Professional Documents
Culture Documents
(Theory) 2016. Modeling Concepts For Intermetallic Titanium Aluminides
(Theory) 2016. Modeling Concepts For Intermetallic Titanium Aluminides
a r t i c l e i n f o a b s t r a c t
⇑ Corresponding author.
E-mail address: mechanik@unileoben.ac.at (F.D. Fischer).
http://dx.doi.org/10.1016/j.pmatsci.2016.01.001
0079-6425/Ó 2016 Elsevier Ltd. All rights reserved.
56 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.1. Ti–Al alloys as structural materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.1.1. A short summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.1.2. Aspects of constitution and properties of c(TiAl)-based alloys as structural materials . 59
1.1.3. Aspects of microstructure and heat treatments of c(TiAl)-based alloys as structural
materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1.1.4. Applications of c(TiAl)-based alloys in jet engines and automotive engines . . . . . . . . . 60
1.1.5. Some further remarks concerning c(TiAl)-based alloys as structural materials . . . . . . . 61
1.2. Aspects of modeling of c(TiAl)-based alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2. Constitution and morphology of Ti–Al alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.1. Crystallographic data of major constituents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.2. Constitution and morphology of the lamellar microstructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.3. Constitution of the modulated microstructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3. Deformation mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.1. Deformation behaviour of multi-phase alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2. The deformation mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.1. The elastic deformation state. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.2. Dislocation core structures and glide in c(TiAl) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.3. Dislocation glide in a2 (Ti3Al) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2.4. Dislocation glide in b/B2-phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2.5. Mechanical order twinning in c(TiAl) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.2.6. Combination of dislocation glide and twinning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4. Modeling concepts of Ti–Al alloys. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1. Metalphysical modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1.1. The concept of thermal and athermal stresses applied to c(TiAl)-based alloys . . . . . . . 78
4.1.2. Dislocation mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.1.3. Jog dragging and work hardening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.1.4. Flow behaviour in reversed straining – the Bauschinger effect. . . . . . . . . . . . . . . . . . . . 95
4.1.5. Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.2. Continuum-mechanical modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.2.1. An extended constitutive plasticity law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.2.2. Modeling of stress/strain curves of PST crystals and polycrystals. . . . . . . . . . . . . . . . . 102
4.2.3. Structural stability and conversion of the lamellar microstructure. . . . . . . . . . . . . . . . 106
4.3. Thermodynamical modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.3.1. General aspects for modeling the developing microstructure . . . . . . . . . . . . . . . . . . . . 109
4.3.2. Modeling of the aða2 Þ ! c transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.3.3. Modeling of the massive a ! cm transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.3.4. Modeling the b ! a transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.3.5. Modeling of precipitation in Ti1x Alx N . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.3.6. Modeling of excess vacancy annihilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.4. Combined continuum-mechanical and thermodynamical modeling . . . . . . . . . . . . . . . . . . . . . 116
4.4.1. General aspects of phase transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.4.2. Modeling the formation of deformation twins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5. Some final comments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 57
1. Introduction
Fig. 1. Variation of (a) specific yield strength and (b) specific elastic modulus with temperature of selected structural materials
in comparison with c(TiAl)-based alloys. IN 625 and IN 718 represent Ni-base superalloys, whereas IMI 834 is a so-called near-a
Ti-based alloy. Reproduced from [4] by permission of Wiley-VCH Verlag GmbH & CoKGaA.
58 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
temperatures. The variation of the specific yield strength and the specific elastic modulus with tem-
perature for various c(TiAl)-based alloys in comparison with Ti- and Ni-based alloys is shown in Fig. 1.
Particularly at temperatures between 600 °C and 800 °C c(TiAl)-based alloys are superior to Ti-based
alloys in terms of their specific strength. Compared to the heavier Ni-based alloys, their specific
strength is at least similar. In the last two decades, high-strength c(TiAl)-based alloys, so-called 3rd
generation alloys (see Section 1.1.2) have been developed, which exhibit considerably increased speci-
fic yield strength levels. These alloys are characterized by a high content of b-stabilizing alloying ele-
ments, such as Nb and Mo [3,4]. At room temperature strength levels >1000 MPa can be achieved by
appropriate thermo-mechanical processing. It is also important to note that the high temperature
properties, such as creep resistance, were considerably improved, thus extending the application
range of c(TiAl)-based alloys. In 1999 the first commercial application of c(TiAl)-based alloys was
announced using a 2nd generation alloy (see Section 1.1.2). Mitsubishi has implemented cast Ti–Al
turbocharger wheels in one of their sports cars [5]. Shortly after that, in 2002, the serial production
of wrought processed high-performance c(TiAl) valves for racing car application has started using a
2nd generation c(TiAl)-based alloy [4]. Only few years ago, an US aero engine manufacturer
announced the initiation of investment – cast 2nd generation alloy Ti–Al blades in the low-pressure
turbine. Certification and flight tests with Ti–Al equipped engines have been conducted successfully
and in the meantime the regular service for passenger transportation and carriage of freight has begun
already [6]. In September 2014, the Airbus 320neo (here, neo stands for new engine option) completed
its maiden flight whereby for the first time the jet engines were equipped with forged Ti–Al turbine
blades of a 3rd generation alloy.
Fig. 2. Mid-section of the binary Ti–Al phase diagram and representative microstructures obtained by means of heat treatments
within the a and (a þ cÞ phase field conducted on a Ti–46.5Al–4(Cr, Nb, Ta, B) alloy (in at.%). Note that the left half of the
microstructural image represents a light-optical microscope image, whereas the right half is a scanning electron microscope
image taken in back-scattered electron mode, i.e. c(TiAl) appears dark, whereas a2 (Ti3Al) shows a light contrast. Heat
treatments: little above the eutectoid temperature (Teu) ? near gamma (NG) microstructure; between Teu and a-transus
temperature Ta ! duplex (D) microstructure. The volume fraction of lamellar grains depends on the heat treatment
temperature relative to Teu and Ta; just below Ta ! nearly lamellar (NL) microstructure. A NL microstructure exhibits a defined
volume fraction of globular c-grains; above Ta ! fully lamellar (FL) microstructure. Reproduced from [4] by permission of
Wiley-VCH Verlag GmbH & CoKGaA.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 59
zero volume fraction of b-phase allowing a controlled adjustment of the mechanical properties [4].
Both TNM and TNB alloys exhibit improved strength properties and oxidation resistance when com-
pared with conventional 2nd generation c(TiAl)-based alloys [8]. For example, room temperature ten-
sile strengths in the range of 800–1100 MPa were reported, along with plastic tensile strains >2%. The
high contents of Nb and Mo impedes diffusion processes in these alloys, thus decreasing the climb rate
of dislocations. Whereas this behaviour is of advantage with regard to creep and thermal stability, the
kinetics of phase transformations and recrystallization processes are slowed down. As a consequence
the parameters of heat treatments and hot working operations must be carefully adapted. Improve-
ments in creep strength have also been achieved by C additions in the range of 0.2–0.4 at.%, where
C forms dispersed Ti3AlC precipitates of the perovskite type, e.g. see [3,4,9].
1.1.3. Aspects of microstructure and heat treatments of c(TiAl)-based alloys as structural materials
Four different important types of microstructures obtained in engineering c(TiAl)-based alloys are
summarized in Fig. 2. The shown examples were adjusted in a Ti–46.5Al–4(Cr, Nb, Ta, B) alloy (in at.%)
by means of heat treatments conducted on a fine grained starting material, see [4] and references
therein. Generally, the influence of microstructure on mechanical properties of c(TiAl)-based alloys
can be summarized as follows: coarse-grained fully lamellar and nearly lamellar microstructures exhi-
bit relatively high fracture toughness and creep resistance, but poor tensile ductility and strength
especially at room temperature [1–4]. For details concerning the constitution and morphology of
the lamellar microstructure see Section 2.2. Comparably fine-grained equiaxed near gamma and duplex
microstructures with only small amounts of lamellar colonies show low fracture toughness and creep
resistance but moderate tensile ductility and strength at room temperature and elevated tempera-
tures. This inverse correlation between tensile properties and resistance to fracture (fracture tough-
ness) requires a careful selection of the microstructure which makes microstructural optimization
important for achieving balanced engineering properties. For engineering TNM-type alloys the opti-
mum balance between fracture toughness and creep resistance on one side and room temperature
tensile ductility/strength on the other side is expected for microstructures composed of relatively
small lamellar colonies (50–100 lm in diameter) exhibiting narrow lamellar spacing and a small vol-
ume fraction of globular c and b grains [4].
announced. Mitsubishi has implemented TiAl turbocharger wheels in their Lancer 6 sports car [5]. The
turbine wheels were produced by means of the LEVICAST process, a modification of the lost wax pre-
cision casting method, as well as by centrifugal casting.
The next generation of aircraft engines is targeted to exhibit higher efficiency which leads to
reduced kerosene consumption, significant lower emission of CO2 and NOx as well as to a noticeable
reduction of engine noise. Until 2020, according to the guidelines set by the Advisory Council for Aero-
nautics Research in Europe (ACARE), emission of CO2 and NOx and engine noise should be reduced by
50%, 80% and 50%, respectively, taking the current status of technology as reference value, see www.
acare4europe.com. Such ambitious targets can only be realized by applying improved engine design
concepts in combination with the use of advanced materials. There are at least three major payoff areas
for c(TiAl)-based alloys in advanced aero engines and gas turbines [4,6,10]: (a) c(TiAl) has a specific
elastic stiffness 50% greater than structural materials commonly used in aircraft engines. The higher
specific stiffness ðE=qÞ also shifts acoustically excited vibrations towards higher frequencies, which
is usually beneficial for structural components, e.g., compressor and turbine blades and parts within
the exhaust nozzle area. (b) The good creep resistance of advanced c(TiAl)-based alloys in the temper-
ature regime of 600–800 °C enables the substitution of Ni-based alloys (twice the density as c(TiAl)-
based alloys) for certain applications. (c) The high fire resistance of c(TiAl)-based alloys (nearly as
resistant as Ni-based alloys) can enable the substitution of heavy and expensive fire-resistant designed
Ti-based alloys in some components. Only few years ago, General Electric announced the initiation of
investment cast blades of a 2nd generation c(TiAl)-based alloy in the low-pressure turbine (LPT) of their
GEnx engine because this section of the engine offers the highest weight reduction potential [6]. Since
then certification and flight tests with c(TiAl)-based alloy equipped engines have been conducted suc-
cessfully and in the meantime the regular service for carriage of freight and passenger transport has
begun already. Meanwhile also other major aero engine manufacturers have announced in press
releases and their homepages the future use of c(TiAl)-based alloy in their next jet engine generation.
Depending on engine type 100–150 LPT blades per stage are required. Advanced engine concepts, such
as geared turbofans (GTF), as reported in [10,11], however, involve higher mechanical loading of the
LPT blades, thus demanding the use of high-strength advanced c(TiAl)-based alloys. Start of regular
air traffic is expected for end of 2015 or beginning of 2016. The aircrafts will be equipped with two
GTF engines which contain forged TNM turbine blades.
Table 1
Crystallographic data of stable and metastable constituents present in modern Ti–Al-alloys.
using advanced metallurgical methods – a factor, which is decisive for these specific materials to be
economically competitive with other state-of-the-art materials. In current c(TiAl)-based alloys bal-
anced properties like room temperature ductility, fracture toughness, high-temperature strength,
creep and oxidation resistance can be achieved.
Although the experimental elucidation of the materials behaviour, aspects of microstructure–prop-
erty relationships as well as manufacture and process technology have been accompanied by model-
ing activities. However, no comprehensive review paper on the modeling concepts for intermetallic
titanium aluminides is available. The present review should close this gap and will show which mod-
eling concepts have been employed and how fundamentally modeling has contributed to the under-
standing and development of this innovative class of structural materials.
Since modeling can be considered as generally accepted tool to understand the development of the
microstructure as well as its deformation and stress state, modeling is applied, too, in a constantly
increasing amount to Ti–Al structures and components. However, compared, e.g., to steels, the hetero-
geneous microstructure of Ti–Al alloys, consisting of ordered constituents, makes modeling to a signif-
icantly more demanding task. We consider it, therefore, as necessary and first step to report on
modeling to explore the microstructure and the observed deformation mechanisms. The Ti–Al litera-
ture abounds with experimental results about the microstructure, deformation and fracture. A full dis-
cussion of this data is not possible within the scope of the present article. Instead, special attention
will be given to areas where relatively recent work has in some way changed the perspectives. The
accent is principally on two-phase or multi-phase alloys based on c(TiAl), which are particularly
demanding for modeling studies. The investigated alloys are binary alloys as well as 2nd and 3rd gen-
eration alloys (see also Tables 2 and 4 below). Separate sections will address the deformation issues in
these alloys in order to provide a comprehensive context for the modeling case studies described
below. In order to do this in a concise manner, it is necessary to omit many details, both, with respect
to experiments and modeling, but these may be found in the cited papers and reviews.
The next two chapters are devoted to these topics and basic state-of-the-art-knowledge on the Ti–
Al alloys microstructure and the deformation mechanisms activated there by an external stress state.
There is good consensus in the Ti–Al literature that the mechanical properties of c(TiAl)-based
alloys are largely governed by the constitution and microstructure. Modern alloys, which usually
are 3rd generation alloys, are multi-phase assemblies involving the phases c(TiAl), a2 (Ti3Al), b/B2,
and an orthorhombic phase (oP4) with B19 structure, for a reviews see [3,12–15]. The crystallographic
data of these constituents are listed in Table 1. The volume fractions of these phases are in principle
determined by the phase diagrams available in the literature. However, the equilibrium constitution
often cannot be achieved within the processing constraints. Thus, several metastable phases such as b/
B2 or B19 are only kinetically stabilized. It should be noted, however, that this complex constitution
may lead to peculiar microstructures, a fact that will be considered in Section 2.3.
A particular microstructural form are lamellar colonies (e.g. Fig. 2) that, in the simplest case, consist
of c(TiAl) and a2 (Ti3Al) layers. It is well documented in the literature [20] that such a multilayer sys-
tem could exhibit extraordinary mechanical properties when the layer thickness is small enough. This
aspect provides a significant potential for the alloy design. Indeed, the alloy performance concerning
strength, ductility and toughness can be tailored by the volume content of lamellar colonies. The
deformation and fracture behaviour of the lamellar morphology has attracted extensive research;
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 63
Fig. 3. Fine lamellar morphology in a modern high-Nb containing alloy with composition Ti–45Al–7.5Nb–0.2C (at.%) (TNB-V2);
extruded, nearly lamellar microstructure. (a) Low magnification, high-resolution micrograph imaged along h1 1 0ic. The insert
shows the fast-Fourier transform (FFT) of the micrograph. Note the fine lamellar spacing that is also manifested in the streaking
of the reflexes in the FFT. (b) Higher magnification of a part of the figure.
for a review see [3]. The collected data has shown that the relevant micromechanisms are governed by
various structural details, which will be briefly reviewed in order to lay the necessary background and
to identify the challenges for realistic modeling.
The lamellar morphology results from the precipitation of c-lamellae in either a disordered a or a
congruently ordered a02 matrix, following one of the transformation paths [13] as
a ! a02 ! a2 þ c or a ! a2 þ c: ð1Þ
Within the lamellar morphology the two phases are crystallographically aligned so that close-
packed planes and directions match, as described by the Blackburn orientation relationship [21] as
0 k1 1 2
f1 1 1gc kð0 0 0 1Þa2 and h1 1 0 : ð2Þ
c a2
the parent a=a2 grain. There are four types of lamellar interfaces: the a2 =c interface and three distinct
c=c interfaces that are specified by rotations of 60°, 120° and 180° around h1 1 1ic or ½0 0 0 1a2 . The
individual c-lamellae are further subdivided by order domain boundaries; however, in contrast to
lamellar interfaces, the habit plane of the domain boundaries is not usually {1 1 1}. A particular form
of the lamellar morphology is the polysynthetically twinned (PST) crystal, which is comprised of a sin-
gle set of lamellae [22], i.e., it does not contain colony boundaries.
The theoretical estimates of interfacial energies indicate that the lamellar 180° true twin boundary
has by far the lowest energy followed by the 120° c=c, the 60° c=c and the a2 =c boundaries; for
reviews see [3,23]. This finding is in good agreement with the experimentally observed relative occur-
rence of the different c=c interface types. It should be noted that the interface energies could be min-
imized by a rigid-body translation of the adjacent lattices relatively to each other along certain fault
vectors. There are also other factors that may favour the channelling of the deformation between the
lamellae. The bonding of some lamellar interfaces is such that a fault translation reduces their inter-
facial energy [23–25]. This means that the dislocations may dissociate at the interfaces into widely
separated partials, which reduces the glide resistance during soft mode deformation. The reduction
of the complex stacking fault (CSF) energy is such that the stress to move an ordinary dislocation at
the interface could be about one order of magnitude smaller than in the bulk [24,25]. On the other
hand, screw dislocations gliding in the hard mode on {1 1 1} planes inclined to the lamellar interfaces
could reduce their energy by cross-gliding onto the interface. These dislocations become immobilized
and thus harden the hard deformation mode.
The relative volume fraction of a2 - and c-lamellae does not only depend on the alloy composition
according to the phase diagram but also on the cooling rate after solidification. The decomposition of
the parent a/a2 -phase into lamellar (a2 þ cÞ is sluggish and often cannot be established within the
constraints of standard processing routes. Thus, the volume fraction of c-phase is less than the equi-
librium value, which provides a significant chemical driving force for structural changes.
Among the various interfaces occurring in the lamellar morphology only the 180° true twin bound-
ary is fully coherent because the adjacent lattices are symmetrically oriented. In all the other cases the
matching is imperfect. The misfit parameter De defined along a certain crystallographic direction is
De ¼ ða2 a1 Þ=a1 . The quantities a1 and a2 are the lattice constants of the two crystals parallel to
the interface. The misfit parameter is typically De = 0.01–0.02, slightly dependent on alloy composi-
tion. The mismatch strains at the 60° and 120° c/c interfaces are pure shear strains and solely caused
by the tetragonality of the c-phase. It might be expected that the misfit situation changes with tem-
perature because the thermal expansion coefficients of the c-phase in the a and c directions differ sig-
nificantly. Analytical modeling by Hazzledine [26] has shown that the situation at the a2 =c interface is
more complex. There is a biaxial tension in c and a biaxial compression in a2 because the atomic spac-
ings in a2 are larger than in c. Furthermore, the c-phase does not possess a triad axis along h1 1 1i,
while the a2 -phase has hexagonal symmetry along [0 0 0 1]. Hence, the c-phase has to be sheared to
three-fold symmetry to match the six-fold symmetry of the a2 -phase. This accommodation produces
shear strains of opposite signs in the adjacent a2 - and c-lamellae. Because of the differences in the
elastic constants the mismatch strains are unequally shared between the two phases.
Up to a certain point, the mismatch strains could be solely taken up by elastic distortion, i.e., the
lamellae are uniformly strained to bring the atomic spacings into registry. However, this homoge-
neous straining raises the total energy of the system and introduces high coherency shear stresses
of the order sc;max ¼ lDe, with l being the shear modulus. Thus, elastic misfit accommodation is only
possible when the lamellae are very thin. Hazzledine [26] has determined by analytical modeling the
critical thickness kLc of the lamellae above of which the coherency is lost. The predicted values are
kLc 6 8 nm for the mismatched c=c interfaces and kLc 6 0:8–3.9 nm for the a2 =c interfaces, depending
on the volume content of a2 -phase. Investigation by high resolution electron microscopy has essen-
tially confirmed these values [3].
At thicker lamellae, the mismatch is partly taken up by disconnections, meaning defects that exhi-
bit both dislocation and step-like character [3,27,28]. Propagation of a disconnection along the inter-
face structurally accomplishes the phase transformation. In a sense, a disconnection concentrates the
strain mismatch, and thus often acts as nucleation centre for stacking faults or mechanical twins. In
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 65
spite of this misfit accommodation significant coherency strains remain at the semicoherent interfaces
[29–31]. The resulting coherency stresses are comparable with the shear stress level which has to be
externally applied during yielding of the material and significantly affect the deformation behaviour.
The most important structural parameters of the lamellar morphology are:
In spite of these uncertainties, the data illustrates that already the specification of a polycrystalline
fully lamellar alloy involves a wide length scale of structural parameters, which makes modeling very
difficult. The problems are exacerbated in duplex alloys, which contain lamellar and equiaxed con-
stituents. Thus, for micromechanical modeling the PST crystal is often considered. This material sys-
tem has the advantage that the spurious influence of the colony size on the yield stress can be
eliminated. For technical applications, microstructures are preferred that are comprised of a mixture
of lamellar colonies and equiaxed grains. These are specified with decreasing content of lamellar colo-
nies as fully lamellar, nearly lamellar, duplex, or equiaxed [3].
