C for 7 days (2) or at 1008C for 4 days (3) or at 1508C for 14 days (4).
CHARAI et al.: CLUSTERS, ZONES AND PHASES 2759
e�ect were only observed for ageing times less than
about 2 d at 1708C, 4 h at 1908C or 2 h at 2008C.Figure 9 demonstrates that each heat treatment
yields DSC curves which correlate with the meta-
stable phases present before the run as well as their
evolution during the run. It illustrates the strong in-
¯uence of sample preparation before ageing at
2008C for 4 h (curve w, water quenching to 208C;curve a, water quenching and ageing for 7 d at RT;
curve d, direct cooling from 525 to 2008C). Quite
generally, preageing enhances the attainment of the
equilibrium state with some S0 still being present,
while direct cooling promotes S0 formation. Curve
d ' refers to an alloy directly cooled to 2408C and
Fig. 8. (a) DSC scans, at 20 K/min, of alloys A (1) and B (2) water quenched from 5258C. (b) DSCcurves of alloy A quenched and aged for 4 days at 208C (1), 100 (2), 150 (3), 170 (4) or 1908C (5).
2760 CHARAI et al.: CLUSTERS, ZONES AND PHASES
kept there for 5 h; the presence of S0 and S ' beforethe DSC scan was checked by HREM. Curve d isof particular interest as the exothermal e�ect whichaccompanies S ' and S formation during heating is
only slightly smaller than the last dissolution e�ect.In agreement with isothermal studies, it con®rmsthat mainly the S0 phase, besides some GPB zones,is present after direct cooling from 525 to 2008Cand 4 h ageing. S0 gets dissolved rather than trans-formed into S ' and S, which would yield anexothermal heat e�ect.
4. DISCUSSION
4.1. Clusters and GPB zones
The combined HREM and calorimetric studies
allow us to propose a new precipitation sequence inAl alloys containing low amounts of Cu and Mg.Inhomogeneities in solute atom and vacancy con-
centrations may favour the appearance of two sortsof pre-precipitates at low temperatures, ellipsoidalMg-rich clusters containing Cu atoms andvacancies, and Cu±rich monolayer-thick GPB zones
containing some Mg atoms. It is well establishedthat the binding energy between Mg atoms andvacancies is higher than that between Cu atoms and
vacancies [27, 28]; the activation energy for Mg dif-fusion in Al is also lower than for Cu di�usion [see29]. This suggests that Mg±Mg clusters are the ®rst
to appear followed by Cu±Cu and Cu±Mg co-clus-ters. After appropriate heat treatments, these clus-ters induce an initial hardening e�ect for which the
large size of the Mg atom seems to be a crucial fac-tor. In a previous publication [10], the authors con-cluded from EDS analyses that clusters (shown inFig. 3 of the cited article and erroneously taken for
GPB zones) are richer in Mg than in Cu. Further
investigations of a signi®cant number of these clus-ters show that the Cu/Mg ratio is not actually con-stant. This variation may be related to the vacancy
concentration present in such clusters. More workis underway to con®rm this interpretation.More slowly, monolayer GPB zones appear
which resemble those in binary Al±Cu alloys butmay also contain some Mg. Their formation at308C is accompanied by a very strong heat e�ect
(Fig. 6) and is a nucleation and di�usion controlledgrowth process [30]. At 2008C, they are only 4 21 nm wide, which explains their di�cult or imposs-
ible detection at lower temperatures. Yet the X-raydi�raction study by Silcock [1] at RT con®rmedtheir existence and indicated probable order.
Lattice parameter measurements by Rajan [31]performed on a similar alloy during RT ageing alsosuggest that up to 1 h, Mg atom clustering prevails
and that at longer times mainly Cu atom clusteringoccurs. DSC studies on quenched solid solutions[Fig. 8(a)] show a ®rst heat evolution probably due
to cluster formation, the beginning of which is lost,followed by a strong heat output of the same orderas that measured during isothermal ageing and
which corresponds to GPB zone formation.The hardening processes of three Al±Cu±Mg
alloys with di�erent Cu and Mg contents also show
a two-step precipitation process at 608C [32]. In theinitial stage of ageing (up to 1 h), mainly the Mgcontent controls the increase in the yield strength.
