47
Landolt-Börnstein
New Series IV/11A2
MSIT®
Al–Cu–Mg
Aluminium – Copper – Magnesium
Günter Effenberg, Alan Prince
†
, updated by Nathalie Lebrun, Hans Leo Lukas, Mireille G. Harmelin
Literature Data
This system was previously evaluated by [1991Eff]. Their evaluation has been used by two groups as the
basis for thermodynamic assessments and phase diagram calculations [1993Zuo, 1996Zuo, 1997Che] and
[1998Buh, 2003Jan]. Some experiments have been performed to support these calculations [1995Hua,
1995Kim, 1995Soa, 1998Fau] and [1999Fau]. The equilibria in the Al-Cu-Mg system are complicated by
the existence of four ternary phases. There is need for experiments to clarify the ternary equilibria involving
the three Laves phases, �1-3, which have clearly been identified as three separate phases. The �1 phase with
a Cu2Mg type structure is a solution phase of the binary Cu2Mg compound with replacement of the Cu
atoms by Al along the 33.3 at.% Mg section. At a composition close to the Cu3Mg2Al formula, the �1 phase
melts congruently at ~910°C. Further replacement of Cu by Al stabilizes the �2 phase with a MgNi2 type
structure and then the �3 phase with a MgZn2 type structure. A variety of polytype structures with different
atom layer stacking sequences have been observed between the MgNi2 and MgZn2 type phases. The �2-3
phases appear to be formed by peritectic reaction and each Laves phase is associated with a region in which
it forms as the primary phase on solidification of melts. Four additional ternary compounds have also been
studied extensively. The S phase is based on the CuMgAl2 composition, V on Cu6Mg2Al5 and Q on
Cu3Mg6Al7. These three phases exist over very limited homogeneity ranges. The T phase has a broad range
of homogeneity. A formula (Cu1-xAlx)49Mg32 is derived from the crystal structure [1952Ber], but also some
mutual replacement between Mg and Cu+Al takes place.
The liquidus projection, presented by [1952Ura], does not include the monovariant curves associated with
the L + �1 �2 and L + �2 �3 peritectic reactions. The Laves phase �1 is the predominant primary phase,
but also the regions for primary solidification of (Al) and (Mg) are relatively large. Six pseudobinary
reactions have been identified experimentally, and the pseudobinary reaction e3 (Table 2b) has been
suggested. The invariant reactions associated with the primary (Al) phase region are well characterized by
numerous workers. The invariant reactions associated with the primary V, Q and T phase regions have been
elucidated by Russian workers, summarized by [1952Ura]. The liquidus surface across the Mg2Al3, T and
Mg17Al12 phase regions is exceptionally flat and ranges in temperature from 420 to 475°C. [1952Ura] gave
a complete reaction scheme. The thermodynamic calculations referred to above in principle reproduce this
reaction scheme, but differ in some details.
Binary Systems
Assessments of the Al-Cu system by [2003Gro], of the Al-Mg system by [2003Luk] and of the Cu-Mg
system by [2002Iva] are accepted. They are based on [1994Mur, 1998Liu] for Al-Cu, [1982Mur, 1998Lia1]
for Al-Mg and [1994Nay] for Cu-Mg. The thermodynamic data set of the COST 507 action [1998Ans,
1998Buh] was updated recently in some details [2003Jan]. It was used for the calculated figures and the
reaction scheme presented in this assessment. The homogeneity ranges of the phases Mg2Al3, � and
were
simplified to stoichiometric phases. �1 and �2 were treated as a single phase, �. �1 and �2 were also not
distinguished and called �.
