schmidt1999
schmidt1999
org
Email alerting service Receive free email alerts when new articles cite this article - sign up in the box at the top right-hand
corner of the article or click here
1. Introduction
Due to their superior mechanical properties at elevated temperatures, single-crystal
Ni-base superalloys are widely used in high-temperature applications, such as gas
turbine blades. The high creep resistance of these materials is related to their micro-
structure, which initially consists of a highly regular arrangement of cuboidal pre-
cipitates with an ordered structure (the γ 0 -phase) in a matrix with a disordered
structure (the γ-phase). This microstructure is known to undergo drastic changes
when the material is subjected to an external uniaxial load at high temperatures.
The microstructural evolution during this coarsening process is called ‘rafting’ and
has been experimentally observed to result in γ 0 precipitate morphologies that are
either plates perpendicular to, or rods parallel to, the load axis. For an overview of
experimental observations, see Chang & Allen (1991). While the creep resistance of
these rafted microstructures can be either enhanced or diminished as compared with
the initial morphology (see, for example, Müller (1993), and references therein), it
is clear that it must be related to the microstructural changes. The prediction of
Proc. R. Soc. Lond. A (1999) 455, 3085–3106 c 1999 The Royal Society
Printed in Great Britain 3085 TEX Paper
Downloaded from rspa.royalsocietypublishing.org
the rafting behaviour of these materials has therefore been, and continues to be, a
subject of interest.
Generally, the rafting direction has been found to depend on the sign of the exter-
nal load, the sign of lattice misfit between the γ and γ 0 -phases, and the difference
between the elastic constants of the two phases. Classical rafting models are based on
the calculation of the strain energy associated with a coherently misfitting inhomoge-
neous inclusion subject to an external load. Tien & Copley (1971) compared approx-
imations of the strain energy of three distinct morphologies (cube, plate, rectangular
parallelepiped) to explain their experimental observations. Using Eshelby’s solution
for ellipsoidal inclusions and the ‘equivalent inclusion method’ (Eshelby 1957) and
assuming isotropic phases, Pineau (1976) constructed a map for the prediction of
the rafting behaviour by comparing strain energies of a spherical, a plate-like and a
rod-shaped morphology of a single inclusion. These results are particularly attrac-
tive because they provide simple criteria for the prediction of rafting in terms of two
dimensionless quantities, a load parameter and the ratio of Young’s moduli of the two
phases. Pineau’s well-known conclusion is that the rafting direction is reversed when
this ratio passes through unity. The cubic anisotropy of the phases has been taken
into account by Miyazaki et al. (1979), who considered an isolated ellipsoidal particle
of variable aspect ratio. From energy versus aspect ratio plots, the most favourable
shape is then inferred. Only one set of material parameters has been used in this
work, which does not allow us to draw any conclusions from their results. Chang
& Allen (1991) used a different formula for the numerical evaluation of the Eshelby
tensor and presented energy versus aspect ratio plots for various material parame-
ters. However, the inclusion’s elastic constants were taken to be a simple multiple of
the matrix’s elastic constants, i.e. Cij (precipitate) = αCij (matrix), and, therefore,
the anisotropy of the two phases was constrained to be equal. Their result was that
the rafting behaviour is reversed if the sign of either the external load or the misfit
changes, or if the proportionality factor between the elastic constants passes through
1. Focusing on shape bifurcation phenomena, Johnson et al. (1988) considered an
inhomogeneous ellipsoidal particle in an orthotropic system and used a Landau-type
expansion of the energy, which requires that the difference in the elastic constants
be small. They extracted two dimensionless parameters from their expansion and
derived relations between them that indicate which type of shape transition is to
be expected at what particle size. One of their findings is that the rafting direction
should be reversed if the sign of [[C11 − C12 ]] changes. All of the above-mentioned
investigations examined an isolated particle and thus neglected particle interaction,
which is considered an open question in these studies.
Also based on the assumption of linearly elastic material behaviour and fully coher-
ent interfaces are the Monte Carlo-type simulations of the coarsening process pub-
lished by Gayda & Srolovitz (1989), Lee (1996) and Laberge et al. (1997). In these
works, the two-phase structure is broken into a large number of elementary particles,
to each of which one of the two phases is assigned. In the Monte Carlo procedure, an
exchange of two of these elementary particles is either accepted or rejected, depend-
ing on the change of elastic energy associated with this exchange. The changes in the
elastic energy are calculated either with a finite-element method in an approximate
manner (Gayda & Srolovitz 1989) or through the assumption of a pairwise interac-
tion potential resembling Cauchy’s elasticity model (Lee 1996; Laberge et al. 1997).