However, for long ageing times, the in¯uence of theCu content becomes dominant.The formation of ternary compounds in the Al±
Cu±Mg system hints at a strong interaction betweenCu and Mg atoms. A reduction in strain energyshould result when these two solute atoms occupy
Fig. 9. DSC curves obtained at 20 K/min on solid solutions of alloy A having undergone: w, waterquenching and ageing for 4 h at 2008C; a, water quenching, ageing for 7 days at 208C and for 4 h at
2008C; d, direct cooling to 2008C and 4 h hold; d ', direct cooling to 2408C and 5 h hold.
CHARAI et al.: CLUSTERS, ZONES AND PHASES 2761
adjacent positions, i.e. for an ordered precipitate,
since the lattice expansion due to the atomic size of
Mg is counterbalanced by a contraction of similar
magnitude due to Cu [33]. Hence, it is not surpris-
ing that in the present HREM studies clusters with
ordering parallel to h011iAl matrix planes and GPB
zones with ordering on alternating {002}Al planes
are detected. Further investigations are being car-
ried out in order to check the degree of order in the
clusters.
In the temperature range between 120 and 2808C[Figs 8(a) and (b)], growth or dissolution of clusters
and GPB zones as well as their partial transform-
ation into the S ' and S0 phases, respectively, is
expected to occur. These overlapping phenomena
lead to complicated DSC curves as actually
observed in alloy B and in the alloy studied by
Ringer et al. [11]. It is worth mentioning that in
Al±1.7 at.% Cu [34], DSC dissolution curves on
alloys aged after quenching showed several humps,
but only one if the alloy was reverted before being
aged at 70 or 1008C, as the low vacancy concen-
tration did not favour size evolution of GP zones
and their partial transformation into y0 during heat-
ing. The DSC curves of alloy A have a simpler
form in the above temperature interval and may be
interpreted as being mainly due to GPB zone dissol-
ution and their partial transformation into S0. Thepresence of clusters and S0 at 2358C is con®rmed by
HREM.
GPB zone dissolution is clearly detected in iso-
thermal and non-isothermal calorimetric measure-
ments. If formed at RT, a strong initial
endothermal e�ect which lasts for more than 1 h is
observed at 1908C (Fig. 7, curve 2) or above 1208Cduring continuous heating [Fig. 8(b), curve 1].
Larger zones formed at 100 [Fig. 8(b), curve 2] or
1508C (curve 3) start to dissolve at increasingly
higher temperatures. A comparison of Figs 8(a) and
(b) shows that zones formed during DSC heating
after water quenching attain sizes greater than those
formed at RT.
A mean activation energy of 67 2 0.5 kJ/mol
(0.69 eV) is derived for GPB zone formation in
quenched alloy A from DSC runs at 0.5±208C/min.
For comparison, the value for the migration energy
of vacancies in pure Al is 0.6620.05 eV according
to [35]. The result compares well with Horiuchi and
Minonishi's value of 65 kJ/mol obtained form elec-
trical resistivity measurements on an Al±1.8 at.%
Cu±1.6 at.% Mg alloy [36] and with Starink's result
for industrial Al±Si±Cu±Mg alloys [37]. DSC scans
at 5±208C/min performed by Jena et al. [38] on
quenched Al±0.7 at.% Cu±0.9 at.% Mg solid sol-
utions yielded 56 kJ/mol. It may be noted that in
their study the total heat e�ect accompanying GPB
zone formation also diminished with increasing
heating rate and that the following dissolution heat
was about 20% greater, as in the present alloy [39],
mainly as a result of the initial heat loss due tocluster formation.
Air cooling and, to a stronger extent, addition ofZr which possesses a solute-vacancy binding energyof the same order as Mg (about 0.25 eV) decrease
the isothermal formation rate of GPB zones, as lessvacancies are available for the di�usion of soluteatoms. For the same reason, GPB zone formation
during continuous heating of water quenched solidsolutions is shifted to higher temperatures in alloyB. The enthalpy values between 0.5 and 96 h are
greater for alloy B than for A, as less heat lossoccurs in B during the ®rst 0.5 h ageing. The high-est value attained is about ÿ350 J/mol.
