Acta Materialia 107 (2016) 484e507

Contents lists available at ScienceDirect

Acta Materialia
journal homepage: www.elsevier.com/locate/actamat

By invitation only: overview article

Failure of metals II: Fatigue

Pineau a, David L. McDowell b, c, Esteban P. Busso e, *, Stephen D. Antolovich b, c, d
Andre
a
MINES ParisTech, Centre des Mat
eriaux, CNRS UMR 7633, BP 87, 91003 Evry, France
b
School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, GA, USA
c
George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, USA
d
School of Mechanical and Materials Engineering, Washington State University, Pullman, WA, USA
e
Scientific Directorate, ONERA, B.P. 80100, 91123 Palaiseau, France

a r t i c l e i n f o a b s t r a c t

Article history: In this interpretive review, fatigue in metallic systems is considered primarily from the perspective of
Received 23 February 2015 interactions between the microstructure, the deformation mode and the mechanical state at both low
Received in revised form and high temperatures. In Part 1 the development and early propagation of cracks is considered in terms
26 May 2015
of the basic damage mechanisms and the relative size of the crack with respect to applicable micro-
Accepted 31 May 2015
structural feature(s). In this section, a multistage grain scale approach to microstructure-sensitive fatigue
Available online 21 July 2015
crack formation and growth is presented which uses Fatigue Indicator Parameters (FIPs) to correlate
these processes. Various FIPs parameters are discussed in terms of their indication of the state of fatigue.
Keywords:
Local and global fatigue approaches
The development and early crack propagation is considered in the context of microstructure and notches,
Fatigue in metallic alloys and probabilistic aspects of the notch fatigue problem are discussed. These features are integrated into a
Multistage fatigue systematic approach for the selection of fatigue resistant microstructures for given applications. In Part 2,
Environmentemicrostructure interaction attention is focused on Ni-base superalloys and the interaction between oxidation, creep and micro-
Oxidation structure (including coatings) in the formation and propagation of cracks. This part of the overview
Superalloys addresses both experimental and modelling aspects. Methodologies based upon fundamental physical
Intergranular cracking, thermomechanical processes are presented for understanding and predicting the development and propagation of fatigue
fatigue
cracks, including effects of sequential oxide type formation and of creep on either restraining or accel-
erating damage by oxidation. The variable fatigue resistance of discs in jet engines is seen to depend
upon the variability of microstructure and its influence on the severity of creep/oxidation interactions.
All of these factors are considered in the practical case where both temperature and loading parameters
vary simultaneously (thermomechanical fatigue). A physics-based life prediction model considering the
interactions of deformation and environmental damage is reviewed in terms of its applicability to life
prediction of components.
© 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. General introduction bridge local and global fields, and can incorporate aspects arising
from materials processing, manufacturing, and the environment,
In this overview, the physical understanding arising from the amongst others [1e3]. This work will show that one of the most
study of fatigue phenomena in metallic alloys at both low and high significant factors which has greatly contributed to our current
temperatures as well as the associated microstructure-inspired understanding of fatigue behaviour in metallic materials at low and
modelling approaches are discussed. The emphasis of this selec- high temperatures has been the development of sophisticated
tive overview of fatigue is on approaches which establish links testing, characterisation and visualisation techniques at the nano-
between microstructure, environmental and geometrical effects. meter and micrometer length scales. This has enabled a great deal
The understanding of such complex and coupled phenomena is of insight and understanding about the mechanisms by which
crucial to enable the structural integrity assessments of engineer- defects nucleate and grow. This understanding goes hand-in-hand
ing components since they convey the information necessary to with the establishment of key computational modelling ap-
proaches, such as those based on molecular dynamics and crystal
plasticity approaches, which can now increasingly be validated at
the relevant length scales by direct, in situ measurements. For
* Corresponding author.

http://dx.doi.org/10.1016/j.actamat.2015.05.050
1359-6454/© 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 485

instance, the relatively high level of uncertainty and the statistical arbitrary length corresponding to a transition from a regime of
nature associated with cyclic plastic strain localisation and cracking growth, which is in turn conditioned by the notch root stress and
processes at the subgrain scale can now be described more accu- strain fields [8]. Variability of fatigue response is pronounced,
rately from the understanding gained through a combination of in particularly in the HCF regime. Explicit consideration in these re-
situ experiments and multi-scale modelling calculations. lations of mesoscale microstructural attributes such as grains and
The paper is divided into two parts. Part 1 provides an overview grain boundaries within the framework provided by dislocation-
of microstructure-sensitive modelling approaches for low temper- based crystal plasticity has been extremely limited, and the
ature fatigue. It addresses the dominant mechanisms of mechanisms are not completely understood. It is known that
microstructure-scale cyclic slip, crack formation, and early growth probability distribution functions of local (i.e. grain level) driving
processes in polycrystalline engineering alloys ranging from ferritic forces associated with fatigue crack nucleation and growth in the
and duplex steels, to aeb Ti alloys and Ni-base superalloys. It starts HCF regime exhibit greater values for the largest defect size, that is,
with a refinement of the typical definition of the stages of fatigue a thicker tail. Since design for low probability of fatigue failure in
crack formation and growth. It is then followed by a discussion of actual components is of great practical concern, the role of micro-
microstructure-sensitive growth relations, and strategies for structure deserves therefore attention by the materials science and
computing surrogate driving force measures using crystal plasticity fatigue design communities. The focus of this article is placed on
based finite element approaches and methods to account for sta- microcrystalline materials rather than ultrafine grained or nano-
tistical distributions of driving forces and fatigue life on micro- structured materials, which are considered in Overview III.
structure randomness. Finally, the complexities and anomalies
associated with the growth of fatigue cracks of lengths no greater 2.1. Local and global perspectives of early stage fatigue phenomena
than just a few grains e known as microstructurally small cracks
(MSC) e are discussed both in terms of existing experimental evi- Microstructure has a dominant influence in the early stages of
dence and new methods for modelling their 3D growth behaviour. fatigue crack formation and growth under HCF conditions, specif-
In the second part of this paper, the effects of high temperature ically when crack size is on the order of the grain size and spacing
on fatigue life as influenced by the phenomena of creep, fatigue, of microstructure barriers such as grain or phase boundaries. This
oxidation and, importantly, their interactions are considered. The is a particularly important point since it is well known that the
importance of and interest in this subject is manifested in part by majority of the total fatigue life is typically spent in this regime. In
recent extensive reviews (c.f. [4e7]). Here, interactions are exam- LCF, microstructure morphology of grains/phases influences the
ined in terms of fundamental physical and mechanical processes mean (50% probability) fatigue response. In HCF or VHCF, micro-
which may occur throughout the bulk of the material or which are structural variability contributes significantly to the wide vari-
limited to the crack-tip region. Fundamental approaches are used to ability of total life. The design focus of many practical applications
show when such interactions either lower (the usual situation) or is placed on minimum fatigue life. Unfortunately, physically-based
extend (less frequent) the life. While the emphasis is on g0 - models for formation and growth of small fatigue cracks are rela-
strengthened Ni-base superalloys, other important materials such tively less developed than long crack growth models, and are
as g00 or solid solution alloys are also considered. The importance of material specific. Congruent with the goals of Integrated Compu-
microstructural instability (e.g., coarsening of g0 precipitates, tational Materials Engineering [9] and the notion of materials
oxidation-induced dissolution of g0 ) on life is also emphasised. This design, it is essential to consider approaches that include micro-
includes the microstructural degradation of either coated or un- structure attributes as design variables for improving HCF resis-
coated Ni-base superalloys caused by surface oxidation, and the tance [10,11].
resulting effect in terms of crack initiation and propagation In traditional mechanics of fatigue crack initiation [1,12], the
behaviour. Finally, all of these phenomena are considered as they “local approach” refers to estimating notch root fields given remote
affect the thermo-mechanical fatigue (TMF) life. All these high applied loading. In contrast, global approaches use stress-life or
temperature fatigue mechanisms are examined in great detail in strain-life relations based on the remote applied loading with some
order to motivate the formulation of physically-based, micro- correction accounting for stress concentration at notches. In
structure-sensitive methods for modelling fatigue damage pro- micromechanics, local response typically relates to heterogeneity of
gression and interaction with the environment. stress and strain fields at the scale of periodic (e.g., composites) or
random microstructure. Micromechanics typically assumes that
notches induce longer range stressestrain field gradients than
2. Part 1: microstructure-sensitive modelling of low
those associated with local microstructure heterogeneities; in other
temperature fatigue
words, a scale separation is assumed between the gradients asso-
ciated with notch root fields and these short range spatial corre-
Historically, local stress-based criteria for High Cycle Fatigue
lations. In this way, representative volume elements (RVEs) of
(HCF) and Very High Cycle Fatigue (VHCF) or plastic strain-based
periodic or random microstructures can be assumed to represent
criteria for Low Cycle Fatigue (LCF) have been applied to estimate
material behaviour at a point within a notch root gradient field.
fatigue life based on transfer of results from experiments on ide-
The local approach to fracture, as reflected in the foundational
alised laboratory specimens to structural components. In so doing,
works of Pineau et al. [13e15], pertains to the quantitative incor-
it is conventionally assumed that the total life is:
poration of microstructure effects and failure micro-mechanisms in
estimating fracture toughness or resistance to propagation of long
NT ¼ Ni þ Np : (1)
fatigue cracks. They are considered to be sufficiently long when
Here, NT is the total fatigue life (cycles to failure at specimen or compared to underlying microstructure scales. In this case, the role
component level). The initiation life Ni corresponds to the devel- of microstructure in influencing fatigue damage evolution within
opment of cracks that are substantially longer than intrinsic the cyclic damage process zone near the crack tip is of key concern,
microstructural features such as grains or phases, and Np is the along with corresponding shifts of mean da/dN versus DK behav-
number of cycles spent in crack propagation. The initiation crack iour with microstructural changes. In contrast, the present over-
size, ai, is based either on crack lengths measured in smooth view focuses primarily on the variability of the fatigue crack
specimens, typically in the range of 0.5e1 mm, or on some non- formation and early growth for a given microstructure, facilitating
486 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

comparison among microstructures in the micro-structurally small process zone). Physically small cracks are sufficiently large to smear
crack (MSC) growth regime where remote applied DK cannot be out the influence of microstructure on the crack growth rate;
used as a meaningful driving force for the local response. This conventional homogeneous fracture mechanics applies, although
complicates localeglobal approaches, as evidenced by the historical elasticeplastic fracture mechanics may be necessary depending on
emphasis on highly empirical stress- and strain-based approaches anisotropy and the ratio of the crack length to cyclic plastic zone
for fatigue crack initiation. Such approaches offer fundamental size. NPSC is further decomposed according to NPSC ¼ NPSC 0 þ NPSC00 ,
challenges in the transition from small crack to long crack growth reflecting the fact that a given definition of initiation often implies
behaviour, requiring the introduction of special schemes to an artificial partition of the growth history into parts associated
distinguish initiation and propagation regimes [16] that often do with conventional “initiation” and “propagation” regimes. These
not depend on microstructure. parts are most often distinguished by definition rather than phys-
ics. In most engineering applications, operational definitions of
2.2. Role of microstructure in multistage fatigue initiation crack length are too long to lie within the MSC regime.
Fatigue crack initiation approaches that relate NPSC0 to remote
In the absence of cracked or debonded second phase particles or driving force (e.g., macroscopic plastic strain or stress range) via
pre-existing crack-like defects, the process of fatigue crack forma- some parametric scaling law (e.g., Coffin-Manson) implicitly
tion involves the localisation of irreversible plastic strain within parameterise the growth relations and assume similitude among
persistent slip bands (PSBs) that normally lead to either the for- geometries. This is not a valid assumption if the initiation crack
mation of intrusions/extrusions at surface grains or their length is defined on the order of microstructure scales or, in cases of
impingement on grain boundaries. Through continued cyclic small notches, particularly those with characteristic scales
loading, these slip band mechanisms lead to lattice/interface approaching the level of microstructure. By logical extension, it
disruption and to the formation of a crack of either a transgranular may be expected that such scaling relations for fatigue life based on
or intergranular nature. We employ the term fatigue crack “for- remote fields will be most accurate for LCF conditions for which the
mation” rather than “nucleation” to refer to the processes by which contribution of crack nucleation and MSC growth regimes to total
cracks are manifested at the scale of individual grains/phases in fatigue life is negligible. Moreover, in the HCF regime there is no
cyclically deformed metals. This definition may combine elements way to avoid the issue of microstructure, as even the fatigue limit in
of classical pre-and post-critical embryonic nucleation, along with stress-life approaches reflects a threshold for cracks with some
limited crack growth in the initial nucleant grain/phase. The more characteristic length related to microstructure; the fatigue limit is
commonly used term in engineering practice “crack initiation” re- microstructure and process-history sensitive.
fers to the combined processes of crack formation and growth to Early mesoscopic approaches to characterising distributions of
some predefined crack length that encompasses a very large driving force to support probabilistic fatigue crack formation for
number of grains/phases. metallic polycrystals can be traced to early works such as those of
Microstructure-sensitive modelling must address key aspects of Provan [21,22] and Sakai et al. [23]. The importance of interaction of
physical fatigue processes that range from the sub-micron micro- small cracks with microstructure had been established in the 1980s
structural scale to that of the components. It is apparent that even [24,25]. Hoshide and Socie [25] employed slip band cracking
HCF failure is governed by distributed micro-plasticity at the scale models and crack growth relations to model distributed crack for-
of individual grains [17,18]. Moreover, small cracks with size on the mation within a polycrystal using a simple Sachs assumption for
order of short range microstructure heterogeneity (grains, phases) intergranular interactions.
propagate below driving force levels for long crack thresholds, and More specific focus on the behaviour and mechanics of fatigue
can be arrested by interaction with strong microstructure barriers cracks in the MSC regime drew attention to the fact that conven-
early in their growth history. Such micro-structurally small cracks tional linear elastic fracture mechanics (LEFM) did not apply in
are subject to the varying heterogeneity and anisotropy of micro- many cases. McDowell [26] reviewed the status of microstructure
structure, giving rise to scatter in any definition of fatigue crack sensitivity to formation and growth of small cracks in the HCF
initiation, as well as a lack of applicability of long crack fracture regime. Miller [27,28] discussed thresholds based on MSC in-
mechanics. To facilitate a more detailed breakdown of the stages of teractions with microstructure and emphasised the importance of
fatigue crack formation and growth, we consider a more detailed cyclic plastic strain in driving small cracks via microstructure-
decomposition [19,20] of Eq. (1). Here, coupled growth relations. Finite element simulations were used
to explore effects of crystallographic orientation on variability of
Ni ¼ Nnucl þ NMSC þ NPSC0 MSC growth rate in crystals using DCTD concepts [29e32].
(2)
Np ¼ NPSC00 þ NLC Building on earlier developments of crystal plasticity [33e37]
and void nucleation, growth and interface decohesion [38e42], a
natural progression led to the application of emerging computa-
where, Nnucl is the number of cycles to nucleate a crack (for- tional micromechanics to early stages of fatigue outlined in Eq. (1),
mation of a crack embryo and sub-critical propagation to size of which constitutes a multistage formulation. McDowell and co-
stable nucleus), which is essentially a solid state phase transition, workers introduced a multistage approach to microstructure-
and NMSC, NPSC and NLC represent the number of cycles to propagate sensitive fatigue crack formation and growth that employs Fa-
the crack(s) in the regimes of micro-structurally small, physically tigue Indicator Parameters (FIPs) to correlate crack nucleation and
small, and long crack growth, respectively. Normally, the micro- growth processes that have been studied at the grain scale [43e45],
structurally small crack regime is approximately 3e10 times the along with heuristic relations for micro-structurally small crack
grain or phase size/spacing; it is characterised by a potentially growth based on the cyclic crack tip displacement, DCTD, outlined
pronounced dependence of the fatigue crack growth rate on in Fig. 1 [19,20]. The DCTD is comprised of contributions from cyclic
microstructure attributes such as grain size, second phase particle crack tip opening (DCTOD) and sliding (DCTSD) displacements. As is
size and spacing, etc. It also depends on the amplitude of the the case with the nucleation driving force, it is presumed that the
applied stress, since dominant dependence is manifested in cases DCTD (or a surrogate FIP measure) is computed from
where the size of the cyclic crack tip plastic zone is on the order of microstructure-sensitive crystal plasticity relations for cracks of
the grain size, interphase spacing, etc. (i.e., mesoscale damage various scales relative to microstructure. The da/dN versus DCTD
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 487