The lamellar morphology exhibits a marked plastic anisotropy concerning the orientation angle U
of the lamellae with respect to the sample axis, which was early recognized on PST crystals by Fuji-
wara et al. [32]. Both the yield and fracture stresses are low when deformation occurs in the plane
of the lamellae (soft mode) and are high for a deformation across the lamellae (hard mode). The duc-
tility is high when the tensile axis lies close to the lamellar plane and is low when the axis is nearly
normal to the lamellar plane [32–34]. In broad terms, the anisotropy of the yield stress can be rational-
ized on the basis of factors that limit the slip path of the dislocations, i.e., by classical Hall–Petch argu-
ments. During soft mode deformation (U 45°) dislocation glide is relatively easy because it is only
impeded by the widely separated domain boundaries. If U is close to 0° or 90°, deformation occurs
along the maximum shear stress, i.e., on the slip planes and slip directions oriented closest to the
45° direction from the deformation axis. Thus, shear must be traversed through the closely spaced
lamellar boundaries. The barrier strength of the lamellar boundaries to dislocation glide originates
from different factors, which in detail depend on the interface type and character of the incoming
dislocations [35,36]. These are:
Table 2
The effect of structural parameters on the room temperature yield stress ry of c(TiAl)-based alloys. Analysis performed according
to the Hall–Petch model; kL (lamellar spacing), kD (domain size).
Overcoming this kind of glide barriers essentially provides an athermal stress, for definition see
later, Section 4.1.1. This concept has been used by Umakoshi et al. [33,37] to derive empirical relation-
ships by relating the flow stress of differently oriented PST crystals as the relevant structural param-
eter. In the framework of the Hall–Petch model the yield stress was described as
k
r ¼ r0 þ pyffiffiffiffi : ð3Þ
D
The structural parameter D determines the length of the dislocation slip path. The parameter ky is a
material constant that basically measures the difficulty with which slip penetrates from one grain or
lamella to the next. The stress contribution r0 is independent of D. In the soft 45° orientation the rel-
evant structural parameter is the separation distance kD of the domain boundaries, whereas in the
hard orientations, U ¼ 0 and U ¼ 90 , the dislocation slip path is essentially determined by the lamel-
lar spacing kL . The results performed by this analysis are listed in Table 2, together with those of poly-
crystalline lamellar alloys [38–40]. The variation of ky clearly reflects the different barrier strength of
the interfaces. Slip transfer through the widely separated domain boundaries is relatively easy
(U ¼ 45 ), compared with translamellar shear (U ¼ 0 and U ¼ 90 ). However, deformation in the hard
orientations seems to be markedly affected by the slip modes of a2 -lamellae. In the 0° orientation,
glide of the a2 -phase is accomplished by prismatic slip, which is relatively easy. On the contrary, in
the 90° orientation pyramidal slip has to be activated in the a2 -phase, which requires an extremely
high stress. Thus, although the a2 -lamellae may widely be separated in the PST crystals investigated,
the difference in the ky values observed for the two hard orientations probably reflects the different
slip modes with which the a2 -lamellae are intersected. In supporting this hypothesis, Umakoshi
et al. [33,37] have shown that the flow stress of PST crystals in the 90° orientation substantially
increases with increasing volume fraction of a2 -lamellae. This interpretation is also consistent with
the temperature dependence of the Hall–Petch parameters [41]. In the soft 45° orientation, the
Hall–Petch parameter ky is almost independent of the test temperature up to 900 °C, as it is expected
for athermal glide barriers. However, in the 90° orientation, ky goes through a maximum of
ky ¼ 1:4 MPa m1/2 at 500 °C and then decreases to a very low value of ky ¼ 0:1 MPa m1/2 at 900 °C.
The maximum of ky corresponds to the maximum in the temperature profile of the critical resolved
shear stress of a2 single crystals. These factors also explain why the variation of the yield stress is
not symmetrical as U decreases or increases from 45°. Taken together, there is good evidence that
Fig. 4. Modulated laths (T) sandwiched between c(TiAl) lamellae imaged by diffraction contrast; cast c-Md alloy.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 67
the a2 =c interface provides the highest barrier strength for incoming dislocations, when compared
with the c=c interfaces. This can be rationalized by the fact that overcoming the a2 =c interface is asso-
ciated with all the energy dissipating processes listed above. Owing to this situation, the incoming
shear is often elastically transferred through a2 -lamellae [3,42]. At thin a2 -lamellae, the stress state
could be high enough to active dislocation sources at the exit site of the lamella. This ‘‘elastically-med
iated” slip transfer preferentially occurs in front of incoming twins.
Recently a novel type of alloys with a modulated microstructure, designated as c-Md, has been dis-
covered in Nb-bearing titanium aluminide alloys [43–45]. The characteristic feature of the c-Md alloy
is the existence of ‘‘modulated” laths, which are comprised of B19-phase that is interspersed with
slabs of b/B2 and a2 -phase. In conventional TEM bright-field imaging the modulation is manifested
by strong fluctuating strain contrast (Fig. 4). High-resolution examination has revealed [43] that the
modulation occurs at the nm-scale, with diffusive (i.e., unsharp) boundaries between the individual
constituents. Thus, when compared with conventional (a2 þ cÞ alloys, the modulation provides an
additional structural feature that refines the material. It should be noted that the modulated structure
can be established in both as-cast and wrought processed alloys.
The B19 structure is closely related to the orthorhombic phase with the ideal stoichiometric com-
position Ti2AlNb with oC16 (Cmcm) structure. At first glance, it appears that the B19-lamellae are
formed by an additional ordering reaction of the a2 -phase, and there is good evidence from high-
resolution electron microscopy that this indeed occurs. However, the B19-phase could also be formed
by a decomposition reaction of the B2-phase. First principle calculations [46,47] have shown that the
B2-phase existing in Ti–Al alloys is unstable under tetragonal distortion, if the alloys contain supersat-
urated solutions of transition metals (Nb, Zr, V). This shear instability was attributed to the anomalous
(negative) tetragonal shear modulus. Specifically, the transformation of B2 into several orthorhombic
structures such as B19 and B33 is energetically favourable, according to [47] as
The B2 ? B19 transformation can be accomplished via h0 1 1i{0 1 1} shuffle operations and does not
require long-range diffusion. The modulation is probably another consequence of the above described
phase instability in Nb bearing and Al-lean Ti–Al alloys. The TEM evidence is that in the c-Md-alloy the
B19-phase is formed as a domain structure, regardless of whether the parent phase was B2 or a2 . For
example, the formation of the superstructure in the a2 -lamellae may start at different places giving
rise to different B19 domains separated by antiphase boundaries. If the parent phase is B2, then dif-
ferent B19 domains can be formed because all h0 1 1i{0 1 1} shuffle operations are crystallographically
equivalent.
The modulated laths are usually sandwiched between c-lamellae. The mismatch between the con-
stituents is partially taken up by disconnections, meaning interfacial defects that exhibit both
dislocation- and step-like character. The lattice mismatch is concentrated at the disconnections, thus,
the interfaces are often facetted with atomically flat terraces parallel to f1 1 1gc in between the ledges.
There is also good evidence that large coherency stresses are present at the interfaces; for details see
[44,45,48]. When compared with conventional duplex (a2 þ cÞ-alloys of the 3rd generation, the char-
acteristic features of modulated alloys are:
As will be described in Section 3.1, these attributes are beneficial for the performance of the c-Md-
alloys. It should be noted that the modulation was established without adding grain refining agents,
such as borides or silicides, which could be harmful for fracture toughness.
68 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
3. Deformation mechanisms
With respect to the deformation mechanisms one can consult the pioneering papers of Blackburn
[17], Shechtman et al. [49], Greenberg [50], Lipsitt et al. [51] and Yamaguchi and Umakoshi [22]. How-
ever, since remarkable research has been done during the last 25 years, we consider it as necessary to
give an overview on the current knowledge including the findings of the last two decades. Further-
more, the mechanics and physics behind the deformation mechanisms are presented with a certain
view on applying the knowledge to modeling of deformation processes, which may also be coupled
with thermodynamic processes. First of all we report some aspects on the deformation behaviour of
multi-phase alloys.
There is good consensus that deformation is mainly confined to the majority c(TiAl)-phase, regard-
less of whether the alloy has a lamellar or equiaxed microstructure [37,52–56]. Most investigations
Fig. 5. Deformation structure observed in the c-Md alloy after room temperature tensile deformation to stress r = 1100 MPa
and strain e = 2.5%. High-resolution micrograph demonstrating twin nucleation at an interface between a modulated lath (T)
and adjacent c phase. The figure below shows the arrowed region in higher magnification.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 69
have shown that in the c-phase of two-phase alloys, glide of ordinary dislocations are the primary
deformation mode, followed by mechanical twinning [37,53–58]. This twinning activity is a remark-
able difference to the deformation of single-phase c-alloys, which show a general reluctance to twin.
The observation suggests that particular twin nucleation mechanisms occur in multi-phase alloys, a
subject that will be considered in Sections 3.2.5 and 4.4.2.
In the c-phase of two-phase alloys superlattice slip is obviously difficult to activate. This imposes
serious implications on the deformability of the c-phase, as c component glide of the tetragonal unit
cell is retarded. In this respect, the enhanced activity of twinning of the c-phase is certainly important,
as it could provide c component shear and reduces the required number of dislocation slip systems.
However, in grains or lamellae that are unfavourably oriented for 1/2h1 1 0] glide or mechanical twin-
ning, significant constraint stresses can develop due to the shape change of deformed adjacent grains.
The constraint stresses can be high enough to activate glide of superdislocations or even fracture. To
date, there is no satisfactory explanation available for the difficulty of superlattice slip in the c-phase
of two-phase alloys.
As mentioned above, of the two constituents of (a2 þ cÞ-alloys, the a2 -phase is more difficult to
deform. a2 -grains or lamellae that appear virtually free of defects are often sandwiched between heav-
ily deformed c-phase. In most investigations, only localized prism glide of 1/3h1 1 2 0i superdisloca-
tions ahead of blocked twins was observed. There are mainly two factors that could be responsible
for the observed strain partitioning.
First, Ti–Al alloys contain a significant amount of interstitial hardening elements, like oxygen,
nitrogen and carbon, which preferentially partition to the a2 -phase [59]. Thus, the c-phase of two-
phase alloys is almost devoid of precipitates [60]. Another factor is certainly the strong plastic aniso-
tropy of Ti3Al, which will be addressed in Section 3.2.3. There are certain grain orientations for which
the a2 -phase is difficult to deform. For example, a2 -grains or lamellae that are loaded parallel to the
[0 0 0 1] direction cannot deform plastically without glide of hci-type dislocations on pyramidal planes.
These a2 grains are certainly constrained by the surrounding deformed grains. The resulting stresses
can easily exceed the fracture stress; this could be a reason for premature failure.
When compared with conventional duplex (c þ a2 Þ-alloys of the 3rd generation, the modulated
multi-phase alloys (Section 2.3) are certainly advantageous for deformability because a dispersion
of shear directions is provided in a relatively small volume element of the phase compound, which
in monolithic a2 - and c-lamellae is not available. This relaxes the need of independent slip systems,
which in terms of the von Mises condition are required for the general plasticity of polycrystalline
material, see also Section 3.2. Furthermore, deformation of these small volume elements seems to
be also supported by the mismatch structures and according coherency stresses. A particular feature
Fig. 6. Stress/strain curve of a room temperature tensile test performed on the c-Md alloy. The test involves incremental
unloadings and stress relaxations for measuring the internal stress and the stress rate sensitivity, respectively.
70 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
is the emission of superlattice intrinsic stacking faults (SISF) and embryonic twins at misfit disloca-
tions or interfacial steps. Thus, mechanical twinning is a prominent deformation mechanism in the
modulated alloy. These features are demonstrated in Fig. 5. Also, due to the modulation, the volume
content of a2 -lamellae is depleted. This seems to be beneficial for ductility because the a2 -phase is dif-
ficult to deform and prone to cleavage fracture along its basal plane. Taken together, these attributes of
modulated alloys lead to an outstanding balance of strength, tensile ductility and toughness, com-
bined with good temperature strength retention. A load–elongation trace of a room temperature ten-
sile test is shown in Fig. 6. The combination of tensile strength and ductility demonstrated in the test is
unique in the TiAl literature. More details about the mechanical properties are described in [3,43–45].
The Miller indices with mixed parentheses hu v w] and {h k l) were introduced by Hug et al. [72,73]
in order to differentiate the first two (equivalent) indices from the third one that is fixed in the L10
structure. The dislocation Burgers vectors are illustrated in Fig. 7. Contrary to the 1/2h1 1 0] and
Fig. 7. Potential dislocation slip systems in c(TiAl). (a) Atomic arrangement of the L10 structure of c(TiAl) with one of the four
octahedral {1 1 1} planes imaged opaque. (b) Illustration of the Burgers vectors bh1 1 0 = 1/2h1 1 0], bh1 1 2 = 1/2h1
1 2],
1, and bh1 0 1 ¼ h1 0 1
bh0 1 1 ¼ h0 1 situated on the (1 1 1) plane. Figure adapted from [3,55], changed.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 71
Fig. 8. Shear vectors corresponding to perfect and partial dislocations. The figure shows the three-layer ABC of atom stacking on
the (1 1 1) plane. The (1 1 1) slip plane is imaged opaque so that the atoms below the slip plane appear darker. The vectors b1, b2,
b3, and b4 represent shear displacements (partial dislocations), which are associated with distinct planar faults; see
accompanying text. Figure adapted from [3,55], changed.
1] dislocations, there is only one octahedral plane available for glide of an 1/2h1 1 2]
h0 1 dislocation. In
principle, shear may also occur along h1 0 0i, which is another translation vector in the L10 structure
but not situated on the close-packed octahedral planes, thus h1 0 0i shear is only seldom observed.
The potential cross-slip planes of the perfect dislocations are the close-packed or relatively close-
packed planes that intersect the primary (1 1 1) plane so that the line of intersection is parallel to
the Burgers vector. For example, the ordinary screw dislocation 1=2½1 1 0 gliding on the (1 1 1) plane
is capable to cross-slip onto (1 1 1), (1 1 0) and (0 0 1) planes. The ½0 1 1 superdislocation may cross-slip
from the primary (1 1 1) glide plane onto (1 1 1Þ,
(1 0 0) and (0 1 1) planes, respectively. A possible low-
index cross-slip plane for the bh1 1 2 = 1/2h1 1 2] superdislocation is {1 1 0). The driving forces for cross-
slip of the 1/2h1 1 0] dislocation were considered by Sun [74] and Jiao et al. [75] by determining the
dislocation line tension. The calculation, based on anisotropic elasticity, revealed that the line tension
is lowest for the {1 1 0) planes, intermediate for the {1 1 1} planes and highest for the (0 0 1) plane;
hence cross-slip is expected to occur on {1 1 0) planes. However, it should be noted that, apart from
these factors, the capability of a dislocation to cross-slip depends on its core structure and glide resis-
tance on the cross-slip plane. The ordinary dislocation has a compact core and is expected to cross-slip
easily. As will be shown below, the superdislocations are dissociated and thus have to constrict before
cross-gliding.
The shear displacements b1, b2, b3, and b4 shown in Fig. 8 produce planar faults, which are unique
to the L10 structure. A complex stacking fault (CSF) is produced, when the top layer of the {1 1 1}
planes and all planes above the top layer are shifted along b1 = 1/6h2 1 1] or b2 = 1/6h1 2
1]; equivalent
displacements are 1/6h1 4 5] or 1/6h4 5 1]. This displacement results in a local hexagonal stacking
sequence, i.e., the Al atoms in the top layer C lie directly above the Ti atoms in the bottom layer A.
The CSF destroys the chemical environment of the first neighbours for the atoms in the fault plane
and is thus expected to have a very high energy.
A displacement along b3 = 1/6h1 1 2] generates a superlattice intrinsic stacking fault (SISF) with an
ABCBCA stacking of {1 1 1} planes, which does not change the nearest-neighbour coordination; equiv-
alent displacements are along 2b1, 2b2, 2b3, and 1/6h1 5 4], respectively. A repetition of this shear
on consecutive (1 1 1) planes produces a true twin, a mechanism that will be considered in the next
section. The displacement along b3 generates a superlattice extrinsic stacking fault (SESF), which
can formally be described by an ABBCAB stacking. As with the SISF, the SESF does not change the
nearest-neighbour coordination.
72 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
The displacement b4 = 1/2bh0 1 1 generates an antiphase boundary (APB), which can also be pro-
duced by the displacements 3b1 or 3b2, i.e., by shear along 1/2h1 2 1]. The APB destroys the nearest-
neighbour coordination of the L10 superlattice and affects the directional bonding of the Ti and Al
atoms, which implies high formation energy. While the SISF and CSF are specific to the {1 1 1} planes,
an APB can be introduced on any crystallographic plane [76]. In c(TiAl) the most important APB planes
are {1 1 1} and {0 1 0), [83].
The surface energy C associated with the formation of planar faults is important for various disso-
ciation reactions of the dislocations. Thus, much effort has been spent in estimating the planar fault
energy by simulations, for reviews see [22,23,77,78] and weak-beam transmission electron micro-
scopy [79–82]. A detailed account on the planar faults in intermetallic alloys is given in reviews by
Paidar and Vitek [83] and Yoo and Fu [23]. A collection of recent TiAl data is provided in [3]. The cal-
culated fault energies are generally larger than those determined from the dissociation width of
superdislocations. According to a recent analysis the surface energies were ranked as [82] C(CSF)
> C(APB) C(SISF). Atomistic simulations suggest that the APB energy for the {0 1 0) plane is lower
than that of the {1 1 1} plane. Thus, cross-slip of the APB from the {1 1 1} planes onto the {0 1 0) planes
is thermodynamically justified.
Theoretical studies using the so-called ‘‘layered Korringa–Kohn–Rostoker (LKKR) method” have
been performed in order to predict the effects of alloy composition on the fault energies [19]. Such
information is difficult to gather experimentally, but important for understanding the activation of
shear mechanisms that involve planar faults, as twin nucleation. In binary alloys, containing 48–
51 at.% Al, the APB and SISF energies weakly increase with increasing Al content. At a given Al concen-
tration, Nb additions reduce the APB and SISF energies. Cr additions appear to have little influence on
the planar fault energies in Ti-rich alloys, while they slightly reduce the energy for the APB on {1 1 1}
planes in Al-rich alloys.
Due to this variety of planar faults, dissociation in c(TiAl) is complex and may result in planar and
non-planar core configurations, for a reviews see [3,72,80,84,85]. The ordinary dislocation may reduce
its energy by planar dissociation on its {1 1 1} glide plane into Shockley partials, according to 1/2
1 0] ? 1/6[2
[1 1 1] + CSF + 1/6[1
2 1].
The fully dissociated state of the [0 1 1] superdislocation involves
four partial dislocations, an SISF, an APB, and a CSF, according to
! 1=6½1 1 2
½0 1 1 þ SISF þ 1=6½1
2 1
þ APB þ 1=6½1 1 2
þ CSF þ 1=6½1
2 1:
ð5Þ
The large differences in the planar fault energies imply that Shockley partials bordering the SISF are
widely spread in the {1 1 1} plane, whereas the separation width of the APB- and CSF-coupled partials
is small. This reasoning was largely confirmed by weak-beam and high resolution TEM observations.
In particular, the core of the ordinary dislocation was found to be compact. The planar dissociations
described above result in core configurations, which in principle are glissile. However, the superdislo-
cations are also liable to adopt sessile non-planar core configurations that are immobile. Several mod-
ifications have been proposed, which differ in the core configuration that produces the pinning effect.
The common feature of all core modifications is that the screw component of the leading superpartial
constricts and locally cross-slips from the original (1 1 1) slip plane onto a conjugated octahedral plane
or a cube plane. The dislocation then re-dissociates on a parallel or oblique {1 1 1} plane.
Woodward and MacLaren [86] have suggested that the mechanism is thermodynamically driven
because the APB energy for the cube plane is lower than that of the octahedral plane. There is also
a moment of the torque force between the leading and trailing superpartials due to elastic anisotropy,
which pushes the superpartials off the primary (1 1 1) glide plane and thus promotes cross-slip [23].
The stresses associated with the torque are substantially larger than the shear stresses resulting from
the externally applied load [23,87]. This finding implies that components of the stress tensor that
exert no net resolved shear stress on the perfect dislocation can significantly affect the fine structure
of the dislocation and thus the flow behaviour [84].