Solid Phases
There are four well-defined ternary phases, designated in the literature as Q, S, T and V phases. It is quite
interesting to note that all ternary compounds in the Al-Cu-Mg system are formed at maxima of three-phase
equilibria involving the liquid phase, except the V phase, which is formed in a four-phase peritectic reaction
(P1). In addition the section at 33.3 at.% Mg contains a complex series of ternary Laves-Friauf phases that
are designated as �1, �2, �3, 5L, 6L, 9L and 16L in this assessment, Table 1. The Q phase is based on the
chemical formula Cu3Mg6Al7 [1947Str, 1951Mir1] and has a very limited homogeneity range. The S phase
48
Landolt-Börnstein
New Series IV/11A2
MSIT®
Al–Cu–Mg
has been extensively studied [1936Lav1, 1937Nis1, 1938Pet1, 1938Pet2, 1940Kuz, 1941Obi, 1943Per,
1944Lit, 1946Pet, 1946Ura, 1947Str, 1949Mir]. It also has a limited homogeneity range, based on the
chemical formula CuMgAl2. Its structure was determined by [1943Per] and confirmed by [1949Mir]. The
T phase has been equally thoroughly investigated [1919Vog, 1923Gay, 1935Lav, 1937Nis1, 1940Kuz,
1943Gue, 1944Lit, 1946Pet, 1946Ura, 1948Str, 1949Ura1, 1949Ura2, 1950Phr, 1952Ber, 1966Aul,
2000Tak] and a variety of chemical formulae assigned to it. From the crystal structure determined by
[1952Ber], the formula (Cu1-x Alx)49Mg32 is adequate. It is found that very few Al atoms occupy site A,
which is the center of an isochahedral cluster being almost empty [2000Tak]. The V phase has a small
region of homogeneity centered on the Cu6Mg2Al5 formula [1936Lav1, 1936Lav2, 1937Sch, 1943Gue,
1947Str, 1948Str, 1949Sam, 1949Ura1, 1951Mir3, 1952Ura] although other chemical formulae have been
quoted in the literature. Its structure was determined by [1949Sam] with the ideal formula Cu6Mg2Al5. New
recent results using DSC and EDS/WDX techniques [2001Fau] confirmed small solubility ranges of the Q
and S phases. Moreover, the solubility domain of the V phase seems to be parallel to the Al-Cu binary edge
[2001Fau]. Addittional experiments are needed to confirm it.
The Laves-Friauf phases, although well studied, have not been integrated experimentally into the ternary
equilibria in a satisfactory manner. The �1 phase with a Cu2Mg type structure is based on the Cu2Mg binary
compound with a substitution of Al atoms for Cu to form a solid solution series. At a composition close to
Cu3Mg2Al, the �1 phase melts congruently [1936Lav1, 1952Ura]. With further replacement of Cu by Al on
the 33.3 at.% Mg section, an MgNi2 type phase is stable, �2. There is general agreement between [1953Kle,
1965Sli, 1977Kom, 1981Mel1] and [1981Mel2] on the extent of the �2 phase region. Earlier work did not
detect �2 [1934Lav, 1943Gue, 1949Ura1] or regarded it as stable at high temperature only [1936Lav1]. The
MgZn2 type structure, �3, is formed with further substitution of Cu atoms by Al. The results from the
different workers are summarized in Fig. 1. Polytype structure Laves phases with variations in the layer
stacking sequences have been studied by [1962Kom, 1977Kit, 1977Kom] and [1981Mel1]. They are
located between �1 and �2, but their ranges of stability could not exactly be separated from those of �1 and
�2. [1998Che] proposed a “new intermetallic compound Mg1.75Cu1.0Al0.4” at a composition, where
[1991Eff, 2000Fau] and the calculations [1997Che, 1998Buh, 2003Jan] assume two phases, �1 and (Mg).
The characteristics of this “new phase”, however, clearly identify it as the �1 phase [2000Fau]. The presence
of (Mg) and �1 phases were confirmed by [2000Fau] who made XRD experiments on alloys having the
same composition as those reported by [1998Che]. Most probably the also present (Mg) phase was not
detected in the X-ray patterns of [1998Che] due to line broadening by cold deformation.
Pseudobinary Systems
A number of pseudobinary systems have been reported. The calculation [2003Jan] found 13 maxima of
three-phase equilibria, but some of them are less than 1 K above an adjacent four-phase equilibrium and
must be taken as tentative only. The (Mg)-�1 section is a pseudobinary eutectic [1932Por, 1933Bas,
1934Por, 1949Ura2], e13, Table 2. The (Al)-S section contains a pseudobinary eutectic e14 [1937Nis1,
1946Ura, 1948Bro, 1952Han]. The calculated temperatures [2003Jan] of both equilibria are far below those
given by [1946Ura] and accepted by [1991Eff]. The sections Mg2Al3-T, e19, and Mg17Al12-T, e16, are also
pseudobinary eutectic sections at Cu contents below the beginning of the primary Q phase region.