indicate consideration of the role of microstructure at the meso-

scale involving grain and phase distributions rather than details of
subgrain dislocation structures and resulting irreversibility of cyclic
dislocation glide that have commonly been studied in reference to
LCF [49]. Of course, HCF involves cyclic plastic strain localisation but
with a high degree of heterogeneity of slip with regard to intrinsic
microstructure or extrinsic features such as inclusions [17,18,50].
McDowell has clarified the significant limitations of the micro-
mechanics concept of a statistically homogeneous representative
volume element (RVE) in HCF applications since scatter is observed
at the scale of laboratory specimens. It is not appropriate to select a
volume element size for computation that simply achieves first
order convergence of the average stressestrain behaviour or elastic
stiffness since differences in spatial correlation lengths for fatigue
crack formation and early growth can be very large. In view of this
fact, and considering the limitations of large scale modelling of tens
of thousands of grains using polycrystal plasticity with sub-grain
Fig. 1. Heuristic multistage modelling framework for nucleation and growth of fatigue
mesh resolution, it is necessary to build up statistics based on
cracks. Cycles NMSC and NPSC, respectively, represent the number of cycles to propagate
the crack(s) in the regimes of micro-structurally small (normally 3e10 times the grain multiple computational realisations [51,52] of statistical volume
or second phase size/spacing) that affects retardation and undulation of the crack elements (SVEs).
driving force, see inset figure (lower right) from Künkler et al. [48,40], physically small A significant body of microstructure-sensitive fatigue modelling
crack (PSC) growth, and long crack (LC) growth. In the PSC regime, the crack is suitably
work has evolved since the late 1990s. The multistage fatigue
long to be treated using conventional LEFM, but is still below the size considered
amenable to a definition of the initial crack length for propagation analyses using
model in Eq. (1) was first introduced in the DoE USCAR program
LEFM. from 1995 to 2000 [53,54]; finite element modelling of micro-
structure was employed to model formation and early growth of
inter-dendritic cracks in A356-T6 casting alloys, culminating in the
relation in Fig. 1 is considered primal, whether applied to MSCs or formulation of a probabilistic framework for fatigue response based
to long cracks. The irreversibility factor G depends on composition, on various distributions of microstructure attributes at five
nanoscale strengthening factors, and dislocation substructure [45]. different length scales [55]. Tryon and Cruse [56e58] related
It is directly related to irreversibility of slip in the damage process mesoscopic finite element modelling of microstructure to distri-
zone. Mughrabi [46] has described how slip irreversibility in fatigue butions of fatigue responses of polycrystalline microstructures,
crack formation can depend on the applied strain amplitude; R- albeit within the context of an RVE concept that has the afore-
ratio effects are likely to emerge as well. Micromechanical studies mentioned limitations.
can be introduced to capture DCTD interactions with microstruc-
ture [26,29e32]. In general, due to shear localisation at the tips of 2.3. Nucleation and early growth of microstructurally small fatigue
cracks on the order of grain size and lack of similitude, use of cracks
macroscopic, homogeneous fracture mechanics approaches to es-
timate DCTD is inappropriate [26]. The form shown in Fig. 1 does 2.3.1. Modelling approaches at the grain scale
not rely on applicability of small scale yielding or conventional We focus here on microstructure-sensitive modelling of fatigue
LEFM concepts. As the crack lengthens, conventional da/dN versus crack nucleation and growth of fatigue cracks up to a size at which
DKeff relations may be used, with the crossover defined by the the microstructure plays a less pronounced role in affecting vari-
maximum of MSC/PSC and long crack growth rates, as indicated in ability of subsequent crack growth. Typically, this implies a crack
Fig. 1. length of less than three grains/phase domains in a polycrystalline
In the HCF regime, crack nucleation and micro-structurally small and/or multi-phase alloy system [59]. Some of the potentially
growth phases dominate the fatigue lifetime; cyclic plastic defor- important microstructure attributes that influence fatigue crack
mation is highly heterogeneous within the microstructure. formation include, at the mesoscale, microstructure unit size,
Accordingly, variability and size effects are most pronounced in morphology, and crystallographic orientation [59e63]. At smaller
HCF. Experiments under these conditions are very time-consuming length scales, the development of lattice curvature, dislocation
and fraught with difficulty since crack formation is a rare-event substructures, and local grain boundary (GB) structures have pro-
phenomenon. Although naturally-occurring cracks behave differ- found effects on cyclic deformation and fatigue. The relative role of
ently from machined micro-notches (e.g., FIB machined) under HCF these smaller scales versus that of intergranular correlations is not
conditions due to differences in microstructure-scale roughness yet fully resolved, and depends on material and load history.
and plasticity [47], it is very difficult to identify the location of such McDowell [64] and McDowell and Dunne [65] reviewed crystal
small cracks a priori within a given field of view, and access to plasticity-based fatigue approaches that emphasise the role of
subsurface growth is not provided by classical surface observation microstructure in forming cracks at the scale of grains; in particular,
techniques. HCF requires a focus on extreme value statistics of they considered the role of computable Fatigue Indicator Parame-
potential sites for micro-plastic strain localisation and fracture that ters (FIPs) based on cyclic plastic strain range and cumulative
drive crack formation and early growth. Moreover, small fatigue ratchetting deformation as a means of correlating the local
cracks under HCF conditions may arrest at the first few micro- microstructure with the most likely locations of fatigue crack for-
structure barriers, leading to a fatigue limit with a threshold that mation and subsequent early growth. The rationale for mesoscopic
differs from that of the long crack growth threshold [26e28]. FIPs based on the deformation behaviour of individual grains cor-
McDowell [19,20] discussed the importance of mesoscopic responds well with the need to understand the driving forces
computational simulations to quantify the role of microstructure behind crystallographic fatigue crack nucleation and early growth
from the perspective of a frequency distribution of driving forces. (cf. [45,46,48,66] of microstructurally small cracks.
He introduced the term qualifier “microstructure-sensitive” to The computation of FIPs depends on the slip behaviour of the
488 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

polycrystalline material in view of the underlying structure as promoting planar localisation of subsequent dislocations on the
represented by the crystal plasticity constitutive equations same glide plane. The interested reader may consult extensive
employed. The kinematic decomposition of the deformation discussion in this regard appearing in Ref. [67], as well as numerous
gradient into parts is classical, comprising elastic deformation plus references cited in Sections 1.3.2, 1.3.3, and 1.4 to follow. Section
rigid lattice rotation Fe , and plastic deformation, Fp , i.e., 3.1.2 also employs a crystal plasticity framework for modelling
environment-microstructure interactions.
F ¼ F e Fp (3) The nucleation life term in Eq. (1) is perhaps the most chal-
lenging to model, as it is material and mechanism specific,
The plastic velocity gradient Lpo in the undeformed reference
depending on composition, microstructure, and environment. It
configuration is given by
remains the element with perhaps the greatest uncertainty
X
Na owing to the small scales involved and to difficulties with iden-
p
L0 ¼ F_ ðFp Þ1 ¼ g_ ðma0 5na0 Þ
p a
(4) tifying nucleation sites. The works of Essman et al. [68] and
a¼1 Differt et al. [69] remain definitive (cf. [49]) in capturing the
essence of physical observations relative to crystallographic PSB-
where mao and nao are unit vectors defining the slip and the slip induced cyclic strain localisation and fatigue cracking in quanti-
plane normal directions in the reference (and isoclinic, interme- tative microstructure-dependent models. Early micromechanics
diate) configuration, respectively, Na is the number of active slip models were based on crack nucleation in slip bands [70], and
systems, and g_ is the shearing rate for the slip system a. The
a
dislocation approaches continue to be extended [71e73] to Stage
plastic strain rate is the symmetric part of the plastic velocity I MSC growth. Recently, Sangid and co-workers [74e78] devel-
gradient, while the plastic spin that drives lattice rotation is oped an energy-based approach informed by atomistic simula-
defined by the skew-symmetric part. In the intermediate config- tions for modelling fatigue crack formation associated with PSB/
uration, the second PiolaeKirchhoff stress, T, satisfies the matrix and PSB/GB interactions. The predictions were consistent
hyperelastic relation, with some of the observations made by Zhang et al. [79] and
involved modelling a number of random polycrystalline aggre-
T ¼ C : Ee (5) gates of a Ni-base superalloy to predict the minimum number of
cycles to form a fatigue crack at the scale of large grains and
where C is the tensor of elastic moduli and Ee is the Green strain
grain clusters, as well as the scatter of HCF response. Suffice it to
tensor. The resolved shear stress, ta , serves as the driving force for
say that the incorporation of multiple dislocations into a GB is
slip on the system a, and is given by
highly complex in terms of build-up of residual Burgers vector, as
are sequential processes of emission of dislocations elsewhere
ta ¼ T : ðma0 5na0 Þ (6)
along the interface. Parametric atomistic studies of GB structure
Finally, the Cauchy stress in the current configuration is pushed and dislocation nucleation processes have illustrated such
forward from the intermediate configuration, i.e., complexity and sensitivity to structure and macroscopic GB de-
grees of freedom [80].
1 Crack nucleation models that address detailed slip band pro-
s¼ Fe TðFe ÞT (7)
detðFe Þ cesses can be considered as belonging to a class of bottom-up
models. By and large, bottom-up approaches to crack formation
with rigid body lattice rotations at finite strain accounted for within at the scale of nucleant grains/phases do not attempt full field
Fe . polycrystalline solutions but are rather based on apparent Schmid
Various physically-based phenomenological forms for the ki- factors that do not consider intergranular stress and strain field
a
netic equation for the slip system shearing rate g_ ¼ ĝ
a
interactions [74,78] or realistic dislocation substructures (cf.
ðt  B ; S ; qÞ are used for different alloy systems, as physically
a a a [81,82]). Since inelastic strain at the grain level under HCF condi-
required by the dependence of initial flow stress and rate of work tions is highly dependent on local stress, the estimates of grain-to-
hardening on temperature q, reflecting processes of dislocation grain interactions are deemed relevant. The work of Kirane and
multiplication, interaction and annihilation during inelastic defor- Ghosh [83], who employed a nonlocal slip band model for fatigue
mation. These forms are in fact too varied to list here in compact crack formation in realistic 3D microstructures, offers an example
fashion. Internal state variables Ba and Sa correspond to slip system of how bottom-up and top-down methodologies can be combined.
back stress and mesoscopically average slip system resistance, Similarly, approaches based on cohesive zone elements (cf. [81,84])
respectively. Their evolution equations are key to the description of incorporate interactions with the surrounding microstructure, such
cyclic plasticity at the level of grains. Additional variables are as 2D grain structures [46] or dislocation pile-ups. However, typi-
introduced to reflect microstructure attributes at finer scales such cally such approaches are not intimately coupled to the physics of
as precipitate size and spacing, as well as the evolving mean free slip band irreversibility in 3D microstructures. Indeed, the coupling
path(s) of gliding dislocations due to their interactions. The elastic of driving forces for slip localisation and slip-band methods with
moduli C are fully anisotropic and differ among high symmetry the surrounding 3D microstructure and remote applied loading is
(e.g., fcc, bcc) and low symmetry (e.g., hcp) crystals. It is worth exceedingly complex and is subject to considerable uncertainty.
noting that elastic anisotropy is vitally important for high cycle Unfortunately, the role of uncertainty quantification in bottom-up
fatigue simulations where the applied peak strains are typically modelling approaches for fatigue crack nucleation and early
below the macroscopic flow stress or yield point. Moreover, the growth has not been closely examined to date, and remains an open
inelastic response is fully anisotropic, as is evident from the fore- issue of high significance to practical applications. Accordingly, the
going shearing rate relations for each potential slip system: each application of scaling relations for nucleation lifetime expressed in
crystal class has different numbers and types of potentially active terms of mesoscopic cyclic plastic strain fields is likely to remain for
slip systems (12 for fcc, 24 for bcc, and at least 24 for hcp). Kinetics the foreseeable future a dominant practice in multistage fatigue
and hardening mechanisms vary widely; for example, in a-Ti-5/6Al modelling. Surrogate measures for driving force of micro-
the first glide dislocation from a grain boundary source results in a structurally small cracks and their interactions with GBs will be
breakdown of short range order, creating a softening effect and discussed later in Section 3.4.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 489

2.3.2. Macroscopic/top-down modelling approaches Considering the PFS in Eq. (8) as a slip band-based FIP, an
An alternative top-down route to link the macroscale response impingement FIP [64] is given by:
to the grain scale can involve correlations of grain-scale computed  
slip fields with fatigue crack nucleation and early growth p smax
Pimpinge ¼ gnet 1 þ KGB GBn (9)
[19,20,64]. Its physical basis resides on the use of the Coffin- sy
Manson relation as a scaling relation for fatigue crack initiation
p
(Ni term in Eq. (1)) for various definitions of initial crack size; here where gnet is the nonlocal directional plastic shear strain that im-
the exponent of the scaling relation is nearly invariant whereas the pinges on a GB facet, smax GBn is the corresponding normal stress
coefficient changes. Moreover, fatigue crack nucleation approaches component acting on the facet, and the coefficient KGB incorporates
based on slip-matrix decohesion [70] also reduce to forms of the influence of the normal stress. The work of Zhang et al. [79] on
similar power law dependence on cyclic stress and/or plastic strain fcc Cu single crystals, bicrystals and micro-crystalline polycrystals
range. Elasticeplastic fracture approaches [26e28,85e88] also with low and high angle grain boundaries revealed that fatigue
incorporate dependence on the cyclic plastic strain range. For cracks can nucleate along slip bands, deformation bands, high angle
instance, recent crystal plasticity work by Castelluccio and McDo- grain boundaries, and twins. Similar conclusions were obtained on
well [59] indicates intimate correlation of the cyclic crack tip other fcc alloys with relatively low stacking fault energies (SFE)
displacement for fatigue cracks in single crystals with the cyclic such as AleCu and CueZn. Cracking due to slip band impingement
plastic shear strain averaged within bands adjacent to the crack was most common for twin boundaries in low-SFE materials and
(specifically with the FatemieSocie parameter defined in Eq. (8) high angle GBs. Subsequent post-embryonic growth can occur via
below). Given this convergence of forms among scales and pro- either transgranular or intergranular modes. The study of Li et al.
cesses, top-down approaches rely on computable quantities of in- [92] on several bicrystal interfaces in Cu further concluded that the
terest, such as FIPs [64], averaged within a grain rather than at propensity to nucleate cracks at GBs or along persistent slip bands
individual integration points within sub-grain domains. These FIPs depends on the magnitude of the residual GB dislocations and the
are then employed in multistage fatigue relations (e.g., Eq. (2)) for GBs interfacial energy. The coupling of the computed FIPs with a GB
both the nucleation and MSC regimes. For example, the Fate- network encompassing low and high angle GBs as well as twin
mieSocie [89,90] shear-based parameter has been shown to boundaries, needs to distinguish between the probability of crack
correlate multiaxial fatigue crack initiation data at the grain scale formation (e.g., nucleation and early growth within the nucleant
and above [88] in both LCF and HCF regimes very well, in addition phase) among a distribution of such interfaces, via extreme value
to distinguishing between the different multiaxial fatigue cracking statistics. The micro-structurally small fatigue crack growth
modes (i.e., Types A and B) introduced by Brown and Miller [91]. behaviour is more complex from a statistical viewpoint, since the
This parameter is defined as connectivity of resistant and compliant interfaces among favour-
ably oriented grains in the network must be considered so as to
p   incorporate them within a growth law. Accordingly, we argue in
Lgmax smax
PFS ¼ 1 þ K0 n (8) favour of simulations of statistical ensembles with realistic GB
2 sy character distributions that embed easy slip transmission through
low angle boundaries versus intergranular growth paths due to GB-
where Dgpmax =2 is the nonlocal maximum cyclic plastic shear strain PSB interactions.
averaged over a finite volume of material (subgrain scale), smax n and The aforementioned work of Castelluccio and McDowell [59]
sy are the maximum stress normal to the plane of Dgpmax =2 and demonstrated how Eq. (8)’s FIP, when averaged along bands with
cyclic yield strength, respectively, and the constant K 0 mediates the width on the order of a few microns, serves as an effective surrogate
influence of the normal stress. The spatial volume for the nonlocal measure for the cyclic crack tip displacement of crystallographic
averaging of the driving force may be defined according to the fatigue cracks in either single crystals with otherwise homoge-
nature of the simulation, and is desirable both for purposes of neous deformation or along slip band-matrix interfaces. It is simi-
numerical regularisation (mesh insensitivity) and targeting length larly feasible to employ an intergranular FIP based on slip band
scales associated with crack embryos (e.g., slip band width/spacing, impingement as a surrogate for the cyclic opening displacement at
inclusion size, etc. As illustrated in Fig. 2, McDowell [19,20,64] high angle grain boundaries that resist slip transfer. In such cases,
discussed the need to augment an intragranular (slip band- provision for excitation of dislocation sources on the other side of
related) FIP of the type given in Eq. (8) with consideration of a the interface must be made since slip activation can promote
slip band impingement FIP based on directional accumulation of continued transgranular growth. Moreover, since existing cracks at
slip (dislocation pile-ups or plastic strain ratcheting) acting on GBs. grain or phase boundaries (including PSBematrix interfaces) must
typically grow to the scale of the nucleant grain/phase to arrest, this
FIP serves as a useful surrogate measure for correlating cycles of
crack advance at the mesoscale of grains.
Since computation of the slip response at the scale of the grain
requires larger scale polycrystal simulations, interactions arising
from various spatial correlations of microstructure at and above the
grain scale are naturally incorporated. Calibration of such top-down
approaches with fatigue experiments for polycrystals is necessary,
preferably involving observations of early stage fatigue processes at
scales of a few grains. It is possible to employ such an approach,
albeit with higher levels of uncertainty due to the limited amount
of information generally available from this type of experiments.
Musinski and McDowell [93] showed that, once calibrated to a
Fig. 2. Fatigue Indicator Parameters serving as surrogate measures for fatigue crack
formation and early growth: (a) along transgranular slip bands (e.g., PSBematrix in-
given Ni-base superalloy microstructure, such a top-down
terfaces) and (b) as a result of the impingement of slip bands or dislocation pile-ups approach is capable of describing the variability of HCF life with
against GBs [64]. microstructure. Furthermore, the transition from LCF strain-life
490 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