When the APB is completely on the cube plane, the dissociated configuration is known as a
Kear–Wilsdorf lock [88]. As the partial dislocations involved have no common slip plane, the non-
planar cores are considered as immobile with respect to glide. This pinning of screw dislocations leads
to a strengthening of the material. Since cross-slip is thermally activated, the number of cross-slipped
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 73
segments increases with increasing temperature. The associated dislocation locking can give rise to an
anomalous increase of the yield strength with temperature and strong work hardening and eventually
makes the strength properties dependent on the deformation pathway.
0i
1/3h1 1 2 0}, prismatic slip;
{1 0 1
0i
1/3h1 1 2 (0 0 0 1), basal slip;
21
1/3h1 6i 1 1}, slip on pyramidal planes of the second order or p2 planes;
{1 2
6i
1/3h1 2 1 0 1}, slip on pyramidal planes of the first order or p1 planes.
{2 2
Investigations performed on Ti3Al single crystals have shown that the activation of prismatic slip is
easiest, followed by basal and pyramidal slip; the ratio of the critical resolved shear stresses (CRSS) scr
are approximately [91–93]
dislocation core into three {1 1 0} planes having the h1 1 1i zone. Hence, simple calculations following
elasticity theory are inadequate, and much of the knowledge about the dislocation core structure in
b.c.c. metals was deduced from atomistic simulations. Such studies, performed by Vitek et al. [108]
have justified the non-planar core extension. In particular the model explains the indeterminate nat-
ure of the slip plane, the high Peierls stress of b.c.c. metals, and that the dislocation structure after
deformation tends to consist of straight screw dislocations. A very important consequence of the
non-planar cores is that the dislocation behaviour is prone to non-glide components of the stress ten-
sor [109]. The self-trapping of the screw dislocations and their release by thermal nucleation of
double-kinks is a commonly held concept to explain the rapid increase of the flow stress of b.c.c. met-
als at low temperatures [110].
There are even more restrictions for the deformation of the B2-phase, as the symmetry of the b.c.c
structure is reduced by the ordering. The shortest translation vectors are, in the order of increasing
magnitude, h1 0 0i, h1 1 0i and h1 1 1i. With rare exceptions, only dislocations with h1 0 0i and h1 1 1i
Burgers vectors have been observed after low temperature deformation. The commonly observed slip
planes of the h1 0 0i dislocations are {0 1 1} and {0 0 1}, whereas the h1 1,1i dislocations normally slip on
{1 1 0}, {1 1 2} and {1 2 3} planes, see [111]. Fully ordered B2 compounds apparently cannot twin on the
usual b.c.c. twinning system. Ab-initio calculations of Nguyen-Man and Pettifor [46] have shown that
the B2-phase is prone to decomposition into structurally related stable and metastable phases. In this
context, pseudo-twinning of B2 compounds has been discussed, which can formally be described as a
martensitic transformation from the simple cubic structure Pm 3m to an orthorhombic structure of
space group Cmmm, i.e., the product structure is not crystallographically identical to the parent struc-
ture. Debates remain, whether or not such a mechanism occurs under usual stress levels. Thus, several
details of the deformation behaviour of the b/B2-phases remain to be elucidated.
Table 3
Crystallographic elements of mechanical twinning in the L10 structure of c(TiAl); g twinning shear strain, kT = c/a; (after Yoo
[23,106]).
Mode K1 K2 g1 g2 g
pffiffiffi
Primary twinning (1 1 1)
(1 1 1Þ
1/2[1 1 2] 1/2[1 1 2] 2k2T 1 = 2kT
pffiffiffi
Complementary or anti-twinning (1 1 1) (0 0 1) 1
1/2[1 2] 1/2[1 1 0] 2=cT
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 75
Fig. 9. Atomic structure of a deformation twin observed in a Ti–48.5Al–0.37C (at.%) alloy after room temperature compression
to strain e = 3%. (a) High-resolution electron micrograph along the common h1 1 0] direction. c designates the matrix and c the
M T
twin. White lines indicate the trace of the (1 1
1Þ planes. Note the one-plane step in the (1 1 1) interface marked by the arrow,
which is associated with a Shockley partial dislocation. (b) Higher magnification of the area marked by the arrow in (a). The red
bars mark the position of the interfacial plane; Letters S and F denote the start and finish, respectively, of the Burgers circuit
constructed around the Shockley partial dislocation. The resulting Burgers vector is b = 1/6[1 1 2]. Figure adapted from [3,55],
changed.
Thus, there are only four order twinning systems available in c(TiAl), contrasting to twelve in f.c.c.
metals. Twinning shear is unidirectional, i.e., the operating twinning systems vary with the sense of
load and loading direction. Taken together, these factors restrict the operation of mechanical twinning
in TiAl to certain crystal orientations. The Schmid factors being relevant for mechanical twinning were
calculated by Sun et al. [112].
The activation of mechanical twinning is important for the plasticity of Ti–Al-alloys because it
provides auxiliary deformation modes with shear components in the c direction of the c(TiAl) unit cell.
In two-phase alloys the twinning propensity seems to be enhanced by the relatively low Al content,
the presence of substitutional alloying elements like Mn or Nb, a lamellar microstructure, and high
deformation temperature [3,105,106]. However, the mechanistic details behind these factors are
not completely clear. Thus, much effort has been invested for understanding twin generation.
partial dislocation loop on a
The initial stage of twin nucleation is the passage of a first 1/6h1 1 2]
f1 1 1gc plane. The loop involves an intrinsic stacking fault, which increases the surface energy (i.e.
the twin/matrix interface); hence a large part of the work to form a twin goes into creating its
76 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
boundaries [105,106]. However, once the first loop has nucleated, succeeding loops are easier to form
because these loops only propagate the coherent twin/matrix interface but do not create new interface
area. For this reason, separate consideration is usually given to the nucleation of a small twinned vol-
ume, the embryonic twin, and to its subsequent growth into a large twin. A three-layer fault is the
smallest possible twin, which reproduces the L10 structure.
The homogeneous shearing of a lenticular-shaped twin volume would require unreasonably high
stresses of five to ten percent of the shear modulus [113]; hence heterogeneous nucleation at a defect
site is more likely. In accordance with this concept, twin nucleation in c(TiAl) has been associated with
energetically feasible dislocation reactions [105] and structural heterogeneities which serve as nucle-
ation sites and stress risers. In Section 4.4.2 a case study on twin nucleation will be presented that may
typify these processes.
The lateral growth of an embryonic twin is considered as a coordinated movement of Shockley par-
tial dislocations through successive {1 1 1} planes, as proposed for cubic and hexagonal metals
[105,106,113,114]. Most explanations of the origin of such a coordinated motion invoke the existence
of a so-called pole dislocation, being partly or totally of screw character. The screw component of the
pole dislocation is perpendicular to the twin plane and is equal to the interplanar spacing of the {1 1 1}
planes. A twinning partial dislocation winds around the pole dislocation in a fashion similar to a stair-
case. In doing so, the partial dislocation not only produces a stacking fault, but also climbs up the pole
dislocation to the next layer. Repetition of this mechanism forms a thick twin. A general problem with
this classical pole mechanism is that it cannot accommodate the high propagation rate of twins that
are usually observed.
In this context, based on molecular dynamic simulation, Xu et al. [115] have proposed that twins in
c(TiAl) could be formed by five coordinated 1/6h1 1 2i shears on adjacent (1 1 1) planes, in the fashion
of a synchro-shear mechanism [116]. The five shear steps are equivalent to a composite dislocation
that is comprised of a 1/6[5 1 4] superpartial at the twin/matrix interface and an ordinary 1/2[1 1 0]
dislocation on the next (1 1 1) plane. It should be noted that 1/6h5 1 4] superpartials have been
observed as one dissociation component of h1 0 1] superdislocations (Section 3.2). The total Burgers
vector of the synchro-shear mechanism is b = 2/3[2 1 1]. The associated twinning shear g = 2p 2 is four
times that of conventional true twinning. The advantage of this mechanism is that it does not require
twinning dislocations to multiply by a pole mechanism. Xu et al. [115] have suggested that this type of
twin formation could effectively accommodate stress concentrations or high strain rates. However, no
experimental evidence for the mechanism has yet been observed. The twins observed in TiAl are gen-
erally very thin; this suggests that the operation of pole sources is limited to only a few turns. Thus, at
least at low temperatures, the final thickness of the twins seems to be related to the geometrical
extension of heterogeneous nucleation centres [117,118].
A striking feature is the increase of the twinning propensity with temperature because twinning is
usually observed under conditions where dislocation glide is difficult, i.e. at low temperatures and
high strain rates. In this context, Yoo [119] has investigated the climb expansion of Frank partial dis-
location segments as the nucleation stage of a pole mechanism. The analysis suggests that a high
supersaturation of thermal vacancies, y0 =yeq 0 > 13, is required in order to drive the mechanism solely
by chemical driving forces. The quantity yeq 0 is the equilibrium molar fraction and y0 the actual molar
fraction of vacancies. This appears unrealistic in terms of the existing diffusion data of TiAl [120]. At
7
1500 K, the equilibrium molar fraction of vacancies is thought to be yeq 0 = 10 to 106 [106]. However,
diffusion could be supported by the presence of TiAl antisite defects, which in Ti-rich alloys are formed
in order to accommodate the off-stoichimetric deviation. A TiAl antisite defect is a Ti atom situated on
the Al sublattice. Microanalysis performed on two-phase alloys with the base-line composition Ti–
47Al has shown that the concentration of TiAl antisite defects in the c-phase is in the order of 102
[117,121]. At such a high concentration, the antisite defects form a percolating substructure of anti-
structural bridges along which diffusion can occur. The point to note is that the migration energy of
TiAl antisite defects is expected to be significantly lower than the self-diffusion energy of Ti deter-
mined at high homologous temperatures [120]. Thus, via anti-structural bridges high diffusion rates
can be accomplished at moderately high temperatures, where the conventional vacancy mechanism
is still ineffective. Under the same conditions, diffusion in Al-rich TiAl is probably very sluggish
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 77
because the migration energy of AlTi antisite defects is significantly higher. Thus climb-induced twin
nucleation is expected to be less efficient in Al-rich TiAl. In PST crystals, no significant change of the
deformation mode was observed up to 800 °C, i.e., the relative contributions of dislocation glide and
mechanical twinning remained constant [122]. This might be a consequence of the slow growing pro-
cess of PST crystals, which probably leads to a near-equilibrium phase constitution and a correspond-
ingly low density of TiAl antisite defects. Thus, as with Al-rich TiAl, twin nucleation is not supported by
dislocation climb. Taken together, these factors may account for the observed differences in the twin-
ning behaviour of Ti- and Al-rich Ti–Al-alloys. Clearly the problem could be a challenge for further
experimental investigations and simulation studies.
3.2.6.2. Deformation channeling in PST crystals. A remarkable phenomenon is the so-called deformation
channelling between the lamellar boundaries, which was recognized on PST crystals for sample
orientations around U ¼ 0 [125,126]. Due to this effect, plastic shear is confined to an appropriate
78 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
combination of ordinary slip and mechanical twinning. The important point to note is that this com-
bined shear occurs on a common slip plane with a near-zero Schmid factor, although h1 0 1] superdis-
location systems with a large Schmid factor are available. The fact that these superdislocation systems
were not activated indicates that glide of superdislocations is difficult. The close coupling of ordinary
slip and mechanical twinning was also suggested by atomistic simulations, which showed that ordi-
nary dislocations and mechanical twins could propagate with nearly the same velocity [126].
In an analytical study, Paidar et al. [127] has simulated the situation, in which a colony is loaded
parallel to the lamellae (U = 0°), while deformation along the lamellae and perpendicular to the load-
ing axis is blocked. Under these conditions significant constraint stresses develop, in addition to the
externally applied stress. The trends are such that the shear stresses on the slip systems with the high-
est Schmid factor are reduced, whereas those of the other systems are increased. Thus, in a constrained
colony supplementary glide and twinning modes could be activated. Taken together, the observations
indicate that the deformation modes operating in a certain c domain of the lamellar microstructure
cannot be solely predicted on the basis of Schmid factor considerations. This important finding has
not yet adequately been considered in modeling studies.
The various modeling concepts described in the literature will be classified into ‘‘metalphysical
modeling”, ‘‘continuum-mechanical modeling” and ‘‘thermodynamic modeling”, according to the
underlying assumptions and simplifications.
In the following sections we have selected some phenomena, which are of high practical relevance
and need a reasonable amount of theoretical work to be understood. Specifically the lamellar
microstructure due to simultaneous existence of two phases does not allow a straight forward adapt-
ing of concepts that are already available for structural materials as Fe-, Al- and Ti-based alloys. We
would like to emphasize that several results reported in this chapter are new and very often not pub-
lished up to now. Insofar the text may be sometimes more demanding for the reader than modeling
concepts, which are already well established. We ask here for the understanding of the reader and
report a selection of several topics we consider as very relevant (and often new) for c(TiAl)-based
alloys.
4.1.1. The concept of thermal and athermal stresses applied to c(TiAl)-based alloys
Ti–Al-alloys exhibit a wide range of strength properties, depending on alloy composition and
microstructure. For example, the room temperature yield stress varies between 350 MPa and
1110 MPa, for a review see [3]. There are certainly several sources that may contribute to the yield
stress. Concerning a strain-controlled test with a constant strain rate a commonly held concept in
crystal plasticity is to subdivide the yield stress r of a material into internal and effective stress parts
according to [128] as
r ¼ rl þ r ðe_ ; TÞ: ð8Þ
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 79
Table 4
Composition and constitution of the alloys investigated [129].
The internal (or athermal) stress rl results from glide obstacles with long-range stress fields,
which can only be overcome mechanically. Thus, rl is almost independent of temperature and strain
rate, respectively, and partially takes up the total applied stress r. Prime examples of long-range (or
athermal) glide obstacles are the grain and phase boundaries occurring in multi-phase assemblies.
The effective (or thermal) stress part r arises from glide obstacles with short-range stress fields,
which can be overcome with the aid of thermal activation [128]. Thus, r depends on strain rate e_ and
temperature T. Typical examples of short-range obstacles are impurity-related defects or mono-
atomic jogs dragged behind moving screw dislocations.
The internal stress (back-stress) rl is opposite to the applied shear stress and must be overcome for
slip to continue. Thus, knowledge about the relative magnitude of rl and r is vital in order to assess
the deformability of a material. Given this importance, internal stresses produced during tensile defor-
mation were recently determined by the incremental unloading technique (dip testing) [129]. The
operative concept is that slip in the reverse direction can take place if the applied stress is reduced
or removed. This anticipated dislocation behaviour was indeed recognized in a TEM in situ study
[129]. The internal stress was determined as the critical stress at which the inelastic sample relaxation
is reversed, i.e., going from the tensile direction into the compression direction. The flow behaviour of
the sample after unloading was monitored in the strain-controlled relaxation regime. The investigation
involved a wide range of alloy composition (Table 4) with a corresponding variation of microstructure
and strength. Fig. 10 demonstrates the tensile characteristics of the alloys investigated. The observed
strength properties illustrate the trends in alloy design, which have been described in several articles
[3,7,130]. A typical example of dip testing is demonstrated in Figs. 11 and 12. Relaxation from a stress
rx above rl is thought to result in normal stress decay with time. Unloading to a stress level rx < rl
should give rise to an anomalous relaxation, i.e., to an increase of stress with time. Ideally, there is a
small stress range rx around rl , where no stress relaxation occurs. The unloading steps (9–12)
(Fig. 11) were used in order to iteratively approach the rl level. Stress reduction to level (9) is followed
by a normal force C decay with time (Fig. 12a), whereas an anomalous increase of F with t occurred
after stress reduction to level (12) (Fig. 12c). These observations suggest that the rl level is between
levels (10) and (12). This is confirmed by the stress reductions (10) and (11), which exhibits neither
normal nor anomalous relaxation (Fig. 12b); thus, these two normal forces flank the rl level.
The internal stresses measured on alloys #2-7 and #9 at room temperature are summarized in
Fig. 13. The data set fulfils relatively well a linear function between rl and r with a slope of
m = 0.8 and a relatively small intercept. Thus, for all the materials investigated, rl is always a large
proportion of the total applied stress r. A thorough discussion of the mechanical data, supported by
electron microscope examination of the defect structure, has shown that the most important sources
80 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 10. Stress/strain curves measured by room temperature tensile testing. The numbers refer to the alloys listed in Table 4.
The experiments involved stress relaxations and incremental unloadings to measure the activation volumes and the internal
stresses, respectively. The alloys indicated by numbers are defined in Table 4. Figure from [129], permission from Elsevier.
Fig. 11. Normal force/strain curve of a strain-controlled tensile test performed on alloy #3 (Table 4) at room temperature. Note
the well-expressed yield drop (1) at the beginning of flow. The experiment involves several stress reduction tests that are
indicated by numbers (2)–(16), the details of which are demonstrated in Fig. 12. Figure from [129], permission from Elsevier.
of the internal stresses are the constraint stresses arising from misfitting interfaces (dislocations are
incorporated in the interfaces reducing the energy of the system), the lack of independent slip systems
in the majority phases, and the elasto-plastic co-deformation of the different alloy constituents; see
also Section 4.1.4. The constraint stresses are expected to act as constant long-range retarding forces
that impede dislocation motion.
A theoretical investigation by Shoykhet et al. [131] has shown that the coherency stress in lamellar
alloys is inversely proportional to the lamellar spacing kL . Thus, with continued diminution of kL to the
nanoscopic scale, the constraint effects in controlling the internal stresses seemingly obtain increasing
importance. For example, in the design of alloys #3 and #6, the thickness of the lamellae was reduced
in order to maximize the yield stress. The lamellar colonies of alloy #6 are comprised of platelets of
c; a2 and B19-phase. The average lamellar spacing is about kL = 11 nm, which is close to the coherency
limit kLC . Thus, the constraint stresses produced at misfitting interfaces could be relatively close to the
maximum coherency shear stress sc;max ¼ lDe ¼ l=100 to l=50 Section 2.2). This constraint effect is
certainly the most important mechanism behind the high strength of the TNB-alloys (see Section 1.1.2)
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 81
Fig. 12. Normal force reductions involved in the tensile test shown in Fig. 11. (a) Normal force reduction to level (9). Note the
normal force decay with time t. (b) Normal force reduction to level (11). Note the fairly constant force F. (c) Normal force
reduction to level (12). Note the anomalous increase of F with t. Figure from [129], permission from Elsevier.
and c-Md-alloys. The great variety of structural details associated with the constraint effects is
demonstrated in Fig. 14. The micrograph shows the reaction front of the B19 ? c transformation in
the TNB-V2-alloy (alloy #6). There are various terminated lamellae and disconnections seen in the
micrograph, associated with local strain contrast. The c-phase consists of twin related variants with
ABCA and CBAC stacking sequence, respectively. Stacking of the lattice in different directions intro-
duces strain fields of opposite sign, thus reducing the total strain energy and making the B19 ? c
transformation easier. This fact, together with the small thickness of the c-lamellae allows a largely
homogeneous misfit accommodation between the B19 and c-phase. Nevertheless, significant con-
straint stresses are probably left at the growth front and manifest themselves through a significant
distortion of the lattice planes. The compressed image (Fig. 14b) shows this feature more clearly.
82 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 13. Correlation of the internal stress rl with the yield stress r measured at room temperature and e = 1–2%. The stress
immediately prior to the individual dip tests is r. The symbols correspond to those used in Table 4. The error bar represents the
main standard deviation for measurements performed on four to six individual samples of the same alloy. Figure from [129],
permission of Elsevier.
The micrograph illustrates that even the plastic state in multi-phase Ti–Al alloys even at the nano-
scopic scale is very heterogeneous.
Several attempts have been made to describe the yield stress of c(TiAl)-based alloys in terms of a
Hall–Petch relationship, for a review see [3]. For multi-phase alloys, such an assessment appears
unsatisfactory, since the constitution and microstructure cannot be described by a single length
parameter. Furthermore, in the Hall–Petch model, pile-up of dislocations at grain boundaries is envi-
sioned as a key mechanistic process underlying an enhanced resistance to plastic flow. However, there
are very few specific instances of piled-up dislocations published in the Ti–Al literature, but plenty of
examples showing evenly distributed ordinary screw dislocations. Owing to this situation, the evolu-
tion of internal stresses upon deformation was addressed by finite-element simulation [132,133]. The
simulations suggest that a polycrystalline lamellar material deforms very inhomogeneously with a
strong tendency to develop shear bands, lattice rotations and internal buckling of the lamellae. Even
under compression, very high tensile hydrostatic stresses are generated at the triple points of colony
boundaries. Collectively these findings indicate that the constraint phenomena discussed above are
the dominant factor in determining the internal stress within multi-phase alloys based on c(TiAl).
