M. Bambach et al.

,QWHUPHWDOOLFV  ²

the fabrication of compression specimens, the powder has been pro- DIL805A/D/T. The specimens were deformed up to a true strain of 0.8.
cessed at 1250 °C for 4 min by SPS with the lowest possible punch force Compression tests were conducted in an argon atmosphere. The speci-
to obtain partially sintered material for the examination of the micro- mens were heated with a heating rate of 10 K/s and held for 3 min to
structure in the necks between powder particles. This sample's micro- obtain a homogeneous microstructure before compression testing.
structure was investigated by SEM (Zeiss Gemini 1550) using energy Molybdenum plates of ∼1 mm thickness were placed between the
dispersive x-ray analysis (EDX) and electron backscatter diffraction specimen and the Si3N4-punches. After the compression tests, the
(EBSD) for composition and phase analysis, specifically in the region of samples were quenched to preserve the high-temperature micro-
the necks. EBSD maps were acquired using the Oxford-INCA software, structure.
while probing the γ-TiAl, α2-Ti3Al and bcc-Ti phases, respectively. In all
cases, the bcc-Ti phase was not present in the microstructure. 2.5. Microstructure analysis

2.2. Thermodynamic computations To investigate the microstructure evolution after the SPS process,
metallographic examinations were carried out on sections parallel and
Thermodynamic calculations were performed with Thermo-Calc perpendicular to the load direction. The sections were prepared in ac-
using a proprietary database for the quinary alloy system Ti-Al-Nb-C-B cordance with applicable standard procedures (grinding followed by
based on [45]. The purpose was to check which phases would appear in vibration polishing) and analyzed using scanning electron microscopy,
the neck regions at processing temperatures of 1250 °C and 1260 °C SEM (Tescan Orsay Holding, Czech Republic). The initial micro-
respectively. An isopleth section for Nb = 5.5 at.%, C = 0.0 at.% and structure of the SPS samples was homogeneous and consisted of ultra-
B = 0.0 at.% was also calculated. Boron and carbon were omitted for fine globular α- and γ-phase grains with an average grain size of about
the sake of simplicity, having little if any influence on the phase 5 μm.
boundaries at T < 1400 °C.
2.6. Dynamic material modelling (DMM)
2.3. Specimen manufacturing
The concept of processing maps, based on the principles of dynamic
Ingots with the composition Ti–45Al–5Nb–0.2B–0.2C (in at.%) were materials modeling (DMM), was first proposed by Prasad [47] for
used as raw material for the preparation of pre-alloyed powder. The modeling of hot forging behavior of a Ti-6242 alloy. In spite of a critical
ingots were gas atomised at Helmholtz-Zentrum Geesthacht, Germany paper by Montheillet et al. [48], the concept of DMM persists to be used
using the EIGA process as described in Ref. [46]. A powder particle size for workability assessment of metallic materials. DMM gives a physical
fraction between 45 and 63 μm was separated by sieving and used for interpretation to the strain rate sensitivity of flow stress, m, as a power-
the further processing steps. The powder was consolidated by means of partitioning factor between heat generation during plastic flow and
field-assisted sintering technique (FAST), also known as spark plasma microstructural change. Although the power-partitioning concept was
sintering (SPS), using a KCE-FCT HP D10-SD machine (FCT System criticized [48], the applicability of DMM has been validated by ex-
GmbH, Germany). tensive microstructural investigations for a wide range of materials,
The sintering process was performed using graphite tools with a including Al, Co, Ni and superalloys, Mg, Zr, Ti and Ti-Al, stainless
diameter of 40 mm. The loading chamber in the graphite tools unit was steels and several metal matrix composites [49]. Processing maps have
filled with the powder. For technological reasons, a graphite film was been employed for several decades as effective tools for design and
placed between the powder and the die and punches. The sintering optimization of metalworking parameters.
process was conducted in vacuum at a sintering temperature of 1250 °C A processing map is capable of predicting either ‘safe’ or ‘‘unsafe’
for 4 min and under a compaction pressure of 35 MPa. Samples with hot deformation domains for a given material. Dynamic recovery, dy-
dimensions Ø 40 × 15 mm were produced (Fig. 1a). The deformation namic recrystallization, and superplastic deformation are characterized
samples used in the compression tests (Ø 5 × 8 mm) were cut from the as safe mechanisms. However, several instability mechanisms such as
SPSed specimens using electrical discharge machining (Fig. 1b), and the flow localization, adiabatic heating, dynamic strain aging, void for-
surfaces in contact with graphite tools were removed prior to com- mation, and different types of cracking processes are conducive to the
pression tests. Therefore, carbon contamination from the graphite tools appearance of damage domains in processing maps [49,50]. A proces-
plays no role for the obtained results. sing map is built by superimposition of a power dissipation map and an
For comparison, cylindrical compression specimens (Ø 5 × 8 mm) instability map. At a given temperature in the hot working regime, the
were machined from extruded reference material and compression- externally applied power is consumed through two major paths: gen-
tested in the same way as the specimens produced by SPS. This material eration of heat due to plastic flow and dissipation by microstructural
was produced from the same powder using extrusion. changes. The instantaneously dissipated total power is determined by
Ref. [51]:
2.4. Hot compression tests
ε˙ σ