regime to a Basquin’s law type regime is predicted simply by virtue 1D nonlocal FIPs in the neighbourhood of inclusions in Al alloys.
of the heterogeneity of the cyclic micro-plasticity within the Furthermore, the effects of interactions of nearest and 2nd nearest
polycrystal. neighbour grains on FIPs within a favourably oriented grain for
Recent papers have also focused on correlating computed FIPs crack formation is very important [96], particularly for low sym-
with measured locations of fatigue crack formation within poly- metry hcp systems that exhibit hard and soft grain orientations
crystals, using reasonably representative crystal plasticity based [83,97e99].
finite element models of the microstructure (cf. [65]). For example, Przybyla and McDowell have employed extreme value statistical
Sweeney et al. [94] performed a detailed comparison between ex- approaches combined with 3D crystal plasticity models and
periments and crystal plasticity models of fatigue crack formation nonlocal FIPs to quantify the variability of fatigue crack formation
in single edge four point bending specimens of a polycrystalline and early MSC growth for complex alloy systems, including Ni base
ferritic steel. There, the grain morphology and crystallographic superalloy IN100 [100] and dual phase Tie6Ale4V [101]. Nonlocal
orientation distribution at the notch root were accounted for. As maximum FIPs compiled from a large number of RVE samples fit
illustrated in Fig. 3, the location of crack formation was well pre- the Gumbel extreme value distribution, thus enabling a comparison
dicted and their results confirmed that that the effective plastic of the minimum fatigue lifetime response among microstructures
strain per cycle is a better FIP than cumulative effective plastic (cf. Fig. 4). Marked spatial correlation functions (radial distribution
strain over many cycles. functions) were used to quantify the coupling of specific micro-
Bridier et al. [95] performed a study comparing relative levels of structure attributes (e.g., grain size, orientation, phases) at sites
basal, prismatic and pyramidal slip system activation in Ti-6Al-4V with high FIP values to express likelihood for the most detrimental
using a crystal plasticity constitutive model, emphasising the microstructure aspects associated with MSC formation. This type of
importance of slip system softening in the primary alpha phase of extreme value computational framework has also been used to
this particular alloy system associated with breakdown of short model the transition from surface to subsurface fatigue crack for-
range order. The crystal plasticity model was validated with direct mation with a decrease in either applied stress or strain amplitude
experimental measurement of the cyclic stress relaxation behav- in the VHCF regime [102,103]. The work of Wen and Zabaras [104]
iour of the polycrystalline specimen under strain control with considered the distributions of maximum FIPs and microstructure
positive mean strain. Single slip was typically observed at low variability in a two-phase Ni-base superalloy using a rigorous sta-
applied strain amplitudes, and fatigue cracks were associated with tistical approach.
a combination of relatively high Schmid factors along with a tensile
stress normal to the basal plane, which supports the use of slip 2.3.3. Probabilistic approaches to account for coupled notch-
band-based FIPs as in Eq. (8) [89]. microstructure effects
A distinct advantage of top-down approaches to fatigue crack Inspired by Pineau’s early works on probabilistic fatigue asso-
formation is that various FIPs have increasingly been used as ciated with microstructure effects near the surface and at notches
mesoscopic indicators to relate to processes of fatigue crack for- [105,106], Salajegheh and McDowell [107] introduced a probabi-
mation and growth of MSCs. We referred earlier to the work of listic strategy to model the transition from surface to bulk of HCF
Castelluccio and McDowell [59] in modelling single crystals with crack formation promoted by the interaction of debonded primary
and without slip bands subjected to cyclic shear with and without inclusions with microstructure in IN 100 Ni-base superalloy. Pra-
superimposed cyclic normal stress to the crack plane. There, an sannavenkatesan and co-workers [1,2] considered the effect of shot
essentially linear correlation was found between the cyclic range of peening-induced surface residual stresses, pores, and hard and soft
crack tip displacement and the nonlocal FatemieSocie FIP. Such a primary inclusions in martensitic gear steel in the context of 3D
relation holds in the case of localised plasticity and supports the polycrystal plasticity and nonlocal FIPs. Effects of inclusion fracture/
statistical assessment of fatigue resistance of different micro- debonding and initial subsurface residual stress following carbur-
structures using FIPs [60]. FIPs have also been used to quantify the isation and tempering were coupled via computation as a function
influence of twins in cyclic plastic shear strain localisation and of surface proximity to estimate the likelihood of crack formation. It
crack formation [61], suggesting that slip intensification could was also found that polycrystal plasticity was necessary to accu-
promote early crack formation. The importance of solving the field rately model residual stress relaxation, which occurs primarily
equations for 3D microstructure to compute FIPs has been during the first few fatigue cycles [3].
emphasised by Hochhalter and co-workers [62,63], amongst Microstructure-sensitive computational approaches can also be
others, who used crystal plasticity simulations to compare multiple used to explore the combined effects of microstructure and

Fig. 3. Example of (a) the specimen notch and FE model, and (b) monotonic and effective plastic strain after several cycles for a four-point bending experiment on a polycrystalline
ferritic steel [94], comparing site and orientation of fatigue crack formation with computed FIPs; the effective plastic strain after several cycles tends to localize towards the
observed site of crack formation.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 491

Simulaons – Physically Based Models

Distribuon of FIPs

Grain Size
FS Parameter

Digital Samples
Experimental Calibraon-
Validaon

Rank-Order Range of
Microstructures (for Project Fague
Materials max./min. life or min. Potency via FIPs
Improvements variability)

Stochasc Variaon of Properes and

Microstructure Parameters

Fig. 4. Using microstructure-sensitive simulations to ‘‘close the loop” in rank ordering microstructures for fatigue resistance, including design for desired failure probability [65].

notches. In conventional local approaches, the fatigue notch factor validation of MSC growth models due to the strong role of micro-
is employed as a means of estimating the notch root cyclic strain structure. Recent advances have exploited coupling of electron
fields, thus the effects of the notch root acuity relative to the backscatter diffraction (EBSD), scanning electron microscopy
microstructure are also implicitly accounted for. However, it is (SEM), and digital image correlation (DIC) techniques [110e112] to
generally estimated empirically and requires costly experimental characterise fatigue crack formation in relation to grain size, crys-
protocols. Owolabi et al. [108] developed a probabilistic method- tallographic orientation, and nearest neighbour grain interactions.
ology applied to highly stressed notch regions following a FIP-based The reader is referred to a recent review for more details [113].
approach that employed probabilistic meso-mechanics to define a Fatigue cracks form either in favourably oriented grains or in
microstructure-sensitive fatigue notch factor. Musinski and cooperation with strongly localised slip field. Such tools offer po-
McDowell [93] relied on a 3D crystal plasticity FE model to assess tential for validation of grain scale crystal plasticity models, as well
the fatigue life scatter of notched specimens with different notch as FIPs that correlate with fatigue crack formation and early growth.
root radii sizes and random microstructures of equiaxed IN100 and In addition, a large number of observations of formation and
predicted the number of cycles required to propagate a MSC. The growth of fatigue cracks from free surfaces have been possible us-
crack growth rate was assumed to be proportional to the value of ing in situ loading stages [114] with microscale specimens and
FatemieSocie FIP assessed from polycrystal simulations without MEMS devices [115,116]. Techniques such as high energy X-ray
cracks, along with the crack length dependence proposed by She- diffraction microscopy (HEDM), X-ray diffraction contrast tomog-
noy et al. [109]. A probabilistic approach based on a transitional raphy (DCT), also known as micro-CT (mCT) [117,118], and syn-
crack length from a MSC growing from the notch root up to a long chrotron X-ray methods [119] can provide more information to
crack operating under small scale yielding. This enables the study of validate 3D MSC models for growth in the initial set of surrounding
the fatigue life variability under HCF and VHCF conditions, with the grains. FIB-machined starter notches/cracks are not fully satisfac-
results characterised using a cumulative distribution function. tory replacements for naturally occurring fatigue cracks. Never-
theless, they are often required to introduce cracks within a limited,
2.4. Microstructurally small fatigue crack growth beyond the first designated field of view in combination with such advanced 3D
grain or phase imaging techniques.
As previously mentioned, Navarro and de los Rios [71] pio-
The preceding discussion about FIPs has focused on the forma- neered the application of 2D distributed dislocation models to
tion (nucleation and MSC growth) of cracks within the nucleant describe the growth rate and GB interactions of MSCs under fatigue
grain or phase. The subsequent behaviour of MSCs in the first few conditions based on the cyclic crack tip displacement range. Ben-
grains is decidedly more complex since the crack front samples a nett et al. [30e32], and Kübbeler et al. [120] employed a 2D
highly heterogeneous field of grains/phases, is not self-similar, and computational method to model plastic localisation in MSC growth
has no unique correlation with the remote applied loading. It is involving multiple grains during the transition between Stage I and
therefore essential to appeal to microstructure-sensitive modelling Stage II. Hilgendorff et al. [121] relied on 2D boundary element
to describe the fatigue life stage associated with MSC crack growth methods to model slip band impingement on grain boundaries and
beyond the first grain or phase, i.e. NMSC in Eq. (2). shear stress fields within grains. These 2D studies reinforced the
Experimental evidence of anomalous behaviour of MSCs in HCF utility of the cyclic crack tip displacement in an explicit micro-
was reviewed as early as 1986 [26]. Much progress has been made structure to capture essential features of MSC growth. Christ et al.
since regarding in situ experiments of grain scale inelastic defor- [122] have recently provided a thorough review of micro-
mation and growth of MSCs, although 3D measurements are still mechanical modelling of microstructurally small fatigue cracks
limited. Historical observations of MSC growth behaviour have growth, distinguishing three classes of models, namely those (a)
focused on surface measurements, which are of limited utility in based on an empirical treatment of grain boundary effects, (b) that
492 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

incorporate plastic deformation mechanisms at the crack tip (e.g.,

[48]), and (c) that are based on discrete dislocations approaches.
Discrete dislocation modelling is complicated by the uncertainty
associated with 3D dislocation structures, the incorporation of
grain boundary defects and their slip transfer reactions, amongst
others. Moreover, slip morphology can be complex in high stacking
fault energy materials that exhibit cross slip or for dislocations in
bcc structures with extended cores. Studies within the category (b)
are generally based on an extension of existing methods and con-
cepts to the 3D growth of MSCs (cf. Ko € ster et al. [123]).
Multiple finite element modelling strategies have explored the
growth of MSCs in the first few grains with and without explicit
provision for crack extension. Most have been applied to idealised
2D microstructures. Castelluccio and McDowell [124] introduced
3D crystal plasticity models and an algorithm for MSC growth in a
3D field of grains based on the FatemieSocie FIP averaged within
bands that circumscribe crystallographic slip planes within grains.
Here, crack growth was considered via the reduction of the elastic
stiffness in elements associated with the crack plane. Crack size
within individual grains was defined according to Murakami’s
Fig. 6. Comparison of crack length versus cycles for MSC growth between coarse and
approach of the square root of cracked cross-sectional area
refined meshes for 50 simulated realisations of a RR1000 15 mm mean grain size disc
[125,126]. As the cracks extended along crystallographic bands that bore microstructure [128]. Coarse and refined mesh simulations overlap and share
spanned the entire grain cross-section, the FIP decreased following similar minimum nucleation lives, indicating mesh insensitivity of the nonlocal (band)
a trend similar to the relation of Hobson et al. [127], and then averaging FIP approach.
continued to grow through several grains. The algorithm for 3D
growth through a field of grains makes use of redistributed cyclic
stressestrain fields computed over just a few stable cycles after model the growth of long cracks, such as enrichment functions and
crack front extension through each grain, as Fig. 5’s top view of a mesh free methods. It is worth noting that the work in Ref. [128]
projected planar representation of the 3D crack progression among can be expanded to incorporate microstructurally small crack in-
grains shows [128]. Analytical MSC growth relations based on the teractions with twist and tilt GBs in more detail [129e131] as well
band-averaged FIP serving as a surrogate for cyclic crack displace- as the effect of load amplitude and ratio dependence on slip irre-
ments are employed to extend the crack across each grain. In Fig. 5’s versibility and crack growth rate and direction.
case, the crack extends into the interior first (surface at left), and We conclude this section by highlighting that the formation and
then along the surface as shown in Fig. 6 for the RR1000 Ni-base PM growth of microstructurally small fatigue cracks can be enhanced
superalloy with a disc bore microstructure having a unimodal mean by the shear strain localisation induced by the presence of coherent
grain size of about 15 mm [128]. Note that this approach can predict S3 twins and GBs, as reported by Zhang et al. [79]. Such observa-
the crack evolution within grains for a given microstructure tions are consistent with those of Phung et al. [50] and the exper-
instantiation. Castelluccio and McDowell [128] also employed the imental study of Miao et al. [132], which suggest that
evolution of the FIP in a mesoscale MSC growth approach to esti- crystallographic crack growth in Ni-base superalloys occurs on
mate the fatigue resistance of bimodal microstructures. The 3D systems parallel to S3 twins but offset by some distance
growth of MSCs is arguably too complex to model the detailed (50e200 nm) by virtue of cyclic plastic shear strain intensification,
processes of crack front extension through individual grains in a effectively a strain concentration effect. From a mesoscale model-
continuous fashion. Moreover, the grain-scale plasticity is highly ling perspective, the use of the FIP as a surrogate measure of the
anisotropic and large in scale when compared to the crack size, cyclic crack tip displacement range for transgranular fatigue
which complicates the application of emergent approaches to growth seems to be consistent with these findings.