More details on this subject are provided in a recent publication [129].
e_ ¼ e_ 0 expðDG=kTÞ: ð9Þ
The total stress r measured in a constant strain rate test, given by Eq. (8) and Eq. (9), can be
extended to
MT
r ¼ rl þ r ðeÞ; r ðeÞ ¼ ðDF þ kT lnðe_ =e_ 0 ÞÞ: ð10Þ
V
The quantities involved in Eqs. (9) and (10) are, see, e.g., [128,134,135]:
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 83
Fig. 14. High-resolution micrograph of the lamellar microstructure imaged down h1 1 0]c//[0 1 0]B19; alloy #6. (a) Reaction front
of the B19 ! c transformation occurring within a lamellar colony. Note the formation twin related c variants and various
interfacial steps at the interfaces. (b) A computer-generated projection of the micrograph along the interface produced by
rotating the high-resolution image in (a) by 17° clockwise and adjusting the proportions to foreshorten the length of the
interfaces. Note the distortion of the complementary atomic planes at the reaction front. Figure from [129], permission from
Elsevier.
The thermodynamic glide parameters can be estimated from the temperature and strain rate sen-
sitivity of the flow stress according to well-known methods established in the literature [128]. The
activation volume V can be measured by reversible strain rate cycling tests according to
84 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 15. Temperature profiles of (a) the tensile yield stress r0:2 and (b) of the internal stress rl determined by incremental
unloading. Ti–45.6Al–7.7Nb–0.2C (at.%), nearly lamellar microstructure. Data from [136].
M T kT
V¼ : ð11Þ
ðDr=D ln e_ ÞT
The term R ¼ ðDr=D ln e_ ÞT is the reduced stress increment that occurs upon strain rate cycling at
constant temperature T. The quantity Dr is the difference between the flow stresses r1 and r2 related
to the strain rates e_ 1 and e_ 2 , respectively; D ln e_ ¼ lnðe_ 2 =e_ 1 Þ is the logarithm of the strain rate ratio.
Equivalent to R is the stress rate sensitivity S ¼ ðDr=D lnðr_ ÞÞT , which can be determined from the
kinetics of stress relaxation tests.
For a certain deformation mechanism, the glide parameters have distinct values and could help
identifying the factors controlling the dislocation velocity. However, direct observation of the mech-
anistic details is always important for confirmation. Thus, a wealth of data has been provided by
mechanical testing and electron microscope observations.
Fig. 15 demonstrates the temperature profiles of the total stress r and athermal stress rl that were
determined on the Ti–45.6Al–7.7Nb–0.2C (at.%)-alloy by dip testing [136]. Throughout the entire tem-
perature range rl is always a large portion of the applied stress; however, its contribution decreases
with rising temperature. The behaviour is also typical for other multi-phase alloys [129]. The distinct
difference to single-phase c-alloys is that there is no yield stress anomaly in the stress–temperature
profile. This is probably another manifestation that deformation in multi-phase alloys is mainly
provided by ordinary dislocations, which, due to their compact core structure, are not liable to form
sessile non-planar dislocation locks. The collected literature data has shown that the velocity of ordi-
nary dislocations in this phase is controlled by several mechanisms, which superimpose in certain
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 85
Fig. 16. Possible mechanisms contributing to the yield stress of two-phase titanium aluminide alloys at low to moderately high
strain rates. The situation sketched in the figure corresponds to high Nb containing alloys with the base-line composition Ti–
45Al–(5–10)Nb. Note the high athermal contribution rl to the total yield stress r. The athermal contribution rl mainly results
from the elasto-plastic co-deformation of the various constituents and the fine microstructure. r ¼ r rl is the effective
(normal) stress. SA (strain ageing) refers to the stress contribution resulting from defect atmospheres consisting of TiAl antisite
defects and vacancies. DR designates dynamic recovery and DRX dynamic recrystallization. Figure adapted from [3,55],
changed.
Fig. 17. Jogged ordinary screw dislocations observed in a Ti–48Al–2Cr alloy (at.%). Note the trailing and termination of dipoles
and debris. (1) Terminated debris defect, (2) a dislocation dipole trailed at a jogged screw dislocation, and (3) a screw
dislocation about to multiply by forming single-ended sources at a high jog. Room temperature compression to strain e = 3.2%.
86 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
temperature ranges. However, there are three distinct temperature domains in which a certain
mechanism is predominant; Fig. 16 is a graphical representation of this situation.
4.1.2.2. Dislocation glide (T = 77–573 K). In domain I of the stress/temperature profile (Fig. 16),
deformation seems to be controlled by dislocation glide. The activation parameters determined on
the Ti–45Al–8Nb–0.2C (at.%)-alloy at room temperature are r ¼ 284 MPa, V ¼ 65b , DG ¼ 0:8 eV,
3
DF ¼ 1:53 eV. This data set is typical for other two-phase alloys [3,137]. The activation volume was
referred to the Burgers vector of ordinary dislocations, which mainly contribute to deformation.
Electron microscope examination of the deformation structure has shown that jog dragging and
king bulging of ordinary dislocations on different octahedral planes significantly contribute to the
glide resistance. The typical dislocation structure observed after room temperature deformation is
shown in Fig. 17. More details about jog dragging are provided in Section 4.1.3. A salient feature of
the TEM observations is that the observed obstacle distances are considerably larger than the value
expected from the activation volume. This gives rise to the speculation that the glide resistance in
domain I may in part originate from lattice friction due to the Ti–Al directional bonding [138]. Atomis-
tic calculations by Simmons et al. [139] revealed a high friction arising from a Peierls-type mechanism,
particularly along the screw and 60° mixed dislocation directions. In the TEM investigations the pres-
ence of lattice friction is suggested by the observation that the dislocations in the unloaded TEM sam-
ples are still bowed out in a smooth arc between the obstacles (Fig. 17). This finding indicates that
lattice friction forces occur on all dislocation characters and impede a complete relaxation of the
bowed segments into the geometrically shortest configuration between the obstacles. The Peierls
stress arises as a direct consequence of the lattice periodicity; the relevant structural unit, the kink,
has atomic dimensions. Thus, the activation volume of the Peierls process is very small, typically
V = (1–20)b3 [113,140]. The superposition of lattice friction with any of the above-discussed mecha-
nisms could explain the relatively small activation volumes that have been experimentally
determined.
Several authors [141–143] have revealed that the glide resistance of ordinary dislocations origi-
nates from localized extrinsic obstacles. In Ti–Al alloys, oxygen, nitrogen and carbon are unavoidable
impurity elements, which have a low solubility of about 300 at. ppm in the c-phase [60]. Thus, these
elements can form oxides, nitrides and carbides, and these precipitates may act as localized glide
obstacles. For example, Messerschmidt et al. [142], by performing TEM in situ straining experiments
at room temperature, recognized a dislocation kinematics that was reminiscent of overcoming local-
ized obstacles (presumably Al2O3), which were separated by about 100 nm. However, there is a gen-
eral problem with the hypothesis of localized pinning. Impurity related defects with a spherical shape
would only interact with the hydrostatic stress field of the edge dislocations. There is only a very weak
interaction of spherical defects with the screw dislocations because their strain field is nearly pure
shear. Thus, the question remains, why screw dislocations and not edge dislocations are pinned?
So, all in all, one is left with the impression that external pinning by impurity containing precipitates
is only significant in single-phase c-alloys or, perhaps, in near-stoichiometric alloys with a low
amount of a2 -phase. In Ti-rich two-phase alloys, the c-phase is devoid of precipitates because impu-
rity elements exceeding the solubility limit of the c-phase are taken up by the a2 -phase.
An often overlooked aspect is that antisite defects could act as glide obstacles. In Ti-rich alloys TiAl
antisite defects, meaning Ti atoms situated on Al sites, compensate for the off-stoichiometric deviation
[23]. The TiAl antisite defect has a lower symmetry as the TiAl matrix and hence produces a non-
centrosymmetric distortion [144]. Thus, the antisite defects are expected to interact with both
hydrostatic and shear stress fields. This is important with regard to the dislocation glide because
the TiAl antisite defects will interact with all dislocation characters, i.e., also with screw dislocations.
The ordinary dislocations exhibit remarkable glide anisotropy with regard to the dislocation char-
acter [3]. A dislocation loop is usually elongated along its screw components, which indicates that the
velocity of the screw component v screw is slower than that of the edge component v edge . The observed
loop shape indicates the ratio as
mscrew =medge ¼ 1 : ð3—5Þ: ð12Þ
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 87
Fig. 18. Dynamic strain ageing observed on the alloy Ti–45Al–5Nb–0.2B–0.2C (at.%), TNB-V5, at the deformation temperature
T = 873 K. (a) Load/elongation traces of a strain controlled tensile test involving reversible strain rate changes between
e_ 1 ¼ 2:38
105 s1 and e_ 2 ¼ 20e_ 1 . The total sample elongation etS was used as feed-back parameter. The diagrams show the
evolution of stress with strain e and time t, respectively. (b) Detailed analysis of the flow behaviour at the beginning of
deformation (arrowed in (a)). The diagrams show the flow stress r, the total displacement of the machine drive DlM and the
total sample elongation etS as functions of time t. The strain rate changes are marked by arrows. The sample exhibits serrated
yielding, whereas the strain rates d etS =dt measured in between the strain rate changes are fairly constant. Note the slight
variation of the machine displacement rate dðDlM Þ=dt in the course of heavy serrations (arrow (1) in order to accommodate the
sudden changes of the flow rate of the sample.
88 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 19. Static strain ageing experiment performed on a Ti–48.5Al–0.37C (at.%) alloy under compression. The stress increments
Dra were measured as the difference before ageing and the upper yield point occurring on reloading. Different ageing times ta
between unloading and reloading were applied in order to establish different levels of dislocation locking by defect
atmospheres. Figure taken from [55], permission from Elsevier.
This anisotropy is probably caused by a higher glide resistance of the screw components due to jog
dragging. Jogs in edge dislocations can conservatively propagate together with the dislocations. Thus,
the glide resistance of the edge components is expected to be relatively low. This finding explains the
predominance of ordinary screw dislocations in the deformation structure. However, the glide aniso-
tropy might be another factor that further restrict the available shear modes in c(TiAl). This could be
important for the accommodation of stress concentrations, e.g. at crack tips. The effect has obviously
not yet been considered in modeling studies.
4.1.2.3. Static and dynamic strain ageing (T = 450–750 K). Deformation of Ti-rich alloys in the interme-
diate temperature interval of 450–750 K (domain II in Fig. 16) is characterized by discontinuous yield-
ing, zero or even negative strain rate sensitivity and strain ageing phenomena [117,137]. These
characteristics are usually associated with dislocation locking due to the formation of defect atmo-
spheres [145]. Figs. 18a and b demonstrate the flow characteristics observed during a tensile test per-
formed at T = 650 K on a Ti–45Al–5Nb–0.2B–0.2C (at.%)-alloy (TNB-V5). Under the conditions used in
the experiments, successive locking and unlocking of the dislocations apparently occurs, with the con-
sequence that the stress/strain curve breaks up into serrations. This plastic instability is known as
dynamic strain ageing or the Portevin–LeChatelier effect [145]. These phenomena indicate that in
domain II a new friction mechanism appears, which substitutes for those occurring in domain I. To
make this statement quantitative, static strain ageing experiments were performed for various Ti–
Al-alloys [117]; an example of which is demonstrated in Fig. 19. The samples were deformed to dif-
ferent levels of pre-strain, aged in situ on the load frame for certain ageing periods ta , and then
retested. On reloading distinct yield points Dra occurred, after which the original stress strain curve
was retraced, i.e., there was no permanent hardening effect. This corresponds to a situation in which
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 89
Fig. 20. Possible orientations of the vacancy/TiAl antisite defect complex. The figure shows an ordinary edge dislocation in a
In the dislocation stress field the complex may reorient via the anti-structural bridge mechanism ASB-
projection near to h1 1 2].
2 described in the text. TiAl – Ti atom situated on an Al site, VTi – vacancy on the Ti-sub-lattice. Figure taken from [3], changed.
pinning of dislocations during the ageing period occurred due to formation of a TiAl atmosphere. Upon
reloading an extra stress Dra is required to move the dislocations away from the atmospheres; thus,
this stress increment reflects the degree of dislocation locking. The locking effect strongly increases
with increasing off-stoichiometric deviation to the Ti-rich side. The pinning process is characterized
by a very fast kinetics and a low activation energy of about Q a = 0.68 eV, which is close to the migra-
tion energy of a TiAl antisite defect associated with a Ti vacancy [120]. These two arguments led to the
assumption that dislocation locking occurs by the formation of oriented defect atmospheres [117]. The
relevant defects are TiAl antisite atoms associated with Ti vacancies, which can easily reorient by
vacancy jumps in the dislocation stress field; the crystallographic situation is sketched in Fig. 20. It
should be noted that the described strain ageing phenomena do not occur in Al-rich alloys, where AlTi
antisite defects accommodate the off-stoichiometric deviation. This is because the migration energy of
the AlTi antisite defect is expected to be significantly higher than that of TiAl [117,120]. Thus, atmo-
spheres cannot be formed within the time frame of conventional deformation or ageing tests. Consid-
ering this aspect and the available information about the portioning of third elements, the strain
ageing phenomena in c(TiAl)-based alloys can be reliably predicted.
In an attempt to describe the kinetics of the process it was assumed that the rate of defect reori-
entation dN=dta is proportional to the number N of defects that have not yet reoriented within the age-
ing period t a . Also, the assumption was made that the stress increments Dra ðta Þ are proportional to the
number of reoriented antisite defect/vacancy complexes. The evolution of Dra with t a is then given by
[146]
could be a versatile description of the kinetics for the formation of ordered atmospheres.
There is a good correlation between the ageing increments and the reciprocal activation volume
measured at low temperatures. Thus, at low temperatures, where atmosphere formation is not
90 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 21. Climb of ordinary dislocations observed during in situ heating inside the electron microscope under the conditions
indicated. The TEM foil was taken from a Ti–48Al–2Cr(at.%) compression sample pre-deformed to strain e ¼ 3%. Note the
formation of helical dislocation configurations. Acceleration voltage 120 kV.
possible, the antisite atoms may act as localized obstacles. Whereas defect atmospheres are solely
formed by TiAl antisite defects, both TiAl and AlTi antisite defects may act as localized obstacles.
It should be mentioned that strain ageing phenomena associated with off-stoichiometric deviations
were also recognized in alloys based on Ti3Al [147].
4.1.2.4. Dislocation climb, dynamic recovery, dynamic recrystallization (T = 800–1200 K). In this temper-
ature domain, diffusion becomes gradually important. The activation parameters determined at
T = 1073 K and e_ = 2.38
105 s1 on the Ti–45Al–8Nb–0.2C (at.%)-alloy are [3,137] r ¼ 154 MPa,
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 91
Fig. 22. Registration curve of a tensile test performed on the Ti–45.6Al–7.7Nb–0.2C (at.%) alloy at T = 976. Note the constant
stress level at strains e P 2%. The stress relaxation tests R1 to R20 were implemented for measuring the stress rate sensitivity;
see accompanying text. Data from [136].
V = 14b3 (b = 1/2h1 1 0], M T ¼ 1), DH ¼ 3:3 eV The quantity DH is the activation enthalpy [128,134].
This is indicative of a diffusion assisted climb mechanism because climb of an edge dislocation
depends on the diffusion of vacancies either towards the edge dislocation or away from it. In c(TiAl)
deformation by climb requires both elemental components to be mobile; otherwise large gradients of
the chemical potential are set up. The self-diffusion energy for Ti is Q sd (Ti) = 2.59 eV, and that for Al is
expected to be Q sd (Al) = 3.71 eV [120]. Thus, the measured DH value probably represents an average of
the Ti and Al self-diffusion energies. From theory [3] dislocation climb is associated with V = 1b3 to
10b3, which is relatively close to the observed value. Thus, it is expected that dislocation climb is fully
established at T P 1073 K and e_ = 2.38
105 s1. Lending support was provided by post mortem and
in situ TEM observations [3,148]. Climb was exclusively observed on 1/2h1 1 0] dislocations, which is
plausible because these dislocations have a compact core structure. Fig. 21 demonstrates climb struc-
tures in a Ti–45Al–10Nb-alloy. Similar findings were obtained on other (a2 þ cÞ-alloys, although there
is a significant scatter in the data. This is probably due to other climb-controlling factors as alloy
chemistry, constitution, microstructure, and impurity content.
Dislocation climb certainly reduces the glide restrictions in the c(TiAl)-phase as dislocation prop-
agation is not anymore confined to the slip plane. Thus, the constraint phenomena, which are the
major sources of the internal stress, are greatly reduced. This effect probably gives rise to the transi-
tion from brittle to ductile (BDT) material behaviour, which occurs in domain III. However, due to the
thermally activated character of dislocation climb, the BDT transition temperature is expected to
depend on strain rate.
Dislocation climb is associated with dynamical recovery. Fig. 22 shows the registration curve of a
tensile test performed on the Ti–45Al–8Nb–0.2C (at.%)-alloy at 976 K and e_ ¼ 2:38
105 s1 [137].
The flow stress develops to a constant level, which indicates that the production and annihilation rates
of deformation induced defects are equal. This view is supported by the stress rate sensitivity
S ¼ ½Dr=D lnðr_ ÞT , which in the same test was measured by stress relaxation tests. Upon straining,
this parameter develops to the constant value of S = 80 MPa. The quantity S is related to activation vol-
ume [140] as
kT
V ¼ MT ; ð15Þ
S
giving the value V = 7b3 (with M T ¼ 1 and b = 1/2h1 1 0]) and lending support to the hypothesis that
dynamical recovery occurred.
At further increasing test temperature, deformation is characterized by work softening associated
with a decrease of the stress rate sensitivity. Fig. 23 demonstrates the registration curve of a tensile
92 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 23. Registration curve of a tensile test performed on the Ti–45.6Al–7.7Nb–0.2C (at.%) alloy at T = 1276 K, indicating the
occurrence of dynamic recrystallization. The stress relaxation tests R1 to R30 were implemented for measuring the stress rate
sensitivity (see accompanying text). Data from [136].
test performed at 1276 K. Electron microscope observations have shown that under these conditions
dynamic recrystallization and the phase transformations a2 ! c and B19 ? c occur, depending on
the starting constitution [129].
The new grains are almost free of defects, since the dislocation motion in these grains is easy. This
observation explains quantitatively the relatively large activation volume measured at e ¼ 15% of
about V = 50b3 (b = 1/2h1 1 0]). However, a detailed interpretation is difficult because the observed
phase transformations and perhaps grain boundary sliding could contribute to deformation. More
about this subject will be provided in a forthcoming publication [149].
(i) dislocation accumulation and formation of multipoles due to the elastic interaction of disloca-
tions propagating on parallel {1 1 1} slip planes [55];
(ii) formation of sessile dislocation locks due to non-planar dissociation of superdislocations
[23,55,84];
(iii) multiple slip and intersection of dislocations propagating on oblique slip planes [55];
(iv) simultaneous operation of dislocation glide and mechanical twinning [55];
(v) development of constrain stresses due to the elasto-plastic co-deformation of the different
constituents [129], see also Section 4.1.1;
(vi) jog dragging of ordinary screw dislocations and debris hardening [55,150].
It is expected that the mechanisms (i)–(iv) produce glide obstacles with long-range stress fields
that provide an athermal stress component rl ðeÞ to work hardening see Eq. (8). These processes were
addressed in the cited publications, to which the reader is referred for more details. On the contrary,
jog dragging and debris hardening (vi) are expected to be thermally activated processes, which
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 93
Fig. 24. Work hardening behaviour of the Ti–45.6Al–7.7Nb–0.2C (at.%) alloy with nearly lamellar microstructure. (a) Load/
elongation traces of compression tests performed at room temperature. Stress relaxations and incremental unloading tests were
implemented for determining the stress rate sensitivity and the internal stress, respectively. (b) Dependence of the stress rate
sensitivity S on strain e. Data from [136].
produces a strain dependent thermal stress component r ðeÞ see Eq. (8). In terms of the above mech-
anisms, the strain dependent flow stress rðeÞ may be described with [55,148] as
MT
rðeÞ ¼ r0 þ rl ðeÞ þ r ðeÞ ¼ r0 þ rl ðeÞ þ DF D þ kT ln ðe_ =e_ 0 Þ ; ð16Þ
V D ðeÞ
see Section 4.1.2.1 and compare with Eq. (10). The quantity r0 represents a stress contribution from
dislocation mechanisms operating at the onset of yielding and is considered to be independent of
strain e. It should be noted that r0 may involve thermally activated mechanisms that are represented
by the activation volume V 0 . Possible mechanisms associated with r0 have been discussed in Sec-
tion 4.1.2.2. The quantity V D ðeÞ is the activation volume and DF D the free energy of the deformation
induced thermal process. The total activation volume V is determined by reversible strain rate cycling
tests as was described in Section 4.1.2.1, see also [128,134,110]. It is assumed that 1=V 0 and 1=V D ðeÞ
are linearly additive [56,148] yielding V as
1 1 1
¼ þ : ð17Þ
V V 0 V D ðeÞ
Jog dragging and debris hardening will now be considered in more detail in order to provide the
background information for the case study described in Section 4.3.