Cylindrical compression specimens of 8 mm height and 5 mm dia-

P= ∫ σ . dε˙ + ∫ ε˙ . dσ = G + J
0 0 (1)
meter were machined from the disks produced by SPS. The specimens
were deformed at temperatures ranging from 1150 °C to 1250 °C, and at where σ and ε̇ are the effective stress and effective strain rate, respec-
strain rates from 0.0013 s−1 to 0.1 s−1 using a deformation dilatometer tively. The first and second integrals are called G (content) and J (co-

Fig. 1. TNB-V5 alloy samples produced by

SPS (a) and machined and compressed at
1200 °C/0.01 s−1 to a true strain of 0.8 (b).


M. Bambach et al. ,QWHUPHWDOOLFV  ²

content) and represent the power dissipation through the plastic de- plastic deformation of metals at high temperatures. Work hardening
formation and microstructural dissipation, respectively. The relative describes the increase of stress necessary to continue deformation at
partitioning of power between heat generation and microstructural any stage of plastic deformation due to pinning of moving dislocation
transitions is defined by strain rate sensitivity (of flow stress), m, as by different mechanisms. In contrast, dynamic softening processes lead
follows: to a microstructure with a lower yield point by decreasing the density of
dislocations.
dJ ε˙ dσ d (lnσ )
m= = = At the initial stage of deformation, the flow stress increases sharply
dG σdε˙ d (lnε˙ ) (2) to a peak stress due to work hardening. At a higher strain, the flow
The ‘efficiency of power dissipation (η)’ with respect to a linear stress then reduces gradually to a steady state where the work hard-
dissipator (m = 1) is defined by: ening is balanced by softening mechanisms (dynamic recovery, DRV,
and dynamic recrystallization, DRX). The magnitude of the steady-state
ΔJ
ΔP m /(m + 1) 2m flow stress decreases with increasing temperature and decreasing strain
= = ≡η
( )
ΔJ
ΔP
linear 1/2 m+1
(3)
rate. The steady-state flow stress is reached faster at lower strain rates
compared with the higher strain rates and low deformation temperature
A power dissipation map displays the 3D variation of η as a function which is ascribed to the adiabatic heating [52]. The observed peak of
of temperature and strain rate. flow stress is more pronounced at a strain rate of 0.1 s−1, suggesting
Instability maps are developed based on a continuum instability that the studied alloy is subject to DRX during deformation at a high
criterion derived on the basis of the extremum principles of irreversible strain rate.
thermodynamics when applied to continuum mechanics of large plastic
flow proposed by Ziegler [51]. The instability criterion is given by the 3.2. Kinetic analysis of flow behavior and constitutive equation
dimensionless parameter ξ as follows:
A constitutive equation presents the functional dependence of flow
∂ln(m / m + 1)
ξ (ε˙ ) = +m≤0 stress on hot deformation parameters including strain, strain rate, and
∂lnε˙ (4) temperature. When a material undergoes deformation at elevated
An instability map delineates the dependence of ξ on temperature temperatures, the relationship between flow stress and processing
and strain rate. parameters is expressed by the classical hyperbolic function proposed
The temperature and strain rate regimes where ξ has negative values by Sellars and Tegart [53]:
cause flow instabilities. The continuum instability criterion describes Q ⎞
that the flow localization will take place in a system if the entropy ε˙ = A sinh(ασ )n exp ⎛−
⎝ RT ⎠ (5)
reduction by the system could not be offset against accumulated en-
−1
tropy imposed by plastic deformation. where ε̇ is the strain rate (s ), σ is the stress (MPa), Q is the apparent
The temperature and strain range regimes where the efficiency of activation energy of hot deformation (kJ.mol−1), T is absolute de-
power dissipation takes its maximum without subjecting the workpiece formation temperature (K), R is the universal gas constant
to flow instabilities are considered as an optimum processing window (8.314 J mol−1 K−1), and A, α, and n are material constants in-
for hot workability problems. dependent of deformation temperature. By taking the natural logarithm
from Eq. (5), the activation energy of hot deformation can be obtained
by:
3. Results