Fig. 5. Schematic top view of a 2D projection of the evolution of the crack front along bands within individual grains, and the total life consumed for grains AeE in relation to the
computational time/cycles [128]. The colored grains represent the path followed by the crack through the microstructure. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 493

3. Part 2: experiments and modelling of high temperature

fatigue

3.1. Surface crack oxidation of superalloys

In this section, an overview of the dominant mechanisms of

microstructural degradation and surface crack initiation in un-
coated or coated single crystal and polycrystalline Ni base super-
alloys subject to creep and fatigue loading conditions will be
reviewed in terms of the fundamental deformation modes, envi-
ronmental effects and interactions between environment and
deformation. In particular, the degradation of either coated or un-
coated Ni-base superalloys caused by the oxidation of a surface and
its effect on surface crack initiation and propagation behaviour will
be discussed. This situation applies to relatively high temperatures
typical of those experienced by the hot sections of discs and blades
in turbo-engines. In such a high temperature regime, the micro-
structure is not stable: the g0 precipitates coarsen rapidly, especially
under stress [133e135] and, in the vicinity of free surfaces (e.g.,
notches), there is a depletion of the g0 precipitates due to surface
oxidation [136]. Both mechanisms lead to a local softening of the
material which in turn reduces the local stresses. An example of
precipitate coarsening under cyclic loading is shown in Fig. 7.

3.1.1. Microstructural degradation due to surface oxidation

An important contributing factor to the failure of Ni-based sin-
gle-crystal superalloys operating under complex thermomechan-
ical loading and harsh environments is the localisation of the
inelastic deformation caused by the heterogeneity of the material
microstructure. In two-phase single crystals Ni-base superalloys,
heterogeneities exist at both the microscopic and the mesoscopic
levels (Fig. 8). At the microscale (see Fig. 8(c)), such heterogeneities
result from the presence of the g0 precipitates, while at the meso-
scale (see Fig. 8(b)), the presence of casting porosities introduces an
additional degree of heterogeneity in the local microstructure
[137]. Each of these features has its own particular influence on the

Fig. 8. Relevant material length scales: (a) notch in a CT specimen (macroscale); (b)
micrograph of a typical casting defect (mesoscale); (c) micrograph of the superalloy
microstructure (microscale).

overall behaviour and, as a consequence, the dimensional land-

scape of concern ranges over at least four orders of magnitude.
A further source of heterogeneity at the microstructural length
scale may arise from the microstructural degradation caused by
local oxidation, which becomes significant for uncoated Ni-base
superalloys at temperatures in excess of about 800  C (e.g., see
[138e143]. Generally, oxidation leads to the degradation of mate-
rials in two ways. First, a brittle oxide layer, which is more
vulnerable to cracking, forms on the material surface. Cracks initi-
ating in the oxide may propagate into the substrate and lead to
failure [134,135,144e148]. An example of surface connected oxide
Fig. 7. TEM image showing coarsening of g0 precipitates in Rene
80 during strain- cracking is seen in Fig. 9.
controlled LCF at 982  C. The plastic strain range was 0.196% and the strain rate was A second effect is the microstructural change in the bulk
7.7  105 1/s. Specimen failed at 530 cycles. Note the agglomeration of the large
precipitates, the dislocations at the g/g0 interfaces (determined to be edge in nature)
material beneath the oxide due to oxidation-linked diffusion and
and the relatively g0 precipitate free regions between the precipitates. Image formed interdiffusion processes, which alters the phase composition of
under two beam conditions with g ¼ 〈2 0 0〉 [134]. the material [134,148e150]. In effect, the matrix and g0
494 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

Fig. 11. Oxide-filled crack and depleted g-region surrounding a crack in RENE N4
following exposure to an oxidising environment for approximately 10 h at 1093  C
(adapted from [144]).
Fig. 9. Example of surface-connected, oxide-induced cracking during LCF of Rene
80
fatigued under strain control at 872  C. The plastic strain range was 0.25% and the
strain rate was 8.3  105 1/s. The test was stopped and the specimen removed after a
load drop of 75% (732 cycles) [134]. discs. These components are subject to fatigue cycles at elevated
temperatures (500e600  C) with imposed hold times at maximum
load which is of the order of 1 or 2 h for jet engines but can reach
precipitates both contain elements that are prone to oxidation. several days or months for land-based gas turbines.
Thus not only does an oxide form but, in consuming g0 stabilising Among the elements present in In 718 and other Ni base alloys
elements such as Ni, Al and Ti, there is also a zone, much like the such as In 706 and alloy 625, Nb is largely used to strengthen these
so-called precipitate free zones (PFZ) seen in Al alloys, which is materials. This element combines with titanium and aluminium to
free of g0 but which may contain complex phase mixtures as seen form g00 (DO22) Ni3 (Nb, Ti, Al), an ordered strengthening precipitate
in Fig. 10. [152]. Niobium can also precipitate as niobium carbides (NbC)
The local degradation in the material microstructure modifies when these materials are aged at elevated temperature
the material constitutive behaviour which in turn can lead to a (600e700  C) [152]. The presence of Ti leads to the formation of
reduction in the time to crack initiation [137]. TiNb (C, N) carbides during the solidification of the alloy, which
A micrograph illustrating the depleted g0 region surrounding a produces particles that can be fairly large (10e20 mm). NbC pre-
crack in the Ni-base superalloy RENE N4 at 1093  C is shown in cipitate on the free surface of Ti(CN) carbides leading to a coree-
Fig. 11 (from [144]). It may be noted that the size of the gregion is shell structure, Fig. 12.
on the order of 5 mm. The time at which this micrograph was taken NbC particles oxidise easily and the oxide Nb2O5 which is
is not stated in the paper but we have inferred it to be approxi- formed occupies a volume of about twice that of the NbC particles.
mately 10 h (based on a life of 20,000 cycles with a frequency of This volume expansion produces large local stresses which can be
0.5 Hz). tensile when the particles are located close to the free surface as
schematically shown in Fig. 13. This type of oxidation and particle
arrangement leads to the initiation of cleavage cracks in the TiNb
3.1.1.1. Preferential oxidation of near-surface carbides. Nickel-base su- (CN) particles, Fig. 13 [152]. It produces initial small cracks with a
peralloys like Inconel 718 (In 718) are used extensively for turbine size of the order of 10e20 mm even in the absence of mechanical
discs and other components in industrial and aerospace gas turbines. loading. These short cracks propagate during the fatigue loading
It represents more than 50% of the total production of superalloys. and their growth is accelerated by further oxidation [153]. It should
This explains why considerable effort has been expended in deter- be added that in many cases it has been observed that the fatigue
mining the safe operating life of critical components such as turbine cracks can be initiated at bulge-like features formed by volume
expansion of oxidising NbTi (CN) particles [154,155]. It is easily
understood that this situation implies a size effect: the fatigue life is
a decreasing function of component size as shown by Deyber et al.
[156].

Fig. 10. Illustration of oxidation-induced changes in a Ni-base superalloy. Near surface

region of sample of PWA 1484 which was oxidized at 1050  C for 300 h [151]. The
arrows indicate sliding of cracked oxides. Fig. 12. SEM image of TiN after passivation testing.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 495

Fig. 13. Effect of oxidation on (a) NbC [154] and (b) TiN.

3.1.1.2. Environmental protection of superalloys. The understanding cracking of an Electron Beam Physical Vapor Deposition (EBPVD)
of how microcracks nucleate at the surface of Ni-base superalloy NiCoCrAlY overlay coating after 6000 thermal cycles between
components subjected to thermal fatigue loadings is crucial to 520e1090  C is shown in Fig. 14. It can be seen in Fig. 14(a) that the
predict accurately their operational lifetimes. The effects of oxida- initial coating microstructure is made up of equi-axed grains of b
tion on fatigue life have been investigated experimentally in Ni- (NiAl) and g (Ni solid solution) grains. Fig. 14(b) shows that after
base superalloys under thermo-mechanical loading. MacLachlan 6000 520e1090  C thermal cycles, there is a considerable reduction
and Knowles [147] showed that during low cycle fatigue of the and coarsening of the b phase. Moreover, cracks appear to have
commercial superalloys RR2000 and CMSX4 at 850 and 950  C, initiated from the surface and then propagated into the substrate
fatigue cracks may initiate from oxide spikes on the surface. This [159].
phenomenon was also observed in tests by Ohtani et al. [148] on Generally two different-crack initiation processes are observed
the nickel-base single crystal superalloy CMSX10 at 1000  C as well in this type of coatings: at the surface due to early cracking of the
as by Antolovich et al. on Rene
77 [133], Rene
80 [134,135] and protective oxide scale which forms at the surface, and the metallic
Waspaloy [144] at elevated temperatures. Additional work on the coating-substrate interface region from cracks which originate
effects of oxidation on the fatigue behaviour of superalloys used in from internal Kirkendal type of voids that grow from the original
turbo-engine disc applications can be seen elsewhere [157,158]. coating-substrate interface towards the coating surface. Once the
When the superalloy chemistry is incapable of providing suffi- crack propagates into the substrate, the environmental interaction
cient resistance against high temperature oxidation, corrosion and with the microstructure adjacent to the crack faces becomes anal-
erosion, superalloys are protected by applying either a 50e100 mm ogous to the degraded microstructure of the type shown in Fig. 11. A
thick metallic coating or a thermal barrier coating (TBC) system, typical TBC system consists of three layers of dissimilar materials:
essential to high performance and long life. After a certain number an yttria stabilised zirconia (YSZ) top coat, an intermediate ther-
of either operating hours at high temperatures or thermal cycles, mally grown oxide (TGO) layer, and a metallic layer or bond coat
such coatings progressively degrade chemically and may eventually (BC). A representative microstructure of a Ptealuminide EB-PVD
develop surface cracks which propagate into the single crystal thermal barrier coating system and a CMSX-4 Ni-based superalloy
substrate. In this section, some of the most commonly observed substrate is shown in Fig. 15.
types of damage which develop in protective coatings during ser- The mechanisms of coating failure in TBCs are different from
vice, and which control the fatigue lifetimes of superalloy sub- those on overly coatings and strongly dependent on the interfacial
strates, will be briefly outlined. and microstructural features of the bimaterial system, which are in
A typical example of the initial microstructure and surface turn conditioned by the deposition method used. For instance, the

Fig. 14. (a) SEM micrograph of the initial microstructure of a typical 92 mm thick EB-PVD NiCoCrAlY overlay coating, and (b) degradation after 6000 520e1090  C thermal cycles
showing oxidation and crack propagation into the substrate [159].
496 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

plane stress component or bond-coat ratchetting behaviour

[161e166]. In some TBCs, the top ceramic layer is known to fail by
buckling involving a global in-plane compressive state, as predicted
by either a critical in-plane stress/strain [167] or an energy based
criterion [168e170]. Furthermore, some TBC systems are prone to
developing complex 3D morphologies, such as that due to the
rumpling mechanism which develops under thermal cycling
[171,172]. In all cases, it has been shown that creep properties of
both the TGO and bond-coat layers are required to accurately
predict stress relaxation at high temperatures [173,174]. Finally,
even though most TBC models are based on a 2D idealisation of the
problem, the role of complex interfacial roughness and 3D effects
has only recently been addressed. For instance, recent work by
Maurel et al. [175] has shown that 2D calculations of local stress
Fig. 15. Backscattered SEM image of a section through an EB-PVD TBC system on a fields in rough TBC interfaces tend to underpredict 3D results. Such
CMSX-4 substrate after exposure to air at 1100  C for 1 h [160].
considerations are becoming ever more relevant in view of the
considerable progress which has been made in validating TBC
models with powerful characterisation techniques such as X-ray
interface morphology of Fig. 14’s EB-PVD TBC is fundamentally tomography and piezo-spectroscopy.
different from that produced by other methods such as plasma-
sprayed (PS)-TBCs. In the latter, the high energy associated with
the impinging plasma sprayed particles gives rise to a denser 3.1.2. Modelling environmentemicrostructure interaction
coating than that produced by EB-PVD, containing porosities, so- To account for the effects of oxidation on the deformation and
lidification defects known as splats, and a rough and wavy inter- fracture behaviour of a metallic alloy, a coupled oxida-
face. The presence of such microstructural defects in the ceramic tionedeformation formulation that incorporates both time and
generally introduces a long-term reliability problem as they tend to spatial-dependent behaviour on different length scales is required.
sinter after prolonged exposure at high temperatures leading to an A physics based approach such as that relied on by Dumoulin et al.
increase in the coating elastic stiffness. Microstructural evidence of [137] and Zhao et al. [136] to model such phenomena is based on a
the formation of delamination-type cracks within a combustor multiscale continuum mechanics framework, which can readily be
liner’s PS-TBC after 18,000 h exposure to 520e1100  C thermal implemented within the finite-element (FE) method. Here, the ef-
cycles is shown in Fig. 16 [161]. Such cracks are understood to be fects of microstructure, kinetic processes and thermomechanical
caused by the coalescence of brittle microcracks which form as the loads on the initiation of surface cracks exposed to an oxidising
result of the high thermal and residual tensile stresses induced in environment is considered. Coupled finite element dif-
the ceramic material and TGO at the top of the wavy interface upon fusionedeformation analyses are performed to study crack initia-
cooling. tion in a notched compact tension specimen of a single crystal
Initial numerical studies to predict the effects of oxidation on superalloy. The specimen is assumed to contain a micro-void,
overlay and thermal barrier coatings concentrated on the effect of representing a material casting defect, near the notch. The
the thermo-elastic property mismatch between either the oxide approach also relies on a rate-dependent crystallographic frame-
scale and the substrate, or between the oxide layer and the TBC work to describe the macroscopic (i.e. homogenised) constitutive
constituents. An increased level of sophistication to predict the behaviour of the two-phase single crystal and the degradation of
lifetime of thermal barrier coating systems was achieved when the the superalloy due to AleCreNi diffusion can be illustrated with the
behaviour of the TGO interfaces with either the top ceramic or the aid of the schematic ternary NieCreAl phase diagram shown in
metallic bond coat were taken into account. They can be considered Fig. 17(b).
to be a function of local microstructural and mechanical variables The non-equilibrium conditions associated with the diffusion
such as, among others, oxide thickness and growth strains, out-of- processes across the g  (g þ g0 ) interface have been approximated
as a quasi-equilibrium problem since, experimentally, no discern-
ible transitional region (i.e. gradient in the precipitate volume
fraction) [141] can be seen. Thus, points a and b in Fig. 17(b),
represent the compositions of the g and g0 phases in equilibrium
with each other. Furthermore, point c in Fig. 17(b), which lies on the
tie-line, aeb, represents the initial superalloy composition. As the
Al is supplied to form the oxide, a non-equilibrium state thus exists
in the g þ g0 region of the alloy, leading to the dissolution of the g0
precipitates. The precipitates continue to dissolve as oxidation
progresses, leading to the development of two distinct regions: a
pure g region near the surface, and the unaffected g þ g0 region
within the bulk of the crystal. An illustration of the development of
these two distinct regions is given in Fig. 17(c), which shows typical
Al concentration profiles (from the oxideeg region interface) at two
different times for an AleCreNi alloy. Here, the interface between
the two regions translates away from the oxidizing surface with
time, leading to a thickening of the g region. The corresponding
diffusion path in the ternary phase diagram’s single-phase g region
Fig. 16. Typical mesocracks observed in the YSZ-thermally grown oxide (TGO) inter- would be ced (see Fig. 17(b)). The flux of aluminium from the metal
face region of a PS-TBC from a combustor liner after 18,000 h service [161]. into the oxide, JAl, at time, t, can be expressed as
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 497

Fig. 17. Schematic diagrams of (a) the oxidationediffusion processes at the crack faces, (b) NieCreAl ternary equilibrium phase diagram at the oxidising temperature and (c) one-
dimensional Al concentration profiles at two different times [137].