In fact, the specific modelling of anisotropic elasticity used in Lee (1996) confines
the elastic constants of both phases to obey Cauchy’s rule, i.e. C12 = C44 , and,
in Laberge et al. (1997), small differences between the elastic constants, and equal
anisotropies, in matrix and precipitate are assumed. Neither study presents param-
eter studies that clarify the influence of anisotropy on the rafting process. Ohashi et
al. (1997) derived a map for the prediction of the rafting behaviour based on approx-
imate expressions for the stresses and strains in a periodic cuboid arrangement and
the assertion that rafting takes place in the direction of the matrix channel with the
highest strain energy.
The fact that in most experiments rafting is associated with a finite amount of
creep strain (see, for example, Pollock & Argon 1994) raises the question of whether
purely elasticity-based models are sufficient for the description of rafting. A criterion
for the expected rafting behaviour that identifies the formation of interfacial misfit
dislocations as the relevant mechanism has been proposed by Valles & Arrell (1994).
In a rather heuristic way, this criterion compares the difference in the lattice spac-
ing on interfaces parallel to the load with that perpendicular to the applied load,
assuming a uniaxial state of stress. This criterion implies that the rafting direction
is reversed when the difference of the Young’s moduli of the two phases changes
sign. Veron et al. (1996) take both elastic and plastic effects into account to set up
a dynamic model in which the drastic simplification of assuming a uniaxial state of
stress in the composite is made.
Also motivated by the experimental observation of dislocation networks on the
γ–γ 0 interface, Socrate & Parks (1993) addressed the question of how elasto-plastic
material behaviour in the matrix affects the microstructural evolution. Their analy-
sis is based on the numerical evaluation, via the finite-element method, of Eshelby’s
‘force on an interface’ for a two-dimensional periodic arrangement of cuboidal parti-
cles. From the profiles of this generalized traction τn along the interface, the rafting
direction is then inferred in much the same way as we shall proceed in this paper. It
is shown that the incorporation of inelastic material behaviour in the matrix signifi-
cantly changes the expected rafting behaviour, their results implying that only the
sign of misfit and applied stress determine the rafting direction. Two of the examples
studied therein will be discussed later. Also using the concept of a thermodynamic
force on an interface, Nabarro et al. (1996) derived criteria for the prediction of the
rafting direction that are based on the assumption of high γ 0 volume fractions and
small differences in the elastic constants. They confirmed the result of Johnson et al.
(1988) and also gave an explanation for the effect of plastic straining in the matrix
as observed in the calculations of Socrate & Parks (1993).
In this context, Paris et al. (1997) conducted experiments that were aimed at
clarifying whether rafting can be detected in the elastic regime. Their conclusion
was that while there are good reasons to believe that the rafting process is plasticity
driven, ‘the results alone. . . do not allow for a discrimination between “elastic” and
“plastic” rafting models’.
It is the purpose of this paper to contribute to this discussion by supplying results
of an elasticity-based model for the prediction of the rafting direction at the onset of
rafting that provides analytical results and reduces simplifications and assumptions
regarding the particle shape, particle interaction and elastic constants to a minimum.
The outline of the paper is as follows. In § 2, some basic equations are recalled
for completeness and clarity. In § 3, a generally valid formula for Eshelby’s gen-
eralized traction is derived, which allows distinction between the influence of the
δw
x1
details of the elastic fields and the eigenstrain. In § 4, it is shown that the problem
of an inhomogeneous inclusion of arbitrary shape under external load is equivalent
to an inhomogeneous inclusion without external load and a modified eigenstrain. It
is further shown that the driving force for this equivalent problem coincides with
the driving force for the original problem. Together with certain symmetry consid-
erations, these results are then used in § 5 to construct a morphology map for the
two-dimensional case of a plane-strain state. The same arguments are then extended
to the three-dimensional case in § 6. A discussion and a summary of the results and
of some experimental observations are presented in §§ 7 and 8.
2. Basic equations
S
Consider a finite domain K = K− K+ , in which the region K− exhibits an eigen-
strain ε0 (figure 1). The stresses in K obey Hooke’s law,
(
C + , in K+ ,
σ= − 0 (2.1)
C ( − ), in K− ,
where = 12 (∇u + ∇uT ) is the total strain, and u is the displacement measured
from the stress-free state of the matrix (+) phase. For a coherent interface, the
displacement is continuous,
[[u]] = 0, on F, (2.2)
where F = ∂K− and [[·]] = (·)+ − (·)− denotes the jump of a quantity across F.