[1943Gue, 1949Ura1, 1949Ura2, 1951Mir2] and [2003Jan] are in agreement on the nature of these two
sections, Table 2. The investigation [1951Mir2] of the region of primary solidification of Q led to the
conclusion that the T phase is formed by peritectic reaction with Q at p13, Fig. 2. A pseudobinary reaction
was indicated by [1946Ura] who found a maximum on the curve U8U11 corresponding with the peritectic
formation of S by reaction of liquid with a Laves phase. [1949Ura1, 1949Ura2] and [1952Ura] refer to the
cubic Cu2Mg type phase �1 or to a composition CuMgAl. They take no account of the �2 and �3 Laves
phases. The calculation of [2003Jan] gives �2 as the Laves phase participating in this reaction, p10, which
is also favoured by [1991Eff]. [1938Pet1] regarded the CuAl2-S section as a pseudobinary, but later work
has disproved this assumption.
49
Landolt-Börnstein
New Series IV/11A2
MSIT®
Al–Cu–Mg
Invariant Equilibria
Table 2 lists the invariant reactions following from the thermodynamic calculation of [2003Jan] for the
Al-Cu-Mg ternary system and may be read in conjunction with Fig. 2. The reaction scheme, following from
this calculation is given in Fig. 3. In this calculation, �1 and �2 as well as �1 and �2 were considered as single
phases and called � and �, respectively. The ternary eutectic reaction E5 has been widely studied, Table 3.
The flat nature of the liquidus surface near to E7 has led to a considerable scatter in quoted compositions
and temperatures, Table 4. The reaction has normally been quoted as a ternary eutectic reaction and this is
accepted. The transition reaction U16 has also been widely studied, Table 5. The work of [1946Ura,
1949Ura2] and [1948Bro] rests on an examination of a greater number of alloys than other work and
allowed a more precise determination of the liquid composition at U16. Ternary eutectic reactions in
Mg-rich alloys occur at E6 and E9. The reaction temperature at E6 is 1°C [1932Por, 1933Bas, 1934Por] or
2°C [1949Ura2] below the binary Cu-Mg eutectic temperature. The ternary eutectic E9, Table 6, was
initially regarded as involving a Laves phase, but the work of [1951Mir2] indicates that this eutectic
involves the Q phase, which was not detected by the previous workers. Faudot et al. [1998Fau, 1999Fau]
confirmed the eutectic, Table 6. The ternary eutectic reaction at E9 was found by [1949Ura2] at 423°C, what
agrees well with that calculated by [2003Jan], 424°C. The reaction at U13 was regarded as a transition
reaction by [1937Nis1, 1952Han], as calculated by [2003Jan], whereas [1946Ura] and [1949Ura2]
considered it to be a ternary peritectic reaction, L+�1+ST. [1951Mir2] gives it as L+�1+SQ. There is
doubt about this reaction on two counts. The Q phase lies virtually on the L-�1 tie line [1952Ura] and it is
unlikely that the Laves phase is �1. For the reactions U15 and U18 [2003Jan] reproduced those given by
[1951Mir2] with 3°C deviation. For U18 Faudot et al. [1998Fau] gave 427°C as calculated by [1998Buh,
2003Jan]. But later [1999Fau] found it at 451°C with a more Al-rich liquid, Table 6. The transition reaction
at U17 was given by [1951Mir2] as L+�1(Mg)+Q, but the work of [1981Mel2] indicates �3 as the reactant
rather than �1, whereas [1998Fau, 1999Fau, 2003Jan] assume �2. The reactions in the Cu-rich corner have
been little studied. In Table 2 are given those calculated by [2003Jan]. [1949Ura2] assumed an eutectic
instead of U2 and a transition reaction instead of E4. The temperatures of the invariant equilibrium in this
area calculated by the two groups [1997Che] and [1998Buh, 2003Jan] deviate up to 20°C. The regions of
primary solidification of the Laves phases �1, �2 and �3 have not been experimentally defined, but the
calculation [2003Jan] gives them as shown in Fig. 2. [1997Che] did not distinguish these Laves phases.
