Hundred years of precipitation hardening

Erhard Hornbogen
*
Institute for Materials, Ruhr-University Bochum, Universitatsstr. 150, 44801 Bochum, Germany
Abstract
About 100 years ago, precipitation hardening was discovered and the rst structural aluminum alloys with considerable strength
were developed. Our present understanding of the phenomenon in based on the application of dislocation theory as well as reaction
kinetics, especially the role of metastable phases and lattice defects in nucleation from supersaturated solids. The prerequisites for
optimum precipitation hardening can be precisely dened. However, the progress in the development of new alloys has been slow in
the recent years. The original alloy is still in use today. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Precipitation hardening; Nano-dispersoids; Critical particle size; Coherency; Metastability; High strength; Incipient microstructure
1. A discovery
Around the turn of the last century, 1901, Dr. Alfred
Wilm started work to improve the strength of Al-alloys
which then left much to be desired. He joined a gov-
ernment laboratory close to Berlin. With a background
in Chemistry from Berlin and G ottingen Universities he
knew all about the physical metallurgy of his days. He
knew that steel can be hardened if the right composi-
tions were cooled fast enough. So he mixed a very large
number of Al-alloys and quenched them at dierent
rates. To his great frustration many alloys became even
softer the more rapidly they were quenched. The saga of
his discovery tells that Wilm was a devoted researcher,
but a man with other hobbies too. He conducted many
quenching experiments with AlCu-alloys with small
amounts of additional elements on a Saturday morning.
The sun was shining and so he remembered his other
hobby, interrupted his work and went out sailing on the
Havel-river for the whole weekend. On Monday morn-
ing, rst hardness measurements, then tensile tests were
resumed of the quenched alloys. To his greatest aston-
ishment, hardness had increased considerably during the
two days and so did the tensile properties. The experi-
ments were repeated systematically, by 1906 an alloy of
Al with 3.55.5% Cu, plus less than 1% Mg and Mn was
patented [1]. This alloy is still in use 100 years later. The
industrial rm which bought the patent invented the
trade name. Soon Zeppelins were ying, constructed of
DURALUMIN. Wilm was not an eager writer. His rst
publication on age-hardening is dated 1911 [2]. He did
not know that he had invented the rst nano-technol-
ogy. Soon afterwards, 1919, he took up another hobby.
He retreated from research and became a happy farmer
until the end of his life in 1937.
The scientic community was rst baed by the
phenomenon. But already by 1919 an American had the
right answer: precipitation from solid state [3]. Merica
was not believed at rst, because nothing could be seen
in the microscope. Also electrical conductivity did not
show the expected increase. During aging at ambient
temperature conductivity even decreased. But it did not
take long until precipitation was acknowledged as the
cause of hardening at least for elevated aging tempera-
tures. Inspired by Ostwald's step rule [4], Wassermann
introduced in 1935 the idea of intermediate, metastable
phases which could be recognized by X-ray diraction
[5,6]. For the explanation of hardening by aging at
ambient temperatures, the discovery of coherent Cu-rich
zones in 1937 was required [7,8]. The direct observation
of these GuinierPreston zones, as well as all the details
of other dispersoids and their interfaces in the nm-range,
had to wait until the advent of transmission electron
microscopy in 1959 [9,10].
2. Hardening
The quantitative model for the explanation of the
hardening eect did however not have to wait for the
Journal of Light Metals 1 (2001) 127132
www.elsevier.com/locate/lightmetals
*
Fax: +49-234-32-14235.
E-mail address: erhard.hornbogen@ruhr-uni-bochum.de (E. Horn-
bogen).
1471-5317/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 1 4 7 1 - 5 3 1 7 ( 0 1 ) 0 0 0 0 6 - 2
direct observation of the relevant microstructure. Three
requirements had to be fullled in Orowan's theory of
1948 (Fig. 1) [11]:
1. a two-phase microstructure of the dispersoid type
which impedes motion of dislocations in the matrix,
2. particle diameters and spacings not much larger than
atomic size b: S
p
100b,
3. particles strong enough to promote looping of dislo-
cations without being sheared: d
c
< d
p
.
The force F which a particle is able to sustain depends
on its intrinsic strength and on the particle diameter.
Bypassing must take place for
F > Gb
2
= [N[: (1)
For coherent, ordered particles (c
APB
energy of the
antiphase domain boundaries) F = Cc
APB
d
p
.