Mechanical equilibrium demands
div σ = 0, in K± ,
(2.3)
[[σ]]n = 0, on F,
where n is the unit normal on F pointing from K− to K+ , along with the boundary
condition
σν = 0, on ∂K, (2.4)
where ν is the unit outward normal on ∂K. The potential of the system is the strain
energy and is given by†
Z Z
1 + 1
Π= · C dV + ( − 0 ) · C − ( − 0 ) dV. (2.5)
2 K+ 2 K−
where τn = n · [[P ]]n is the ‘driving force’, and P = W 1 − ∇uT σ is Eshelby’s energy
momentum tensor, in which W is the strain energy density. With the continuity
condition (2.3)2 , the driving force becomes
∂u
τn = [[W ]] − t · , (3.2)
∂n
which can be specialized further when the jump conditions (2.2) and (2.3)2 are used
to calculate the jump of the normal derivative of u to yield (Schmidt & Gross 1995,
1997)
from which, by comparison with (2.1), it readily follows that the prime quantities
are the solution for the problem without external load if the ‘equivalent eigenstrain’,
given by
0 0
ε = ε0 − [[S]]σ ∞ , (4.6)
is used. The true fields for the original problem then follow from (4.2).
This procedure is somewhat similar to Eshelby’s well-known ‘equivalent inclusion
method’ (Eshelby 1957). But, unlike the latter, it can be applied to arbitrary inclusion
shapes, because it does not reduce the original problem to a homogeneous inclusion
without external load, but to an inhomogeneous inclusion without external load.
the second term being the potential of the external load. A lengthy calculation (see
Appendix A) yields the following expression for the potential,
Π = 0Π + ( 12 σ ∞ · [[S]]σ ∞ − σ ∞ · ε0 ) VK− − 12 σ ∞ · ε∞ VK , (4.8)
| {z }
λ
in which
Z Z
0 1 0 1
Π= ε · C + 0ε dV + ( 0ε − ε0 ) · C − ( 0ε − ε0 ) dV, (4.9)
2 K+ 2 K−
is the potential of the equivalent problem, and VK− , VK is the volume of the inclusion
and the whole body, respectively. Note that the potential of the original and the
equivalent problem differ only through terms that are independent of the inclusion
shape, but depend only on the volume of K− and K. From (4.8), the change of the
potential due to a change in particle morphology for the external load case is given
by
δΠ = δ 0Π + λδVK− , (4.10)
from which it follows that, if the volume of the inclusion remains constant during
the morphology change, then
Z
δΠ = δ 0Π = − 0
τn δw dA, (4.11)
F
i.e. the driving force for the original problem coincides with the driving force for the
equivalent problem.
It should be emphasized that the results (3.6), (3.7) and (4.8) remain unchanged
when there is more than one inclusion, because their derivation does not assume that
the domain K− is simply connected. Moreover, they are not affected by the geometry
of the body K and, thus, remain unchanged when considering an infinitely extended
matrix.
When the lattice misfit is purely dilatational, ε0 = ε∗ 1, and both phases are
orthotropic, as is the case in Ni-base superalloys, then, by virtue of (4.6), a uniaxial
stress σ0 in the x1 -direction leads to the equivalent eigenstrain,
0 0
ε11 = ε∗ − [[S11 ]]σ0 ,
0 0
ε22 = 0ε033 = ε∗ − [[S12 ]]σ0 , (4.12)
0 0 0 0 0 0
ε12 = ε13 = ε23 = 0,
where Sij are the compliances in the Voigt notation.
x2
x1
where ∆τn = τnA − τnB is the difference of the driving force at the symmetry-related
points A and B, and A denotes the part of the interface comprising points ‘of type A’
(figure 2). Since the potential must decrease for the actual morphology change (δΠ <
0), the important conclusion that can be drawn from (5.3) is this: if the ϕ-weighted
average of ∆τn is positive, then δ w̄ must be positive and the particle elongates
parallel to the load direction, while in the opposite case it elongates perpendicular
to the load direction. The next step is to express the difference of the driving force,
∆τn , in terms of the equivalent eigenstrain (4.6) by using the representation (3.6).