Liquidus Surface
A liquidus projection, Fig. 2b, is taken from the calculation of [2003Jan] with some minor modifications on
the edges according to the binary systems accepted in this assessment. It should be compared with the
projection, Figs. 2 and 2a, deduced also from the calculations of [2003Jan]. The liquidus in the ternary
diagram was also calculated by [2001Che, 2002Che] using the multicomponent phase diagram calculation
software PANDAT. [1999Xie] also studied the liquidus projection in the Al rich corner. Results are in
agreemnt with those calculated by [2003Jan]. According to the liquidus of the binary systems accepted in
this assessment, the liquidus projection was modified at the edge boundaries. The liquidus isotherms
reproduce fairly well those assessed by [1991Eff]. The primary (Al) region has been widely studied with
general agreement on the form of the liquidus. The isotherms for the region of primary solidification of the
series of Cu-rich Al-Cu phases are uncertain.
Isothermal Sections
The calculated 400°C isothermal section calculated by [2003Jan], Fig. 4, agrees with Fig. 4 of [1991Eff]
except the broadening of the homogeneity range of �1 near 25 at.% Al, which in calculation needs to model
an anomaly in the Gibbs energy description at that composition, but there is no other evidence for an
anomaly. The phase Mg2Al3 is simplified as a stoichiometric phase as well as the CuMg2, �, � and
phases.
The solubility of Cu and Mg in Al-rich alloys at 460°C was determined by [1944Lit] and [1947Str], Fig. 5.
[1944Lit] also produced data for 375°C. The results of [1932Dix] agree with the solubilities given in Fig. 5.
[1946Pet] found lower Mg solubilities but used fewer alloys. [1955Zam] published solubility curves with a
50
Landolt-Börnstein
New Series IV/11A2
MSIT®
Al–Cu–Mg
series of cusps that cannot be reconciled with the alloy constitution. The solubilities of Mg and Cu in (Al)
reported in the accepted binary systems have also been taken into account in Fig. 5. The calculated solvus
isotherms of [1986Cha] and [2003Jan], Fig. 6, are in good agreement with [1944Lit] and [1947Str].
[1957Rog] reported the solubility of Al and Mg in (Cu), Fig. 7. No comparable work has appeared. In this
area the calculation is less reliable, as it cannot be based on adjacent experimental data. More extensive
isothermal sections were determined by [1946Pet] at 400°C in the region from Al to S and T. [1949Mir]
reported on the S phase region at 420°C, [1949Ura1] on the T and �1 phase region at 400°C, [1951Mir1]
on the Q phase region at 400°C, [1952Ura] on an almost complete isothermal section at 400°C and
[1981Mel2] on the region from 33.3 to 100 at.% Mg at 400°C. [1944Lit] and [1947Str] studied the 460°C
isothermal section from Al to the �, S, Q and T phases.
Temperature – Composition Sections
The liquidus and solidus of the Al rich alloys along the isopleth Al-Cu0.5Mg0.5 were calculated by
[1997Che, 1999Xie, 2000Lia] using thermodynamic descriptions. The measured solidus data found by
[1988Mur] was found to be � 0.5 at.% higher than the model-calculated values, while the measured liquidus
is in good agreement with the model-calculation. The inaccuracy for the solidus is explained by
microsegregations occuring in ternary Al-Cu-Mg alloys [1999Xie].
Several isopleths were calculated by [1997Che, 2003Jan] from thermodynamic descriptions. Figs. 8 and 9a,
9b, 9c show isopleth sections at 33.3 at.% Mg and x mass% Al (x =60, 70 and 95.5) respectively. The
calculated isopleth, taken from [2003Jan] and reported on Fig. 8, is in agreement with the experimental data
reported by [1936Lav1] and [1953Kle]. The calculated isopleths reported on Figs. 9a, 9b and 9c are taken
from [1998Buh] and describe quite well the experimental information reported by [1937Nis1, 1937Nis2,
1952Han] and [1946Ura]. The calculated isopleths at 37 at.% Al (Fig. 10a) and 43.75 at.% Al (Fig. 10b)
show the �2 and the Q phases formations respectively [2003Jan].