 12 where l and hab account for the nature of the dislocation distri-
kp
gAl ¼ ð1  VAl Þ (10) butions and cross hardening effects, respectively, whereas ba is the
t
magnitude of the Burgers vector. It is worth noting that recent work
where kp is the parabolic rate constant for oxidation, C IAl the Al has shown that the coefficients hab can be inferred directly from
concentration at the alloyeoxide interface, VAl the partial molar discrete dislocation dynamics (DDD) calculations [178]. The
volume of Al, and l a material-dependent parameter. For CMSX4, evolutionary behaviour of the overall dislocation density for each
l ¼ 0.125 mm3/mg and 1/VAl ¼ 81.8 at.% [137]. slip system is given by a KockseMecking [179] type-relation:
The macroscopic mechanical constitutive behaviour of the su- " #
peralloy was formulated using a finite deformation, rate-dependent 1 1
r_ T ¼ a _a
a
 dD rT jg j (13)
hyperelastic crystallographic framework (see [136,176]), as outlined ba laapp
in Section 2.3.1. The system shearing slip rate in the present case is
related to the resolved shear stress, ta , through the exponential A similar hardening-dynamic recovery time-dependent
function proposed by Busso and McClintock [177]: expression is used for the back stress tensor from which the
equivalent value Ba needs to be obtained (see [136]). It is worth
   
F jta  Ba j  Sa m=m0 p q noting that the multiscale nature of the formulation is incorporated
g_ ¼ g_ 0 exp  0 1  〈
a
〉 signðta  Ba Þ into Eq. (13) by making an explicit link between the initial dislo-
kq t̂
cation density and the dislocation mean free path with the features
(11)
of the microstructure at the micro-scale, namely the precipitate
size, D, and the mean dislocation spacing, la, i.e.,
which accounts for the absolute temperature (q, K) and the stress
dependence of the activation energy. In Eq. (11), F0 represents the X 1=2
Helmholtz free energy of activation at 0 K, k the Boltzmann con- la ¼ hab rbT (14)
̇ ̂
stant, g o a reference slip rate and, t , the maximum glide resistance
at which dislocations can be mobilised without thermal activation. Thus, for the current material
Furthermore, m and m0 are the shear moduli at q and 0 K, respec-
h
tively. The exponents p and q describe the shape of the energy ra0 ¼ r a0 ðD=Dm ; yf ; qÞ; (15)
barrier vs. stress profile associated with interactions between dis-
locations and obstacles. The average slip resistance in this case is
related to the overall dislocation density, rbT , by the relationship laapp ¼ la f ̂ s ðD=Dm ; yf ; qÞ (16)
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u0 1 where Dm is a reference particle size (¼500 nm) and vf the pre-
u X N
au cipitate volume fraction
S ¼ lmb t@
a
hab rT A b
(12)
b¼1
As discussed in Zhao et al. [136], the single crystal behaviour
described by the equations outlined in this section can account for
mechanical effects of the local dissolution of g0 precipitates linked
498 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

Fig. 18. (a) Distribution of the [010] stress component along a CT specimen symmetry line for L/d ¼ 10 at K ¼ 35 MPa at different times, (b) distribution of [010] stress when
effects of oxidation are accounted for, and comparison of (c) [010] stress with and without oxidation after 400 h, and (d) equivalent inelastic strain, with and without oxidation after
400 h [136].

to surface oxidation, the extreme case being when a pure g phase crystallographic axis and the crack lies on the (010) plane. To ac-
results, that is, when yf ¼ 0% in Eqs. (15) and (16). It should be count for the effects of a typical casting defect, such as that shown
noted that the introduction of microstructural related parameters in Fig. 8(b), on notch behaviour, a spherical void was assumed to
such as done here should also be possible in alternative single have a diameter of d ¼ 15 mm and to be located in the mid-plane at
crystal approaches found in the literature (e.g., see [180,181]). a distance L so that L/d ¼ 10 from the notch. To solve the overall
The above coupled diffusion-crystal plasticity formulation can oxidationedeformation problem, the fully coupled
be combined with a damage model to study the initiation of massediffusionemechanical analysis was performed whereby the
microcracks from the surface of existing internal voids at the aluminium concentration profile determined from a mass diffusion
mesoscale. Such an approach was followed in Zhao et al. [136] to analysis was used to identify the material behaviour at each inte-
investigate environmental effects on the crack tip stresses and gration point. If the concentration was above the upper solubility
strains in an oxidising superalloy as well as on the local damage limit of aluminium in g, the material properties were assumed to be
responsible for surface crack initiation. Typical finite element re- those of (g þ g0 ); otherwise they are those of pure g.
sults obtained on a compact tension specimen are given in Figs. 18 As oxidation of the superalloy leads to the formation of a soft g
and 19. The CT specimen contains a semi-circular notch with radius, -region adjacent to the notch root, the local reduction in strength of
R ¼ 0.1 mm, and the notch length to specimen width ratio, a/W, is this region results in the redistribution of stress (and strain) in the
0.5. The mid-plane of the specimen is normal to the [0 01] vicinity of the notch. These effects are shown in Figs. 18(a) and (b),

Fig. 19. Effect of oxidation of the superalloy CMSX-4 on the time to crack initiation under creep loading conditions at 950  C: dependence on (a) stress intensity factor K and on (b)
the oxidation rate constant kp with K ¼ 35 MPa m1/2 [136].
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 499

which reveal the evolution of the stress distribution in the notch the time to crack initiation for the single crystal alloy was also
vicinity for L/d ¼ 10 with and without oxidation. In Fig. 18, stress proposed in Zhao et al. [136] to account for the strong sensitivity of
and strain are plotted as a function of distance, r, from the notch the result to void location relative to a stress concentration such as
root; a distance r/L ¼ 1 corresponds to the void surface (see inset to a notch.
Fig. 18(a)). Note that the peak stress is initially located close to the
notch root (r/L ¼ 0) but following the redistribution of stress due to 3.2. Stress assisted intergranular oxidation of superalloys
creep (in Fig. 18(a)) and creep and oxidation (in Fig. 18(b)), the peak
stress shifts towards the void, (r/L ¼ 1). A direct comparison be- The effect of intergranular oxygen diffusion on the fatigue crack
tween the stresses with and without oxidation is provided in growth behaviour of polycrystalline Ni base superalloys used in gas
Fig. 18(c). It may be seen that due to oxidation and the formation of turbine disc applications, in particular In 718, will be considered in
the low strength g layer, the stress is significantly reduced in the this section. The temperature conditions of interest are those
vicinity of the notch (from 400 to 50 MPa at the notch root after encountered in gas turbine disc applications, typically ~600  C. The
400 h) and slightly increased near the void (from 460 to 500 MPa emphasis of this section will be on “creep-fatigue-oxidation” in-
after 400 h). The peak inelastic strain is still located in the vicinity of teractions which are important when these materials are subject to
the void (increasing from 0.39 to 0.47 at r/L ¼ 1) and crack initiation a dwell time, th, at maximum load during cycling loading. The main
is again predicted to occur at the void. mechanism to be discussed is stress-assisted intergranular fracture
It should be noted that in some cases the redistribution of driven by a combination of high oxygen pressure and local stresses
stresses predicted in Figs. 18 and 19 may be significantly influenced at critical strain rates. Recent results from creep-fatigue tests on In
by morphological changes in the g0 precipitate which depend upon 718 over a wide range of temperatures (550e700  C) and dwell
the difference in the lattice parameters of g and g0 as well as the times (90e3600 s) reported by Pierret et al. [186] and by Gustafsson
sense of the applied stress. Under uniaxial loading, the morphology et al. [187,188]. Ho€ rnqvist [189], Viskari et al. [190,191] show that
changes from cubes to plates which are either parallel or perpen- the dwell time effect on fatigue crack growth rates varies as t1/4h at
dicular to the sense of the applied stress depending upon the relatively low temperatures and dwell times. In contrast, tests
simultaneous signs of the stress (tension vs. compression) and the carried out at higher temperatures and longer dwell times revealed
difference in lattice parameters of the g and g0 [182e184]. For cyclic that fatigue crack growth rates are purely time-dependent, viz.
loading with plastic deformation, coarsening is very rapid resulting proportional to th. The most important operational effect is that, if
in a somewhat globular microstructure. During continuous strain the local stresses are not relaxed sufficiently fast by the time the
cycling, such as that which occurs at the crack tip, the stresses go material is exposed to high temperatures and loaded in fatigue,
from tension to compression and a “globular” g0 structure results. oxygen is able to diffuse along grain boundaries which may lead to
Effectively, the coarsened g0 structure is trapped between two their embrittlement. The regime associated with a slope of ¼ might
equilibrium morphologies. The effects of mismatch are discussed be related to the fact that oxygen ingress takes place not only at the
further below in the section on TMF. grain boundary but also along the free surface [192]. The regime
The effect of oxidation on the predicted time to crack initiation is where the slope is equal to 1, i.e. where the FCGR is purely time-
shown in Fig. 19. Here, the time to crack initiation, ti, was calculated dependent, has been discussed recently [187,188,186]. It will be
assuming that initiation occurs when the critical equivalent in- shown that the deleterious dwell time effect on fatigue crack
elastic strain, εc, is reached at a small distance, r, ahead of the crack growth behaviour can be largely reduced when stress relaxation
tip. The expression for ti originally proposed for isotropic materials occurs rapidly during the dwell period by applying either a small
by Shih [185] in terms of the stress intensity factor, K, adapted to under-load or a temporary temperature increase (~25  C) at the
deal with the single crystal anisotropy, was used. The results show onset of dwell period. This effect can also be achieved by devel-
that the magnitude of the reduction in life due to oxidation de- oping beneficial microstructural modifications. The real operating
pends on the oxide growth rate constant kp and the loading level, conditions in a jet engine mimic, to some degree, these conditions
quantified by the stress intensity factor, K, for a given void location. resulting in a large decrease in crack propagation rates thus
Lower values of K and higher values of kp lead to a greater reduction providing an added measure of durability and safety to the engines.
in the predicted life time. Fig. 19(a) shows the predicted time to In the preceding sections it has been shown that there is a dynamic
microcrack initiation as a function of K, for L/d ¼ 10 with kp ¼ 2.84 interaction between oxide-induced microstructural changes and
10e11 mg2/mm4/s (the value for CMSX4). As expected, the time to the redistribution of stresses and strains as well as a trade-off be-
crack initiation is reduced if the effects of oxidation are included. tween stress relaxation and the potential for environmental dam-
Fig. 19(a) also shows that the effect is reduced with an increase in age. In this section some experimental results are reprised which
load, since the life-time is significantly shortened and that in turn broadly reinforce the ideas presented in the preceding section.
reduces the time for oxidation to occur. The sensitivity of the crack In some cases, high primary creep rates have been shown to be
initiation time to kp, the oxide growth rate constant, is shown in beneficial on fatigue crack propagation. A turbine disc fabricated
Fig. 19(b) for L/d ¼ 10 and K ¼ 35 MPa m1/2. from the Ni-base superalloy N18 and given a standard heat treat-
Under mechanical loading, significant inelastic strain was found ment was studied with one of the goals being to sort out the effects
to accumulate at the void surface. Using a strain-based failure cri- of the environment and creep as well as their interactions in either
terion, implemented at the meso-scale, the time to crack initiation increasing or decreasing the fatigue crack propagation (FCP) life
under creep loading was predicted. As a result of the diffusion and [193]. See also [194,5,195] for additional discussions of these re-
interdiffusion processes linked to notch surface oxidation, the sults. The heat treatments resulted in coarse precipitates in the
phase composition of the material is altered locally and introduces interior and much finer precipitates in the surface region. The stress
a pure g phase in the notch area. The presence of the g region in- relaxation rates were studied and it was found that the creep or
tensifies the strain accumulation and leads to earlier crack initia- relaxation rate was higher for specimens taken from the interior
tion. The results reported by Zhao et al. [136] predicts that the crack than it was for those taken from the surface region. The FCP rates, as
initiation time is strongly dependent on the void location but is less shown in Fig. 20, were very dependent upon the state of the ma-
sensitive to environment (i.e. whether oxidation occurs) in mate- terial. When loaded with a 10 s ramp, held for 300 s and then
rials with a relatively low magnitude of the oxide growth rate unloaded in 10 s (10e300e10), or when loaded with a 10e10 cycle
constant, kp. Finally, a probabilistic framework for the prediction of under vacuum, there was no difference in behaviour. Also, for both
500 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

Fig. 20. Effect of cycle type and environment on the crack growth rate of N18 super-
alloy taken from the interior (rapid relaxation) and the surface (high creep resistance).
A comparison of the results demonstrates that stress relaxation is beneficial in
reducing the environmental effects [193]. Fig. 22. Effect of PO2 on the FCP rate for different K levels [198].

microstructures tested under a 10e10 cycle in air there was no on the mechanical state, increasing with K levels. In Fig. 23 the
difference in the FCP rates. Comparison of these results leads to the effect of cycle type on the oxidation behaviour is shown.
conclusion that creep effects are either absent (e.g., 10e10 cycle) or Longer cycle times lead to higher plateau growth rates for a
not very damaging. However for hold time testing carried out in air, given mechanical state (i.e. K level). Additional studies applied an
the difference in the FCP rate was highly dependent upon the oxygen pressure pulse (60 s duration and 100 MPa pressure)
microstructure and very large. The interior material, which showed throughout the 10e300e10 vacuum fatigue test. The 60-s pressure
the most rapid stress relaxation, was seven times more fatigue oxygen pulse allows the more damaging spinels to form but does
resistant than was the surface material where stress relaxation was not provide enough time for passivation by the more protective
very slow. Clearly stress relaxation at the crack tip was actually Cr2O3 to occur. Under these conditions the crack growth rate
beneficial, making oxidation less damaging. This could be explained accelerated, regardless of where in the cycle the oxygen partial
by either a reduction of the oxidation rate, a reduced crack tip stress pressure pulse, so long as the local crack tip strain rate remained
due to relaxation or both. positive. This shows that the damaging process of oxygen is
Additional studies on the interaction between crack tip stresses continuous. As the strain rate decreased (e.g., late in the hold
and oxidation effects were carried out on In 718 [196]. Based on period), the damaging effect was progressively reduced. When the
thermodynamic calculations, it was recognised that friable Ni, Fe crack tip strain rate was negative (e.g., start of unloading) the
spinel-type oxides could form rapidly at high partial pressures but deleterious effect of oxygen was virtually eliminated. It is thus
that as the diffusion distance to the oxide/metal interface at the concluded that any heat-treatment which develops a microstruc-
crack tip increased, then the partial pressure would be reduced and ture giving rise to more rapid relaxation at the crack tip (i.e. more
a more thermodynamically stable, mechanically protective Cr- rapid primary creep) would reduce the severity of environmental
oxide could form. Some of the important results are shown in attack in In 718. The direct ageing treatment (i.e. ageing directly
Figs. 21 below. after forging and usually denoted as DA) produces coarser pre-
This figure illustrates the fact that the minimum cracking rate is cipitates and a somewhat increased creep rate which would tend to
reached very soon for low partial pressures of O2, (PO2) and with reduce environmental attack. Further examples of load relaxation
increasing PO2, the time increases. This happens since the damaging effects are for the Ni-base superalloy N18 are seen in Fig. 24 for N18
Ni and Fe oxides are stable for a longer time until the partial with similar behaviour being observed for In 718 [193,196]. Fig. 24
pressure at the metal oxide interfaces becomes low enough (due to demonstrates that upon decreasing the load for crack propagation
the tortuous path to the crack tip) to stabilise the protective Cr with a hold time, the effect of the hold time is reduced until a
oxides. This phenomenon has been further studied by Molins et al. minimum crack propagation rate plateau occurs for a load reduc-
[198]. Fig. 22 illustrates that the transition from passivating to non- tion on the order of 20%.
passivating occurs at a fixed PO2, but after that transition the rate
increases with increasing applied stress intensity parameter K. That
figure suggests that there is an upper plateau which is dependent

Fig. 23. FCP rates at a constant mechanical state but for different loading profiles. The
Fig. 21. Transition time to passivation as a function of oxygen partial pressure showing transition is independent of the load and the plateau depends upon the cycle character,
the type of stable oxide [197]. with longer cycle times being most damaging [198].
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 501

Fig. 24. Effect of load reduction on the crack growth rate of the superalloy N18 during
hold time fatigue crack propagation [193].