For the uniaxial load case it follows from (4.12) that the equivalent eigenstrain can
be written as
0 0 α 0 ε∗ − [[S11 ]]σ0
ε = ε̂ , where α = ∗ , ε̂ = 0ε022 . (5.4)
0 1 ε − [[S12 ]]σ0
Using a matrix representation for the fourth-rank tensor Ξ (i.e. Ξik ↔ Ξiikk ) and
exploiting its symmetries, the driving force reads
τn = 12 ε̂2 (α2 Ξ11 + Ξ22 + 2αΞ12 ), (5.5)
and, consequently,
∆τn = 12 ε̂2 [α2 Ξ11
A A
+ Ξ22 A
+ 2αΞ12 − (α2 Ξ11
B B
+ Ξ22 B
+ 2αΞ12 )]. (5.6)
As pointed out in § 3, the tensor Ξ is completely determined by the particle shape and
the material constants. It follows from the cubic symmetry of these that Ξijkl (x1 , x2 )
is invariant with respect to exchanging the indices 1 and 2. Hence,
B A B A B A A
Ξ11 = Ξ22 , Ξ22 = Ξ11 , Ξ12 = Ξ21 = Ξ12 , (5.7)
where the last equation follows from the symmetry (3.8) of Ξ itself. Inserting (5.7)
into (5.6) and rearranging terms yields
∆τn = 12 ε̂2 (1 − α2 )(Ξ22
A A
− Ξ11 ). (5.8)
This representation now allows for the important conclusion that the ϕ-weighted
average must change sign for α = ±1, and, hence, the rafting direction is reversed
for these α. Because neither the details of the elastic field nor those of the function ϕ
enter into the ratio of equivalent eigenstrain α, this conclusion holds independently
of the particle shape, possible particle interactions (i.e. the volume fraction), and
the nature of the morphological change ϕ. The direction itself follows from the sign
of the ϕ-weighted average of the driving force difference (denoted by ∆τ̄n ). Granted
A A
that Ξ̄22 − Ξ̄11 is positive (for which we will give some arguments later), the rafting
direction is dictated by
|α| < 1 ⇒ k rod parallel to the load direction,
(5.9)
|α| > 1 ⇒ ⊥ plate perpendicular to the load direction.
To visualize this result, let us examine the problem parameters for which α = const.
It is convenient to express the three independent elastic constants of the phases, C11 ,
C12 , C44 , through three coordinate invariant quantities µ̄, A, ν̄, via
2 + A 1 − 4ν̄ 2A 1 − 4ν̄ 2A
Cij = µ̄cij , c11 = 2 − , c12 = − , c44 = ,
1 + A 1 − 2ν̄ 1 + A 1 − 2ν̄ 1+A
(5.10)
where µ̄ is the average of the shear modulus over all directions in the (x1 , x2 )-plane,
A is the Zener–anisotropy ratio defined in the usual way, and ν̄ is an equivalent
Poisson number (see Appendix A). The compliances for a state of plane strain are
then given by
1 1+A
Sij = sij , s11 = 18 (3 + A − 4ν̄), s12 = 18 (1 − A − 4ν̄), s44 = .
µ̄ 2A
(5.11)
1 − (s+ −
11 − s11 /ω)p
α= , (5.13)
1 − (s+ −
12 − s12 /ω)p
where
µ̄− σ0
ω= , p= , (5.14)
µ̄+ µ̄+ ε∗
are the dimensionless average stiffness ratio and load parameter, respectively. From
(5.13) it follows that, in a p–ω diagram, lines α = const. are generally hyperbolas
obeying
p ω
− 1 − 1 = 1, (5.15)
p∗ ω∗
or, equivalently,
pω ∗ ωp∗
ω(p) = , p(ω) = , (5.16)
p − p∗ ω − ω∗
where p∗ and ω ∗ are the asymptotes of the hyperbola and are given by
1−α s− −
12 − αs12
p∗ = , ω∗ = + . (5.17)
s12 − αs+
+
12 s+
12 − αs12
where
β ± = (1 − 2ν̄ ± )/(1 + A± ). (5.21)
ω 1*
ω
ω 2*
0
0 p* p
2 0
p
Figure 3. Load parameter–stiffness ratio map with different regions for the expected rafting
behaviour; p∗2 and ω2∗ are the asymptotes of the hyperbola.
