Acta mater. Vol. 46, No. 11, pp.

3893±3904, 1998
# 1998 Acta Metallurgica Inc.
Published by Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S1359-6454(98)00059-7 1359-6454/98 $19.00 + 0.00

THE PRECIPITATION SEQUENCE IN Al±Mg±Si ALLOYS

G. A. EDWARDS1, K. STILLER2, G. L. DUNLOP1{ and M. J. COUPER3
CRC for Alloy and Solidi®cation Technology (CAST), Department of Mining, Minerals and Materials
1

Engineering, The University of Queensland, QLD 4072, Australia, 2Chalmers University of Technology,
S-41296 Gothenburg, Sweden and 3Comalco Research Centre, 15 Edgars Rd, Thomastown, VIC 3074,
Australia

(Received 11 September 1997; accepted 7 February 1998)

AbstractÐFine-scale precipitation that occurs during age hardening of Al alloy 6061 has been studied
using di€erential scanning calorimetry (DSC), atom probe ®eld ion microscopy (APFIM) and transmission
electron microscopy (TEM). It was found that the precipitation sequence is: independent clusters of Mg
and Si atoms 4 co-clusters that contain Mg and Si atoms 4 small precipitates of unknown structure 4 b0
needle-shaped precipitates 4B ' lath-shaped precipitates and b' rod-shaped precipitates. A new structure is
proposed for the b0 precipitate. It was found that the Mg:Si ratio in the intermediate precipitates and co-
clusters was close to 1:1. # 1998 Acta Metallurgica Inc.