The critical particle diameter follows as:
d
c
= C
Gb
2
c
APB
(2)
and C is a geometrical factor given by the shape of the
particle. The other extreme case is a non-coherent par-
ticle with a high shear modulus G
b
. A dislocation has to
be recreated in order to shear the particle. Its critical
diameter is consequently very small:
d
c
=
4pbG
a
G
b
: (3)
The coherent case would be representative for a d
/

Al
3
Li particle or a H
//
zone in AlCu-alloys. Hard non-
coherent particles are Si or H-Al
2
Cu. The condition
d
p
> d
c
limits the applicability of Eq. (4) which expresses
the increase in yield stress due to dispersoid particles
Dr
p
(see Fig. 1) [11,12]:
Dr
p
=
Gb
S
p
=
Gb
S d
p
=
Gb
d
p
f
1=3
p
; (4)
where S is the spacing of the particle center. S ~ S
p
for
small volume fractions f K1%, which are customary in
Al-alloys. These relations provide a good estimate of the
amounts of strengthening which can be expected in Al-
alloys, and dene d
c
as an important microstructural
parameter (Figs. 2, 3 and Table 1).
3. Nano-dispersoids
The understanding of precipitation hardening denes
which microstructure has to be aspired for: an even
dispersoid of nanometer-sized particles (Fig. 1):
Fig. 1. Types of two-phase microstructures: dispersion, net, cell, du-
plex.
Fig. 2. Hardening Dr
p
of Al depending on volume fraction f and di-
ameter d of particles.
Fig. 3. Force F exerted on particle by looping dislocation, schematic.
128 E. Hornbogen / Journal of Light Metals 1 (2001) 127132
d
c
< d
p
MIN: (5)
Classical nucleation theory implies a decreasing critical
nucleus size d
N
with increasing driving free energy g
ab
(or undercooling) for a reaction a
ss
a b. Increasing
interfacial c
ab
energy increases d
N
, and therefore lowers
the probability of nucleation:
d
N
~
c
ab
g
ab
=
J m
2
J m
3

: (6)
No Al-based alloy system is known in which a stable
phase can form coherently with the fcc matrix (as c c
/
in Ni-based alloys). The interfacial energy and therefore
the nucleation barrier is high for the formation of
equilibrium phases from an Al-based supersaturated
solid solution a
ss
: c
fc
< c
sc
< c
nc
(fully-, semi-, non-co-
herent: fc, sc, nc).
In addition, many equilibrium phases show a large
size of the elementary cells a
0
as H-Al
2
Cu. It has
therefore to be considered that d
N
a
0
for very large
undercoolings (Fig. 4).
d
N
(T) a
0
~
c
aa
g
ab
: (7)
This shows why Al-alloys are not good candidates
for the formation of ne dispersoids of non-coher-
ent equilibrium phases by homogeneous nucleation
(Fig. 5).
The smallest elementary cell of a hard equilibrium
phase b provides the diamond structure of Si. The
mechanism of the precipitation reaction is shown in
Fig. 6. The microstructure and the well-known quench-
sensitivity of AlSi provide evidence that Si particles are
not homogeneously nucleated. Heterogeneous nucle-
ation, in this case vacancy-aided, determines the size and
site of particles.
4. Competitive evolution of microstructure
The nucleation problems of the equilibrium phases in
Al-alloys have two consequences: (a) defect-aided nu-
cleation is frequent, and (b) the formation of less stable
Fig. 4. Nucleus size d
N
as a function of undercooling T
e
T = DT, a
0
size of the elementary cell.
Fig. 5. Typical incipient microstructure formed by competitive nucle-
ation mechanisms, schematic.
Table 1
Critical particle diameters d
c
(shearingbypassing)
Examples Coherency d
c
(nm)
Si, Ge n, p 2 Diamond structure
H
/
-Al
2
Cu T-Al
2
CuLi n, p 310 Intermetallic compounds
Zn n, p 20 Non-coherent, solid solutions
Al
3
Li c >50 Coherent ordered, depending on
coherency stress, c
APB
E. Hornbogen / Journal of Light Metals 1 (2001) 127132 129
but more coherent phases is favored. Only a few systems
(-Si, -Ge, -Zn) are not able to form metastable inter-
metallic compounds.
The structure of the a=b-interface, and of a particular
lattice defect, leads to favorable combinations: non-co-
herent +grain boundary, semi-coherent +dislocation,
etc. A scheme can be set up for the possible total number
of reactions N, with the number of phases N
ph
, the
number of varieties of defects N
d
, and the possibility of
discontinuous or continuous mode N
m
:
N = N
ph
N
d
N
m
: (8)
Table 2 shows the typical example of the AlCu-alloys
with N > 10. Out of the large number of options only a
few will occur in a special situation which depends, for
example, on the mechano-thermal past history. What is
obtained in the early stages of aging is an incipient
microstructure usually far away from thermodynamical
equilibrium and from the structure most desirable for
maximum hardening.