Note that p0 can be positive or negative and is very large in magnitude for only
slightly differing material parameters. The restrictions (5.12) on the elastic constants
imply that
p∗2 > 8/(3(1 + A+ )) > 83 , ω2∗ > 0, |p0 | > p∗2 . (5.22)
The rafting direction indicated by the k / ⊥ symbol in figure 3 follows from a limit
analysis for p, ω → ∞ (which we shall omit here for brevity) by making use of the
A A
restrictions (5.12); it relies on the assumption that ∆Ξ̄ = Ξ̄22 − Ξ̄11 is positive for
all material parameters. While we did not succeed in rigorously proving this, there
are some arguments in support of the assertion. When the particle shape is circular,
∆Ξ̄ reduces to
+
∆Ξ̄ = det Ω −1 C44 (2n21 − 1)[x− − (E11 + E12 )(x− − x+ )][y − − (E11 − E12 )(y − − y + )],
± ± ± ±
where x± = C11 + C12 > 0, y ± = C11 − C12 > 0. (5.23)
Because the acoustic tensor Ω is positive definite for a linearly elastic material due
to the positive definiteness√of C, the determinant of its inverse is positive. C44 is
obviously positive and for 2/2 < n1 < 1, which is true for point A (figure 2), the
same holds for (2n21 − 1). The two remaining terms in the square brackets can be
bounded by x− and x+ , y − and y + (which are all positive), respectively, if
0 < E11 + E12 < 1 and 0 < E11 − E12 < 1. (5.24)
This is obviously true, because (E11 + E12 )ε0v is the dilatational strain inside the
inclusion due to a dilatational eigenstrain ε0v , which must be greater than zero and
less then ε0v . A similar argument holds for the difference E11 − E12 . Figure 4 shows
numerically calculated results for Ξ22 − Ξ11 as a function of the arclength for dif-
ferent particle shapes and arbitrarily chosen material parameters. (The details of
the numerical technique have been published in Schmidt & Gross (1997).) It can be
seen that this quantity is strictly positive, also for the case where periodic boundary
conditions have been imposed on the edges of an elementary cell. A more convincing,
physical argument for the positiveness of ∆Ξ̄ will be given in the discussion.
(a)
0.3
A
B
C
0.2
Ξ 22 – Ξ 11
A B C
0.1
D E F 0
0 1/16 1/8
s/s tot
(b) (c)
0.15
D A
E B
F C
0.1
Ξ 22 – Ξ 11
Ξ 22 – Ξ 11
0.05
0
0 1/16 1/8 0 1/16 1/8
s/s tot s/s tot
Figure 4. Ξ22 − Ξ11 versus relative arclength for different particle shapes and different material
parameters. (b) ω = 1.5, A+ = 2, ν̄ + = 0.3, A− = 3, ν̄ − = 0.3. (a), (c) ω = 0.5, A+ = 3,
ν̄ + = 0.3, A− = 2, ν̄ − = 0.2.
x2
B'
B
C'
x1
a
x3 A'
C
Figure 5. Symmetry-related points in the three-dimensional case. The points represent the
projection of the respective points on the interface onto the face of the cube.
B B'
A
C'
x3 x1
Figure 6. Assumed morphology change that leads to rafting. Sections in the (x2 , x3 )-plane
(left) and in the (x1 , x2 )-plane (right). The arrow indicates the load direction.
Contrary to the two-dimensional case, the second α value depends on the nature of
the function ϕ and, more importantly, on the shape of the particle and the interaction
with other particles. However, α1 is the same as in the two-dimensional case and
it turns out that the corresponding curves in the p–ω diagram are also identical.
Equations (5.13)–(5.17) are still valid, except for compliances to be used there. The
representation (5.10) is used again for the stiffnesses, but the compliances for the
three-dimensional case are different from those given in (5.11) due to the plane-
strain assumption that was used there. For the three-dimensional case they are given
by
1
Sij = sij ,
µ̄
2
(6.6)
1 (1 + A) 1 (1 − A − 4ν̄)(1 + A) 1+A
s11 = , s12 = , s44 = .
2 1 + 3A + 4ν̄ 4 1 + 3A + 4ν̄ 2A
Inserting (6.6) into (5.17) for α = 1 leads to the same result for the asymptotes as
before:
p∗1 = 0, ω1∗ = (1 + A− )/(1 + A+ ). (6.7)
The asymptotes of the second direction-separating curve are given by (5.17) with
α2 = θ̄ − 1:
2 − θ̄ s− − −
12 + s12 − θ̄s12
p∗2 = + , ω2∗ = . (6.8)
s+
12 + s+
12 − θ̄s12 s12 + s12 − θ̄s+
+ +
12
This second line of direction reversal is no longer a simple hyperbola, because θ̄ is,
in general, a function of ω that enters into the Eshelby tensor in (3.7). However, the
general features of the second transition line are retained: from (5.16)2 , there is a
value ω ∗ for which p becomes infinite and vice versa.
It is interesting to note that the intersection point (p0 , ω1∗ ) of the curves for α1 = 1
and α2 = θ̄ − 1 is independent of θ̄ and, therefore, independent of the details of the
particle shape and particle interactions. As before, p0 follows from equating the ω
values for α1/2 ,
1 (3 − 2β + )(3 − 2β − )
p0 = , (6.9)
1 + A+ β+ − β−
with β ± from (5.21). From (6.9), the position of the second region-separating line in
the p–ω diagram can be estimated. As before, p0 can be positive or negative and is
large in magnitude for only slightly differing materials.