Jogs in ordinary screw dislocations can easily be formed by cross-gliding because the dislocations
have a compact core and several cross-slip planes are available. Cross-slip is certainly initiated by fluc-
tuating internal stresses. The density and magnitude of these fluctuations are expected to increase
94 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
with strain. Hence, the density of cross-slip events is expected to increase with strain. The dislocation
segment constituting the jog has edge character and is unable to move conservatively together with
the screw dislocation. The jog is connected to the moving dislocation by two lengths of edge disloca-
tions of opposite sign forming a dislocation dipole. Because of the mutual attraction of the positive and
negative edge dislocations forming the dipole, a dipole may break up in a row of prismatic loops.
It has been shown that small debris can act as obstacles and cause dislocation pinning [55].
However, the interaction forces between dislocations and small debris defects are highly localized
[151–153]. Thus, it is expected that the dislocation can overcome the debris with the help of thermal
activation.
At sufficiently high stress, movement of a mono-atomic jog may occur and will leave behind a trail
of vacancies or interstitial atoms depending on the sign of the dislocation and on the direction the dis-
location is moving (Fig. 17). The steady state configuration of the jogged dislocation is characterized
by so-called osmotic forces at the jogs balancing line tension forces from the bowed-out adjacent dis-
location segments. In principle, during deformation a screw dislocation can acquire both vacancy- and
interstitial-producing jogs. However, the stress required for the forward motion of the jogged disloca-
tion producing point defects is proportional to the formation energy of the relevant point defects.
Molecular dynamic simulations of Wang et al. [154] have shown that the formation energies for all
the stable interstitial configurations in TiAl are two to three times those of vacancies. Thus, it is unli-
kely that jog dragging generates interstitial atoms. It is more likely that interstitial jogs conservatively
glide along the screw dislocation and recombine with vacancy producing jogs. Because of the rela-
tively small formation energy for vacancies and the atomic dimensions of the activation complex,
thermal activation is expected to support jog dragging.
By virtue of these characteristics, a jog dragging mechanism should be manifested by an increase of
the reciprocal activation volume with strain. As demonstrated in Fig. 24b, this is indeed a salient fea-
ture of room temperature work hardening of Ti–Al-alloys; the stress rate sensitivity S is proportional
to the reciprocal activation volume. Another consequence that might be expected from jog dragging is
an increase of the vacancy concentration above that of thermal equilibrium, which in turn enhances
the diffusivity. Since the density of cross-slip events is expected to increase with strain, the related
dragging mechanism should be manifested by an increase of the reciprocal activation volume with
strain.
The mechanisms discussed certainly give rise to a substantial strain energy stored in the material.
Thus, the cold worked condition is thermodynamically unstable relative to the undeformed state.
Static recovery performed on various c(TiAl)-based alloys exhibit the following characteristics:
(i) a significant amount of the yield stress produced by work hardening is lost by annealing;
(ii) recovery occurs already at low homologues temperatures of only 0.38T m , where conventional
bulk diffusion is still ineffective (T m is the absolute melting temperature);
(iii) the recovery kinetic is very fast;
(iv) recovery of the flow stress is accompanied by a nearly complete recovery of the reciprocal acti-
vation volume to the value measured at the beginning of the pre-deformation.
These findings are probably directly related to the recovery of the dislocation dipoles and debris,
which are associated with the strain dependence of the activation volume. Dislocation debris has a
very small defect volume, which can easily be annealed out. Likewise, the dislocations of opposite
Burgers vectors constituting a dipole can easily annihilate by a combination of cross-glide and climb.
This view is supported by TEM in situ heating experiments that have been performed on samples pre-
deformed at room temperature [55,148]. The investigation has demonstrated that the vacancies
supersaturated in the matrix condense onto screw dislocations, causing the climb into helices. There
is also good evidence that the point defects also form additional sinks by the spontaneous nucleation
of dislocation loops. It should be noted that dense debris structures were also observed after room
temperature low cycle fatigue of the alloy Ti–45.6Al–7.7Nb–0.2C (at.%) [148,155]. As with monotonic
deformation, the fatigue debris structure could easily be annealed out at moderately high tempera-
tures. It should be noted, however, that recovery at these moderately high temperatures could also
affect the local stress state due to constraints developed during work hardening due to the mismatch
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 95
of the mechanical properties of the constituent phases. This might be expected because the climb pro-
cess may reduce the strain mismatch produced during the elasto-plastic co-deformation of the indi-
vidual constituents. The observed thermal instability of the work hardening may have several
implications for the processing and performance of c(TiAl)-based alloys. A significant amount of prior
recovery lowers the driving force for static recrystallization, which might be important for the evolu-
tion of the microstructure upon annealing. Recovery and relaxation occurring in a component, part of
a closed load train, is primarily responsible for loss of residual stress, because the associated plastic
strain partially releases the elastic strain stored in the load train. This could be a problem in bolted
fasteners. Also, the early recovery is certainly detrimental for surface hardening, which is currently
explored for c(TiAl)-based alloys to produce a crack initiation resistant outer layer [156].
Fig. 25. The Bauschinger effect in TiAl based alloys. Stress/strain curves of tensile tests performed on as-received and pre-
compressed material, with the pre-compression strain epc indicated. The tensile strains of the pre-compressed specimens are
plotted as accumulated absolute strains ea involving the pre-compression strain epc and the tensile strain et measured after
strain reversal. The green line indicates Hooke’s law with an elastic modulus of 175 GPa.
shown that the internal stresses easily relax upon unloading. In the extreme, unloading after
elastic–plastic deformation can largely invert the internal stress pattern that was formed during
yielding. This is a good precondition for the tensile segment (backward straining) of the Bauschinger
test because the residual stress in the grains that had been deformed during forward glide acts now in
the same direction as the applied stress. Furthermore, a high density of mobile dislocations and
dislocation is available. Taken together, these effects allow plastic deformation to occur almost at
the beginning of strain reversal. Since these deformation characteristics cannot be captured by the
classical isotropic hardening law, the use of a non-linear kinematic hardening approach seems to be
more appropriate for modeling strain hardening phenomena. More details about the Bauschinger
phenomenon are provided in [160].
4.1.5. Fracture
4.1.5.1. General aspects on modeling. Modeling of a cracked specimen is now, more or less, a standard
task in computational fracture mechanics. Also, established concepts exist not only for stationary
cracks but also for propagating cracks. The situation becomes much more difficult, if one has to deal
with an inhomogeneous specimen, since in this case the crack driving force is also affected by the ther-
modynamic forces acting on inhomogenities as interfaces. This means that the classical J-integral must
be supplemented by an inhomogeneity term. Although continuum mechanics offers the concept of
Configurational Forces [161–163] to evaluate this inhomogeneity term, its computational realization
is still demanding and a topic of research and development for several groups. The situation becomes
even more complicated for a very fine microstructure, as typically occurring in c(TiAl)-based alloys.
Therefore, modeling of crack propagation in c(TiAl)-based alloys can still be considered as research
in progress. Most of the current literature on fracture in c(TiAl)-based alloys reports on experiments
and their metalphysical interpretation. Therefore, we add a section on fracture within the context of
metalphysical modeling. Continuum mechanical modeling of the fracture process in lamellar (or, more
generally, inhomogeneous) material systems is going to become a research topic for the near future.
4.1.5.2. A short review on fracture studies for Ti–Al specimens. Earlier concerns about the brittleness of
c(TiAl)-based alloys have initiated intensive research, which is documented in various review papers
[3,23,164–167]. There is good metallographic evidence that the individual constituents are prone to
cleavage fracture on low index planes. At the atomic scale, the crack propagates by stretching and
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 97
Table 5
Cleavage strength Gc (ideal cleavage energy) and K 1c (critical stress intensity factors for a mode-I crack) from Yoo et al. [168].
pffiffiffiffiffi
Alloy Cleavage plane (h k l) Gc (J m2) Cleavage direction [u v w] K 1c ðMPa mÞ
TiAl (1 0 0) 4.6 [0 0 1] 0.94
[011] 0.89
(0 0 1) 5.6 [0 1 0] 1.04
[110] 1.04
(1 1 0) 5.3 [0 0 1] 1.03
0]
[1 1 0.86
(1 1 1) 4.5 0]
[1 1 0.90
[1 1 2] 0.94
breaking individual bonds between the atoms. These mechanisms are now accessible to atomistic sim-
ulation, either by means of empirical potentials or through ab-initio quantum mechanical calculations
[23,168–170]. Yoo et al. [168] determined the ideal cleavage strength for different intermetallic com-
pounds. The data obtained for TiAl and Ti3Al are listed in Table 5. The ideal cleavage energy (Griffith
energy) Gc was defined as the total surface energy Cs of the two cleaved planes, i.e., Gc ¼ 2Cs . From the
Gc values, the critical stress intensity factors K 1c for a mode-I crack, designated also as fracture tough-
ness, can be calculated, according to standard fracture mechanics. The values obtained are listed in
Table 5.
Based on the calculated cleavage energies it may be expected that the ð1 1 1Þc and ð1 0 0Þc planes are
the cleavage planes of c(TiAl). Panova and Farkas [171] have essentially confirmed these findings by
atomistic modeling. The anisotropy among the Gc values can be attributed to the atomic composition
of the cleavage planes [172]. Experimental evidence of cleavage-like fracture in c(TiAl)-based alloys
was obtained by transmission electron microscope observations on crack tips [173–175]. The investi-
gations were performed on cracks that originated from perforations of thin TEM foils; thus the cracked
sample geometry was not defined. On the other hand the concept provides information not accessible
with other techniques. The cracks follow the f1 1 1gc planes at the atomic scale according to crystal-
lography, even after deflection at interfaces; thus giving strong evidence of cleavage fracture on the
close-packed, i.e. widely spaced planes. It should be noted that the low toughness values listed in
Table 5 for the c-phase have been confirmed by fracture mechanical tests [176–179].
A similar picture holds for a2 (Ti3Al) as indicated by the low cleavage energy of the ð0 0 0 1Þa2 plane.
Fracture toughness tests performed by Umakoshi et al. [180] on Ti3Al single crystals have, indeed,
shown that the material is prone to cleavage fracture on the ð0 0 0 1Þa2 basal and 1 0 1 2
a2 pyramidal
planes. The authors noted that brittle fracture never occurred on the prism planes, which is reasonable
because Ti3Al single crystals are very ductile when pure prismatic slip is activated. Research performed
by Yakovenkova et al. [181] on Ti3Al single crystals revealed cleavage on the ð0 0 0 1Þa2 basal plane and
1g , f1 0 1
on f1 0 1 2g and f1 0 1 3g pyramidal planes, when the samples were oriented for basal slip.
a2 a2 a2
As the f1 1 1gc cleavage plane coincides with the dislocation slip and twin habit planes, blocked slip
or twinning can easily nucleate cracks. For example, when translamellar shear is blocked at an a2 -
lamellae, shear cracks could be produced in the c- and a2 -lamellae or along the a2 =c interface. Simi-
larly, dislocation pile ups can produce stress concentrations that affect crack propagation in various
different ways. For example, edge dislocation pile ups would produce stress concentrations that would
tend to push a pure mode-II crack, as the piled up edge dislocations are confined to their slip plane.
Edge dislocation pile ups may also coalesce into a crack nucleus, as early proposed by Zener [182].
Screw dislocation pile ups would promote a pure mode-III crack [183]. Once nucleated, cleavage
cracks may grow extremely fast to its critical dimension unless no other toughening mechanism is
available.
Information about crack tip plasticity in c(TiAl)-based alloys is rare because most investigations
have been performed by scanning electron microscopy (SEM), which is not capable of imaging the
98 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
structural details at the appropriate size scale. The few TEM studies available [173,174] have recog-
nized limited crack tip plasticity due to dislocation and twin emission. The shape of the plastic zone
was found to be largely determined by the crack orientation with respect to the few easy slip and
twinning systems available in the L10 structure. Thus the development of crack tip plasticity is largely
governed by the crystallography and slip geometry of the L10 structure. This finding coincides with
atomistic studies of Panova and Farkas [171], according to which the amount of crack tip plasticity
in c(TiAl) strongly depends on the crystallographic orientation of the crack front. For crack propaga-
tion on the (1 1 1) plane, extensive emission of Shockley partial dislocations on the inclined (1 1 1Þ
plane is expected for a crack front parallel to h1 1 0]. For an equivalent crack no significant dislocation
generation was recognized when the crack front was parallel to [1 1 2].
Whether a propagating crack becomes immobilized depends on the competition between cleavage
decohesion, crack deflection, and crack tip shielding or blunting. Stable crack growth requires the plas-
tic zone to keep up with the cleavage crack, which requires continuous dislocation multiplication with
an adequate rate. TEM observations [173] have revealed that the dislocations in the plastic zone mul-
tiply via multiple cross-slip, a process which needs some glide or time to take place. In view of the high
dislocation glide resistance at room temperature (Section 4.1.2.2), it seems to be unlikely that large
plastic zones can solely be formed by dislocation multiplication. In particular, plastic zones consisting
of dislocation assemblies may easily be outrun by fast cracks. However, as recognized by TEM analysis,
the semicoherent lamellar interfaces provide a high density of misfit dislocation, which subjected to
crack tip stresses serve as dislocation sources. Glide of these dislocations can locally shield and blunt a
translamellar crack. This could be another factor that contributes to the relatively high fracture resis-
tance of lamellar alloys.
For crack tip plasticity, the activation of mechanical twinning might be more important because the
growth rate of twins is often a significant fraction of the elastic shear wave velocity. Fig. 26 demon-
strates in some detail mechanical twinning in front of a crack tip [175]. The crack originated from
the perforation of the thin foil, was probably subjected to mixed mode loading, which involved a sig-
nificant mode-II shear component. The overall shape of the crack indicated crack deflection at the c=c-
lamellae interfaces. The crack became immobilized at position (1) within a c grain. Fig. 27 shows the
atomic structure at the very front of the crack tip region. In response to the crack stress field, a narrow
twin is formed in front of the crack tip. The width of the twin gradually becomes narrower with
increasing distance from the crack tip, as may be expected from the decay of the crack stress field. This
suggests that nucleation of the twin took place sequentially as the crack advanced. Twinning seems to
have preceded crack propagation as indicated by the twin configurations seen near to the crack wake.
The crack wakes are facetted, which may indicate another mechanism of energy absorption that
occurred during crack propagation. Micromechanical modeling of the cleavage modes in c(TiAl)
[184] has shown that the shear stress acting on a [1 1 0](1 1 1) mode-II crack can be shielded by
Fig. 26. Association between fracture and mechanical twinning in a duplex Ti–46.5Al–4(Cr, Nb, Ta, B)(at.%) alloy. Crack
propagation along f1 1 1gc planes and deflection at lamellar interfaces. The crack was immobilized at position (1). Details in
front of the crack are shown in Fig. 27. Figure taken from [3], permission from Wiley-VCH Verlag GmbH & CoKGaA.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 99
Fig. 27. High-resolution image of detail 2 in Fig. 26. (a) Twin formation ahead of the crack tip. (b) Higher magnification of the
arrowed region 3 in (a). Figure taken from [3], permission from Wiley-VCH Verlag GmbH & CoKGaA.
1/6h1 1 2](1 1 1) twins. However, unstable crack propagation is expected, when a mode-I component is
superimposed. Generally, mixed-mode crack loading tends to concentrate crack tip plasticity onto the
(1 1 1) plane. Yoo et al. [47] have suggested that this strain localization may support translamellar frac-
ture along {1 1 1} planes across all types of c=c-lamellae interfaces.
There are various interactions between propagating cracks and pre-existing dislocations, which
could provide some potential to toughen the material. For example, crack deflection at dislocations
was recognized at the atomic scale [3,173]. Such a mechanism makes the crack path more tortuous
and is expected to enhance the toughness. Apart from this interaction between the crack tip and dis-
location stress fields, additional energy dissipation might be expected when a crack intersects screw
dislocations. At each screw dislocation the fracture surface receives a step whose height equals the
Burgers vector of the dislocation. The formation of such steps requires a substantial surface energy,
which could stabilize crack growth.
Based on these considerations an experiment was designed in order to assess the effect of deforma-
tion induced defect structures on the crack resistance [175]. Pre-deformation by compression clearly
increases the resistance of the material against crack propagation, which is manifested by higher frac-
ture toughness and larger work of fracture. The full potential of this treatment remains still to be
confirmed.
At the mesoscopic scale, the path of the crack during fracture is usually controlled by the
microstructure, the dominant factors being lamellar colonies, equiaxed grains, grain size, and textur-
ing due to mechanical working. The effects are well documented in several reviews [3,164–168] and
need not to be repeated here.
An interesting problem that appeared recently is the fracture behaviour of the c-Md-alloys with a
modulated microstructure (see Section 2.3) [44]. The modulation remarkably affects crack propaga-
tion as illustrated in Fig. 28. The crack is deflected at a bundle of modulated lamellae. The modulated
laths obviously formed crack-bridging ligaments; one of them is seen to be disrupted. Thus, the crack
path becomes microscopically very tortuous with the production of highly distorted material along
the fracture surface. Finally, the crack is arrested at another modulated lath. This process is apparently
accompanied by crack tip plasticity, as indicated by twin nucleation. These features contrast with the
observations made on conventional duplex or lamellar alloys of the 3rd generation, which revealed
pronounced cleavage fracture on f1 1 1gc planes. Thus, the modulated laths seem to toughen the mate-
rial in a fashion that is reminiscent to that of ductile reinforcement in a brittle material. This point of
100 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 28. Transmission electron micrograph of a crack C in a TiAl alloy (c-Md) containing modulated laths (T). Note the crack
deflection at a bundle of modulated lamellae, the blunting of the crack tip and the emission of a mechanical twin (arrowed).
Figure taken from [3], permission from Wiley-VCH Verlag GmbH & CoKGaA.
view is confirmed by fracture tests that were performed to determine the fracture toughness. Fig. 29a
demonstrates a load–deflection trace of such a test. The characteristic features are pop-in events,
which apparently reflect the different crack growth resistances of the individual constituents. These
sudden drops of the bending force might be related to intralamellar plate failure when the crack prop-
agates through regions that are almost free of modulated lamellae. The toughness values determined
on extruded material of the c-Md-alloy are significantly higher than those determined for conven-
tional duplex alloys. A unique feature of the c-Md-alloy is the existence of a maximum in the tough-
ness/temperature profile (Fig. 29b), which could be explained as follows. The B19-phase present in the
modulated laths can apparently further transform into the c-phase, as has been discussed in [44]. This
transformation was frequently observed in deformed samples. Consequently, it might be speculated
that the process is stress-induced and also occurs at crack tips. The transformation is expected to
be thermally activated, which could be the reason for the toughness maximum. Such transformations
have long been known to take place in metals (austenite ? martensite in ausformed steels [185]) and
ceramics (partially stabilized zirconia [186]) and, likewise, may toughen the c-Md-alloy. The problem,
although demanding, could be a subject for modeling.
Methods based on elasticity theory have been very successful in describing defect structures with
long-range stresses, e.g., of dislocations or of incoherent twin boundaries. However, realistic materials
involve relevant structural parameters spanning a wide length scale which cannot be covered solely
by standard elasticity theory. For example, the mechanical properties of lamellar TiAl alloys are deter-
mined by several length parameters, ranging from the nanometer scale (lamellar spacing) up to about
1 mm (colony size). The case studies described in the following sections may illustrate the challenges
associated with the modeling of such a multiscale material.
Fig. 29. Fracture behaviour of the c-Md alloy. (a) Load against deflection trace from a three-point bending test performed at
room temperature on a chevron-notched specimen of the c-Md alloy, in order to determine the fracture toughness. Extruded
material, crack propagation perpendicular to the extrusion direction. Note the pop-in event occurring upon loading (arrowed).
(b) Dependence of the fracture toughness on test temperature of the extruded alloys c-Md and Ti–45Al–8Nb–0.2C (TNB-V2,
comp.in at.%) for crack propagation perpendicular to the extrusion direction. Figure taken from [3], permission from Wiley-VCH
Verlag GmbH & CoKGaA.
sa unit vector in slip direction, ma unit vector normal to the slip plane, as
sa ¼ Pa : r: ð18:2Þ
For the shear strain rate c_ a the term
102 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
s 1=m
c_ a ¼ c_ a;ref signðsa Þ a ð18:3Þ
scr;a
is introduced with c_ a;ref being a reference shear rate and scr;a the actual CRSS. The ratio sa =scr;a , together
with the exponent 1=m, represents the shear strain rate sensitivity of the material. For m tending to
zero, 1=m ! 1, the shear strain obtains the value ca;ref signðsa Þ showing no rate sensitivity.