1. INTRODUCTION 0208C and 1208C on thermal analysis scans

Al±Mg±Si alloys are an important group of alloys
that are widely used in both cast and wrought
form. The alloys are age hardenable, and are routi-
nely heat treated to the T6 condition to develop
adequate strength. The age-hardening response of
the alloys is very signi®cant and hence control of
precipitation during heat treatment is critical for
attaining optimal alloy performance. The import-
ance of the alloy group and the considerable age-
hardening response have motivated numerous stu-
dies of precipitation in these alloys. However, due
to the complexity of the precipitation sequence, and
considerable diculties in the analysis of precipi-
tates in this system, the subject remains controver-
sial.
In a recent study [1] it was proposed that the pre-
cipitation sequence was:
clusters of Si atoms !GP-I zones
! GP-II- zones=b 0 0 ! b 0 ! bMg2 Si:
However considerable confusion remains concern-
ing the initial, ``clustering'' processes, the structure
of b0 and the GP-I zones, and the compositions of
all of the intermediate precipitates. There is also
evidence to suggest that several other precipitates
may form in addition to those listed above [2±4].
Clustering of solute atoms has been proposed
to explain indirect observations of a precipitation
reaction that occurs early in the ageing sequence.
The reaction releases heat that may be detected
using thermal analysis techniques, producing exo-
thermic peaks at temperatures in the range between
{To whom all correspondence should be addressed.
3893
[1, 4±7]. It also alters the electrical resistance of the
alloys and can therefore be detected using resistivity
measurements [8±13]. An increase in resistance is
observed over a similar temperature range to that
for the exothermic peak.
Dutta and Allen [1] proposed that the observed
e€ects result from clustering of Si atoms. However
this is at variance with observations for Al±Si
alloys, in which neither the exothermic peak nor the
increase in resistivity are observed [4, 10, 14]. It is
more plausible that Mg atoms move to clusters of
Si atoms that form immediately after quenching,
producing the observed thermal and resistance
e€ects [4, 7, 10, 12]. In Al±Si alloys it is thought that
Si atoms accompany vacancies as they condense
during and immediately after quenching [15] and it
is likely that this process also occurs in Al±Mg±Si
alloys. Gupta and Lloyd [4] have reported that the
exothermic peak may in fact consist of two overlap-
ping peaks, indicating the possible presence of a
second precipitation reaction.
Uncertainties also exist concerning the precipi-
tates that form in the next stage of precipitation.
Smith [16] and Cordier and Gruhl [17] reported
that the precipitates were spherical in shape whereas
Dutta and Allen [1] claimed that their shape could
not be resolved. Dutta and Allen designated the
precipitates GP-I zones although no evidence con-
cerning the crystal structure of the precipitates has
been obtained. The compositions of all these early
clusters and precipitates are unknown.
The b0 precipitate has been the subject of much
interest as this precipitate is associated with peak-
aged conditions. The precipitates, which are needle
shaped and aligned along h100iAl, were ®rst discov-
ered by X-ray analysis [18±20], and were sub-
3894 EDWARDS et al.: THE PRECIPITATION SEQUENCE
sequently observed by TEM [16, 17, 21±25]. They
range between 200 and 1000 AÊ in length, and are
060 AÊ in diameter [21]. They produce streaks par-
allel to h100iAl directions on di€raction patterns
obtained with the beam parallel to h001iAl [21].
This precipitate is the predominant precipitate in
peak-aged alloys [23±25]. Various crystal structures
have been proposed for the b0 precipitate: mono-
clinic with a = c = 6.16 AÊ, b = 7.1 AÊ and
b = 828 [26]; monoclinic with a = 3.0 AÊ, b = 3.3 AÊ,
c = 4.0 AÊ, g = 718 [26]; hexagonal with a = 7.05 AÊ
and c = 4.05 AÊ [22]. There has been no attempt to
determine the composition of these precipitates.
b' precipitates form after b0 precipitates in the
ageing sequence. They are rod shaped and are
aligned along h100iAl [16, 21±25]. The structure has
been determined by Jacobs [22] to be hexagonal,
with a = 7.05 AÊ and c = 4.05 AÊ. This has been
con®rmed by subsequent HRTEM and
microdi€raction [3, 27]. The composition of this
phase has also been analysed but the results are
somewhat inconclusive. Lynch et al. [27], using
EDS, determined that the Mg:Si ratio (atomic) was
much less than the 2:1 ratio in the equilibrium
phase and possibly as low as 0.44. Matsuda et
al. [28] also used EDS to obtain a Mg:Si ratio of
1.68.
Several other intermediate precipitates have also
been observed to form together with b' [2, 3]. One
of these is known as B' which has a crystal structure
that may be described as either hexagonal, a = 10.4
AÊ, c = 4.05 AÊ, or base-centred orthorhombic
a = 18 AÊ, b = 10.4 AÊ, c = 4.05 AÊ [2]. B' is lath-
shaped with rectangular cross sections, and lies
along h100iAl. A ``type C'' precipitate has also been
observed via HREM [3]. Again the compositions of
these phases have not been investigated.
The equilibrium Mg2Si phase forms as platelets,
lying in {100}Al planes [22]. The structure has been
well characterized as the f.c.c. anti-¯uorite structure
with a = 6.39 AÊ. This precipitate contributes little
to the strength of Al±Mg±Si alloys [21].
The present investigation was performed to
further clarify the precipitation sequence in Al±
Mg±Si alloys with regard to both the crystal struc-
tures and compositions of the precipitating phases.
2. EXPERIMENTAL
Al alloy 6061 was supplied by Comalco Research
Centre in the form of as-extruded bar. The compo-
sition of the alloy was (wt%): 0.80 Mg, 0.79 Si, 0.18
Cu, 0.22 Fe, 0.01 Ti. Di€erential scanning calorime-
try (DSC) was carried out on a Perkin±Elmer
DSC7 instrument, using a heating rate of 58C/min.
Specimens for DSC were solution treated for 1.5 h
at 5308C in an argon atmosphere, and then
quenched into iced water. Heating in the DSC com-
menced within 10 min of quenching.
Thin-foil specimens for TEM were prepared by
electropolishing in a 1:4 solution of nitric acid:
methanol, at ÿ308C and 08 V, using a Tenupol jet
polisher. Transmission electron microscopy was per-
formed using JEOL 4000FX and 1210 instruments
operating at 400 kV and 120 kV, respectively. High