4.2. S0 phase
Monolayer GPB zones may agglomerate andform the S0 phase, just as in binary Al±Cu alloys
GP zones transform into the y0 phase. This point isexperimentally corroborated by the match betweenthe two corresponding di�ractograms: the streaks inthe di�ractograms corresponding to GPB zones
evolve towards dots with increasing thickness of S0but stay at the same position with respect to thealuminium matrix spots. The present HREM stu-
dies con®rm the existence of this metastable phasealready discovered by [3±6, 40] but describedtherein with di�erent lattice parameters. It has a
monoclinic structure, is semi-coherent with thematrix and does not represent a variant of S(orthorhombic) as proposed by Ringer et al. [11];
no possible projection and corresponding di�ractionpattern of S ®ts with our edge-on HREM images ofS0.The appearance of the S0 phase during continu-
ous heating causes an exothermal heat e�ect whichoverlaps the endothermal GPB zone (and cluster)dissolution in Figs 8(a) and (b). This was already
suggested by the authors earlier [29]. S0 seems toappear the earlier in the precipitation process thelower the previous ageing temperature, hence the
smaller the GPB zone size is. This may be under-stood by considering that the thickening process ofzones depends on the supply of solute atoms andnot on their extension in the {100}Al planes.
Hence, earlier dissolution of zones attaining subcri-tical sizes enables transformation of others into S0.The coexistence of GPB zones, S0, clusters, S ' or S
after 4 h of ageing at 2008C is possible for thermo-dynamic and kinetic reasons.No direct transformation of the S0 phase into the
S ' or S phase is expected to occur from structuralconsiderations: the structures and the epitaxial re-lationship with the matrix are di�erent. The inter-
facial energy will hinder this transformation. Curvesw and d in Fig. 9 con®rm that reversion of S0 pre-cedes new S ' formation. A comparison of these twocurves also indicates that S0 appears preferentially
2762 CHARAI et al.: CLUSTERS, ZONES AND PHASES
after direct cooling to the ageing temperature. Itseems that during the slow air cooling, GPB zone
formation, a necessary step before transformationinto S0, is favoured with respect to clustering.
4.3. S ' and S phases
According to the present HREM investigation,strain contrast accompanies the presence of semi-coherent S ' particles whereas it is absent in the
case of S. Hence, a distinction between these twophases is reasonable. New lattice parameters forS ' are proposed. They vary with the size (degree
of coherency) until they reach the value for S.This may be the origin of discrepancies encoun-tered in the literature concerning the exact lattice
parameters.The high temperature exothermal DSC e�ects
observed in Figs 8(a) and (b) are mainly due toS ' and S precipitation followed by the dissolution
of these phases at still higher temperatures. Theirslight asymmetry may be taken as indication ofan energetic di�erence between these two isomor-
phous phases resulting from di�erent strainenergy contributions. In fact, DSC measurementsare very sensitive and allow a distinction between
spherical and ellipsoidal GP zones in Al±Znalloys [41]. The ®ner grain sizes in the Zr con-taining alloy may promote the occurrence of the
equilibrium phase on grain boundaries. This mayexplain the more distinct separation into S ' andS peaks which characterizes alloy B. Lower heat-ing rates also develop this feature in both alloys.
The form of the isothermal heat ¯ow (curve 1in Fig. 7) indicates that after dissolution of someGPB zones (and probably also some clusters),
two phases appear, S0 and S '. DSC runs per-formed on samples aged for di�erent periods at1908C show that transformation into S0 is ®n-
ished in about 4 h. Hence the heat ¯ow mini-mum around 7 h is due to the precipitation ofS ' (and not S0 as erroneously proposed in [29])which transforms gradually into S at still longer
ageing times at 1908C. Its activation energy wasfound to be 13425 kJ/mol [29], which agrees wellwith the value of Jena et al. [38] derived from DSC
measurements on a similar alloy (130 kJ/mol) andmatches the activation energy for volume di�usionof Cu in Al in the presence of an equilibrium
vacancy concentration [42].Structural similarities suggest that ellipsoidal clus-
ters may transform directly into the S ' phase.
Calorimetry can only prove that there is no in¯u-ence of GPB and S0 sizes on the kinetics of S ' pre-cipitation and that re®nement of S ' occurs after lowtemperature ageing:
. In Fig. 7, the absolute value of the heat ¯owminimum increases without changing its positionwith respect to time: see curves 2 (7 d at RT) and
3 (4 d at 1008C).. Increasing the S0 particle size does not promote a
direct transformation of S0 into S ' during heat-ing, as concluded from Fig. 9, curves d (4 h at2008C) and d ' (5 h at 2408C).
. A comparison of curves 1 in Figs 8(a) and (b)shows that the S ' and S precipitation peak ishigher and narrower after RT ageing following
quenching.