It is an important fundamental question, which
principle is controlling the structure which we nd in the
peak-aged (PA) condition? Observed is the formation of
phases which can form most quickly for dierent
reasons: nucleation or growth. Thus, the initial rate of
precipitation from the matrix a
SS
a is maximized.
This in turn implies a high initial rate of change in free
energy, and raise in thermal entropy S
th
: dissipative
structures:
dS
th
dt
= max : (9)
This implies that secondary reactions are required
to approach the nal equilibrium: dissipative
equilibrium structure. Such reactions are of great im-
portance for the thermal stability of microstructures, if
the alloy is used at elevated temperatures, following the
primary heat treatment. Such secondary reactions are
well known in most precipitation hardening Al-alloys.
Concerned are the phases which have been formed fast
initially, but which are not stable. Connected with the
transition to more stability is usually a loss of coherency
with the fcc matrix. Fig. 7 shows the in situ transition of
semi-coherent H
/
to stable non-coherent H-Al
2
Cu. The
clue is the creation of the interfacial defects by the
growing metastable phase, which is required to aid nu-
cleation of the next stable phase [13].
With respect to precipitation hardening, it has to be
noted that incoherent, hard particles can be formed by
in situ transition without pre-existing defects. However,
this process is connected with an anomalous coarsening
eect (as compared to the Ostwald ripening of a non-
transforming phase). Only a very small portion (the
largest particles of a size distribution) will transform. In
their environment all the smaller particles are dissolved
subsequently. Consequently, they are no longer able to
transform into the stable phase which forms a coarse
dispersoid.
5. Eects of third or trace elements
Favorable alloying addition may either raise peak
hardness or stabilize a microstructure against overaging.
The latter can be based on a retardation of an in situ
transformation of a metastable phase. This is shown for
the well-known eect of Ag on AlCu or AlCu, Mg-
alloys. The retardation can have two reasons (Fig. 8). If
the solubility diers in the two phases, diusion gradi-
ents build up at the interfaces. Ag shows no solubility in
H-Al
2
Cu, except in H
/
or S and in the fcc matrix.
Table 2
Options for nucleation mechanisms in AlCu alloys (defects and metastable phases)
Defects geom.
dimension
0-d 1-d 2-d
Crystal structure Stability Coherency with fcc No defect Vacancy v Dislocation d Grain boundary b
Cluster, GP-zones | c + + + +
H
//
| c + + ++ +++
H
/
| pc +++ ++ + ++
H max nc +++ +++ ++ +
c coherent; pc partially c; nc non c; + low activation barrier for nucleation; +++ high activation barrier for nucleation.
Fig. 6. Formation of dc-Si from the fcc-A (Si)-solid solution. The
volume change favors vacancies as nucleation cites.
130 E. Hornbogen / Journal of Light Metals 1 (2001) 127132
Consequently it is expelled from stable H as it grows.
There is evidence that strong segregation in the newly
forming non-coherent interphase plays an important
role in strongly retarding the growth of the stable phase.
This in turn preserves the much ner dispersoid of the
less stable phase S for longer periods and at higher
temperatures [1416].
A dierent alloying eect is based on the empirical
observation that all ``successful'' alloys, i.e. those
showing considerable hardening are ternary alloys
containing one species of atom larger, a second one
smaller than the atomic volume of aluminum: ()-rule
(Table 3). The explanation for this eect goes back to
the earliest stage of the reaction when a critical nucleus
is formed (Eqs. (6) and (7)). Atoms of dierent sizes
reduce the local strain energy and therefore favor the
formation of larger clusters (as compared to statistical
uctuations) (Fig. 9). Chemical attraction of atoms
should have the same eect. These clusters should reach
the critical size d
N
(Eq. (6)) with more ease. Therefore
nucleation will shift somewhat from vacancy aided to a
more homogeneous mode. This is shown in Figs. 6(c)
and 9 for AlSi +Ge alloys. Si and Ge comply with the
()-rule and are chemically indierent to each other.
Further exploitation of this rule seems to be a vehicle for
the systematic search for new precipitation hardening
alloys for aluminum and other metals as well.