Some further simplifications allow us to actually draw the curve for α2 . When the
particle has a spherical shape, the Eshelby tensor becomes symmetric, i.e. Eijkl =
Eklij , and the value of θ for x1 = x2 = 0 becomes
(E11 + 2E12 )(x− − x+ ) − E11 (y − − y + ) + y − − x−
θa = − , (6.10)
(E11 + 2E12 )(x− − x+ ) − E12 (y − − y + ) − x−
where x± , y ± are given by (5.23). For the case of an isotropic matrix material, the
Eij are available in closed form, by making use of Eshelby’s equivalent inclusion
3D
2D
0
–20 –10 0 10 20 30
p
Figure 7. Rafting maps for the two- and three-dimensional cases of an isotropic matrix (exact).
1
E ij = δ ij
exact
0
–20 –10 0 10 20 30
p
Figure 8. Comparison of morphology maps for an isotropic matrix with a single ellipsoidal
particle (exact) and the approximation f → 1.
24
ν– +
20
16
p*
2
12
4
0 1 2 3 4
A+
Figure 9. Dependence of the p asymptote on the matrix anisotropy for ν̄ + = 0.25/0.3/0.35.
2
ν– – ν– +
ω *2
0
0 1 2 3 4 0 1 2 3 4
A– A+
Figure 10. Dependence of the ω asymptote on the matrix and precipitate anisotropy for
ν̄ ± = 0.25/0.3/0.35. Values that are not varied in the plots are A± = 2, ν̄ ± = 0.3.
100
50
ν– –
p0 ν– +
0
ν– – ν– +
– 50
– 100
0 1 2 3 4 0 1 2 3 4
A– A+
Figure 11. Dependence of the intersection point p0 on the matrix and precipitate anisotropy for
ν̄ ± = 0.25/0.3/0.35. Values that are not varied in the plots are A+ = 2, ν̄ + = 0.25 (left) and
A− = 3, ν̄ − = 0.35 (right).
7. Discussion
Based on the concept of an ‘equivalent eigenstrain’ that reduces the inclusion prob-
lem with external load to a problem without external load but with a modified
eigenstrain, two criteria for the reversal of the rafting direction have been derived in
terms of the ratio of equivalent eigenstrains present in the two coordinate directions.
The first criterion states that the rafting changes from ‘rod parallel to x1 -axis’ to
‘plate perpendicular to x1 -axis’ when the ratio ε01 /ε02 drops below 1. This makes clear
physical sense: because the strain energy arises from the adjustment of the misfitting
lattice planes along the interface, the particle shape will evolve so as to diminish the
interface area along which the larger of the two mismatches (i.e. eigenstrains) must
be accommodated. The second direction reversal can also be understood by this line
of reasoning: because the strain energy is insensitive to the sign of the mismatch to
be adjusted, there exists a second, negative, ratio of mismatches where this direction
reversal occurs. In the two-dimensional case, this ratio is equal to −1, because the
two directions are, in a way, equivalent, while in the three-dimensional case, this
ratio depends on the details of the elastic field.
These two α values divide the load parameter–stiffness ratio plane into regions
where one of the two rafting directions is predicted in much the same way as ‘Pineau’s
map’ for the isotropic case (Pineau 1976). Pineau’s map predicts different rafting
directions for E − /E + ≷ 1, where E ± is Young’s modulus of the respective isotropic
phase. Writing our corresponding result (6.7)2 in terms of the Voigt constants yields
− −
1 + A− C11 − C12
ω≷ ⇐⇒ + + ≷ 1, (7.1)
1 + A+ C11 − C12
which can still be expressed differently, when using the Young’s moduli and Poisson
numbers belonging to one of the crystal directions, to read
− −
E[100] /(1 + ν[100] )
+ + ≷ 1. (7.2)
E[100] /(1 + ν[100] )
This is consistent with Pineau’s finding, because he assumed identical Poisson num-
bers in the two phases. It also makes clear that it is incorrect to use the ratio
of Young’s moduli only for the prediction of the rafting direction. It is interesting
to note that only two of the three independent elastic constants of an orthotropic
material with cubic symmetry enter condition (7.1); inequality (7.2) shows that the
relevant constants can be determined from a single tensile test.
Criterion (7.1) has also been found by Johnson et al. (1988), who assumed small
differences in the elastic constants and considered an isolated ellipsoidal particle. This
result did not give any hints for what the influence of particle interaction might be.
Although not rigorous, the arguments from the beginning of this section still strongly
suggest that the volume fraction should not affect an elasticity-based prediction of
the rafting behaviour. Therefore, if the rafting behaviour were found experimentally
to depend on the volume fraction, this would be a strong argument in support of
plasticity-driven rafting mechanisms.