resolution dark-®eld electron microscopy
(HRDEM) was used to image the crystal structures
of individual ®ne-scale precipitates. This technique
is further described elsewhere [29]. Compositions of
the intermediate phases were mainly analysed by
atom probe ®eld ion microscopy (APFIM). The ac-
curacy of this technique in determining Mg:Si ratios
in small precipitates was con®rmed using parallel
electron energy loss spectroscopy (PEELS) and
energy dispersive spectroscopy (EDS) on selected
precipitates. These analyses are reported in detail
elsewhere [30]. Details of the atom-probe instrument
have been reported elsewhere [31]. APFIM data
were obtained using a specimen temperature of
about 25 K, a pulse fraction of 20% and a vacuum
of less than 1.5  10ÿ10 mbar. These conditions were
chosen to minimize DC evaporation or retention of
atoms. Concentration pro®les were derived from
atoms collected from a cylinder of material approxi-
mately 16 AÊ in diameter by maintaining the e€ec-
tive radius of the probe aperture at about 8 AÊ. The
pro®les were plotted by dividing the data into
groups of 75 or 100 ions, obtaining the numbers of
solute atoms in each group and then plotting these
numbers vs the total number of ions collected. The
detector eciency was kept constant during acqui-
sition by varying the channel plate voltage [32]. The
mean separation method [33, 34] was applied to
selected pro®les to test for clustering or otherwise
of solute atoms.
3. THE PRECIPITATION SEQUENCE
3.1. Di€erential scanning calorimetry
A typical DSC curve is shown in Fig. 1. The
number and positions of the peaks are similar to
results obtained by Dutta and Allen [1], although
the peaks are shifted to lower temperatures in the
present work. This is most likely due to the slower
heating rate that was used in the present investi-
gation (58C/min, cf. 108C/min for Dutta and Allen).
The broadness and asymmetry of the initial peak
suggests the presence of two overlapping peaks.
However, these are not as clearly resolved as in the
work by Gupta and Lloyd [4].
3.2. Exothermic peak 1 on DSC scans
Extensive bright-®eld and dark-®eld TEM as well
as SADP from specimens that had been heated at
58C/min to 1008C, near the end of the ®rst DSC
peak, provided no evidence of any precipitation,
which is contrary to the results that were obtained
by Dutta and Allen [1]. Some faint re¯ections were
 EDWARDS et al.: THE PRECIPITATION SEQUENCE
Fig. 1. Di€erential scanning calorimetry curve for alloy 6061. The number and positions of the peaks
are similar to results in the literature.
present on SADPs but these were also present in
as-quenched specimens and their intensity dimin-
ished after ion-beam thinning. Therefore it is
thought that these re¯ections resulted from a sur-
face layer of oxide or other material on the thin foil
specimens. APFIM concentration pro®les from spe-
cimens that had been given the same heat treatment
are shown in Fig. 2. The pro®les clearly show the
presence of three types of precipitate: clusters of Si
atoms; clusters of Mg atoms; and co-clusters con-
taining both Mg and Si atoms.
To further investigate clustering of atoms,
APFIM was performed on specimens that had been
isothermally aged for 0.5 h, 8 h and 72 h at 708C.
Concentration pro®les from these specimens are
shown in Fig. 3. The pro®les from samples aged for
0.5 h at 708C [Fig. 3(a)] showed some indication of
clustering, but this was by no means conclusive.
The mean separation method was applied to these
pro®les. The clustering parameter was 2.5 for Mg
and 5 for Si. The high value for Si is strong evi-
dence of clustering of Si. The lower value for Mg is
less conclusive but is still indicative that clustering
of Mg atoms also occurred. A contingency
table [35, 36] (Table 1) for Mg and Si was derived
to test for the co-segregation of these elements.
Using 75 ions/group and a 3  3 table (equals 4
degrees of freedom), the calculated value for w2 was
2.17. When compared with the chi-squared distri-
bution, the null hypothesis of independence was
accepted with a signi®cance level of 70%. This is a
very high signi®cance level, indicating that there is
no clear evidence in the data for the co-segregation
of Mg and Si.
Clusters or precipitates that contained both Mg
and Si were clearly present in specimens that had
been aged for the longer times of 8 h and 60 h at
708C [Fig. 3(b) and (c)]. The numbers of atoms
of each alloying element detected in the analysed
precipitates are presented in Table 2. After ageing
3895
for 8 h, most of the precipitates that formed had
Mg:Si ratios close to 0.7. Some precipitates con-
tained signi®cantly more Mg or Si. After ageing for
60 h the Mg:Si ratio was close to one in about half
of the precipitates that were examined. The remain-
ing precipitates contained a predominance of either
Mg or Si.
Cu was present in only some of the precipitates,
and the Cu content in all of the data was much
lower than the bulk analysis of 0.18 wt%. This
suggests that the Cu atoms did not greatly in¯uence
the formation of the Mg/Si co-clusters.
3.3. Exothermic peak 2a on DSC scans
Figure 4(a) shows a TEM bright ®eld micrograph
from a specimen that had been heated to 2158C in
the DSC, close to the shoulder of the second peak
in the DSC scan. Very ®ne precipitates are imaged
by strain ®eld contrast. The precipitates produced
no detectable di€raction e€ects on SADPs, however
they were visible in dark ®eld micrographs that
were obtained with the objective aperture posi-
tioned between the aluminium re¯ections with the
beam, B, parallel to [001]Al. Dark-®eld high resol-
ution micrographs showed no clear structure in the
precipitates.
To examine these precipitates with APFIM, spe-
cimens were isothermally aged for 10 min at 1758C,
which produced precipitates that were identical to
those in Fig. 4(a). An APFIM concentration pro®le
that was obtained from such a specimen is pre-
sented in Fig. 4(b). Two precipitates that clearly
contain both Mg and Si are apparent in the pro®le.
The numbers of Mg and Si atoms in these precipi-
tates are presented in Table 2. The Mg:Si ratio in
these precipitates was close to 1, which is consistent
with the observations for extended ageing at 708C.
The structure of these small precipitates remains
unknown. The fact that no structure was visible
using dark-®eld HREM does not necessarily indi-
3896 EDWARDS et al.: THE PRECIPITATION SEQUENCE