The CRSS is assumed to evolve as
X XZ
s_ cr;a ¼ hab ðcÞ c_ b ; c¼ jc_ a jd~t; t ¼ 0 : c ¼ 0: ð18:4Þ
b a t
The hardening functions hab can be found by adaption to experiments; the summation (indices a; bÞ
is performed over all slip systems in a representative volume element, e.g., a grain; for details the
reader is referred to [187,188] and corresponding literature. As discussed in Sections 3.2.2 and 3.2.6
also superdislocations may be activated to satisfy the von Mises criterion. This kind of slip systems
(max 8 are possible) can be activated, however, in c-lamellae with a significantly higher CRSS than
the CRSS for ordinary slip systems (where only 4 exist). The continuum-mechanical framework for
superdislocations is the same as that outlined above. As already mentioned, mostly twinning is acti-
vated, at least one twin system (of 4 possible twin systems), to fulfil the von-Mises criterion enforcing
5 independent slip systems. The reader may also consult Figs. 7 and 8 which clearly show the individ-
ual slip and twin systems and the according Burgers vectors and displacements, respectively. Our
group formulated an incremental algorithm similar to that for crystal plasticity for twinning, see
[188–190], as following:
pffiffiffi
The twinning shear g has the value 2=2 (a rather large value which should be met in several cases
by a large strain setting). To a twin system b with the twin (volume) fraction f b an average twinning
shear g b ¼ f b g is addressed. For the shear strain rate g_ b the term
sb 1=n
g_ b ¼ g f_ b with f_ b ¼ f_ b;ref ð18:5Þ
sT;cr;b
is introduced, with f_ b;ref being a reference twin fraction rate and sT;cr;b the quasi-critical twinning shear
stress (or, in other words, the shear stress necessary to the formation of a twin). Again, the ratio of sb
(the Schmid stress, see Eqs. (18.1) and (18.2) to sT;cr;b , together with the exponent 1=n, represents the
twin rate sensitivity of the material. For n tending to zero, 1=n ! 1, the twin fraction obtains a fixed
value f b;ref . As outlined in Section 4.4.2 of this paper, atomistic studies have shown that sT;cr;b can be
assumed as fixed value, so no hardening rule is applied. Finally, it should be emphasized that twinning
is unidirectional, i.e., if sb 6 0, then f_ b ¼ 0.
Finally, the total rate e_ in of the irreversible (plastic) deformation in a representative volume ele-
ment (e.g., a grain) follows as
!
X X X
e_ in ¼ 1 fb c_ a Pa þ cT f_ b Pb : ð18:6Þ
b a b
This type of a constitutive law for an elastic–plastic lamellar microstructure has found many appli-
cations, e.g., for modeling of PST crystals; relevant results are reported in the next section.
experimental working as well as modeling the PST microstructure is an ideal selection allowing com-
fortable meshing with finite elements. In this section only modeling concepts, published in the last
two decades, are shortly discussed in chronological order. Obviously the first group, which modelled
the stress–strain curve for a PST single crystal for various orientations of the lamellae as well as poly-
crystals, was Kad and co-workers, see the papers [191–195] in the years 1995–1997. In their pioneer-
ing work they considered four ordinary dislocation slip systems and eight superdislocation slip
systems in the c-phase with several sets of CRSS-values, taking into account the Hall–Petch effect,
see Eq. (3) and the according text with respect to the application of this effect in a lamellar structure.
However, Kad et al. did not take into account the role of the twin systems. With respect to the a2 -
phase all three types of slip systems were taken into account by Kad et al. see Section 3.2.3. Due to
computational restrictions they worked with a two-dimensional idealization, replacing a three-
dimensional slip system by three in-plane slip vectors. It should be mentioned that Kad investigated
also polycrystals as an ensemble of PST crystals, however, also in a two-dimensional setting. Single
PST crystals were also a subject of modeling by the Leoben group, see [188,196] in 1997. This group
has been obviously the first one, which incorporated also the twin systems as outlined in Sections 3.1
and 3.2.6 in a three-dimensional model. In an adjunct work Schlögl et al. [132] presented already yield
Fig. 30. Stress strain curves of a Ti-46.5 at.% Al-4 at.% (Cr, Nb, Ta, B) alloy with near-c microstructure obtained by uniaxial
compression tests applying three different strain rates. (a) Experimental data and (b) predicted by the simulation. Reprinted
from International Journal of Plasticity, 19, Marketz, W.T., Fischer, F.D. & Clemens, H., Deformation mechanisms in TiAl
intermetallics-experiments and modeling, 281–321, 2003 with permission from Elsevier.
104 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 31. Comparison of RMS-voltage in AE measurements applying three different compressive strain rates. The observed
maxima lie in a zone of low strain but a strong deviation from the elastic behaviour in the stress–strain curves. Reprinted from
International Journal of Plasticity, 19, Marketz, W.T., Fischer, F.D. & Clemens, H., Deformation mechanisms in TiAl
intermetallics-experiments and modeling, 281–321, 2003 with permission from Elsevier.
loci for PST crystals. A further remarkable step in modeling PST crystals was done by Uhlenhut,
Lebensohn and co-workers, [197,198] in 1998. These authors studied in detail the slip and twin sys-
tems in the c-phase and introduced a morphological characterization of a sequence of c- and a2 -
lamellae, i.e. layers, distinguishing longitudinal systems (with the slip plane normal vector m parallel
to the layer normal n and no confinement with respect to dislocation motion), mixed systems (with a
confinement, e.g., a domain boundary) and transversal systems (with the slip plane normal vector m
not parallel to the layer normal n). This kind of characterization can be of some practical use. Some
years later the Leoben group took two steps forward in their works [189,199] in 2003. They used a
fully three-dimensional model consisting of 64 c-grains and the a2 -phase in the amount of 5% (to
10%) volume fraction randomly located at the grain boundaries and triple joints. First, they studied
the development of twins by comparison with acoustic emission measurements, concluding that
twinning is the major deformation mechanism for a global strain up to approx. 2% and that the a2 -
phase plays a rather marginal role. The reader is provided with both an experimental and predicted
(by simulation) stress–strain diagram, which shows, by proper adaption of the material properties,
a good agreement, see Fig. 30a and b. A continuous acoustic emission (AE) monitoring was accom-
plished with a root mean square (RMS) voltmeter. Fig. 31 reproduces the RMS-voltage measurement,
which shows its maximum exactly in a zone of low strain but strong deviation from the elastic beha-
viour in the stress–strain diagram and can be interpreted as the first indication for a relationship
between AE and twinning, for details see Section 2.4 of [189]. Second, they studied (as obviously
the first ones) the (power-law-) creep behaviour of a designed fully lamellar c(TiAl)-alloy (i.e. an
ensemble of PST crystals) and compared it with experiments at 700 °C and 800 °C. They could show
that the secondary creep rate decreases significantly with decreasing lamellae interface spacing. As
further remarkable contribution the work by Brockman [200], which appeared also in the same jour-
nal (IJP) and the same year 2003 as [199], shall be considered in more detail to illustrate the status of
such modeling concepts. Brockman [200] worked, more or less, with the same concept for crystal
plasticity as used in the previously mentioned papers, but with a new concept for the shear strain rate
c_ a (compared to the viscoplastic-concept, see Eq. (18.3)) by engaging of a yield condition defining the
plastic strain rate in the sense of rate-independent plasticity. The twin systems however, were not
explicitly dealt with (in contradiction to [199]). The investigation involved an array of 512 randomly
oriented grains. The volume fraction of lamellar colonies was 0.96, with an assumed ratio of a2 and c
lamellae of 1:5. The width of the colonies was about 145 lm. The remaining volume consisted of c
grains filling the space between the colonies. These structural parameters are representative for nearly
lamellar alloys, which combine high creep resistance and toughness with reasonable room
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 105
temperature ductility. The model predicted the smooth load–elongation response in the micro-yield
region of the polycrystal due to localized flow in favourably oriented lamellae, which has often been
recognized in compression or tensile test. The modeling revealed a remarkable heterogeneity in the
deformed state of the polycrystal. For example, at a nominal macroscopic strain of e ¼ 0:2%, about
5% of the grains remain elastic, while in 80% of the colonies the plastic strain is significantly smaller
than 1%. Close to the yield point, at e ¼ 1% strain, nearly all colonies have yielded. However, the defor-
mation is still inhomogeneous because 28% of the grains exhibit a plastic strain of less than 1%. Colo-
nies that contain undeformed material usually have their lamella planes nearly perpendicular to the
compression axis. The heterogeneity of straining is accompanied by an adequate variation of the inter-
nal stress state. It should be noted that it is extremely difficult, if not impossible, to gather such infor-
mation experimentally. The results can certainly be generalized, in order to rationalize other
deformation phenomena occurring in other multiphase alloys. For example, the evolution of high
internal stresses upon deformation may also explain the deformation effects that occur after strain
reversal, which were described in Section 4.1.4. Three years later (2006) a further modeling paper
by Werwer and Cornec [201] appeared.Theseauthors modelled also single PST crystals and assigned
an increasing role to superdislocations by assuming a relatively low yield stress for superslip when
compared with ordinary slip. Of course, it is possible to adapt the remarkable set of input data in such
a way that a certain (usually reported as very good) agreement with experiments can be obtained,
although, according to the knowledge and practical experience of the authors, the activation of
superdislocations can scarcely, or even not, been observed by inspecting deformed specimens, see also
the discussion of the role of superdislocations in Section 3.2.6. Bieler et al. [202] published in 2009 in
Int. J. Plasticity a crystal plasticity modeling paper, where dislocation slip and twinning are described
in an equivalent way as by the Leoben group [189,199] with a new hardening law for scr and sT;cr . The
main goal of this paper was, however, the understanding of formation of microcracks at grain bound-
aries in Ti–Al-alloys. Finally, the very extensive work on TiAl modeling by Zambaldi [64,203] in the
years 2010 and 2011 shall be appreciated, reviewing all the modeling efforts starting with 1993 up
to 2010. Compared to the previous literature, similar steps as in the past were done in [64,203].
The same (quasi-)viscoplastic formulation for crystal plasticity is used as outlined in Section 4.2.1.
Twin systems are interpreted as unidirectional slip systems equivalent to those due to dislocations.
It should also be mentioned that Zambaldi et al. [64,204], studied the indentation in a c(TiAl) single
crystal experimentally as well as performed a detail simulation with the above mentioned crystal
plasticity concept. This work can be considered as a remarkable progress of modeling practical appli-
cations of a Ti–Al-alloy.
In closing this section it might be summarized that extensive modeling, combined with structural
and mechanical characterization, have significantly improved our knowledge about the deformation
behaviour of two-phase titanium aluminide alloys based on c(TiAl) and a2 (Ti3Al). The simulations con-
sider all the potential slip and twinning systems of the majority c(TiAl)-phase, which carries most of the
deformation. It is now possible to predict the mechanical anisotropy of lamellar structures and to
describe the mechanical response of alloys with technically relevant microstructures. The models
explain why these alloys deform very inhomogeneously and provide a qualitative description of the
heterogeneity of the deformed state, which experimentally is difficult to assess. The details hinge a
bit on the assumptions made about the relative magnitude of the critical resolved shear stresses that
are relevant for the individual dislocation and twinning systems. However, it is difficult to design
experiments clearly answering this question. As well known, methods based on elasticity theory
diverge close to the dislocation core region and, thus, cannot be used to describe the strain and stress
field there. For overcoming this shortcoming, quasi-continuum methods were developed, in which ato-
mistic calculations are applied to small regions with large atomic displacements. A continuum approx-
imation is adopted where atomic displacements are small, assuming that the continuum description
provides almost the same result as a full atomistic simulation. These ‘hybrid’ models have been suc-
cessfully applied for determining various defect properties in c(TiAl)-based alloys, involving solid solu-
tion effects, the site occupancy of alloying elements or details of non-planar dislocation core spreading.
In particular, it has been demonstrated that the forces produced by the dislocation core and their cou-
pling to the external stress state can dramatically affect the flow behaviour; for reviews see [23,144].
106 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
With help of such models, further information might be gathered about the Peierls stress for dif-
ferent shear processes. As already outlined in Section 4.2.1, deformation twining plays an important
role in plasticity of c(TiAl)-based alloys. Particularly the mechanisms controlling twin boundary
motion are of immediate relevance. Here we refer to Section 4.4.2, where the immediate interaction
of continuum modeling and atomistic calculations is already practiced. There is also a remarkable
uncertainty about the effect of other constituents, such as b/B2, which in modern alloys are imple-
mented for extending processing windows. However, the presence of these b.c.c.-phases seems to
be harmful for the low temperature ductility and toughness and high-temperature creep resistance.
Understanding of the deformation behaviour of these phases as constituents of a multi-phase com-
pound is still at its infancy and needs further investigation by both experimental and theoretical
investigations.
traverse the whole work piece, often resulting in premature failure. Understanding of this phe-
nomenon is thus important for optimizing hot working procedures.
A previous experimental study [209] has shown that the recrystallization behaviour of the lamellar
morphology is very anisotropic with respect to the loading direction. The volume of recrystallized
grains is highest if the loading direction is parallel to the lamellar planes, intermediate if the loading
direction is perpendicular to the lamellar planes, and smallest if the loading direction is at 45° to the
lamellar planes. Thus, the recrystallization behaviour is obviously triggered by heterogeneities in the
deformed state of the lamellar morphology. These findings have motivated the following modeling
studies on the mechanical stability of lamellar specimens. Particular emphasis was paid on the beha-
viour of lamellar grains under compression parallel to the lamellar planes, which apparently easily
recrystallize. Deformation in this orientation apparently involves an instability, which is manifested
by kinking of the lamellae and is reminiscent of buckling of load-carrying structures.
In the modeling procedure, one has to deal with compression of elastic lamellae (a2 (Ti3Al), volume
fraction 0.10–0.15), embedded in a rather soft matrix c(TiAl) with a shear modulus Gm and a yield
stress sm in longitudinal shear. If a compressive load acts on the lamellae package ða2 c a2 . . .Þ
in the direction of the lamellae, then a stability loss may occur with a kink appearing in the lamellae
package. Mathematically, this phenomenon can be described by an eigenvalue problem with an eigen-
mode having an infinite wave length, so it can only be described by two parallel straight lines con-
nected to the left and right end of a sigmoidal kink. Furthermore, the kinks in the individual c-
lamellae are located in a kink band across the ensemble of lamellae, which is inclined by a certain
angle to the longitudinal (i.e. lamellae) direction, which may change with ongoing loading. A theoret-
ical solution concept for the problem still exists, however, with simplifying assumptions. In reality, we
have an ensemble of grains interacting with that grain, whose lamellae orientation coincides with the
loading direction. Then a rather sophisticated numerical modeling concept must be installed, which is
described in [210] and in detail in [211]. Of course, not the ‘‘classical” eigenvalue problem is solved for
such a setting but a ‘‘stress-controlled” problem, starting from imperfections of the material as well as
of the topology (as initial shape deviations of the lamellae). At least a quantitative picture of reality
can be offered by such a model. Concluding and repeating from [206], one has to keep in mind that
two different types of nonlinear deformation must be distinguished, the ‘‘material” and the ‘‘struc-
tural” deformation. The material deformation is simply the deformation under compression as in clas-
sical mechanics of materials. The structural deformation consists of two deformation modes. The first
deformation mode can be considered as the a/m instability mode, i.e., the above discussed kinking,
which belongs to a deformation mode with an infinite wave length. Such an eigenmode showing kinks
occurs at a rather high value of the compressive longitudinal stress which lies in the range of the shear
modulus, if an elastic body is considered, and in the order of magnitude of the initial shear stress for
yielding in elastic–plastic bodies. Imperfections like geometrical faults of the layered structure inside
the colonies as well as grain boundaries trigger a deformation state tending to the kinking. In other
words, any imperfection (as in the case of a spatially varying geometry) transfers the eigenvalue
problem to a so-called ‘‘stress-controlled” problem enforcing a large strain setting as continuum-
mechanical description. The strongest effect occurs for imperfections being affine to the first buckling
eigenmode. As second deformation mode the kinks of the individual lamellae arrange in typical kink
bands which look like, with respect to their shape, as shear bands. Therefore, it is natural that
imperfections favour initially kink type deformation patterns, which evolve into band-like shear con-
centration zones. It should be mentioned that such a phenomenon is also well known in geology of
layered rocks.
The modeling has revealed that the conversion of the lamellar microstructure is a highly localized
process, as has been observed experimentally. TEM observations have shown that the tendency to
instable lamellae buckling increases, if there is an inhomogeneous distribution of a2 - and
c-lamellae within the colonies. The process starts with local bending. From the curvature and thick-
ness of the lamellae local strains can be produced, which are often larger than 10%. Fig. 32 shows
an initial stage of lamellae kinking. The mechanism is associated with the formation of dense disloca-
tion structures and of sub-boundaries, as demonstrated in some detail in Fig. 33. Subsequent rotation
and coalescence of these sub-grains occur apparently in such a way that kinking of the lamellae is
accomplished.
108 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
Fig. 32. Kinked lamellae in Ti–47Al–5Nb–0.2B–0.2C at.% (TNB-V5). The lamellar ingot was subjected at 1270 °C to a
compression stress r = (25.4 ± 12.7) MPa fluctuating with 30 Hz. The symbol C indicates the compression axis, which is parallel
to the image plane. The dislocations involved in sub-boundaries and mentioned in the text are imaged end-on and thus appear
as contrast dots. Figure from [212], changed.
Fig. 33. Atomic structure of one of the sub-boundaries (marked with 1) in Fig. 32 formed at a kinked lamella. (a) mixed ordinary
dislocations situated in the sub-boundary and (b) one of the dislocations shown in higher magnification. The compressed image
below shows the extra plane of the dislocation more clearly. Figure from [212], changed.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 109
Due to kinking, the lamellae are reoriented with respect to the deformation axis, which may sup-
port shearing along the lamellar interfaces. In the regions of highest local bending, spheroidization and
dissolution of the a2 -phase occur, combined with dynamic recrystallization. This suggests that both
the non-equilibrium constitution and the local stress state provide the driving thermodynamic force
for the observed phase transformation and recrystallization. Indeed, nucleation of nanometer size
grains has frequently been observed at kinked lamellae [208]. These factors apparently support the
formation of shear bands, which consist of extremely fine grains. More experimental details about
these processes are provided in another review [212].
This combination of modeling combined with TEM analysis has certainly extended the basic
knowledge about the recrystallization of lamellar microstructures under hot working conditions.
The findings have also motivated the development of novel hot-working procedures for overcoming
the problems associated with the morphological anisotropy of the recrystallization in lamellar alloys.
For example, it has been demonstrated that the superposition of different shear process, as achieved
by simultaneous torsion and compression, leads to a more homogeneous microstructure [212]. How-
ever, for industrial practice, there is a need for quantitative models, which are capable to predict the
effect of processing parameters on the material to be produced, in order to reduce the amount of
expensive and time consuming parametric studies. It must be admitted that, because of the complex-
ity of the processes involved in hot working, we are far from this goal. In particular, the effect of the
alloy constitution and starting microstructure on the hot working behaviour has not been quantita-
tively considered. Also the spurious influences of alloy element segregation, work piece size, friction
between the work piece and the dies as well as hot-corrosion need to be considered in the future.
A coordinate system in the parent (a2 -phase) is implemented with the 1- and 2-axis in the
0
directions 1 0 1
a and 1 2 1 0 a , respectively, and the 3-axis normal to the base plane as
2 2
½0 0 0 1a2 . According to the Blackburn orientation relations, Eq. (2), the 1-axis corresponds to
112 , the 2-axis to 1 1 0 and the 3-axis to ½1 1 1 .
c c c
110 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
A transformation (strain) tensor eT (using small strain setting) is introduced. The parent phase (i.e.
a2 Þ must be subjected to eT to become compatible with the undistorted product phase (i.e. cÞ. Two
physical processes must be reflected by eT , namely a shearing process transferring the stacking
sequence of the parent phase to the product phase and a diffusion process adapting the stoichiom-
etry ða2 ) Ti : Al ¼ 3 : 1; c ) Ti : Al ¼ 1Þ. Details on the formulation of eT can be taken from
[216], Section 4 and the Appendix there. The following non-zero components of eT write with
the quantities u; v ; w; s as
rffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 1 2 2 pffiffiffi ac 2 ac cc 1
u¼ ac þ 2 cc ; v¼ 2 ; w¼ ; us ¼
pffiffiffi ; ð19Þ
3 aa2 aa2 ca2 ac 2 þ 2 cc 2 8
The above mentioned diagonal terms are premultiplied by ð1 Dath T s Þ, where Dath T s ¼ 3:103
represents the thermal mismatch between both phases at T s ¼ 750 C.