The second direction-separating line, which corresponds to the negative eigenstrain
ratio, has also been found by Pineau (1976). These transition lines become impor-
tant only for small lattice misfit, very high external stresses, and/or large elastic
mismatches. When p is large enough, changing its sign does not affect the rafting
direction, which must then be perpendicular to the load axis. This is in line with the
experiments by Tyapkin et al. (1976), who observed oblate particles in both tensile
and compressive tests on an alloy with a very small misfit.
Socrate & Parks (1993) consider two examples where an elasticity-based numerical
analysis contradicts the experimental observations. For the alloy CMSX3 studied by
Pollock & Argon (1994), elastic constants and misfit are quoted in Socrate & Parks
(1993) as
+ + +
C11 = 202 GPa, C12 = 139 GPa, C44 = 95 GPa
and
− − −
C11 = 179 GPa, C12 = 120 GPa, C44 = 88 GPa, ε∗ = −0.38%.
The matrix’s elastic constants are assumed identical to those of the γ–γ 0 mixture and
the precipitates’ elastic constants were extrapolated with an assumed temperature
dependence (Pollock & Argon 1994). For a tensile load, σ = 50 MPa, rafting into
plates perpendicular to the load direction has been observed in the experiment, while
8. Summary
Criteria for the prediction of the rafting behaviour of Ni-base superalloys under uni-
axial external load have been derived analytically. The derivation is based on ther-
modynamic considerations, using an appropriate representation of Eshelby’s force
on an interface and the concept of a load-equivalent eigenstrain. Linearly elastic
† Oblak & Kear (1968) quote an unpublished work where the constrained misfit in the precipitate has
been measured in thin foils of a two-phase specimen. They then estimated the unconstrained misfit with
the assumption of spherical particle shape, isotropy, equal shear moduli and a Poisson number equal to
1
3
, all of which are obviously strong simplifications.
material behaviour and a cubic symmetry of the crystal structure and the particle
shape/particle arrangement is assumed, and closed-form solutions for the limiting
cases of vanishing and very large volume fractions have been obtained. The result
of our analysis can be put in a load parameter–stiffness ratio diagram in which two
straight lines and a hyperbola separate regions in which one of the two rafting direc-
tions is to be expected. A closer inspection of rafting experiments reported in the
literature shows that, for the very few publications where elastic constants are given,
even small errors in these generally uncertain data can lead to entirely different
results.
The central finding of this investigation is a rigorous proof that the rafting direction
is reversed if [[C11 − C12 ]] changes sign, and a physical argument that this reversal
is independent of the γ 0 volume fraction. It is suggested that if the rafting direction
were found experimentally to depend on the volume fraction, then this would imply
a plasticity-driven morphology change.
I.S. gratefully acknowledges the financial support from the Deutsche Forschungsgemeinschaft
for a postdoctoral fellowship at the University of Cambridge.
In the first line J3 was subtracted and added, so as to produce the potential of the
equivalent problem 0Π. With the definition of the equivalent eigenstrain (4.6), the
with S − 0σ = 0ε − ε0 + [[S]]σ ∞ =⇒
Z
1
= { 0σ · ε0 − 0σ · ε∞ + σ ∞ · 0ε − σ ∞ · ε0 + σ ∞ · [[S]]σ ∞ } dV.
2 K−
Applying the Gaussian theorem and taking into account the continuity conditions
(2.2) and (2.3)2 , J4 becomes
Z Z
J4 = − t · ( u + u ) dA = − {σ ∞ · 0ε + σ ∞ · ε∞ } dV.
∞ 0 ∞
∂K K
Because K = K− ∪ K+ , the last four terms can be expressed as an integral over the
domain K+ only: Z
1
J5 = { 0σ · ε∞ − σ ∞ · 0ε} dV.
2 K+
References
Armstrong, P. E. & Brown, H. L. 1964 Dynamic Young’s modulus measurements above 1000 ◦ C
on some pure polycrystalline metals and commercial graphites. Trans. TMS-AIME 230, 962–
966.
Chang, J. C. & Allen, S. M. 1991 Elastic energy changes accompanying gamma prime rafting
in Ni-base superalloys. J. Mater. Res. 6, 1843–1855.
Davies, R. G. & Stoloff, N. S. 1965 On the yield stress of aged Ni–Al alloys. Trans. TMS-AIME
233, 714–719.
Eshelby, J. D. 1957 The determination of the elastic field of an ellipsoidal inclusion and related
problems. Proc. R. Soc. Lond. A 241, 376–396.
Eshelby, J. D. 1970 The elastic energy momentum tensor. In Inelastic behavior of solids (ed.