Fig. 2. Sections from APFIM concentration pro®les from a specimen that had been heated at 58C/min
to 1008C, near the end of the ®rst exothermic DSC peak. Three types of clusters of atoms were present
(a) clusters of Mg atoms, (b) clusters of Si atoms, and (c) co-clusters that contained both Mg and Si
atoms.

cate that the precipitates do not have their own ameter and orientation of the precipitates were such
de®nite crystal structure, since three beams must be that this did not occur. The fact that the precipi-
present in the objective aperture to image the struc- tates were imaged in dark ®eld indicates that there
ture in dark ®eld. It is possible that the lattice par- was some di€racted intensity from the precipitates.
 EDWARDS et al.: THE PRECIPITATION SEQUENCE 3897

Fig. 3. APFIM concentration pro®les from specimens that had been aged at 708C for (a) 0.5 h, (b) 8 h
and (c) 60 h. Statistical analysis of (a) indicated that there were separate clusters of Si atoms and Mg
atoms in the data. Co-clusters that contained both Si and Mg were present in (b) and (c).

However, this intensity was too weak or too di€use
to detect on the SADPs. It is also possible that the
precipitates are GP zones, or they may not have a
de®ned crystal structure, as was suggested by
Lutts [18]. X-ray analysis by Lutts suggested that
this precipitate structure contains many vacancies.
GP zones in other aluminium alloys, e.g. Al±Cu,
have been found to produce streaks on selected
area di€raction patterns. However streaking from
GP zones would not be expected to occur in Al±
3898 EDWARDS et al.: THE PRECIPITATION SEQUENCE

Table 1. Contingency table that was derived from atom-probe shaped precipitates meant that there was generally
data from specimens aged for 0.5 h at 708C. (a) Experimental
data, (b) expected data for an independent distribution. The analy-
overlap between the Al matrix and the precipitates.
sis showed that there was no clear evidence for the co-clustering of Hence only one view of the precipitate lattice could
Mg and Si be obtained [Fig. 5(c)]. The high resolution images
Mg atoms Si atoms showed that the precipitate lattice had a periodicity
that is de®ned by the vectors a1 and c in Fig. 5(c).
(a)
0 259 144 69 472 A detailed analysis of high-resolution images, the
1 129 65 35 229 streaks in SADPs and the faint, circular re¯ections
e2 51 29 20 100
Total 439 238 124 801
in SADPs led to a proposed base-centred mono-
Si atoms clinic structure for b0, with a = 15.342 0.12 AÊ,
b = 4.05 AÊ, c = 6.83 2 0.15 AÊ and b = 1068
(b)
0 259 140 73 2 1.58 [37]. The unit cell vector a is twice the length
1 126 68 35 of the vector a1 in Fig. 5(c). The orientation re-
e2 55 30 15
lationship between this structure and the Al lattice
was found to be: (001)Al//(010)b0, [310]Al//[001]b0.
There are 12 equivalent orientations of the precipi-
Mg±Si alloys as the atomic scattering factors of tate, four for each h001iAl orientation of the needle
these three elements are all very similar. axes. The precipitate structure is fully coherent with
the Al matrix, and there is signi®cant mis®t along
both sides of the rhombohedral cross-section of the
3.4. Exothermic peak 2b in the DSC scan
precipitate. There is negligible mis®t along the
Specimens that had been heated to 2508C at 58C/ length of the precipitates.
min contained needle shaped precipitates that were An APFIM analysis of a b0 precipitate yielded
clearly delineated by strain-®eld contrast [Fig. 5(a)]. 118 Mg atoms, 97 Si atoms and 3 Cu atoms, giving
SADPs taken with B = [001]Al contained streaks a Mg:Si ratio of 1.2 [30].
and some faint, circular re¯ections in addition to
the oxide re¯ections from the oxide layer [Fig. 5(b)].
The appearance of the precipitates, and the streaks 3.5. Exothermic peak 3 in DSC scans
on the SADPs, are in agreement with previous ob- Figure 6(a) shows a bright-®eld TEM micrograph
servations for b0 precipitates [21]. and SADP (B = [001]Al) from a specimen that was
High-resolution images were only obtained from heated at 58C/min to 3208C, just beyond the third
precipitates that were aligned with the long axis peak in DSC scans. Large rod-shaped precipitates
parallel to the electron beam. For other precipitate that were aligned along h001iAl directions produced
orientations, the small diameters of the needle- mainly strain-®eld contrast in this micrograph.
Streaks and some faint re¯ections were visible on
SADPs.
Table 2. Numbers of Mg, Si and Cu atoms that were contained in Figure 6(b) shows a bright-®eld high-resolution
the clusters of atoms in specimens aged for 8 h and 60 h at 708C, electron micrograph of a lath shaped precipitate.
and for 10 min at 1758C
The cross-sections of the precipitates are generally
No. Mg No. Si No. Cu rectangular so that the precipitates are better
Particle atoms atoms Mg/Si ratio atoms
described as lath shaped rather than needle shaped.
8 h at 708C The long faces of the precipitate cross-sections are
1 9 15 0.6 1 orientated at an angle of about 118 to h001iAl direc-
2 14 19 0.7 0
3 10 13 0.8 1 tions. Four orientations of the precipitates were
4 11 15 0.7 0 observed for each h001iAl direction, giving a total
5 4 16 0.3 0
6 14 6 2.33 1
of 12 orientations.
7 9 15 0.6 0 Fourier transformations of images such as in
8 8 13 0.6 0 Fig. 6(b) indicated that the crystal structure is hexa-
9 4 9 0.4 0
60 h at 708C gonal, with a = 10.5 AÊ20.15 AÊ. This is very close
1 7 21 0.3 0 to the structure of the B' precipitate that was
2 19 17 1.1 0 reported by Dumolt et al. [2], i.e. hexagonal
3 15 16 0.9 0
4 8 8 1 0 a = 10.4 AÊ, c = 4.05 AÊ. The morphologies and
5 18 12 1.5 0 orientations of the precipitates are identical to that
6 8 5 1.6 0
7 7 9 0.8 0
reported by Dumolt et al., therefore this precipitate
8 9 10 0.9 0 is considered to be B'. The hexagonal lattice is not
9 4 11 0.4 0 immediately obvious on the high-resolution image
10 23 8 2.9 2
11 13 8 1.6 0 in Fig. 6(b), however, images taken with a di€erent
10 min at 1758C defocus showed the periodicity more clearly
1 34 31 1.1 0 [Fig. 6(c)]. The hexagonal periodicity was con®rmed
2 24 21 1.1 1
for both small and large precipitates by high-resol-
 EDWARDS et al.: THE PRECIPITATION SEQUENCE 3899