Now the transforming domain is defined with the length l in the 1-direction and the height h in the
3-direction. The extension of the domain in the 2-direction is assumed over a rather large distance
compared to l. Due to the constraint by the parent phase a plane strain configuration can be
assumed as reasonable. The edge ð0 6 g 6 hÞ in the 3-direction with the height h at n ¼ l takes
the role of the moving interface with an average normal velocity V n .
As chemical driving force F chem the difference of the specific Gibbs energy can be taken in the cur-
rent case as Ga2 Gc ¼ DGchem with a value of 480 M J=m3 , which is equivalent to 480 MPa.
The mechanical driving force F mech follows, according to Eshelby (for references and details, see
again [215]), in a material point at the interface between the parent and product phase as
F mech ¼ U 0a2 U 0c ra2 þ rc : ea2 ec =2: ð20:1Þ
The quantities U 0a2 ; U 0c stand for the specific elastic strain energy U 0 on the right and left side, resp.,
of an interface point along 0 6 g 6 h; the quantities ra2 ; rc stand for the stress tensor r and ea2 ; ec
for the strain tensor e, both in the phases a2 and c. Since the velocity V n of the moving edge is
assumed to be uniform, F mech is replaced by on average value Fmech ðh=lÞ which is calculated
numerically.
As further energy term, stored in the system during the formation of the product phase, the inter-
face energy C must be considered; values of C can be found, e.g., in [216,217]. For l h, we have
for the total interface energy 2lC and for its specific value as 2lC=hl ¼ 2C=h.
Both the motion of the interface itself and the local diffusive process, which is necessary to assure
the stoichiometry, are related to two dissipative processes. The corresponding thermodynamic
force is denoted as F c .
The total (average) value F in an interface point ð0 6 g 6 hÞ follows now as
2C
F ¼ DGchem þ Fmech ðh=lÞ þ Fc ; ð20:2Þ
h
V n ¼ MF; D ¼ V n F ¼ MF 2 ; ð21Þ
with the interface mobility M > 0 as material constant, which may increase with an increasing
temperature due to an increasing diffusivity of the atoms. Of course, the according dissipation is
always positive.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 111
The ratio h=l has been a given quantity up to now. However, we can engage again the argument of
maximum dissipation, as in the case of modeling the formation of deformation twins, see
Section 4.4.2 below, to find to an optimal thickness hopt (subscript ‘‘opt” for optimal) of the
c-lamellae. The thermodynamic force F can be rewritten as
2C l
F ¼ ðDGmech F c Þ Fmech ðh=lÞ þ > 0: ð20:3Þ
l h
Obviously the ‘‘driving” (thermodynamic) force (contribution) is DGchem . The ‘‘dragging” contribu-
tion is F c þ Fmech ðh=lÞ þ 2lC hl . The force term in the brackets is the mechanical energy stored dur-
ing the transformation. If one follows the concept of maximum dissipation, then this term has to be
minimized with respect to h=l, see also the short discussion in context with twinning, Section 4.4.2,
below Eq. (41), and leads to a relation for hopt as
sffiffiffiffiffiffiffi sffiffiffiffiffiffiffi
1 Cl hopt 1 C 1
hopt ¼ ; ¼ pffi : ð22Þ
cT;13 2K~ l cT;13 ~
2K l
The quantity K ~ ¼ Fmech 4c2 can be calculated numerically as done in [213], where a ratio of
T;13
3 pffiffiffiffi
hopt =l ¼ 0:473 10 = j was found l ¼ j 104 m , which is in good agreement with experimen-
tal observations.
To summarize, the current modeling concept considers the simultaneous operation of the chemical
and mechanical driving forces that are involved in the a=a2 ! c-phase transformation of Ti–Al alloys.
This accounts for the fact that during the transformation both the stacking sequence and the chemical
composition have to be changed. The model describes the lattice mismatch by a homogeneous eigen-
strain, whose accommodation leads to coherent interfaces but introduces lattice distortions and con-
sequently eigenstresses known as coherency stresses. This is obviously a reasonable simplification of
the experimental observation that, apart from very thin lamellae, the lattice mismatch is concentrated
in disconnections, meaning a defect that has both dislocation and step character. In the framework of
this consideration, the transformation is seen as diffusion-controlled step migration.
Finally it should be mentioned that the current model shows the way to a kinetic law for the
aða2 Þ ! c phase transformation. The alloy performance depends strongly on the kinetics of this
transformation process. However, experimental work is still necessary to obtain proper estimations
for the transformation mobility constant M, see Eq. (21).
in structure to the product phase occurs without a bulk composition change and by short-range
atomic jumps across the parent/massive boundary. Thus, the cooling conditions for optimizing the for-
mation of massively transformed c-phase are very complex.
Dey et al. [213] have considered ‘‘Gamma massive grains” as the forth possible development mech-
anism in case of a aða2 Þ ! c transformation in Ti–Al, occurring by quenching from approx. 1200 °C.
First a tiny (nucleus) cm -grain forms at the grain boundary between two a-grains according to the
Blackburn orientation relations, Eq. (2), with respect to one of the both a-grains. Then a growth mech-
anism occurs generating further cm -grains in twin relation with the former generation. Several opin-
ions exist still now to explain the massive transformation process, see [219], the Introduction and the
References there. Obviously, local diffusion of atoms inside the migrating interface, which extends
now over several atom distances and the formation of a concentration spike in front of the advancing
interface occur (as argued by Hillert three decades ago), see details in [219]. Since, as mentioned
above, a massive phase transformation, combined with a subsequent heat treatment, may remarkably
refine initially coarse-grained c(TiAl)-based alloys microstructures, see, e.g., [220,221], the under-
standing of massive transformation of c(TiAl)-based alloys has become a topic of technological rele-
vance. An extensive research project, both experimental and theoretical, was performed by the
Leoben group, whose results are reported in [219]. Some of the key results of this project are reported
below. However, it should be stated that modeling has been very demanding. Ti–Al–Nb-alloy was
experimentally investigated, which made it necessary to provide the chemical potentials for all three
substitutional elements in both ða2 ; cm Þ-phases; the reader is referred to Section 3 of [219]. Modeling
of the kinetics is based on a so-called ‘‘thick-interface model”, which was developed by Svoboda et al.
[222]. This model describes the interface as material zone of finite thickness between the product
phase (on the left) and the parent phase (on the right). The Thermodynamic Extremal Principle, see
[223], is engaged to find to the fluxes and site fraction profiles of all (three) substitutional elements.
Of course, it is necessary, but also very difficult, to obtain proper values for the diffusion coefficients,
see Table 2 in [219]. The current model reproduces remarkable spikes in the Ti site fraction profile,
however, smaller than 0.5 nm, in the parent ðaÞ-phase in front of the migrating interface. Anyway,
it is questionable to accept such a result, at least in a quantitative way, since with a distance of
0.5 nm one leaves the continuum level (for which the above mentioned model is designed). Unfortu-
nately, due to rather unexpected fluctuations in the chemical composition within in the grains, such
spikes could not be experimentally confirmed. It is also likely that these spikes cannot by fully devel-
oped in reality, so diffusion may be (at least partially) suppressed (quasi-diffusionless transformation).
This fact touches also the question about massive transformation kinetics in a two-phase region.
Modeling of the kinetics of massive transformation leads to a single-phase region, i.e. the parent
ðaÞ-phase transforms in a product ðcm Þ-phase, accompanied by a spike of the molar fraction(s) (in
our case Ti) in the parent phase. However, the above mentioned modeling result (with the very narrow
spike) points towards the formation of a two-phase region, where both phases are still in equilibrium.
Obviously, the diffusion process of Ti is too sluggish (forming a too narrow spike). But enough driving
force is available to allow an interface migration, i.e. the onset of a displacive (instead of diffusive)
phase transformation. Insofar, the modeling can be considered as helpful with respect to the under-
standing of a massive transformation in a two-phase region.
An adjunct problem to the massive transformation is the question, if an autocatalytic process is
possible, meaning that the heat released during a ! cm transformation may trigger the diffusion-
controlled formation of further c-laths. A corresponding study was performed in [224], with the con-
clusion that the so-called ‘‘hot spot effect” dies off too quickly to form c-laths. The stress state induced
by the misfit of the cm -phase in a-phase, however, may provide enough driving force to form a lath. A
corresponding study is still open up to now.
Concluding, the current model describes for the first time the kinetics of the massive transformation
in c(TiAl)-based alloys by considering the chemical potentials for the three substitutional elements in
both the a2 and cm -phases. The interface between the transformed and non-transformed region is
described as zone of finite thickness. It has been demonstrated that the massive transformation could
be slow enough to form thin spikes of the alloying element in front of the advancing interface.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 113
Further modeling could be directed on a better understanding of the effects of minor alloying ele-
ments that have a particularly low solubility within the c-phase, whereas their solubility in the a- or
a2 -phase is two orders of magnitude larger. Such elements, like oxygen, are either unavoidable in
c(TiAl)-based alloys, or are added, like boron, in order to optimize the alloy performance. The extreme
difference in the solubility could largely affect the microstructural evolution following the a ! c
massive transformation.
aspect ratio hk for a precipitate k are assigned. As variables act now the quantities qk ; hk and the con-
centrations cki ðk ¼ 0 for the matrix, k ¼ 1; . . . ; m for the precipitates; the components are numbered by
1 6 i 6 nÞ. The Thermodynamic Extremal Principle delivers finally the time evolution of qk and hk , for
a given alloy, known diffusion coefficients and a given set of initial data for qk and hk .
As material system Ti1xAlxN at 800 °C is selected, see [228] for further details. It should be noted
already here that the concept outlined below can be also used to describe the precipitation behaviour
in carbon-containing c(TiAl)-based alloys. All nitrogen is deposited quickly in two kinds of precipitates
ðk ¼ 1; 2Þ with the chemical composition Ti3AlN as cubic phase and Ti2AlN as hexagonal phase, respec-
tively. The volume misfits as well as the Young’s moduli of the precipitates were found by ab initio
calculations. The elastic strain energy terms are found according to the ‘‘Eshelby-concept”, taking into
account the difference of the elastic properties of the matrix and the precipitates. The chemical driving
force DGchem was used as free parameter, since it was not at disposal, and adapted in such a way that
the predicted and observed kinetics are in agreement. The simulations indicate that the aspect ratios
h1 and h2 reach very quickly their stationary values, leading to thin platelets (oblate spheroids) for
Ti2AlN and ‘‘cigar-type” spheroids for Ti3AlN, both in agreement with experimental observation. The
interface energy does not play a significant role as parameter studies have shown. Finally it should
be mentioned that, after h1 and h2 have reached their stationary values, the kinetics of the system
obeys the established parabolic law, meaning that q1 and q2 grow proportional to the square root
of time.
We would like to add a further contribution to this section, namely the decomposition of Ti1x Alx N,
which is often applied as hard protective coating. Depending on the chemistry (marked by ‘‘x”), a
supersaturation with Al provides for x ¼ 0:61 a chemical driving force DGchem of 28 kJ/mol for the
decomposition into TiN and AlN, for details see [229] and the according ab initio investigations there.
During the decomposition, the amount of n growing particles (assumed as spheres), in a unit volume
consists of ð1 xÞnTiN and xnAlN particles to which an interface energy EC with its specific value C is
assigned as
3C
EC ¼ n 4r2 p C ¼ n ð23Þ
r
with the particle radius denoted as r. The quantity n represents the already transformed volume as
n ¼ nð4p=3Þr3 . Furthermore, the formation of TiN and AlN particles is accompanied by a misfit volume
change. The according elastic strain energy, due to the accommodation of this misfit by the matrix and
the particles themselves, can be calculated engaging a mean field setting. This concept was followed in
[229] and leads to a specific elastic strain energy W 0 which can be described as function of x as
W 0 ¼ 4xð1 xÞW 0max ; ð24Þ
with W 0max 3 kJ=mol. This allows formulating the change DG in the total Gibbs energy as
DG ¼ n DGchem W 0 3C=r : ð25Þ
The dissipation function Q occurs due to diffusion of Ti and Al only in the matrix and can directly by
taken from [226]. If one assumes, for sake of simplicity, the same molar volumes of the matrix and the
precipitates as X, then Q follows as
" #
3Rg T x2 ð1 xÞ2
Q ¼ nr_ 2 þ ; ð26Þ
X Dm
Ti Dm
Al
This initial value problem can be solved analytically and allows the prediction of the evolving
microstructure as function of T and x (note that W 0 is a function of x, see Eq. (24), and the material
parameters; DGchem is considered in this relation as a free parameter constant in time).
We repeat the quantities used in Eq. (28) with y0 as site fraction of vacancies and y0;eq as equilib-
rium site fraction of vacancies in a stress free body; rH represents the hydrostatic stress state and is
defined as positive quantity; for Rg T and X see Section 4.3.5. The rate y_ 0 balances with the flux j0 of
vacancies, employing the divergence operator div in the actual configuration, as
y_ 0 ¼ Xdiv ðj0 Þ þ að1 y0 Þ: ð29Þ
The quantity a represents the rate of vacancy generation/annihilation, derived in [230], Section 5
there, as
a ¼ l0 =K X; ð30Þ
with K being a bulk viscosity term depending on the kind of sources/sinks which are active. If one
assumes a divergence-free flux ðdivðj0 Þ ¼ 0Þ between distinct sources and sinks, one obtains by com-
bining the last two equations
y_ 0 a ¼ l0 =K X: ð31Þ
As second step one has to develop expressions for K for different types of sinks, if vacancy
annihilation is studied. Let us keep in mind that we are especially interested to study the role of
moving dislocation jogs which leave behind a trail of vacancies. Here we refer to the according
metalphysical model in [148,231], where constitutive equations for K for vacancy annihilation at dis-
location jogs with constant jog density H as well as with an evolving jog density at Frank loops and for
vacancy annihilation at grain boundaries are derived. With respect to the c(TiAl)-based alloys listed in
[148] the jog density H (expressed in m3) can be assumed to be constant. Then the product K X fol-
lows as
Rg T y0;eq f
KX ¼ ; ð32Þ
2pa H y D ~ eq
0
with f being a geometrical correlation factor (now for f.c.c. lattice f ¼ 0:7815Þ; a the lattice spacing and
~ eq as temperature-dependent average diffusion coefficient Deq ¼ Pn y Dk;eq . The label k; 1 6 k 6 n,
D k¼1 k
refers to the substitutional elements with the site fractions yk and their tracer diffusion coefficients
Dk;eq . Combining Eqs. (28)–(32) allows formulating an initial value problem for y ~0 ¼ y0 =y0;eq with
~
respect to a dimension-free time s ¼ tC; C ¼ 2paHDeq = y f , as 0;eq
~0
dy
~0 rH X=Rg T ;
~0 ln y
¼ y s ¼ 0 : y~0 ¼ y~0;s ; ð33Þ
ds
116 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
An interesting and very compact equation for y ~0 remains, indicating that y0 decreases exponen-
tially (or even stronger) from y0;s to y0;eq , where the exponent itself is also an exponent, namely
exp ðsÞ. Fischer et al. [148] applied Eq. (34) for a negligible role of rH to describe the flow stress r
during the work hardening recovery process as
with r0 as reference stress at the onset of yielding, rWH as flow stress at the start of the work hard-
ening (or, in other words, the flow stress at the start of the recovery process) and exp ð1=a ~Þ as an
adaption factor to describe the flow stress after a long (i.e. infinite) time period of recovery. The time
law Eq. (35) has helped to fit experimental data nearly perfectly, see [148], Section 3.5 there, and the
examples discussed in Section 4.1.3 of this paper.
The modeling results have rationalized for the first time that work hardening structures of c(TiAl)-
based alloys are highly sensitive to annealing treatments at moderately elevated temperatures. This
fact is, at the first glance, surprising in view of the low diffusivity in c(TiAl). The results particularly
explain the extremely fast kinetics of vacancy annihilation, which is reflected in a significant recovery
of the yield stress. These findings are important for the stability of engineering structures, if deformed
load-bearing components are subjected to moderately high service temperature. The easy recovery is
also a concern for recrystallization procedures, as recovery and recrystallization are competing pro-
cesses. Recrystallization is triggered by heterogeneities in deformed state. The early onset of recovery
reduces the mechanical driving forces for recrystallization, which apparently is a significant factor that
impedes a homogeneous recrystallization.
The results are also of general interest because the generation of vacancy supersaturations by jog
dragging is not confined to c(TiAl)-based alloys but certainly occurs in most materials with high stack-
ing fault energy.
process during the formation of a twin, depending on the motion of atoms and, consequently, the
height (thickness) of the twin. They provided values for critical twinning stresses sT;cr for various
alloys and estimates of the interface energy C, i.e. the energy activated along the boundary of a twin
due to atomic mismatch. Wen and Sun [239] followed this concept and reported recently values for
sT;cr for TiAl-based alloys with different Al contents, ranging from 126 MPa to 1420 MPa (which are
rather surprisingly high values!). In the following context we assume that sT;cr is a known quantity.
For sake of completeness it should be mentioned that most recently a Chinese group [240] repeated,
more or less, the studies by Sehitoglu et al. listed above. However, ab-initio calculations are (at least
nowadays) too time-consuming to model actual configurations of twins, e.g., twins embedded in
grains or twins interacting with defects. Therefore, continuum-mechanical models have been
established, which allow the calculation of interaction energy terms and, finally, also the estimation
of the twin height. Of course, continuum-mechanical models cannot be engaged to calculate such
material-immanent quantities as sT;cr ; however, they allow an estimation. The twin itself is modelled
(usually) as a rectangular elastic band (strip) with the height h and the length l, embedded in a
surrounding elastic (or also elastic–plastic) matrix. Same elastic properties (shear modulus l,
Poisson’s ratio mÞ of both objects are assumed. The presence of twinning is, as mentioned above,
described by a twin eigenstrain tensor eT (with eT;12 ¼ eT;21 ¼ g=2, for the calculation of g see Table 3;
all other tensor components are zero) in a local coordinate system (1-longitudinal direction, 2-height
direction). Of course, the onset of twinning is related to a simultaneous onset of an elastic eigenstrain
energy (sometimes ‘‘self strain energy”), denominated as U B , or its specific value U 0B ¼ U B =V T with the
volume V T ¼ lh for unit thickness in 3-direction. Twins were approximated by an oblate spheroid
with a radius RT ¼ l=2 in most of the previous papers to allow for an ‘‘Eshelby-type” solution with
an elastic (eigen)strain energy density being constant in the twin. A finite element study, discussed
in detail in [241], Section 3 there, and an analytical study [242], performed by Fischer et al. have
brought out that such an approximation by an oblate spheroid does not offer a correct estimation
of U B . It is evident that the elastic strain energy density is concentrated to the ends of twin! The fol-
lowing approximate relation for U 0B can be given, see [242], Section 3 there, and [190], Section 2.1
there, as
h K 1 l
U 0B ¼ 4K g 2 ; ¼ ln : ð36Þ
l l 4pð1 v Þ h
It should be mentioned again that the elastic eigenstrain energy can be expressed analytically for a
single twin as well as sequence of twins, for details see [242].
Two approaches are now possible to predict the formation of a (discrete) twin (band). The first one
is the energy criterion of stability of equilibrium, considering a loaded specimen, where a twin sud-
denly forms, occurring when the associated elastic energy release compensates both the formation
of eigenstrain energy and interfacial energy and some intrinsic dissipation, since twinning is consid-
ered as an irreversible process. The kinetic energy, which develops during the (quick) formation of a
twin, is neglected. This approach was worked out by Petryk for several cases of material instabilities
and was applied to the twin formation in TiAl in [243]. The second approach engages the transforma-
tion condition for a displacive (martensitic) transformation (characterized by a transformation eigen-
strain) of a microregion, as worked out simultaneously, but independently, by Fischer and Reisner
[244] and Levitas [245] in 1998, considering now as microregion the twin band and as transformation
eigenstrain the twinning strain eT . The transformation condition writes as
Z
r : eT dV ¼ W T;diss þ W C þ U B ; ð37Þ
T
with r being the total stress in the microregion (twin band region) before the start of the development
of the twin. As already outlined U B is the ‘‘self strain” energy, i.e. the elastic (eigen)strain energy due to
the formation of a twin, see also Eq. (36). The quantity W T;diss is the energy dissipated during the for-
mation of the twin, which can be expressed by the work done by sT;cr ,
W T;diss ¼ V T sT;cr g; ð38Þ
118 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
based on the assumption that sT;cr remains constant during a displacement of the upper longitudinal
edge in the amount of hg in the 1-direction in relation to the lower longitudinal edge of the twin. Of
course, Eq. (38) must be considered as an estimation of the actual dissipated energy. The quantity W C
represents the energy stored in the interfaces and is expressed with l ¼ V T =h as
W C ¼ 2lC ¼ V T 2C=h ð39Þ
with C being the interface energy; for small values of h=l the contribution of the left and right edge to W C
can be neglected. An estimate for C as 0.06 J/m2 can be taken from [67] as well as from the recently pub-
lished data collection [218]. Note also that no change in the chemical energy is assumed in the case of
twinning. The left side of Eq. (37) can be considered as the (thermodynamical) driving force, or in other
words the thermodynamic force necessary to enable the twin formation, and can be reformulated as
Z
r : eT dV ¼ sT gV T ð40Þ
T
with s
T being the average shear stress in the twin band (since eT is spatially constant in the twin band).