The plateau implies that there is no further crack tip stress

reduction after a macroscopic load reduction of about 20%. Such a
conclusion is supported by FEA calculations [199]. The effect of a
load reduction during hold time fatigue of In 718 was also studied
and the reduction in the FCP rate was even more pronounced than
for N18 calculations [199]. This effect on reducing the hold time FCP Fig. 26. Comparison of model predictions with experiment for FCP rates of In 718
rate through modest load reductions is very fortunate since in jet during continuous and hold time fatigue [201].
engines the operating conditions are such that modest load re-
solution Ni-base alloy [203]. Creep-fatigue studies were carried out
ductions occur during normal operation thus providing for a longer
at 950  C and it was found that the fatigue life was considerably
life and an additional margin of safety.
reduced when a constant strain dwell was added at the peak tensile
The two-oxide mechanism with passivation model is shown in
strain. The majority of the damage was due to extensive cavity for-
Fig. 25 calculations [200]. This study also showed that a small, high-
mation on the interior grain boundaries. Later during the fatigue
frequency superimposed load pulse during a 300 s hold at zero
process, surface cracks formed, presumably due to oxidation; they
stress caused a significant increase in the crack growth rate by
initiated a surface crack which grew rapidly in an intergranular mode
breaking up the non-passivating oxide and allowing for continued
through the pre-cavitated boundaries. It was also found that the
high crack growth rates. It was also pointed out that this process
stress in a strain controlled creep-fatigue test relaxed very rapidly
could be quantified in a fairly straight forward manner. In particular
and at some point, further increases in hold time had no effect on
these authors pointed out that: “…the significance of the influence of
reducing the creep-fatigue life since the stresses were so low. One
environment on crack growth behaviour could be quantitatively
might argue, in agreement with the results previously cited, that the
described on the basis of the build-up rate of the chromia layer taking
lower stresses reduced the effect of oxidation (i.e., very low strain
into account the oxygen partial pressure as well as the degree of
rate) shifting the principal damage mode to creep-cavitation.
chromium depletion from the matrix surrounding the affected grain
boundary paths.”
Following these suggestions, a quantitative model was devel- 3.3. Thermomechanical fatigue of superalloys
oped [201] and applied to several applications [202]. Comparison of
the model predictions to experimental crack growth rates is shown A very important practical application for many of the fatigue
in Fig. 26 and the agreement appears to be quite encouraging. studies reviewed in this article is life prediction and extension of
Similar results were obtained in a study of Alloy 617, a solid aero-engine components. In the literature, one finds that the great

Fig. 25. Schematic of the two oxide passivation model [196].

502 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

Fig. 27. (a) Schematic of a model to predict the OP TMF in a single crystal, and (b) optical micrograph showing slip band impingement on oxide defect in the Ni-base superalloy PWA
1484 [210,212].

majority of fatigue studies have been carried out isothermally. Re- actual physical processes, a model could be developed that was
sults of these and similar isothermal studies are used in material both more accurate and which could be used for extrapolation
selection and life prediction. However, in actual operation, both outside of the calibration range. Based on all of the preceding, and
loading parameters1 and temperatures vary with time. Furthermore, noting that out-of-phase (OP) TMF is generally most damaging in Ni
the loads and temperatures are not necessarily in phase with each base superalloys, the model refined and extended a previous
other which is a very important feature of TMF. In recent years, environmental-based model for LCF in polycrystalline Rene
80
research aimed ultimately at developing a life prediction method- [134]. The basis of the model is shown in Fig. 27 [210,212]. The
ology has been carried out in our laboratories and the evolution of model is based on the hypothesis (supported by experimental ob-
this work has been reported in numerous publications [204e213]. servations) that oxides formed during the high temperature/low
The most important findings are summarized in this section. It is strain portion of the cycle are cracked by impingement of slip bands
appropriate to note that such type of work was motivated, in part, by formed at the low temperature/high strain part. The additional
simple experiments designed to simulate TMF in Rene
80 [213]. In effects taken into account by the model to provide explicit details of
these experiments, a specimen was first cycled at 760  C to about microstructural evolution and damage were: (i) precipitate coars-
half of its projected life (~270 cycles) at Dεp ¼ 0.05%. After stress free ening, (ii) slip band formation and spacing as affected by coars-
cooling, it was then cycled at 25  C at the same strain range. In this ening, and (iii) surface roughness as applied to oxidation rates.
so-called hi-lo (HL) cycle the specimen broke on the first loading The model can also be modified to account for imposed high
cycle at 25  C even though according to isothermal fatigue data, the frequency cycles during the hold portion of a TMF test [211].
sample was predicted to last for hundreds of additional cycles. Post Based on the preceding considerations, the initiation life, Ni, is
mortem examination revealed that slip bands formed at 25  C given by:

1
m ! ðnþdÞ  h
Aro Dgsl
in Qsl þ 2Qox  Bðsmax þ jtmin j n Ra C
Ni ¼ exp A (17)
exp uTeff t 1=3 RTeff nth þ 0:5 8

causing cracking of the previously-formed oxides. Upon reversing where A and u are constants determined from precipitate coars-
the temperature sequence (lo-hi cycle (LH)) at the same plastic ening, t is the total precipitate coarsening time, ro is the initial
strain range, fracture occurred at the predicted number of cycles at equivalent precipitate size, Dgsl in , is the inelastic strain on the slip
the higher temperature for the imposed loading conditions. The system of interest, Teff is the effective temperature (defined below),
specimen behaved as if there was no previous cycling leading to the Qsl is the slip activation energy, Qox is the oxidation activation en-
hypothesis that any prior cyclic damage at 25  C was annealed out. ergy, smax is the maximum tensile stress in the stabilised cycle
To investigate the rates of damage accumulation, the experiment (occurring at the cycle minimum temperature), tmin is the shear
was repeated by cycling for about one quarter of the life at high stress for slip band formation, B is a constant, R is the gas constant,
temperature with continued cycling at 25  C. In this case, the room m is the strain dependency exponent, n is the time exponent
temperature life was over 6000 cycles indicating that damage determined from the non-stressed g0 depletion experiments, d is an
accumulation was both sequence-dependent and non-linear as well. experimentally-determined parameter related to precipitate
The TMF model was first developed using the model proposed coarsening kinetics, n is the cycle frequency, th is a high temperature
by Neu and Sehitoglu [146] (NS model) which is based on a global hold time (if any), h is the exponent determined from g0 depletion
appreciation of damage mechanisms (oxidation, creep and fatigue) surface roughness experiments, and Ra is the specimen surface
and which works reasonably well for practical life prediction. This roughness (arithmetic average) and p is a combined parameter. The
model was adapted for PWA 1484 and provided about a 70% cor- term Teff is the Arrhenius-averaged2 temperature in a TMF cycle
relation for life predictions. However, by including details of the defined as:

1 2
In this part of the article “loading” is used in a general sense and includes all The effective temperature provides a much more accurate assessment of the
modes of loading. Whether strains, loads or stresses are considered, it depends temperature than the arithmetic average temperature for thermally activated
upon the context. processes.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 503

Ott and Mughrabi [214], the alignment of the plates in the direction
of loading is expected to be beneficial for the TMF life. It would be
valuable to carry out similar experiments with an alloy in which the
mismatch parameter was positive to explore this potential source
of improving the TMF life.
To further test the model, it was extended to notches with a
notch severity of Kt ¼ 1.72. Crack tip stresses and strains were
calculated using Neuber’s rule. Agreement of the predicted initia-
tion life with test results was surprisingly good (within about 20%).
A more fully-developed crystal visco-plasticity model, a state-of-
stress dependent damage criterion, and a detailed FE analysis
have the potential to significantly improve the notch predictions
[212]. An idealized path for improving notch fatigue initiation life
predictions is reproduced in Fig. 30 for completeness.

4. Concluding remarks

Microstructure-sensitive computational strategies for esti-

mating the influence of microstructure on fatigue life at ambient
temperatures have been discussed in the context of a multistage
Fig. 28. Life prediction using Eq. (17) for single crystalline, [0 01]-oriented PWA 1484.
fatigue formulation. Decomposition of conventional initiation
Out-of-phase (OP) TMF and bi-thermala (BiF) results are shown for a wide range of
effective temperatures and strain ranges. The data designated as “correlated data”
fatigue life into parts associated with crack nucleation and early
were used to implement the model through Eq. (17) (solid line) [210,212]. aBithermal growth of micro-structurally and physically small cracks is
fatigue (BiF) is a TMF-type of cycle in which half of the hysteresis loop is traversed at introduced and recent developments with regard to bottom-up
the high temperature and the other half at the low temperature. The BiF cycle is useful and top-down modelling approaches have been discussed.
for isolating damage interactions and may be done for either OP or IP cycles.
Emphasis has been placed on top-down approaches which are

Fig. 29. Effect of TMF cycle on precipitate microstructure of single crystalline PWA 1484 along the specimen axis: (a) out-of-phase and (b) in-phase. TMF cycling was performed
using 1.3% mechanical strain. The lowest and highest cycle temperatures were 550  C and 1050  C, respectively. The observed morphological changes are consistent with a negative
mismatch parameter.

particularly adapted to describe early stage fatigue, such as

! Zt2   under high cycles, and are useful to underpin (i) current un-
Q 1 Q derstanding of the fatigue failure of in-service components, (ii)
exp  ¼ exp  dt (18)
RTeff Dt RTðtÞ life prediction schemes, (iii) microstructure selection and design
t1
for application-specific scenarios, and (iv) the development of
where Q is the activation energy for the process, Dt is the cycle time, improved processing methods to enhance fatigue resistance.
R is the gas constant and T(t) is the temperatureetime profile of the The foregoing approaches for modelling formation and early
cycle. growth of fatigue cracks are primarily useful for comparing
Life predictions were made and compared to experiment using microstructures in terms of fatigue resistance. Cyclic plastic
experimentally determined values for single crystalline PWA 1484. strain localisation and cracking processes occurring at the sub-
The results are shown in Fig. 28. grain scale have relatively high uncertainty, which can be
It can be seen that the predictive capability of the model is reduced in time via understanding gained from in situ experi-
excellent, providing a correlation of the data of about 92% accuracy ments and bottom-up modelling and simulation.
using the coefficient of determination (Usually denoted as R2). The In the second part of this overview, the linkage between
precipitate morphology was examined for both IP and OP loading as microstructure, deformation mode, environmental damage,
seen in Fig. 29 [212]. The morphological changes were driven by the compositional modifications and life prediction have been
sign of the stress at the highest temperature and are consistent reviewed and discussed. It was demonstrated how a fundamental
with a negative mismatch parameter.3 Connecting with the work of understanding of these processes and their interactions can be used
to develop accurate life prediction methodologies. It was also
pointed out how this basic understanding is being incorporated
3
The mismatch parameter is denoted as d and is defined as d ¼ (ag0  ag)/aavg, into advanced computational techniques. It is anticipated that it is
the difference between the lattice parameters of the precipitate and matrix in this area that significant future advances will be seen.
respectively divided by the average lattice parameter.
504 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

Fig. 30. Suggested path for the development of improved notch TMF life prediction.

Acknowledgements References

DLM is grateful for the support of the Carter N. Paden, Jr. [1] R. Prasannavenkatesan, C.P. Przybyla, N. Salajegheh, D.L. McDowell, Eng.
Fract. Mech. 78 (2011) 1140e1155.
Distinguished Chair in Metals Processing. AP would like to thank [2] R. Prasannavenkatesan, J. Zhang, D.L. McDowell, G.B. Olson, H.-J. Jou, Int. J.
SNECMA for the support of many of his studies devoted to high Fatigue 31 (2009) 1176e1189.
temperature fatigue of Ni-Base superalloys. SDA is grateful to [3] R. Prasannavenkatesan, D.L. McDowell, ASME J. Eng. Mater. Technol. 132
(2010) 031011.
NASA, AFOSR, GE Aviation, Pratt & Whitney for financial support [4] A. Pineau, S.D. Antolovich, High temperature fatigue, in: C. Bathias, A. Pineau
over many years and to the Ecole des Mines de Paris for making (Eds.), Fatigue of Materials and Structures, ISTE/John Wiley, London, 2011,
it possible to spend several sabbatical visits at the Centre des pp. 1e130.
[5] A. Pineau, S.D. Antolovich, High temperature fatigue of nickel base superal-
Mate
riaux. The support received by EPB from Siemens-UK,
loys e a review with special emphasis on deformation modes and oxidation,
Hitachi, UK’s EPSRC and France’s ANR to study high tempera- Eng. Fail. Anal. 16 (2006) 2668e2697.
ture alloys is greatly appreciated. The authors would also like to [6] A. Pineau, S.D. Antolovich, High temperature fatigue: behaviour of three
typical classes of structural materials, Mater. High Temp. 16 (2015) 298e317.
acknowledge the contributions of their research students over
[7] E.P. Busso, From single crystal to polycrystal plasticity: an overview of main
the years, many of whom occupy important positions in in- approaches, in: George Z. Voyiadjis (Ed.), Handbook of Damage Mechanics:
dustry, academia and in related government laboratories in Nano to Macro Scale for Materials and Structures, Springer (publ.), 2014, pp.
369e394.
Europe and the US.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 505