Fig. 4. (a) Bright-®eld TEM micrograph from a specimen that had been heated at 58C/min to 2158C,
close to the shoulder of the second exothermic peak in the DSC scan, and (b) APFIM concentration
pro®le from a specimen that had been aged for 10 min at 1758C to produce the same precipitates as in
(a).

ution dark-®eld imaging. Most precipitates are 3±5
unit cells thick.
A small proportion of observed precipitates were
of di€erent crystal structure to B' and these had
more equiaxed cross-sections than the B' precipi-
tates [Fig. 6(d)]. The crystal structure of this pre-
cipitate consists of a hexagonal array, with a 0 7.1
AÊ (measured from the image). This structure closely
matches that of b' [22] (hexagonal, a = 7.05 AÊ,
c = 4.05 AÊ).
Hence it is concluded that two phases are formed
during the third peak in the DSC scans: b' and B'.
In this alloy B' is by far the predominant precipi-
tate.
B' precipitates were analysed with APFIM,
PEELs and EDS [30]. The results from the three
techniques were in good agreement and indicated
that the Mg:Si ratio in these precipitates is close to
1.15. In previous work reported by the present
authors [30], precipitates with round cross-sections
that were thought to be b' were actually B'.
4. THE RELATIONSHIP BETWEEN PRECIPITATES
AND AGE HARDENING
4.1. Age hardening curves
Age hardening curves obtained by arti®cially age-
ing alloy 6061 at 1758C and 2008C are shown in
Fig. 7. The age hardening behaviour was in¯uenced
by the ageing temperature in the usual manner: age-
ing at the higher temperature produced an earlier
but somewhat lower peak hardness compared with
ageing at the lower temperature. Ageing at 2008C
produced a much more substantial overageing e€ect
than ageing at 1758C. At 1758C, the time to peak
hardness was 08 h. This is earlier than generally
reported in the literature, although such data are
somewhat varied. Nieh and Karlak [38] reported a
peak at 10 h for ageing at 1778C, and ASTM data
show a peak at 070 h for ageing at 1708C. The ear-
lier time to peak hardness in the present investi-
gation is probably due to the higher than usual
amount of Si in the alloy.
4.2. Underaged microstructures
Figure 8(a) is a bright-®eld TEM micrograph
(B = [001]Al) of a specimen that was aged for
10 min at 1758C. Very ®ne precipitates were visible
on bright-®eld micrographs, but no precipitate
re¯ections were visible on SADPs. High-resolution
dark-®eld imaging highlighted the precipitates
[Fig. 8(b)], but showed no clear structure in most of
the precipitates. Some precipitates did show detail
that matched the b0 structure (arrowed), indicating
that some transformation to b0 had already
occurred at this stage, when the precipitates were
still very small. The bright-®eld and dark-®eld
images show that the predominant precipitate in
this underaged condition was the small precipitate
of unknown structure that formed during peak 2a
in the DSC scans. The hardness curve shows that
these precipitates produced only a small amount of
hardening.
Figure 8(c) shows a bright-®eld TEM micrograph
from a specimen that had been aged for 0.5 h at
1758C. Short b0 precipitates (07 nm long) are pre-
dominant in the microstructure. This ageing treat-
ment substantially increased the hardness of the
alloy, indicating that the b0 precipitates were much
more e€ective in strengthening the alloy than the
small, equiaxed precipitates that preceded them in
the precipitation sequence.
4.3. Peak-aged microstructure
A bright-®eld TEM micrograph from a specimen
that had been aged for 4 h at 1758C (close to peak
hardness) is shown in Fig. 8(d). The predominant
precipitate is b0. This precipitate is thus considered
to be responsible for the maximum age-hardening
e€ect in this alloy. The lengths of the precipitates
ranged from 010 to 15 nm, which is a considerable
3900 EDWARDS et al.: THE PRECIPITATION SEQUENCE

Fig. 5. (a) Bright-®eld TEM micrograph, (b) section from

a SADP (B = [001]) and (c) high-resolution electron
micrograph of a needle shaped precipitate imaged along
its axis, from a specimen that had been heated at 58C/min
to 2508C, close to the end of the second exothermic peak
in the DSC scan.

growth from the average length of 7 nm for these

precipitates after 0.5 h at 1758C.

4.4. Overaged microstructures

Figure 8(e) is a TEM bright-®eld micrograph
from a specimen that had been aged for 72 h at
1758C. There is little obvious di€erence between
this microstructure and the peak-aged microstruc-
ture [Fig. 8(d)], which concurs with the only slightly
decreased hardness at this level of overageing. More Fig. 6. (a) Bright-®eld TEM micrograph from a specimen
extensive overageing for 24 h at the higher tempera- that had been heated at 58C/min to 3208C, close to the
ture of 2008C resulted in the formation of B' pre- third exothermic peak in the DSC scan; (b) and (c)
HRTEM micrographs of B' lath-shaped precipitates
cipitates [Fig. 8(f)]. This indicates that the B' imaged along their axes [optimum and non-optimum defo-
precipitate is largely associated with the reduction cus in (b) and (c), respectively]. (d) HRTEM micrograph
in hardness that occurs upon overageing. of a rod-shaped b' precipitate imaged along its axis.
 EDWARDS et al.: THE PRECIPITATION SEQUENCE 3901