Either clusters or regions rich in Mg and Cu leftbehind after GPB zone and S0 reversion and/or dis-
location loops precipitated during low temperatureageing via the collapse of vacancy clusters [43]could serve as nucleation sites for S '. Further
HREM studies are needed to clarify this point. Itmay be added that dislocations introduced by coldwork after quenching also re®ne and accelerate S 'precipitation during heating [44].
After 4 d ageing at 1508C, some S ' is already pre-sent, hence the amount of S ' precipitation duringheating [curve 3 of Fig. 8(b)] or during subsequent
isothermal ageing at 1908C (curve 4 of Fig. 7) isreduced; the time to reach the heat ¯ow minimumis also shortened in Fig. 7.
5. CONCLUSIONS
The present work opens a new way towardsunderstanding the precipitation phenomena in Al
alloys with low Cu and Mg additions, for whichcontroversial propositions and confusion of phases(clusters taken for GPB zones, S0 taken for S) exist.
It is beyond doubt that rapid initial clustering isresponsible for the strong hardening e�ect observed.We propose that it is governed by vacancyenhanced Mg di�usion and ®nally leads to three-
dimensional ordered clusters which are richer in Mgthan in Cu and probably also contain vacancies.These clusters show ordering parallel to {011}Al
planes and may directly or indirectly (via solute-enriched regions left behind their dissolution) con-stitute nuclei for homogeneous S ' formation. The
latter transforms into the stable S phase.The slower Cu atom di�usion may control the
formation of GPB zones which are ordered onalternating {002}Al planes and are one monolayer
thick such as in binary Al±Cu alloys; they probablycontain more Cu than Mg atoms and appearalready at RT according to a nucleation and growth
controlled process. Their formation kineticsdepends on the excess vacancy concentration; aircooling instead of water quenching and Zr ad-
ditions slow down the di�usion processes. Theirheat contribution is more important than that dueto cluster formation. GPB zones may transform
into the semi-coherent S0 phase whose structure ismonoclinic in agreement with earlier propositions.Both phases show reversion phenomena when trans-ferred to higher temperatures.
CHARAI et al.: CLUSTERS, ZONES AND PHASES 2763
We further propose that a slight distinction maybe made between the semi-coherent S ' and the inco-
herent S phases and give new lattice parameters forthe S0 and S ' phases. RT ageing leads to a re®ne-ment of S ' (S) precipitates at higher temperatures.
Their heterogeneous precipitation on dislocationswas not important for the heat treatments chosen.The present studies illustrate that calorimetric
measurements are sensitive to structural reorganisa-tions at the atomic level. HREM observationscoupled to localized Fourier analysis indicate for
the ®rst time the simultaneous presence of smallpre-precipitates (clusters, GPB zones) and largerphases (S0, S ' and S) at ageing temperatures around2008C. DSC curves enable us to specify the phases
present after di�erent heat treatments and to ob-serve their evolution during heating. Further workis planned in collaboration with scientists presently
engaged in the study of similar compositions.
AcknowledgementsÐWe are indebted to the Centre deRecherches de Voreppe for the fabrication of the binaryand the Zr bearing alloy, and to Austria Metall AG forthe ternary alloy.
REFERENCES
1. Silcock, J. M., J. Inst. Met., 1960±1961, 89, 203.2. Gerold, V. and Haberkorn, H., Z. Metallk., 1959, 50,
568.3. Bagaryatsky, Yu A., Dokl. Akad. Nauk SSSR, 1952,
87, 397 and 559.4. Alekseev, A. A., Anan'ev, V. N., Ber, L. B. and
Kaputkin, E. Ya, Phys. Met. Metallogr., 1993, 75,279.
5. Cusiat, F., TheÁ se de Docteur-Inge nieur, University ofRouen, 1984.
6. Shih, Han.-Cheng, Ho, New.-Jin and Huang, J. C.,Metall. Mater. Trans., 1996, 27A, 2479.
7. Ringer, S. P., Hono, K., Polmear, I. J. and Sakurai,T., Appl. Surf. Sci., 1996, 94(95), 253.
8. Ringer, S. P., Hono, K., Sakurai, T. and Polmear, I.J., Scripta mater., 1997, 36, 517.
9. Ringer, S. P., Sakurai, T. and Polmear, I. J., Actamater., 1997, 45, 3731.
10. Zahra, A. M., Zahra, C. Y., Alfonso, C. and CharaõÈ ,A., Scripta mater., 1998, 39, 1553.
11. Ringer, S. P., Caraher, S. K. and Polmear, I. J.,Scripta mater., 1998, 39, 1559.
12. Zahra, A.-M., Zahra, C. Y., Jeroma-Weiland, G.,Neuer, G. and Lacom, W., J. Mater. Sci., 1995, 30,426.
13. Shchegoleva, T. V. and Shpektor, Ye B., Phys. Met.Metall., 1983, 56, 84.
14. Takahashi, T. and Sato, T., J. Jpn. Inst. Light Metals,1985, 35, 41.
15. Matsuda, K., Gamada, H., Fujii, K., Uetani, Y.,
Sato, T., Kamio, A. and Ikeno, S., Metall. Mater.Trans. A, 1998, 29, 1161.