Q = R. ⎛⎜
∂ lnε˙
⎟
⎞ . ⎛ ∂ ln [sinh(ασ )] ⎞ = R. n. S
⎜ ⎟
3.1. Flow behavior
⎝ ∂ ln [sinh(ασ )] ⎠T ⎝ ∂ (1/ T ) ⎠ε̇ (6)

Fig. 2 shows the true stress-strain curves of SPSed and conventional After extracting the peak stress from the flow curves, the variation
TNB-V5 alloy at different temperatures and strain rates. The flow be- of ln [sinh(ασp)] with ln ε̇ and 1/T is depicted in Fig. 3 for SPSed
havior of the TNB-V5 alloy is sensitive to deformation temperature and sample. Using the average slopes in Fig. 3a and b to obtain the values of
strain rate in the typical way observed for metallic materials, i.e. the n and S, the activation energy is then calculated to be
flow stress declines with increasing temperature at a given strain rate, Q = 380.30 kJ mol−1 for SPSed samples. The value of Q is an indicator
whereas, it increases with strain rate at a given temperature. Work of the degree of difficulty of hot deformation and reflects additional
hardening and dynamic softening are two processes involved in the information on the microstructural evolution during hot deformation.

Fig. 2. True stress-strain curves of SPSed and conventionally extruded TNB-V5 alloys hot compressed at 1150 °C, 1200 °C, and 1250 °C to a true strain of 0.8 at
different strain rates of 1.3 × 10−3 s−1 (a) and 10−2 s−1 (b).


M. Bambach et al. ,QWHUPHWDOOLFV  ²

Fig. 3. The relationship between ln ε̇ and ln [sinh(ασp)] at different temperatures (a), ln [sinh(ασp)] and 1/T at different strain rates (b), and lnZ and ln [sinh(ασp)]
(c) for SPSed samples.

When the value of Q is comparable to the self-diffusion activation en- 3.3. Development of processing maps
ergy, diffusion or dynamic recovery plays a key role in microstructure
evolution during hot deformation. The value of Q obtained in this study 3.3.1. Strain rate sensitivity map
is close to that of the Ti-Al self-diffusion (360 kJ mol−1 [54]), sug- The relationship between the true stress and true strain rate in
gesting that diffusion makes a significant contribution to the hot de- logarithmic scale can be well expressed by a 3-order polynomial fit
formability of the studied TNB-V5 alloy (deformation mechanisms will using the following equation [55]:
be discussed in section 3.3.4).
logσ = a + blogε˙ + c (logε˙)2 + d (logε˙)3 (10)
The temperature-compensated strain rate, known as Zener-
Hollomon parameter Z, is given by: where a, b, c, and d are materials parameters depending on tempera-
ture. The variation of flow stress as a function of strain rate at a true
Q ⎞
Z = ε˙ exp ⎛ = A [sinh(ασ )]n strain of 0.5 is shown in Fig. 4 for SPSed (a) and conventional (b)
⎝ RT ⎠ (7) samples.
By taking partial differential on both sides of Eq. (10), the strain rate
Taking the natural logarithm from both sides of the Eq. (7) presents
sensitivity, m, can be defined as:
a linear relationship between lnZ and ln[sinh(ασp)] as follows:
∂ (logσ )
lnZ = lnA + nln [sinh(ασ )] (8) m= = b + 2c logε˙ + 3d (logε˙)2
∂ (logε˙) (11)