atoms. All three types of cluster were present in the

Fig. 7. Age-hardening curves that were obtained by ageing
at 1758C and 2008C.
5. DISCUSSION
5.1. Clustering of atoms
The APFIM results indicate that three types of
clusters of atoms form in the early stages of ageing
of alloy 6061: clusters of Si atoms; clusters of Mg
atoms; and co-clusters that contain both Mg and Si
alloy after heating to the end of the ®rst exothermic
peak in the DSC scan. However, isothermal ageing
at 708C indicated that the di€erent types of clusters
did not form simultaneously. It appeared that inde-
pendent clusters of Mg and Si atoms formed ®rst,
followed by the formation of co-clusters. The Mg:Si
ratio in the co-clusters was found to increase during
ageing at 708C, ®nally attaining a value close to 1.
It is possible that either or both of the clusters of
Mg atoms and the clusters of Si atoms formed im-
mediately after quenching and did not contribute to
the ®rst exothermic DSC peak. The as-quenched
state of the alloy is very dicult to analyse with
APFIM, since time at room temperature is required
after quenching to prepare specimens. However, the
APFIM results may be related to resistivity and
thermal analysis results in the literature for Al±
Mg±Si alloys, Al±Mg alloys and Al±Si alloys, to
further understand the early clustering of solute
atoms.

Fig. 8. TEM micrographs of precipitate dispersions at di€erent stages of arti®cial ageing. (a) and (b),
10 mins at 1758C, (c) 0.5 h at 1758C, (d) 4 h at 1758C, (e) 72 h at 1758C, (f) 20 h at 2008C.
3902 EDWARDS et al.: THE PRECIPITATION SEQUENCE
As mentioned previously, thermal analysis results
for Al±Si alloys show no exothermic peak in the
early stages of precipitation [4]. Instead, Si atoms
are thought to accompany vacancies when they con-
dense, causing clustering to occur very soon after
quenching. It is likely that this process also occurs
in Al±Mg±Si alloys.
Similarly, resistivity results for Al±Mg alloys
show that an increase in resistivity occurs at
0ÿ 308C [39]. This has been attributed to cluster-
ing of Mg atoms [39] and this clustering was also
observed in the present investigation of an Al±Mg±
Si alloy. Thermal analysis results in the literature
for Al±Mg±Si alloys do not scan below room tem-
perature, thus any exothermic peak associated with
the dissolution of clusters of Mg atoms would not
have been detected.
Thermal analysis results for Al±Mg alloys show
an endothermic peak at 0708C that has been attrib-
uted to the dissolution of the clusters of Mg
atoms [40]. This peak corresponds to a decrease in
resistivity that has been observed at similar
temperatures [39]. It is possible that dissolution of
clusters of Mg atoms also occurs in Al±Mg±Si
alloys, with the dissolving Mg then being free to
combine with the clusters of Si atoms to form co-
clusters. This is consistent with the observations for
isothermal ageing at 708C where separate clusters of
Si atoms and Mg atoms disappeared with time,
being replaced by co-clusters of Mg and Si atoms.
The small endothermic e€ect of the dissolving Mg
may also explain the apparent observation of a
double exothermic peak in some thermal analysis
scans. The combination of the large exothermic
peak, due to the formation of co-clusters, and the
endothermic peak, due to dissolution of Mg, might
give the appearance of a double exothermic peak.
Thus, the proposed initial clustering sequence is
as follows:
Clusters of
Dissolution Formation
Si atoms
Al SSS ! ! of Mg ! of Mg=Si
Clusters of
clusters co-clusters:
Mg atoms
This sequence appears to be consistent with resis-
tivity and thermal analysis observations for Al±Mg
alloys, Al±Si alloys and Al±Mg±Si alloys, and the
APFIM results reported in the present investi-
gation. However, several details remain to be clari-
®ed. For example, the formation of co-clusters may
involve Mg atoms in solid solution as well as those
from the dissolving clusters. Also, Huppert and
Hornbogen [41] have suggested that the co-clusters
may also contain Al atoms. It is not clear whether
the clusters of Si and Mg atoms form during
quenching, or immediately afterwards. The role of
vacancies is also largely unknown, although the log-
arithmic shape of the resistivity curves suggests that
vacancies are trapped by the co-clusters [42]. These
may or may not be released at some later stage in
the precipitation sequence.
The early clustering reactions are of commercial
signi®cance since it has been suggested that ``pre-
ageing'' at room temperature between quenching
and arti®cial ageing is generally detrimental
to the mechanical properties of Al±Mg±Si
alloys [24, 25, 43±45]. Such pre-ageing prior to ther-
mal analysis has been shown to decrease the size of
the initial exothermic peak [5]. The present results
have shown that this peak is due to co-clustering of
Mg and Si atoms; therefore, it is clear that co-clus-
ters form during ageing at room temperature.
Although room-temperature pre-ageing is usually
detrimental to mechanical properties, some studies
have found that if the pre-age is suciently long,
then improved mechanical properties may be
achieved [24, 25, 44, 46, 47]. This has previously been
ascribed to the formation of clusters of Si atoms of
a critical size [24, 25, 45]. However, the present
results suggest that the nature of the clusters (i.e.
co-cluster or individual cluster), rather than their
size, will be critical for the ability of the clusters to
act as nuclei for subsequent intermediate phases.
This is emphasized by the similar compositions of
the co-clusters and intermediate phases. It is
thought that the distribution of hardening inter-
mediate phases will strongly depend upon the distri-
bution of co-clusters. Therefore any low-
temperature ageing treatment should in¯uence the
®nal arti®cially aged strength. It appears likely that
for any low-temperature pre-age to be bene®cial,
sucient time must be allowed to develop a large
number of co-clusters.
The small precipitates of unknown structure
appear to be of similar composition to the co-clus-
ters that precede them in the precipitation sequence.
The transition from co-cluster to small precipitate,
which causes an exothermic heat release, must
therefore involve some kind of ordering, and poss-
ibly the release of Al atoms and/or vacancies from
the co-clusters. The exact details of this transition,
including the structure of the small precipitates, are
very dicult to determine experimentally. The small
precipitates were shown to transform to b0 whilst
still very small. The apparent absence of an
endothermic reaction suggests that the transition
proceeds without the dissolution of the small pre-
cipitates. The rapid transformation from small pre-
cipitate to b0 is consistent with the close proximity
of the corresponding exothermic peaks on thermal
analysis scans and the similarity of the Mg:Si ratios
in both types of precipitate. Thus it seems likely
that the small precipitates nucleate b0 precipitates,
and that the distribution of small precipitates will
signi®cantly a€ect the distribution of b0 precipi-
tates.
The structure proposed here for the b0 precipitate
is di€erent to those proposed previously. The struc-
ture that was obtained from Shergoleva [26] must
 EDWARDS et al.: THE PRECIPITATION SEQUENCE 3903