[8] J.A. Bannantine, J. Comer, J.L. Handrock, Fundamentals of Metal Fatigue [61] G. Castelluccio, D.L. McDowell, J. Mater. Sci. 48 (2013) 2376e2387.
Analysis, Prentice-Hall, Englewood Cliffs, NJ, 1990. [62] J.D. Hochhalter, D.J. Littlewood, R.J. Christ, M.G. Veilleux, J.E. Bozek,
[9] T.M. Pollock, J.E. Allison, D.G. Backman, M.C. Boyce, M. Gersh, E.A. Holm, A.R. Ingraffea, A.M. Manieatty, Model. Simul. Mater. Sci. Eng. 18 (2010)
R. LeSar, M. Long, A.C. Powell, J.J. Schirra, D.D. Whitis, C. Woodward, Inte- 045004.
grated Computational Materials Engineering: A Transformational Discipline [63] J.D. Hochhalter, D.J. Littlewood, M.G. Veilleux, J.E. Bozek, A.M. Maniatty,
for Improved Competitiveness and National Security, National Materials A.D. Rollett, A.R. Ingraffea, Model. Simul. Mater. Sci. Eng. 19 (2011) 035008.
Advisory Board, NAE, National Academies Press, 2008. Report Number: ISBN- [64] D.L. McDowell, in: ASM Handbook Fundam. Model. Met. Process, vol. 22A,
10: 0-309-11999-5. ASM International, 2009.
[10] D.L. McDowell, J. Panchal, H.-J. Choi, C. Seepersad, J. Allen, F. Mistree, Inte- [65] D.L. McDowell, F.P.E. Dunne, Int. J. Fatigue 32 (2010) 1521e1542.
grated Design of Multiscale, Multifunctional Materials and Products, first ed., [66] J. Polak, M. Sauzay, Mater. Sci. Eng. A (2009) 122e129.
Butterworth-Heinemann, Oxford, 2010, ISBN 978-1-85617-662-0. [67] D.L. McDowell, Mater. Sci. Eng. R. Rep. 62 (3) (2008) 67e123.
[11] D. McDowell, G. Olson, Sci. Model. Simul. 15 (2008) 207e240. [68] U. Essmann, U. Go € sele, H. Mughrabi, Phil. Mag. A 44 (1981) 405e426.
[12] S. Suresh, Fatigue of Materials, second ed., Cambridge University Press, [69] K. Differt, U. Essmann, H. Mughrabi, Phil. Mag. A 54 (1986) 237e258.
Cambridge, UK, 1998. [70] K. Tanaka, T. Mura, J. Appl. Mech. 48 (1981) 97e103.
[13] A. Pineau, Anal. Mecan. Fract. 1 (2007) 9e24. [71] A. Navarro, E.R. De los Rios, Phil. Mag. A 57 (1988) 15e36.
[14] A. Pineau, in: D. François (Ed.), Advances in Fracture Research ICF5 Ed., [72] A. Navarro, C. Vallellano, V. Chaves, C. Madrigal, Int. J. Fatigue 33 (2011)
Pergamon, Oxford, UK, 1982, pp. 553e577. 1048e1054.
[15] A. Pineau, in: R.A. Ainsworth, K.-H. Schwalbe (Eds.), Comprehensive Struc- [73] C. Vallellano, A. Navarro, J. Domínguez, Int. J. Fatigue 46 (2013) 27e34.
tural Integrity, vol. 7, Elsevier, 2003, pp. 177e225. [74] M.D. Sangid, H.J. Maier, H. Sehitoglu, J. Mech. Phys. Solids 59 (2011)
[16] D.F. Socie, T&AM Report No. 417, University of Illinois, 1977. 595e609.
[17] R.J. Morrissey, D.L. McDowell, T. Nicholas, Int. J. Fatigue 23 (2001) 55e64. [75] M.D. Sangid, H.J. Maier, H. Sehitoglu, Acta Mater. 59 (2011) 328e341.
[18] R. Morrissey, C.-H. Goh, D.L. McDowell, Mech. Mater. 35 (3e6) (2003) [76] M.D. Sangid, T. Ezaz, H. Sehitoglu, I.M. Robertson, Acta Mater. 59 (2011)
295e311. 283e296.
[19] D.L. McDowell, in: Sidney Yip, M.F. Horstemeyer (Eds.), Handbook of Mate- [77] M.D. Sangid, H.J. Maier, H. Sehitoglu, Int. J. Plasticity 27 (5) (2011) 801e821.
rials Modeling, Part A: Methods, Springer, the Netherlands, 2005, pp. [78] M.D. Sangid, Int. J. Fatigue 57 (2013) 58e72.
1193e1214. [79] P. Zhang, S. Qu, Q.Q. Duan, S.D. Wu, S.X. Li, Z.G. Wang, Z.F. Zhang, Phil. Mag.
[20] D.L. McDowell, Mater. Sci. Eng. A 468e470 (2007) 4e14. 91 (2) (2011) 229e249.
[21] J.W. Provan, in: Advances in Research on the Strength and Fracture of Ma- [80] M.A. Tschopp, D.E. Spearot, D.L. McDowell, Influence of grain boundary
terials, 4th Int. Conference on Fracture, June 19-24, 1977, Waterloo, Ontario, structure on dislocation nucleation in FCC metals, in: J.P. Hirth (Ed.), Dislo-
Pergamon Press, 1977. cations in Solids, A Tribute to F.R.N. Nabarro, vol. 14, Elsevier Publ., 2008, pp.
[22] J.W. Provan, in: G.C. Sih, et al. (Eds.), Defects and Fracture, Martinus Nijhoff 43e139.
Publ, The Hague, 1982, pp. 63e69. [81] V.S. Deshpande, A. Needleman, E. Van der Giessen, Acta Mater. 51 (15)
[23] T. Sakai, K. Tokaji, T. Ogawa, Trans. Jpn. Soc. Mech. Eng. A 57 (539) (1991) (2003) 4637e4651.
1514e1521. [82] S. Groh, S. Olarnrithinun, W.A. Curtin, A. Needleman, A. Deshpande, E. Van
[24] K. Tanaka, Y. Akiniwa, in: Proc ICF7, vol. 2, Houston, TX, March 20e24, 1989, Der Giessen, Phil. Mag. 88 (30e32) (2008) 3565e3583.
pp. 869e887. [83] K. Kirane, S. Ghosh, Int. J. Fatigue 30 (12) (2008) 2127e2139.
[25] T. Hoshide, D.F. Socie, Eng. Fract. Mech. 29 (3) (1988) 287e299. [84] K.L. Roe, T. Sigmund, Eng. Fract. Mech. 70 (2) (2003) 209e232.
[26] D.L. McDowell, Int. J. Fract. 80 (1986) 103e145. [85] B. Tomkins, Phil. Mag. 18 (1968) 1041e1066.
[27] K.J. Miller, Mater. Sci. Technol. 9 (1993) 453e462. [86] T. Hoshide, D.F. Socie, Eng. Fract. Mech. 26 (1987) 842e850.
[28] K.J. Miller, Fatigue Fract. Eng. Mater. Struct. 16 (9) (1993) 931e939. [87] D.L. McDowell, J.-Y. Berard, Fatigue Fract. Eng. Mater. Struct. 15 (1992)
[29] K. Gall, H. Sehitoglu, Y. Kadioglu, ASME J. Eng. Mater. Technol. 119 (2) (1997) 719e741.
171e178. [88] D.L. McDowell, ASM Handbook, Vol. 19 on Fatigue and Fracture, ASM In-
[30] V.P. Bennett, D.L. McDowell, in: K.J. Miller, D.L. McDowell (Eds.), ASTM STP ternational, 1996, pp. 263e273.
1359, 1999, pp. 203e228. [89] A. Fatemi, D.F. Socie, Fatigue Fract. Eng. Mater. Struct. 11 (1988) 149e165.
[31] V.P. Bennett, D.L. McDowell, Int. J. Fatigue 25 (2003) 27e39. [90] A. Fatemi, P. Kurath, ASME J. Eng. Mater. Technol. 110 (1988) 380e388.
[32] V.P. Bennett, D.L. McDowell, Eng. Fract. Mech. 70 (2003) 185e207. [91] M. Brown, K.J. Miller, in: C. Amzallag, B. Leis, P. Rabbe (Eds.), ASTM STP 770,
[33] R.J. Asaro, Adv. Appl. Mech. 23 (1983) 1e115. ASTM, 1982, pp. 482e499.
[34] D. Peirce, R.J. Asaro, A. Needleman, Acta Met. 31 (12) (1983) 1951e1976. [92] L.L. Li, P. Zhang, Z.J. Zhang, Z.F. Zhang, Acta Mater. 61 (2013) 425e438.
[35] R.J. Asaro, A. Needleman, Scripta Met. 18 (5) (1984) 429e435. [93] W.D. Musinski, D.L. McDowell, Int. J. Fatigue 37 (2012) 41e53.
[36] H.E. Deve, R.J. Asaro, Met. Trans. A 20 (4) (1989) 579e593. [94] C.A. Sweeney, W. Vorster, S.B. Leen, E. Sakurada, P.E. McHugh, F.P.E. Dunne,
[37] P.E. McHugh, A.G. Varias, R.J. Asaro, C.F. Shih, Fut. Generat. Comput. Syst. 5 J. Mech. Phys. Solids 61 (2013) 1224e1240.
(2e3) (1989) 295e318. [95] F. Bridier, D.L. McDowell, P. Villechaise, J. Mendez, Int. J. Plasticity 25 (2009)
[38] T.L. Sham, A. Needleman, Acta Met. 31 (6) (1983) 919e926. 1066e1082.
[39] T. Christman, A. Needleman, S. Nutt, S. Suresh, Mater. Sci. Eng. A 107 (1e2) [96] Y. Guilhem, S. Basseville, F. Curtit, J.-M. Ste
phan, G. Cailletaud, Int. J. Fatigue
(1989) 49e61. 32 (2010) 1748e1763.
[40] P. Knoche, A. Needleman, Eur. J. Mech. 12 (1993) 585e601. [97] K. Kirane, S. Ghosh, M. Groeber, A. Bhattacharjee, ASME J. Eng. Mater.
[41] G.L. Povirk, S.R. Nutt, A. Needleman, Scripta Met. Mater. 26 (3) (1992) Technol. 131 (021003) (2009) 1e14.
461e466. [98] M. Anahid, M.K. Samal, S. Ghosh, J. Mech. Phys. Solids 59 (2011) 2157e2176.
[42] A. Needleman, Ultramicroscopy 40 (3) (1993) 203e214. [99] S. Ghosh, P. Chakraborty, Int. J. Fatigue 48 (2013) 231e246.

k, J. Man, K. Obrtlık, Int. J. Fatigue 25 (2003) 1027e1036.
[43] J. Pola [100] C.P. Przybyla, D.L. McDowell, Int. J. Plasticity 26 (2010) 372e394.

k, Mater. Sci. Eng. A 468e470 (2007) 33e39.
[44] J. Pola [101] C.P. Przybyla, D.L. McDowell, Int. J. Plasticity 27 (12) (2011) 1871e1895.
[45] H. Mughrabi, Met. Metal. Trans. B 40 (4) (2009) 431e453. [102] C.P. Przybyla, D.L. McDowell, Acta Mater. 60 (2012) 293e305.
[46] H. Mughrabi, Int. J. Fatigue 57 (2013) 2e8. [103] C. Przybyla, R. Prasannavenkatesan, N. Salajegheh, D.L. McDowell, Int. J. Fa-
[47] R.O. Ritchie, J. Lankford, Mater. Sci. Eng. 84 (1986) 11e16. tigue 32 (2010) 512e525.
[48] B. Künkler, O. Duber, P. Koster, U. Krupp, C.P. Fritzen, H.J. Christ, Eng. Fract. [104] B. Wen, N. Zabaras, Comput. Mater. Sci. 51 (2012) 455e481.
Mech. 75 (2008) 715e725. [105] A. Pineau, Proc. Conf. High Temp. Mater. Power Eng. (1990) 913e934.
[49] H. Mughrabi, Proc. Eng. 2 (2010) 3e26. [106] F. Sansoz, B. Brethes, A. Pineau, Fatigue Fract. Eng. Mater. Struct. 25 (1)
[50] N.L. Phung, V. Favier, N. Ranc, F. Vale s, H. Mughrabi, Int. J. Fatigue 63 (2014) (2002) 41e53.
68e77. [107] N. Salajegheh, D.L. McDowell, Int. J. Fatigue 59 (2014) 188e199.
[51] T. Kanit, S. Forest, L. Galliet, V. Mounoury, D. Jeulin, Int. J. Solids Struct. 40 [108] G.M. Owolabi, R. Prasannavenkatesan, D.L. McDowell, Int. J. Fatigue 32
(13e14) (2003) 3647e3679. (2010) 1378e1388.
[52] D.L. McDowell, S. Ghosh, S.R. Kalidindi, JOM 63 (3) (2011) 45e51. [109] M. Shenoy, J. Zhang, D.L. McDowell, Fatigue Fract. Eng. Mater. Struct. 30 (10)
[53] K. Gall, M. Horstemeyer, D.L. McDowell, J. Fan, Mech. Mater. 32 (5) (2000) (2007) 889e904.
277e301. [110] W. Abuzaid, H. Sehitoglu, Lambros J. Mater. Sci. Eng. A 561 (2013) 507e519.
[54] K. Gall, M.F. Horstemeyer, B.W. Degner, D.L. McDowell, J. Fan, Int. J. Fract. 108 [111] W. Abuzaid, A. Oral, H. Sehitoglu, J. Lambros, H.J. Maier, Fatigue Fract. Eng.
(2001) 207e233. Mater. Struct. 36 (8) (2013) 809e826.
[55] D.L. McDowell, K. Gall, M.F. Horstemeyer, J. Fan, Eng. Fract. Mech. 70 (2003) [112] A. El Bartali, V. Aubin, S. Degallaix, Fatigue Fract. Eng. Mater. Struct. 31 (2008)
49e80. 137e151.
[56] R.G. Tryon, Cruse TA collection of technical papers e AIAA/ASME/ASCE/AHS/ [113] G.M. Castelluccio, W.D. Musinski, D.L. McDowell, Curr. Opin. Solid State
ASC structures, Struct. Dyn. Mater. Conf. 4 (1996) 2159e2166. Mater. Sci. 18 (4) (2014) 180e187.
[57] R.G. Tryon, T.A. Cruse, ASME J. Eng. Mater. Technol. 119 (1) (1997) 65e70. [114] C.J. Szczepanski, S.K. Jha, P.A. Shade, R. Wheeler, J.M. Larsen, Int. J. Fatigue 57
[58] R.G. Tryon, T.A. Cruse, ASME Aeros. Division Publ. AD 55 (1997) 467e474. (2013) 131e139.
[59] G. Castelluccio, D.L. McDowell, Int. J. Fract. 176 (2012) 49e64. [115] T. Straub, E.K. Baumert, C. Eberl, O.N. Pierron, Thin Solid Films 526 (2012)
[60] C. Przybyla, W. Musinski, G.M. Castelluccio, D.L. McDowell, Int. J. Fatigue 57 176e182.
(2013) 9e27. [116] W.L. Shan, Y. Yang, K.T. Hillie, W.A. Jordaan, W.O. Soboyejo, Mater. Sci. Eng. A
506 A. Pineau et al. / Acta Materialia 107 (2016) 484e507