Thermodynamics
[1972Pre] studied the enthalpy of formation of alloys on the 33.3 at.% Mg section. Substitution of Cu by Al
increases the stability of the �1 phase although there is a decrease of stability at a valency electron
concentration of 1.5 (76.9Cu, 17.3Mg). [1987Hoc] calculated the enthalpy of a ternary alloy containing
33.3% “MgAl2”; agreement with [1972Pre] is fair. [1985Kuz] applied a thermodynamic model to predict
the ternary solidus from the ternary liquidus and the binary solidus-liquidus for Al-rich alloys. [1973Dav]
used quasi-chemical regular solution theory to calculate the monovariant curve e2E5 of Fig. 2a. With the
introduction of a ternary interaction parameter the calculated ternary eutectic point E5, Table 3, shows
reasonable agreement with the assessed composition. [1987Lac] calculated the Al-rich region of the phase
diagram using an extended Redlich-Kister formalism. Excellent agreement was obtained with the assessed
liquidus, Fig. 2b. [1985Far] calculated the composition of the ternary eutectic E5, Fig. 2a and Table 2,
assuming both ideal solution behaviour and regular solution behaviour. The calculated eutectic
compositions, 34.4Cu-8.8Mg (mass%) for ideal solutions and 30.3Cu-7.5Mg (mass%) for regular solutions,
approximate to the assessed values. The calculated eutectic temperatures are surprisingly low at 273°C and
271°C, respectively. Recently two groups [1997Che] and [1998Buh, 2003Jan] calculated the whole ternary
system, describing the Gibbs energies of all phases involved by the compound energy formalism. Both
calculations show very similar results, only in the Cu-rich part there is some disagreement of the invariant
temperatures (up to 20°C). The first group also calculated solidification paths using the model of Scheil
[1993Zuo, 1996Zuo].
[1986Che] measured the enthalpy of fusion of the ternary eutectic E5 as 365 J#g
-1 corresponding to 11.8
kJ#mol-1 of atoms. [1986Not] measured the enthalpy of formation of the S phase as -63.2 ± 4.0 kJ#mol-1 of
CuMgAl2. [1995Kim] measured the enthalpy of mixing of ternary liquids by a high temperature calorimeter
at 713°C along three lines with constant Al/Mg ratios up to 40 at.% Cu and along Al/Cu = 13/7 up to 27
at.% Mg. [1995Soa] measured the chemical potential of Mg in ternary melts by an isopiestic method.
51
Landolt-Börnstein
New Series IV/11A2
MSIT®
Al–Cu–Mg
Notes on Materials Properties and Applications
The mechanical properties such as tensile strengh were investigated by [2002Dav] on
0.02Zn-0.05Ti-0.42Mn-0.27Fe-4.5Cu-1.5Mg-Al-0.17Si alloys.
[2002Zhu] reported that a small addition of Ag (< 0.1 at.%) to an Al-Cu-Mg alloy with a high content of Al
promote an increasing strength and creep resistance when compared to Al-Cu-Mg alloys that contain only
the CuAl2 precipitate.
Miscellaneous
[1940Kuz] and [1946Kuz] measured lattice spacings of the (Al) phase along sections from Al with various
Cu:Mg ratios. [1951Poo] measured the lattice spacings of the (Al) phase along sections from 99 at.% Al, 1
at.% Mg to 99.5 at.% Al, 0.5 at.% Cu and from 98 at.% Al, 2 at.% Mg to 99 at.% Al, 1 at.% Cu, Table 7.
A small addition of Mg to Al-Cu alloys accelerates the formation of Guinier-Preston (GP) zones through
the Mg/Cu/vacancy complexes mechanism [2000Hir, 2002Hir].
The crystal structure of a metastable variant of S on aging Al alloys was studied by [1950Bag]. Aging
studies of single crystals of an alloy containing 1.2 at.% Cu, 1.2 at.% Mg [1978Ale] showed S particles to
be coherent with the Al matrix. The effect of aging on mechanical properties of Al-rich alloys have been
reported by [1939Han, 1941Mec] and [1948Sha]. More recent studies on metastable precipitates in (Al) are
from [1990Gar] and [1991Jin].
[1959Pal] prepared thin film Al-rich ternary alloys by evaporation on to Al substrates. The constitution is
claimed to correspond with bulk samples. Ther