be viewed with caution, given the diculties experi- L. Arnberg et al.. Norwegian Institute of Technology
enced by other X-ray analyses, and the fact that and SINTEF Metallurgy, Trondheim, 1992, p. 21.
5. Badini, C., Marino, F. and Tomasi, A., Mater. Sci.
other researchers have been unable to decipher the Eng., 1991, A136, 99.
crystal structure by analysing di€racted streaks. The 6. Papazian, J. M., Metall. Trans., 1988, 19A, 2953.
microdi€raction analysis by Lynch et al. [27] 7. Barczy, P. and Tranta, F., Scand. J. Metall., 1975, 4,
deserves consideration. The alloy used by Lynch et 284.
al. was di€erent from that used in the present inves- 8. Abis, S. and Donzelli, G., J. Mater. Sci. Lett., 1988,
7, 51.
tigation, and the precipitates that were analysed by 9. Lendvai, J., Ungar, T. and Kovacs, I., Mater. Sci.
Lynch et al. were much larger than those analysed Eng., 1974, 16, 85.
here. Therefore it is possible that the precipitates 10. Chatterjee, D. K. and Entwhistle, K. M., J. Inst.
analysed by Lynch et al. may have been di€erent to Metals, 1973, 101, 53.
those analysed in the present investigation. 11. Kovacs, I., Lendvai, J. and Nagy, E., Acta metall.,
1972, 20, 975.
The structure of the b0 precipitate in the alloy 12. Kovacs, I., Lendvai, J. and Nagy, E., Phys. Stat. Sol.,
used in this investigation (alloy Mg:Si ratio 01) 1972, 14, 83.
appears identical to that observed with HREM by 13. Panseri, C. and Federighi, T., J. Inst. Metals, 1966,
Sagalowicz et al. [48] (alloy Mg:Si ratio 01.8). This 94, 99.
suggests that the precipitates do not change with 14. El Sayed, H. and Kovacs, I., Phys. Stat. Sol., 1974,
24, 123.
the Mg:Si ratio in the alloy. 15. Ozawa, E. and Kimura, H., Mater. Sci. Eng., 1971, 8,
It was found that the Mg:Si ratios in the precipi- 327.
tates, throughout the sequence from co-cluster± 16. Smith, W. F., Metall. Trans., 1973, 4, 2435.
small precipitate±b0-B' were close to 1.1. Thus the 17. Cordier, H. and Gruhl, W., Z. Metall., 1965, 56, 669.
atomic structures in these precipitates must be very 18. Lutts, A., Acta metall., 1961, 9, 577.
19. Guinier, A. and Lambot, H., Rev. metall., 1950, 47,
di€erent from those in the equilibrium Mg2Si 709.
phase. 20. Geisler, A. H. and Hill, J. K., Acta Crystallogr., 1948,
1, 283.
21. Thomas, G., J. Inst. Metals, 1961±62, 90, 57.
22. Jacobs, M. H., Phil. Mag., 1972, 26, 1.
6. CONCLUSIONS 23. Skalicky, P. and Oppolzer, H., Z. Metall., 1972, 63.
24. Pashley, D. W., Jacobs, M. H. and Vietz, J. T., Phil.
1. The precipitation sequence in Al±Mg±Si alloys Mag., 1967, 51.
consists of: 25. Pashley, D. W., Rhodes, J. W. and Sendorek, A., J.
Inst. Metals, 1966, 94, 41.
Clusters of 26. Schegoleva, T. V., Phys. Metals Metall., 1968, 25, 56.
Dissolution Formation
Si atoms 27. Lynch, J. P., Brown, L. M. and Jacobs, M. H., Acta
Al SSS ! ! of Mg ! of Mg=Si !
Clusters of metall., 1982, 30, 1389.
clusters co-clusters 28. Matsuda, K., Tada, S. and Ikeno, S., in Proc. 4th Int.
Mg atoms
Conf. on Aluminium Alloys, ed. T. H. Sanders and E.
A. Starke. Georgia Institute of Technology, Atlanta,
Small GA, 1994, p. 605.
b0 0 B0 and b 0 0 bMg2 Si