M. F. Kanninen, W. F. Adler, A. R. Rosenfield & R. I. Jaffee), pp. 77–115. New York:
McGraw-Hill.
Fährmann, M., Hermann, W., Fährmann, E., Boegli, A., Pollock, T. M. & Sockel, H. G. 1999
Determination of matrix and precipitate elastic constants in (γ–γ 0 ) Ni-base model alloys, and
their relevance to rafting. Mater. Sci. Engng A 260, 212–221.
Gayda, J. & Srolovitz, D. J. 1989 A Monte Carlo finite element model for strain energy controlled
microstructural evolution: rafting in superalloys. Acta Metall. 37, 641–650.
Grose, D. A. & Ansell, G. S. 1981 The influence of coherency strain on the elevated temperature
tensile behaviour of Ni–15Cr–Al–Ti–Mo alloys. Metall. Trans. A 12, 1621–1645.
Johnson, W. C., Berkenpas, M. B. & Laughlin, D. E. 1988 Precipitate shape transitions during
coarsening under uniaxial stress. Acta Metall. 36, 3149–3162.
Laberge, C. A., Fratzl, P. & Lebowitz, J. L. 1997 Microscopic model for directional coarsening
of precipitates in alloys under external load. Acta Mater. 45, 3949–3961.
Lee, J. K. 1996 Effects of applied stress on coherent precipitates via a discrete atom method.
Metals Mater. 2, 183–193.
Miyazaki, T., Nakamura, K. & Mori, H. 1979 Experimental and theoretical investigations on mor-
phological changes of γ 0 precipitates in Ni–Al single crystals during uniaxial stress-annealing.
J. Mater. Sci. 14, 1817–1837.
Müller, L. 1993 Modellierung der Eigenspannungen und Dehnungen in der Mikrostruktur ein-
kristalliner Nickelbasissuperlegierungen mit hohem Ausscheidungsanteil. PhD thesis, Tech-
nische Universität Berlin.
Nabarro, F. R. N., Cress, C. M. & Kotschy, P. 1996 The thermodynamic driving force for rafting
in superalloys. Acta Mater. 44, 3189–3198.
Oblak, J. M. & Kear, B. H. 1968 Fringe contrast at coherent precipitate interfaces in a Ni-base
alloy. Trans. Quart. ASM 61, 519–527.
Ohashi, T., Hidaka, K. & Imano, S. 1997 Elastic stress in single crystal Ni-base superalloys and
the driving force for their microstructural evolution under high temperature creep conditions.
Acta Mater. 45, 1801–1810.
Paris, O., Fährmann, M., Fährmann, E., Pollock, T. M. & Fratzl, P. 1997 Early stages of
precipitate rafting in a single crystal Ni–Al–Mo model alloy investigated by a small angle
X-ray scattering and TEM. Acta Mater. 45, 1085–1097.
Pineau, A. 1976 Influence of uniaxial stress on the morphology of coherent precipitates during
coarsening—elastic energy considerations. Acta Metall. 24, 559–564.
Pollock, T. M. & Argon, A. S. 1994 Directional coarsening in nickel-base single crystals with
high volume fractions of coherent precipitates. Acta Metall. Mater. 42, 1859–1874.
Schmidt, I. & Gross, D. 1995 A strategy for determining the equilibrium shape of an inclusion.
Arch. Mech. 47, 379–390.
Schmidt, I. & Gross, D. 1997 The equilibrium shape of an elastically inhomogeneous inclusion.
J. Mech. Phys. Solids 45, 1521–1549.
Socrate, S. & Parks, D. M. 1993 Numerical determination of the elastic driving force for direc-
tional coarsening in Ni-superalloys. Acta Metall. Mater. 41, 2185–2209.
Tien, J. K. & Copley, S. M. 1971 The effect of orientation and sense of applied uniaxial stress on
the morphology of coherent gamma prime precipitates in stress annealed nickel-base superal-
loy crystals. Metall. Trans. 2, 543–553.
Tyapkin, Y. D., Travina, N. T., Kozlov, V. P. & Ugarova, Y. V. 1976 Influence of external loads
on the ageing kinetics and spatial distribution of γ 0 -phase particles in nickel alloys. Fiz. Metal.
Metalloved 42, 1294–1300.
Valles, J. L. & Arrell, D. J. 1994 Monte Carlo simulation of anisotropic coarsening in nickel base
superalloys. Acta Metall. Mater. 42, 2999–3008.
Veron, M., Brechet, Y. & Louchet, F. 1996 Directional coarsening of Ni-based superalloys:
computer simulation at the mesoscopic level. Acta Mater. 44, 3633–3641.