561 (2013) 434e440. life of DA718 for aircraft engine disks in Superalloys 718, 625, 706 and De-
[117] H. Proudhon, A. Moffat, I. Sinclair, J.-Y. Buffiere, Comptes Rendus Phys. 13 rivatives 2005, TMS (The Minerals, Metals & Materials Society), 2005, pp.
(2012) 316e327. 97e110.
[118] M. Herbig, A. King, P. Reischig, H. Proudhon, E.M. Lauridsen, J. Marrow, J.- [157] F. Gallerneau, J.L. Chaboche, Int. J. Damage Mech. 8 (4) (1999) 404e427.
Y. Buffiere, W. Ludwig, Acta Mater. 59 (2) (2011) 590e601. [158] V. Bonnand, J.-L. Chaboche, P. Gomez, P. Kanoute
, D. Pacou, Int. J. Fatigue 33
[119] L. Liu, N.S. Husseini, C.J. Torbet, W.-K. Lee, R. Clarke, J.W. Jones, T.M. Pollock, (2011) 1006e1016.
Acta Mater. 59 (13) (2011) 5103e5115. [159] E.P. Busso, F.A. McClintock, Thermal fatigue degradation of a NiCoCrAlY
[120] M. Kübbeler, I. Roth, U. Krupp, C.-P. Fritzen, H.-J. Christ, Eng. Fract. Mech. 78 overlay coating, Mater. Sci. Eng. A 161 (1993) 165e179.
(2011) 462e468. [160] M.P. Taylor, H.E. Evans, E.P. Busso, E.P. Qian, Creep properties of Pt-aluminide
[121] P.-M. Hilgendorff, A. Grigorescu, M. Zimmermann, C.-P. Fritzen, H.-J. Christ, coatings, Acta Mater. 54 (2006) 3241.
Mater. Sci. Eng. A 575 (2013) 169e176. [161] E.P. Busso, J. Lin, S. Sakurai, M. Nakayama, A mechanistic study of oxidation-
[122] H.-J. Christ, C.-P. Fritzen, P. Ko €ster, Curr. Opin. Solid State Mater. Sci. 18 induced degradation in a plasma-sprayed thermal barrier coating system.
(2014) 205e211. Part I: model formulation, Acta Mater. 49 (9) (2001) 1515e1528.
[123] P. Ko€ster, H. Knobbe, C.-P. Fritzen, H.-J. Christ, U. Krupp, Tech. Mech. 30 [162] E.P. Busso, J. Lin, S. Sakurai, A mechanistic study of oxidation-induced
(2010) 185e194. degradation in a plasma-sprayed thermal barrier coating system. Part II:
[124] G.M. Castelluccio, D.L. McDowell, Int. J. Damage Mech. 23 (6) (2014) life prediction model, Acta Mater. 49 (9) (2001) 1529e1536.
791e818. [163] D.S. Balint, J.W. Hutchinson, An analytical model of rumpling in thermal
[125] Y. Murakami, A. Series, Key Eng. Mater. 51e52 (1991) 37e42. barrier coatings, J. Mech. Phys. Solids 53 (4) (2005) 949e973.
[126] Y. Murakami, Metal fatigue: effects of small defects and nonmetallic in- [164] E.P. Busso, Z.Q. Qian, A mechanistic study of microcracking in transversely
clusions, first ed., Elsevier, Oxford and Boston, 2002, ISBN 0-08-044064-9. isotropic ceramic-metal systems, Acta Mater. 52 (2006) 325e338.
[127] P.D. Hobson, M.W. Brown, E.R. de los Rios, in: K.J. Miller, E.R. de los Rios [165] E.P. Busso, L. Wright, L.N. McCartney, S.R. Saunders, S. Osgerby, J. Nunn,
(Eds.), Behavior of Short Fatigue Cracks, Mechanical Engineering Publica- A physics-based life prediction methodology for thermal barrier coating
tions, London, 1986, pp. 441e459. systems, Acta Mater. 55 (2007) 1491e1503.
[128] G.M. Castelluccio, D.L. McDowell, Mater. Sci. Eng. A 598 (26) (2014) 34e55. [166] J.-R. Vaunois, J.-M. Dorvaux, P. Kanoute
, J.-L. Chaboche, A new version of a
[129] T. Zhai, A.J. Wilkinson, J.W. Martin, Acta Mater. 48 (2000) 4917e4927. rumpling predictive model in thermal barrier coatings, Eur. J. Mech. A Solids
[130] T. Zhai, X.P. Jiang, J.X. Li, H.D. Garratt, G.H. Bray, Int. J. Fatigue 27 (2005) 42 (2013) 402e421.
1202e1209. [167] C. Courcier, V. Maurel, L. Re
my, S. Quilici, I. Rouzou, A. Phelippeau, Interfacial
[131] A. Pineau, Crossing grain boundaries in metals by slip bands, cleavage and damage based life model for EB-PVD thermal barrier coating, Surf. Coat.
fatigue cracks, personal communication, November 2014. Technol. 205 (13e14) (2011) 3763e3773.
[132] J. Miao, T.M. Pollock, J.W. Jones, Acta Mater. 60 (607) (2012) 2840e2854. [168] M. Caliez, J.L. Chaboche, F. Feyel, S. Kruch, Numerical simulation of EBPVD
[133] S.D. Antolovich, E. Rosa, A. Pineau, Low cycle fatigue of Rene
77 at elevated thermal barrier coatings spallation, Acta Mater. 51 (4) (2003) 1133e1141.
temperatures, Mater. Sci. Eng. 47 (1981) 47e57. [169] P.Y. Thery, M. Poulain, M. Dupeux, M. Braccini, Adhesion energy of a YPSZEB-
[134] S.D. Antolovich, S. Liu, R. Baur, Low cycle fatigue behavior of Rene
80 at PVD layer in two thermal barrier coating systems, Surf. Coat. Technol. 202
elevated temperatures, Met. Trans. 12A (1981) 473e481. (4e7) (2007) 648e652.
[135] S.D. Antolovich, P. Domas, J.L. Strudel, Low cycle fatigue of Rene
80 as [170] P.Y. Thery, M. Poulain, M. Dupeux, M. Braccini, Spallation of two thermal
affected by prior exposure, Met. Trans. 10A (1979) 1859e1868. barrier coating systems: experimental study of adhesion and energetic
[136] L.G. Zhao, N.P. O’Dowd, E.P. Busso, A coupled approach to the study of high approach to lifetime during cyclic oxidation, J. Mater. Sci. 44 (2009)
temperature crack initiation in single crystal Ni-base superalloys, J. Mech. 1726e1733.
Phys. Solids 54 (2006) 288e309. [171] V.K. Tolpygo, D.R. Clarke, Surface rumpling of a (Ni, Pt)Al bond coat induced
[137] S. Dumoulin, E.P. Busso, N.P. O’Dowd, D. Allen, A multiscale approach for by cyclic oxidation, Acta Mater. 48 (2000) 3283e3293.
coupled phenomena in multiphase FCC materials at high temperatures, Phil. [172] V. Maurel, R. Soulignac, L. Helfen, T.F. Morgeneyer, A. Koster, L. Re
my, Three-
Mag. 83 (31e34) (2003) 3895e3916. dimensional damage evolution measurement in EBPVD TBC using synchro-
[138] J.M. Martinez-Esnaola, A. Martin-Meizoso, E.E. Affeldt, A. Bennett, tron laminography, Oxid. Met. 79 (3e4) (2013) 313e323.
M. Fuentes, Fatigue Fract. Eng. Mater. Struct. 20 (1997) 771. [173] J. Ro€sler, M. Ba€ker, K. Aufzug, A parametric study of the stress state of
[139] E. Andrieu, A. Pineau, J. Phys. Paris IV (1999), 9, 3. thermal barrier coatings: Part I: creep relaxation, Acta Mater. 52 (16) (2004)
[140] M. Gobel, A. Rahmel, M. Schutze, The isothermal-oxidation behaviour of 4809e4817.
several nickel-base singlecrystal superalloys with and without coatings, [174] E.P. Busso, Z.Q. Qian, M.P. Taylor, H.E. Evans, The influence of bondcoat and
Oxid. Met. 39 (1993) 231e261. topcoat mechanical properties on stress development in thermal barrier
[141] K. Bouhanek, D. Oquab, B. Pieraggi, Mater. Sci. Forum 251e254 (1997) 33. coating systems, Acta Mater. 57 (2009) 2349e2361.
[142] M. Gross, J. Goitia Bigorda, V. Koralik, W. Engel, Oxidation of the turbine [175] V. Maurel, E.P. Busso, J. Frachon, J. Besson, F. N’Guyen, A methodology to
blade material CMSX4 studied by X-ray diffraction, J. Phys. Paris IV (1998), 8, model the complex morphology of rough interfaces, Int. J. Solids Struct. 51
505. (2014) 3293e3490.
[143] M.H. Li, X.F. Sun, J.G. Li, Z.Y. Zhang, T. Jin, H.R. Guan, Oxidation behaviour of a [176] K.S. Cheong, E.P. Busso, Discrete dislocation density modelling of single
single-crystal Ni-base superalloy in air. I: at 800 and 900  C, Oxid. Met. 59 crystal FCC polycrystals, Acta Mater. 52 (2004) 5665e5675.
(2003) 591e605. [177] E.P. Busso, F. McClintock, A dislocation mechanics-based crystallo-graphic
[144] B.A. Lerch, S.D. Antolovich, Fatigue crack propagation behaviour of a single model of a B2-type intermetallic alloy, Int. J. Plasticity. 12 (1996) 1e28.
crystal superalloy, Metall. Trans. A 21A (1990) 2169e2177. [178] D. Devincre, T. Hoc, L. Kubin, Dislocation mean free paths and strain hard-
[145] J. Reuchet, L. Remy, High temperature low cycle fatigue of MAR-M 509 su- ening of crystals, Science 320 (5884) (2008) 1745e1748.
peralloy II: the influence of oxidation at high temperatures, Mater. Sci. Eng. [179] U.F. Kocks, H. Mecking, Acta Mater. 44 (2002) 2599.
58 (1983) 33e42. [180] D. Nouailhas, G. Cailletaud, Comptes rendus de l’Acade
mie des sciences 315
[146] R.W. Neu, H. Sehitoglu, Thermomechanical fatigue, oxidation, and creep: (13) (1992) 1573e1579.
part I: damage mechanisms, Metall. Trans. A 20A (1989) 1755e1767. [181] E.P. Busso, G. Cailletaud, Int. J. Plast. 21 (2005) 2212e2231.
[147] D.W. MacLachlan, D.M. Knowles, Fatigue behaviour and lifting of two single [182] J.K. Tien, S.M. Copley, The effect of uniaxial stress on the periodic
crystal superalloys, Fatigue Fract. Eng. Mater. Struct. 24 (2001) 503e521. morphology of coherent gamma prime precipitates in nickel-base superalloy
[148] R. Ohtani, N. Tada, M. Shibata, S. Taniyama, High temperature fatigue of the crystals, Met. Trans. 2 (1971) 215e219.
nickel-base single-crystal superalloy CMSX-10, Fatigue Fract. Eng. Mater. [183] J.K. Tien, S.M. Copley, The effect of orientation and sense of applied uniaxial
Struct. 24 (2001) 867e876. stress on the morphology of coherent, gamma prime precipitates in stress
[149] J.A. Nesbitt, Solute transport during the cyclic oxidation of NieCreAl alloys, annealed nickel-base superalloy crystals, Met. Trans. 2 (1971) 543e553.
NASA Contractor Report 165544 (1982). [184] A. Pineau, Influence of uniaxial stress on the morphology of coherent pre-
[150] B.A. Lerch, S.D. Antolovich, N. Jayaraman, A study of fatigue damage mech- cipitates during coarsening-elastic energy considerations, Acta Met. 24
anisms in waspaloy from 25  Ce800  C, Mater. Sci. Eng. (1984) 151e165. (1976) 559e564.
[151] R.L. Amaro, Thermomechanical fatigue crack formation in a single crystal Ni- [185] C.F. Shih, Small-scale yielding analysis of mixed mode plane-strain crack
base superalloy, Ph.D. Dissertation, Georgia Institute of Technology, 2010. problems. ASTM STP 560, Am. Soc. Test. Mater. (1974) 187e210.
[152] F. Alexandre, S. Deyber, A. Pineau, Modelling the optimum grain size on the [186] S. Pierret, R. De Moura Pinho, A. Pineau, in: E. Ott, et al. (Eds.), Int. Symp. On
low cycle fatigue life of a Ni-based superalloy in the presence of two possible Superalloy 718 and Derivatives, September 28eOctober 1 2014, Pitttsburgh,
crack initiation sites, Scripta Mater. 50 (2004) 25e30. PA, TMS, 2014, pp. 537e551.
[153] J.Y. Guedou, G. Simon, J.M. Rongvaux, Development of damage tolerant Inco [187] D. Gustafsson, J.J. Moverare, S. Johansson, K. Simonsson, M. Ho €rnqvist,
718 for high temperature usage, in: E.A. Loria (Ed.), Proceedings Superalloys T. Mansson, S. Sjo € stro
€ m, Int. J. Fatigue 33 (2011) 1461e1469.
718, 625, 706 and Various Derivatives, 1994, pp. 509e522. [188] D. Gustafsson, J.J. Moverare, K. Simonsson, M. Ho €rnqvist, T. Mansson,
[154] T. Connolley, P.A.S. Reed, J.M. Starink, Short crack initiation and growth at €stro
S. Sjo €m, Proc. Eng. 10 (2011) 2821e2826.
600  C in notched specimens of Inconel 718, Mater. Sci. Eng. A 340 (1e2) [189] M. Ho €rnqvist, T. Mansson, D. Gustafsson, Proc. Eng. 10 (2011) 147e152.
(2003) 139e154. [190] L. Viskari, S. Johansson, K. Stiller, Mater. High Temp. 28 (2011) 336e341.
[155] F. Alexandre, Probabilistic and microstructural aspects of fatigue crack [191] L. Viskari, M. Ho €rnqvist, K.L. Moore, Y. Cao, K. Stiller, Acta Mater. 61 (2013)
initiation in In 718. PhD thesis, Ecole des Mines de Paris, 2004. 3630e3639.
[156] S. Deyber, F. Alexandre, J. Vaissaud, A. Pineau, in: E.A. Loria (Ed.), Probabilistic [192] J. Reuchet, L. Re
my, Metall. Trans. 14A (1983) 141e149.
 A. Pineau et al. / Acta Materialia 107 (2016) 484e507 507

[193] J.-C. Chassaigne, High temperature crack propagation in Ni-base superalloy STP 1186, ASTM, Philadelphia, PA, USA, 1994, pp. 135e149.
n18 fabricated by PM; study of the coupling of mechanical and environ- [205] R.L. Amaro, Thermomechanical Fatigue Crack formation in a single crystal
mental effects at the head of a crack, Doctoral dissertation, Ecole Nationale Ni-base superalloy, Ph.D. Dissertation, Georgia Institute of Technology, 2010.
Supe
rieure des Mines de Paris, 1997. [206] R.L. Amaro, S.D. Antolovich, R.W. Neu, A. Staroselsky, On thermo-mechanical
[194] A. Pineau, S.D. Antolovich, High temperature fatigue, in: C. Bathias, A. Pineau fatigue in single crystal Ni-base superalloys, Proc. Eng. 2 (1) (2010) 815e824.
(Eds.), Fatigue of Materials and Structures, ISTE/John Wiley, London, 2011, See also http://www.sciencedirect.com/science/article/pii/S18777058
pp. 1e130. 10000895.
[195] A. Pineau, S.D. Antolovich, High temperature fatigue: behaviour of three [207] S.D. Antolovich, R.L. Amaro, R.W. Neu, A. Staroselsky, On the development of
typical classes of structural materials, Mater. High Temp. 16 (2015) 298e317. physically based life prediction models in the thermo mechanical fatigue of
[196] S. Ponnelle, B. Brethes, A. Pineau, High temperature fatigue crack growth rate ni-base superalloysProc. 6th Int. Conf. on Matls. Struct. Micromech. of Fract.,
in Inconel 718: dwell effect annihilations, in: L. Re
my, J. Petit (Eds.), Tem- Brno, CZ, June 28e30, 2010, Key Eng. Mater. 465 (2011) 47e54.
peratureeFatigue Interaction, Elsevier Science Ltd, London, UK, 2002, pp. [208] R.L. Amaro, S.D. Antolovich, R.W. Neu, A. Staroselsky, in: Eric S. Hurdon,
257e266. Roger C. Reed, Mark C. Hardy, Michael J. Mills, Rick E. Montero, Pedro
[197] E. Andrieu, R. Molins, H. Ghonem, A. Pineau, Intergranular crack tip oxidation D. Portella, Jack Telesman (Eds.), Physics-Based Modeling of Thermo-
mechanisms in nickel-based superalloy, Mater. Sci. Eng. A154 (1992) 21e28. Mechanical Fatigue in PWA 1484, Superalloys 2012-12th International
[198] R. Molins, G. Hochstetter, J.C. Chassaigne, E. Andrieu, Oxidation effects on the Symposium on Superalloys, TMS, Warrendale, PA, USA, 2012, pp. 481e490.
fatigue crack growth behavior of alloy 718 at high temperature, Acta Mater. [209] R.L. Amaro, S.D. Antolovich, R.W. Neu, P. Fernandez-Zelaia, W. Hardin,
45 (1997) 663e674. Thermomechanical fatigue and bithermalethermomechanical fatigue of a
[199] S. Ponnelle, Propagation des fissures par fatigue  a haute temperature dans nickel-base single crystal superalloy, Int. J. Fatigue 42 (2012) 165e171.
l’Inconel 718: effets de microstructure et de chargements complexes. PhD [210] R.L. Amaro, S.D. Antolovich, R.W. Neu, Mechanism-based life model for out-
thesis, Ecole des Mines de Paris, Paris, France, 2001. of-phase thermomechanical fatigue in single crystal Ni-base superalloys,
[200] A. Andrieu, Z. Ghonem, A. Pineau, Two-stage crack tip oxidation mechanism Fatigue Fract. Eng. Mater. Struct. 35 (2012) 658e671.
in alloy 718, in: S. Mall, T. Nicholas (Eds.), Elevated Temperature Crack [211] Michael R. Hirsch, R.L. Amaro, S.D. Antolovich, R.W. Neu, Coupled thermo-
Growth (Book No. G00530-1990), ASME, New York, NY, 1991, pp. 25e29. mechanical high cycle fatigue in a single crystal Ni-base superalloy, Int. J.
[201] H. Ghonem, D. Zheng, Oxidation assisted fatigue crack growth behaviour in Fatigue 62 (2014) 53e61.
alloy 718 e part 1. Quantitative modeling, Fatigue Fract. Eng Mater. Struct. [212] R.L. Amaro, S.D. Antolovich, R.W. Neu, Benjamin S Adair, Michael R Hirsch,
14 (1991) 749e760. Patxi Fernandez-Zelaia, Matthew O’Rourke, Alexander Staroselsky, Towards
[202] H. Ghonem, D. Zheng, Oxidation assisted fatigue crack growth behaviour in the development of a physics-based thermo-mechanical fatigue life predic-
alloy 718 e part II. Applications, Fatigue Fract. Eng. Mater. Struct. 14 (1991) tion model for a single crystalline Ni-base superalloy, Mater. Perf. Charact. 3
761e768. (2014) 1e15.
[203] L.J. Carroll, C. Cabet, M.C. Carroll, R.N. Wright, The development of micro- [213] B. Boursier, Evaluation of damage mechanisms in the nickel base superalloy
structural damage during high temperature creepefatigue of a nickel alloy, Rene
80 under low cycle fatigue in the temperature range 75F-1400F, M.S.
Int. J. Fatigue 47 (2013) 115e125. Thesis, University of Cincinnati, 1981.
[204] M.P. Miller, D.L. McDowell, R.L.T. Oehmke, S.D. Antolovich, A life prediction [214] M. Ott, H. Mughrabi, Dependence of the high-temperature low-cycle fatigue
model for thermomechanical fatigue based on microcrack propagation, in: behaviour of the monocrystalline Nickel-base superalloys CMSX-4 and
H. Sehitoglu (Ed.), Thermomechanical Fatigue Behavior of Materials, ASTM CMSX-6 on the g/g0 morphology, Mater. Sci. Eng. A 272 (1999) 24e30.