precipitates 29. Edwards, G. A., Dunlop, G. L. and Couper, M. J., in
! preciÿ ! preciÿ ! preciÿ
of unknown Proc. 4th Int. Conf. on Aluminium Alloys, ed. T. H.
pitates pitates pitates: Sanders and E. A. Starke. Georgia Institute of
structure
Technology, Atlanta, GA, 1994, p. 629.
2. The Mg:Si ratios in the Mg/Si co-clusters, the 30. Edwards, G. A., Stiller, K., Couper, M. J. and
small precipitates, b0 precipitates, and B' precipi- Dunlop, G. L., in Mater. Sci. Forum, 1996, 217±222,
713.
tates are close to 1.1. 31. Andren, H.-O., J. Phys., 1986, 47, C7±483.
32. Lundin, L. and Rolander, U., Appl. Surf. Sci., 1993,
AcknowledgementsÐFinancial support from Comalco 67, 459.
Aluminium Limited, is gratefully acknowledged. CAST is 33. Hetherington, M. G. and Miller, M. K., J. Phys.,
supported by the Australian Governments' Cooperative 1987, 48, C6±559.
Research Centre Program. 34. Hetherington, M. G. and Miller, M. K., J. Phys.,
1988, 49, C6±427.
35. Miller, M. K. and Smith, G. D. W., Atom Probe
REFERENCES Microanalysis: Principles and Applications to Materials
Problems, Materials Research Society, Pittsburgh, PA,
1. Dutta, I. and Allen, S. M., J. Mater. Sci. Lett., 1991, 1989.
10, 323. 36. Hetherington, M. G., Cerezo, A., Hyde, J., Smith, G.
2. Dumolt, S. D., Laughlin, D. E. and Williams, J. C., D. W. and Worrall, G. M., J. Phys., 1986, 47, C7±
Scr. metall., 1984, 18, 1347. 495.
3. Matsuda, K., Uetani, Y., Anada, H., Tada, S. and 37. Edwards, G. A., Dunlop, G. L. and Couper, M. J., in
Ikeno, S., in Aluminium Alloys: Their Physical and Proc. 4th Int. Conf. on Aluminium Alloys, ed. T. H.
Mechanical Properties, Vol. 1, ed. L. Arnberg et al.. Sanders and E. A. Starke. Georgia Institute of
Norwegian Institute of Technology and SINTEF Technology, Atlanta, GA, 1994, p. 620.
Metallurgy, Trondheim, 1992, p. 220. 38. Nieh, G. and Karlak, R. F., Scr. metall., 1984, 18, 25.
4. Gupta, A. K. and Lloyd, D. J., in Aluminium Alloys: 39. Panseri, C., Federighi, T. and Ceresara, S., Trans.
Their Physical and Mechanical Properties, Vol. 2, ed. Metall. Soc. AIME, 1963, 227, 1122.
3904 EDWARDS et al.: THE PRECIPITATION SEQUENCE
40. Osamura, K. and Ogura, T., Metall. Trans. A, 1984,
15A, 835.
41. Huppert, G. and Hornbogen, E., in Proc. 4th Int.
Conf. on Aluminium Alloys, ed. T. H. Sanders and E.
A. Starke. Georgia Institute of Technology, Atlanta,
GA, 1994, p. 628.
42. Kelly, A. and Nicholson, R. B., Progr. Mater. Sci.,
1963, 10, 151.
43. Kovacs, I., Lendvai, J. and Ungar, T., Aluminium,
1974, 50, 578.
44. Muromachi, S. and Mae, T., J. Jap. Inst. Metals,
1974, 38, 131.
45. Hitata, T. and Matsuo, S., Trans. Jap. Inst. Metals,
1972, 13, 231.
46. Tamizifar, M. and Lorimer, G. W., in Aluminium
Alloys: Their Physical and Mechanical Properties, Vol.
1, ed. L. Arnberg et al.. Norwegian Institute of
Technology and SINTEF Metallurgy, Trondheim,
1992, p. 220.
47. Hiroshi, N., J. Jap. Inst. Light Metals, 1973, 23, 23.
48. Sagalowicz, L., Hug, G., Bechet, D., Sainfort, P. and
Lapasset, G., in Proc. 4th Int. Conf. on Aluminium
Alloys, ed. T. H. Sanders and E. A. Starke. Georgia
Institute of Technology, Atlanta, GA, 1994, p. 636.