Table 3
Atomic size ratios of Al and solutes, qualitative
Smaller Larger
Cu Ti
Zn Sn
Si Cd
Ni Ge
Co Li
Fe Mg
Mn Ga
Cr
Fig. 8. Formation of diusion gradients of Ag during the S H
transition in AlCuAg alloy.
Fig. 7. In situ transition H
/
H in AlCu-alloys, TEM. (a) Al 1.3 At%Cu 100 m, 300C (b) Al 1.3 At%Cu 1000 m, 300C (c) Al 1.3 At%Cu 1000 m,
300C (d) Al 2.2 At%Cu 10 m, 300C.
E. Hornbogen / Journal of Light Metals 1 (2001) 127132 131
6. Conclusions
The conditions for easy homogeneous nucleation
(coherency, no change in the type of bonding) are op-
posite to those required for precipitation hardening. The
desired situation can be dened by the following scheme:
MIN d
N
6d
p
Pd
c
MIN
|
MIN:
(10)
It indicates that homogeneous nucleation should create
a ne dispersoid of very small particles, which are
however bigger than the critical size, d
c
for the bypassing
of dislocations. These dispersoids should be as small as
possible. All useful microstructures in high strength Al-
alloys are dissipative structures, which represent some
compromise between these requirements. Reasonable
results are obtained with relatively strong ternary in-
termetallic compounds with partially coherent inter-
faces. An ideal case is the dc-structure of Si, as its
formation is closer to the homogeneous nucleation.
Precipitation hardening Dr
p
is only one strengthening
mechanism, which contributes to the total strength of an
Al-alloy (r
y
, yield stress):
r
y
= r
0
Dr
s
Dr
d
Dr
b
Dr
p
: (11)
Theory of strengthening provides reliable information
on the limits for improvement from Dr
p
(Figs. 2, 3 and
Eq. (4)). Precipitation hardening is neither independent
of the contribution from dislocations Dr
d
(work hard-
ening) nor boundaries Dr
b
. Strengthening of the matrix
Dr
s
seems to be promising for obtaining a still higher
strength. Order hardening (as d
/
-Al
3
Li), or in some cases
even quasicrystalline matrices, may also lead to further
strengthening [1719].
In spite of the high level of scientic understanding,
development of new precipitation hardening alloys has
been sparse. Alfred Wilm's alloy is hardly surpassed,
and it is very much in use today.
References
[1] A. Wilm, DRP 244554 (German patent) 1906.
[2] A. Wilm, Metallurgie 8 (1911) 223.
[3] P.D. Merica, R.G. Waltenberg, J.R. Freeman, Sci. Pap. Bur.
Stand. 337 (1919);
P.D. Merica, R.G. Waltenberg, J.R. Freeman, Trans. AIME 64
(1921) 3;
P.D. Merica, R.G. Waltenberg, H. Scott, Trans. AIME 64 (1921)
41.
[4] W. Ostwald, Z. Phys. Chem. 22 (1897) 289330.
[5] G. Wassermann, Z. Metallk. 48 (1957) 223231.
[6] G. Wassermann, J. Weerts, Z. Metallwirtschaft 14 (1935) 605.
[7] A. Guinier, Nature 142 (1938) 569.
[8] V. Gerold, Z. Metallk. 45 (1954) 593599.
[9] A.K. Nicholson, R.B. Progr. Mater. Sci. 10 (1963) 243.
[10] E. Hornbogen, Nucleation of precipitates in defect solid solu-
tions, in: A.C. Zettelmeyer (Ed.), Nucleation, Marcel Dekker,
New York, 1969, pp. 309378.
[11] E. Orowan, Symp. Int. Stress, J. Met. (1948) 451.
[12] L.M. Brown, R.K. Ham, in: A. Kelly, R.B. Nicholson (Eds.),
Strengthening Methods in Crystals, 1971, p. 331.
[13] E. Hornbogen, Met. Trans. A 9 (1978) 134.
[14] I.J. Polmear, Light Alloys, third ed., Arnold, London, 1995, pp.
41105.
[15] I.J. Polmear, Mater. Sci. Forum 13 (1987) 195.
[16] A.K. Mukhopadhyay, Scr. Mater. 41 (1999) 667672.
[17] B.M. Gable et al., J. Light Met. 1 (2001) 1.
[18] A. Inoue, H. Kimura, J. Light Met. 1 (2001) 31.
[19] E. Hornbogen, E. Starke, Acta Mater. 41 (1993) 1.
Fig. 9. Stress elds connected with clusters of Si, Ge, Si +Ge in Al-
solid solution, schematic.
132 E. Hornbogen / Journal of Light Metals 1 (2001) 127132