Inserting of Eq. (40) into Eq. (37), using of Eqs. (36), (38), (39) and division by gV T yield
sT ¼ sT;cr þ ð2C=lgÞ ðl=hÞ þ 4K g h=l : ð41Þ
The (thermodynamic) dragging force on the right side of Eq. (41) (in the brackets) includes obvi-
ously two contributions which are functions of h=l. Now one can engage the Thermodynamic Extremal
Principle, which, in few words, states that a process develops so that the dissipation obtains a maxi-
mum, see e.g. the overview article [223]. Keeping this fact in mind, the expression in brackets in Eq.
(41) must then obtain a minimum (or, in other words, the stored energy W C þ U B must become a min-
imum). If one assumes that the twin length l as well as the strain energy parameter K are given quan-
tities, then h can immediately be calculated as
rffiffiffiffiffiffiffi
1 Cl
h¼ ð42Þ
g 2K
and s
T as
rffiffiffiffiffiffiffiffiffiffi
2K C
sT ¼ sT;cr þ 4 : ð43Þ
l
Eq. (43) allows estimating of sT;cr , if h can be measured. It should be mentioned that K depends de
facto weakly on h=l, see Eq. (36)2, and can be estimated as 0:6l, so the above mentioned relations for h
and sT change slightly; for details see [190,241]. Furthermore, it should be emphasized that the
concept following the stability of equilibrium by Petryk et al. [243] leads to the same results as those
outlined above due to the transformation condition Eq. (37).
Applying the above mentioned results for a c(TiAl)-based alloy, using for C ¼ 0:06 J m2 ;
pffiffiffi
l ¼ 70 109 N m2 and g ¼ 2=2, we receive with Eq. (42) for h, expressed in nm, the relation
pffi
h 1:2 l for l expressed in lm, which is quite realistic, see, e.g., Fig. 3 in [241] and Fig. 5 of this paper.
pffi
The quantity ðs T sT;cr Þ, Eq. (43), expressed in MPa, follows as ðs
T sT;cr Þ 283= l for l expressed in lm.
It may be of interest to note that, after inserting of h, Eq. (42), into Eq. (41), both contributions to the
stored energy, the elastic eigenstrain energy and the energy stored in the interfaces, obtain the same val-
ues. This means that 50% of the stored energy is due to the interface energy. Therefore, energetic con-
cepts for twin modeling, which do not include the interface energy, as reported, e.g., in [246,247],
cannot represent twinning in a satisfying way.
The concept of applying the transformation condition, later denominated as ‘‘twinning condition”,
as outlined above, has been used for several practical applications, e.g. for the formation of twins due
to an array of misfit dislocations along a c=cT -interface ðcT means ‘‘in twin relation to c”), see [248].
The dimension of these rather embryonic twins can be predicted by the twinning condition. Also
the stress state due to the misfit of a precipitate may trigger twins, e.g. a Ti3AlC perovskite precipitates
in c(TiAl), see [249]. The stress state near a crack tip can trigger a twin, too, which was studied by
Fischer et al. in [250]. Figs. 26 and 27 give a clear evidence of this phenomenon in c(TiAl). Of course,
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 119
the existence of a twin influences then the crack path, see Fig. 26. As further example the formation of
deformation twins in a micropillar, being a PST crystal, is reported in [251] with the dimensions of the
micropillar described in Section 4.2.3. The twinning condition was used to predict distinct twins
formed during compressive loading. The orientation of the twins as well as their thickness could be
predicted in satisfying agreement with the experiment.
When compared with the experimental observations, the described modeling concept, including
the total stress state r, see Eq. (37), reveals on a mechanical basis that mechanical twins are hetero-
geneously nucleated, preferentially at misfit structures associated with lamellar interfaces or precip-
itates. The findings explain why lamellar alloys are prone to twin, whereas single-phase c(TiAl)-alloys
are not. However, there is much room for further modeling because the factors restricting lateral twin
growth are not sufficiently understood. Also, the increase of the twinning propensity with tempera-
ture, which seems to be a specification of the c-phase in (a2 þ cÞ-alloys, needs further elucidation.
Concluding, we would like to refer also to the phase field model as concept to predict the formation
of deformation twins. Recently, such a study appeared for twinning in tantalum, see [252]. However,
we are still not aware on applications of the phase field concept for studying deformation twins in
c(TiAl)-based alloys. This fact does not mean that the phase-field concept cannot be used to model
the microstructure of Ti–Al-alloys, see, e.g., the recent simulation of the development of a PST
microstructure by Teng et al. [253]. The phase-field concept allows generating significantly different
microstructures depending on the energy terms to the gradients of the order parameters. However, if
the main characteristics of a microstructure are known, e.g. the formation and positioning of interfaces
(think on a layered microstructure) and the corresponding interface energy terms, then modeling of
distinct configurations and application of the established methods of continuum mechanics and ther-
modynamics offer, indeed, the straight forward way (also without engaging the phase-field concept).
The worldwide activities in experimental research and modeling have provided a relatively
detailed insight into the physical metallurgy of titanium aluminide alloys based on c(TiAl) and a2 (Ti3-
Al). Essential features of the synthesis, the evolution of constitution and microstructures, and the site
preference of alloying elements are now reasonably well understood. However, while there is cur-
rently great interest in alloys containing phases with b.c.c. structure, the characterization and model-
ing of such alloys are still at their infancy. This is mainly due to the fact that several b.c.c. derivative
metastable phases exist, which are only kinetically stabilized.
At the mechanical side, much effort has been paid on the detailed characterization of the slip geom-
etry in the major phase c(TiAl) and a2 (Ti3Al). Comparatively slip and twinning in the b.c.c. phases
(with non-close packed lattices) are much less documented and understood. Defect properties in c
(TiAl), such as the point defect situation, planar fault energies, dislocation decomposition and dissoci-
ation are well documented and rationalized by atomistic calculations. Based on this data, it is now
possible to rationalize the particular characteristics of the deformation of TiAl alloys, involving the
lack of independent slip systems, violation of Schmidt’s law, anomalous dependence of the yield stress
on temperature, evolution of internal stresses, and recovery phenomena. The analyses demonstrated
that computer simulations are capable of making significant predictions and of revealing hitherto
unknown phenomena. Prime examples are the effects of non-glide stress components on the disloca-
tion motion or the development of high internal stresses during the deformation of multiphase alloys
due to the mismatch of the elastic and plastic properties of the different constituent. It should also be
noted that a detailed analysis of the dislocation dissociation towards the determination of planar fault
energies requires an adequate simulation of the dislocation contrast phenomena based on anisotropic
elastic modeling or even by atomistic calculations.
However, in spite of this detailed knowledge, linking of the individual processes by an adequate
simulation seems to be difficult. Thus, the synergistic effects of the individual processes on the plas-
ticity of TiAl are less understood. For example, it is not yet possible to quantitatively describe the
simultaneous operation of the various dislocation friction mechanism and the evolution of internal
stresses on the yield stress, work hardening and fracture. The problems are even exacerbated for more
complex loading modes such as isothermal and thermo-mechanical fatigue.
120 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124
There is a need for quantitative models which will accurately predict the effects of alloy composition
and processing parameters on the alloy performance. Also there is little linkage to engineering design.
Most of the advantages that have been achieved in processing rely on empirical approaches. This is
mainly due to the complexity of industrial thermo-mechanical processing schedules. Predictions of
processing paths based on physically sound concepts remain a long-term goal and demanding task
for the future.
Acknowledgements
F.A. would like to thank the ‘‘Deutsche Forschungsgemeinschaft (DFG)” for the financial support
provided within the framework of the Projects AP-49/4 to AP-49/6.
Financial support by the Austrian Federal Government (in particular from the Bundesministerium
für Verkehr, Innovation und Technologie and the Bundesministerium für Wirtschaft und Arbeit) and
the Styrian Provincial Government, represented by Österreichische Forschungsförderungsgesellschaft
mbH and by Steirische Wirtschaftsförderungsgesellschaft mbH, within the research activities of the K2
Competence Centre on ‘‘Integrated Research in Materials, Processing and Product Engineering”, oper-
ated by the Materials Center Leoben Forschung GmbH in the framework of the Austrian COMET Com-
petence Centre Programme, Project A1.14 is gratefully acknowledged.
References
[1] Kim Y-W, Morris D, Yang R, Leyens C, editors. Structural aluminides for elevated temperature applications. Warrendale
(PA), USA: The Minerals, Metals and Materials Society (TMS); 2008.
[2] Clemens H, Smarsly W. Adv Mater Res 2011;278:551.
[3] Appel F, Paul JDH, Oehring M. Gamma titanium aluminide alloys – science and technology. Weinheim: WILEY-VCH; 2011.
[4] Clemens H, Mayer S. Adv Eng Mater 2013;15:191.
[5] Tetsui T. Adv Eng Mater 2001;3:307.
[6] Bewlay BP, Weimer M, Kelly T, Suzuki A, Subramanian PR. In: Baker I, Heilmaier M, Kumer S, Yashimi K, editors.
Intermetallic-based alloys-science, technology and applications. MRS Symp Proc, vol. 1516. Warrendale (PA): MRS; 2013.
p. 49.
[7] Appel F, Oehring M, Wagner R. Intermetallics 2000;8:1283.
[8] Pflumm R, Friedle S, Schütze M. Intermetallics 2015;56:1.
[9] Appel F, Wagner R. Mater Sci Eng R 1998;R22:187.
[10] Smarsly W, Baur H, Glitz G, Clemens H, Khan T, Thomas M. In: Hemker K et al., editors. Structural intermetallics
2001. Warrendale (PA): TMS; 2001. p. 25.
[11] MTU Report 1/2013, MTU Aero Engines AG, Germany; 2013. <www.mtu.de/report>.
[12] Murray JL. In: Murray JL, editor. Phase diagrams of binary titanium alloys. Metals Park (OH): ASM; 1987. p. 1.
[13] McCullough C, Valencia JJ, Levi CG, Mehrabian R. Acta Metall 1989;37:1321.
[14] Kattner UR, Lin J-C, Chang YA. Metall Trans A 1992;23:2081.
[15] Kainuma R, Fujita Y, Mitsui H, Ohnuma I, Ishida K. Intermetallics 2000;8:855.
[16] Novoselova A, Malinov S, Sha W, Zhecheva A. Mater Sci Eng A 2004;371:103.
[17] Blackburn MJ. Trans Metall Soc AIME 1967;239:1200.
[18] Duwez P, Taylor JL. Trans AIME/J Met 1952;4:70.
[19] Abe E, Kumagai T, Nakamura M. Intermetallics 1996;4:327.
[20] Lehoczky SL. J Appl Phys 1979;49:5479.
[21] Blackburn MJ. In: Jaffe RI, Promisel NE, editors. The science technology and applications of titanium. Oxford: Pergamon;
1970. p. 633.
[22] Yamaguchi M, Umakoshi Y. Prog Mater Sci 1990;34:1.
[23] Yoo MH, Fu CL. Metall Mater Trans A 1998;29:49.
[24] Fu CL, Zou J, Yoo MH. Scripta Metall Mater 1995;33:885.
[25] Rao S, Woodward C, Hazzledine P. In: Kvam EP, King AH, Mills MJ, Sands TD, Vitek V, editors. Defect interface interactions.
MRS Symp Proc, vol. 319. Pittsburgh (PA): MRS; 1994. p. 285.
[26] Hazzledine PM. Intermetallics 1998;6:673.
[27] Howe JM, Pond RC, Hirth JP. Prog Mater Sci 2009;54:792.
[28] Pond RC, Shang P, Cheng TT, Aindow M. Acta Mater 2000;48:1047.
[29] Hazzledine PM, Kad BK, Fraser HL, Dimiduk DM. In: Miracle DB, Anton DL, Graves JA, editors. Intermetallic matrix
composites II. MRS Symp Proc, vol. 273. Pittsburgh (PA): MRS; 1992. p. 81.
[30] Appel F, Wagner R. In: Murarka SP, Rose K, Ohmi T, Seidel T, editors. Interface control of electrical chemical and
mechanical properties. MRS Symp Proc, vol. 318. Pittsburgh (PA): MRS; 1994. p. 691.
[31] Appel F, Christoph U. Intermetallics 1999;7:1173.
[32] Fujiwara T, Nakamura A, Hosomi M, Nishitani SR, Shirai Y, Yamaguchi M. Philos Mag A 1990;61:591.
[33] Umakoshi Y, Nakano T, Yamane T. Mater Sci Eng A 1992;152:81.
[34] Inui H, Oh MH, Nakamura A, Yamaguchi M. Acta Metall Mater 1992;40:3095.
F. Appel et al. / Progress in Materials Science 81 (2016) 55–124 121
[35] Hazzledine PM, Kad BK, Fraser HL, Dimiduk DM. In: Miracle DB, Anton DL, Graves JA, editors. Intermetallic matrix
composites II. MRS Symp Proc, vol. 273. Pittsburgh (PA): MRS; 1992. p. 81.
[36] Rao SI, Hazzledine PM. Philos Mag A 2000;80:2011.
[37] Umakoshi Y, Nakano T. Acta Metall Mater 1993;41:1155.
[38] Maziasz PJ, Liu CT. Metall Mater Trans A 1998;29:105.
[39] Liu CT, Maziasz PJ. Intermetallics 1998;6:653.
[40] Maruyama K, Yamaguchi M, Suzuki G, Zhu H, Kim HY, Yoo MH. Acta Mater 2004;52:5185.
[41] Umakoshi Y, Nakano T. ISIJ Int 1992;32:1339.
[42] Singh JB, Molénat G, Sundararaman M, Banerjee S, Saada G, Veyssière P, et al. Philos Mag Lett 2006;86:47.
[43] Appel F, Oehring M, Paul JDH. Adv Eng Mater 2006;8:371.
[44] Appel F, Paul JDH, Oehring M. Mater Sci Eng A 2008;493:232.
[45] Appel F, Oehring M, Paul JDH. In: Kim Y-W, Smarsly W, Lin J, Appel F, editors. Gamma titanium aluminide alloys
2014. Warrendale (PA): TMS; 2014. p. 9.
[46] Nguyen-Man D, Pettifor DG. In: Kim YW, Dimiduk DM, Loretto MH, editors. Gamma titanium aluminides
1999. Warrendale (PA): TMS; 1999. p. 175.
[47] Yoo MH, Zou J, Fu CL. Mater Sci Eng A 1995;192–193:14.
[48] Appel F, Paul JDH, Oehring M. Mater Sci Eng A 2009;510–511:342.
[49] Shechtman D, Blackburn MJ, Lipsitt HA. Metall Trans A 1974;5:1373.
[50] Greenberg BA. Phys Status Solidi (B) 1973;55:59.
[51] Lipsitt HA, Shechtman D, Schafrik RE. Metall Trans A 1975;6:1991.
[52] Sriram S, Dimiduk DM, Hazzledine PM. In: Nathal MV, Darolia R, Liu CT, Martin PL, Miracle DB, Wagner R, et al., editors.
Structural Intermetallics 1997. Warrendale (PA): TMS; 1997. p. 157.
[53] Vasudevan VK, Stucke MA, Court SA, Fraser HL. Philos Mag Lett 1989;59:299.
[54] Wiezorek JMK, Zhang XD, Godfrey A, Hu D, Loretto MH, Fraser HL. Scripta Mater 1998;38:811.
[55] Appel F, Wagner R. Mater Sci Eng R 1998;22:187.
[56] Singh JB, Molénat G, Sundararaman M, Banerjee S, Saada G, Veyssière P, et al. Philos Mag 2006;86:2429.
[57] Farenc S, Coujou A, Couret A. Philos Mag A 1993;67:127.
[58] Li YG, Loretto MH. Phys Status Solidi (A) 1995;150:271.
[59] Kaufman MJ, Konitzer DG, Shull RD, Fraser HL. Scripta Metall 1986;20:103.
[60] Menand A, Huguet A, Nérac-Partaix A. Acta Mater 1996;44:4729.
[61] Schafrik RE. Metall Trans A 1977;8:1003.
[62] Tanaka K, Koiwa M. Intermetallics 1996;4:29.
[63] Yoo MH, Fu CL. Metall Mater Trans A 1998;29:49.
[64] Zambaldi CR. Micromechanical modelling of c-TiAl based alloys. Aachen: Shaker Verlag; 2010.
[65] Aoki M, Nguyen-Manh D, Pettifor DG, Vitek V. Prog Mater Sci 2007;52:154.
[66] Vitek V. Prog Mater Sci 2011;56:577.
[67] Fu CL, Yoo MH. Philos Mag Lett 1990;62:159.
[68] Yoo MH, Zou J, Fu CL. Mater Sci Eng A 1995;192/193:14.
[69] Liu YL, Liu LM, Wang SQ, Ye HQ. Intermetallics 2007;15:428.
[70] Fu H, Zhao Z, Liu WF, Peng F, Gao T, Cheng X. Intermetallics 2010;18:761.
[71] Gasik MM. Comput Mater Sci 2009;47:206.
[72] Hug G, Loiseau A, Veyssière P. Philos Mag A 1988;57:499.
[73] Hug G. PhD thesis, Université de Paris-Sud, France; 1988.
[74] Sun YQ. Philos Mag Lett 1999;79:539.
[75] Jiao Z, Whang SH, Yoo MH, Feng Q. Mater Sci Eng A 2002;329–331:171.
[76] Flinn PA. Trans Metall Soc AIME 1960;218:145.
[77] Woodward C, MacLaren JM. Philos Mag A 1996;74:337.
[78] Ehmann J, Fähnle M. Philos Mag A 1998;77:701.
[79] Hug G, Loiseau A, Lasalmonie A. Philos Mag A 1986;54:47.
[80] Court SA, Vasudevan VK, Fraser HL. Philos Mag A 1990;1:141.
[81] Stucke MA, Vasudevan VK, Dimiduk DM. Mater Sci Eng A 1995;192–193:111.
[82] Wiezorek JMK, Humphreys CJ. Scripta Metall Mater 1995;33:451.
[83] Paidar V, Vitek V. In: Westbrook JH, Fleischer RL, editors. Intermetallic compounds. Principles and practice, vol.
3. Chichester: John Wiley & Sons, Ltd; 2002. p. 437.
[84] Vitek V. Intermetallics 1998;6:579.
[85] Veyssière P, Douin J. In: Westbrook JH, Fleischer RL, editors. Intermetallic compounds. Principles, vol. 1. Chichester: John
Wiley & Sons, Ltd.; 1995. p. 519.
[86] Woodward C, MacLaren JM. Philos Mag A 1996;74:337.
[87] Zupan M, Hemker KJ. Acta Mater 2003;51:6277.
[88] Kear BH, Wilsdorf HG. Trans Metall Soc AIME 1962;224:382.
[89] Williams CJ, Blackburn MJ. In: Kear H, Sims T, Stoloff NS, Westbrook JH, editors. Ordered alloys. Baton Rouge (LA): Claitor’s
Publishing Division; 1970. p. 425.
[90] Lipsitt HA, Shechtman D, Schafrik RE. Metall Trans A 1980;11:1369.
[91] Inui H, Toda Y, Shirai Y, Yamaguchi M. Philos Mag A 1994;69:1161.
[92] Umakoshi Y, Nakano T, Takenaka T, Sumimoto K, Yamane T. Acta Metall Mater 1993;41:1149.
[93] Minonishi Y. Mater Sci Eng A 1995;192–193:830.
[94] Court SA, Löfvander JPA, Loretto MH, Fraser HL. Philos Mag A 1990;61:109.
[95] Inui H, Toda Y, Yamaguchi M. Philos Mag A 1993;67:1315.
[96] Wiezorek JMK, Court SA, Humphreys CJ. Philos Mag Lett 1995;72:393.
[97] Legros M, Couret A, Caillard D. Philos Mag A 1996;73:61.
122 F. Appel et al. / Progress in Materials Science 81 (2016) 55–124