16. Radmilovic, V., Thomas, G., Shi¯et, G. J. and StarkeJr, E. A., Scripta metall., 1989, 23, 1141.
17. Alfonso, C., CharaõÈ , A., Zahra, C. Y. and Zahra, A.M., Proc. ICEM 13-Paris, 1994, 2, 689.
18. Gupta, A. K., Gaunt, P. and Chaturvedi, M. C., Phil.Mag. A, 1987, 55, 375.
19. Shchegoleva, T. V. and Buinov, N. N., Sov. Phys.Cryst., 1968, 12, 552.
20. Alekseyev, A. A., Ber, L. S., Klimovich, L. G. andKorobov, O. S., Phys. Met. Metall., 1979, 46, 80.
21. Yan, J., Chunzhi, L. and Minggao, Y., J. Mater. Sci.Lett., 1990, 9, 421.
22. Alekseyev, A. A., Ber, L. S. and Pavlenko, S. G.,Phys. Met. Metall., 1982, 53, 162.
23. Zahra, A.-M. and Starink, M. J., in Advanced LightAlloys and Composites, ed. R. Ciach. KluwerAcademic, Amsterdam, 1998, p. 17.
24. Zahra, A. M., Zahra, C. Y. and Lacom, W., HighTemp.±High Press., 1991, 23, 537.
25. Tite, C. N. J., Gregson, P. J. and Pitcher, P. D.,Scripta metall., 1986, 22, 1005.
26. Kissinger, H. E., Anal. Chem., 1957, 29, 1702.27. OÈ zbilen, S. and Flower, H. M., Acta metall., 1989, 37,
2993.28. Beatrice, C. R. S., Garlipp, W., Cilense, M. and
Adorno, A. T., Scripta metall. mater., 1995, 32, 23.29. Zahra, A.-M., Zahra, C. Y., Lacom, W. and
Spiradek, K., Advanced Aluminium and MagnesiumAlloys, Khan, T. and E�enberg, G., Am. Soc. Met.,1990, pp. 633.
30. Starink, M. J. and Zahra, A.-M., Thermochim. Acta,1997, 298, 179.
31. Rajan, T. V., Trans. Indian Inst. Met., 1973, 26, 72.32. Ratchev, P., Verlinden, B., de Smet, P. and van
Houtte, P., Mater. Trans., JIM, 1999, 40, 34.33. Sen, N. and West, D. R. F., Mechanism of Phase
Transformations in Crystalline Solids. The Institute ofMetals Monograph and Report Series No. 33, 1969,p. 49.
34. Zahra, A., La�tte, M., Vigier, P. and Wintenberger,M., MeÂm. Sci. Rev. MeÂt., 1977, 74, 561.
35. Ono, K. and Kino, T., J. Phys. Soc. Jpn., 1978, 44,875.
36. Horiuchi, R. and Minonishi, Y., J. Jpn. Inst. Met.,1970, 34, 936.
37. Starink, M. J., Doctor's thesis, University of Delft,1992.
38. Jena, A. K., Gupta, A. K. and Chaturvedi, M. C.,Acta metall., 1989, 37, 885.
39. Starink, M. J., Zahra, C. Y. and Zahra, A.-M., J.Therm. Anal., 1998, 51, 933.
40. Cusiat, F., Duval, P. and Graf, R., Scripta metall.,1984, 18, 1051.
41. Degischer, H. P., Zahra, C. Y. and Zahra, A.-M., Z.Metallk., 1982, 73, 635.
42. Bergner, D., Neue HuÈtte, 1984, 29, 207.43. Wilson, R. N. and Partridge, P. G., Acta metall.,
1965, 13, 1321.44. Zahra, A.-M. and Zahra, C. Y., J. Therm. Anal.,
1990, 36, 1465.
2764 CHARAI et al.: CLUSTERS, ZONES AND PHASES