Fig. 3c shows the dependence of lnZ on ln[sinh(ασp)] for the SPSed Fig. 5 illustrates the dependence of m values on deformation tem-
samples. The peak stress is sensitive to both strain rate and deformation perature and strain rate at a true strain of 0.5 for the studied TiAl alloy.
temperature and the value of Z increases with peak stress. The good For the SPSed samples, positive values of m are observed for the
correlation coefficient R2 = 0.937 implies a good linear regression be- entire range of applied deformation conditions. However, regions with
tween peak flow stress and Z values. The stress exponent value, n, negative m-values are visible for the conventional sample, which are
calculated from the slope of the plot in Fig. 3c (n = 1.38) is in good located at the top right corner of iso-contour maps and they occur at
agreement with that obtained from the average slope of the plots in higher values of temperatures and strain rates. The regions with ne-
Fig. 3a (n = 1.28), reflecting a reasonable reliability of the kinetic gative m values are considered instability regions and they could cor-
analysis of the studied SPSed TNB-V5 alloy during the hot deformation respond to flow instability mechanisms including flow localization,
process. shear band formation, dynamic strain aging, and different types of
Based on the results, the constitutive equation of the SPSed alloy cracking processes [49,50]. In contrast to the regions of negative m
associated with peak stress is expressed using the hyperbolic-sine law as values, regions of the highest positive values of m are visible for both
follows: samples, indicating superplasticity of the studied TNB-V5 alloy.

380300
ε˙ = 2.39 × 1014 [sinh(1.63 × 10−2σ )]1.38 exp ⎛− ⎞ 3.3.2. Efficiency of power dissipation map
⎝ 8.314T ⎠ (9)
After determining values of m at different temperatures and strain


M. Bambach et al. ,QWHUPHWDOOLFV  ²

Fig. 4. Flow stress as a function of strain rate at a true strain of 0.5 for SPSed (a) and conventional (b) alloys interpolated and fitted using Eq. (10).

rates using Eq. (11), the efficiency of power dissipation, η, is calculated

according to Eq. (3). Fig. 6 exhibits the variation of η with deformation
temperature and strain rate at a true strain of 0.5 for the SPSed sample.
The contour numbers represent the efficiency of power dissipation in
percent. As shown in Fig. 6, the η is significantly affected by the de-
formation temperature and strain rate. At a given deformation tem-
perature, the values of η monotonously decline with strain rate. A wide
single domain is observed in the range 1150–1250 °C and 1.3 × 10−3-
0.1 s−1 with a peak power dissipation efficiency of 90% occurring at
∼1200 °C and 1.7 × 10−3 s−1. The strain rate sensitivity under these
conditions is ≈ 0.85 and this domain may represent superplastic flow of
the fine equiaxed microstructure [50].
The regions of the high efficiency of power dissipation are favorable
for the hot deformation process; however, they are not necessarily the
best for hot workability since damage can occur in these regions and
render them unsafe for hot working. Thus, the safety of the most effi-
cient power dissipation regions should be verified by considering the
flow instability criterion, which will be discussed in the following
section.

3.3.3. Flow instability map Fig. 6. Power dissipation maps of the SPSed TNB-V5 sample at a true strain of
0.5. Contour numbers represent percent efficiency of power dissipation.
The dimensionless instability parameter, ξ, is calculated by sub-
stituting Eq. (10)–(4) as follows:
Fig. 7 displays the variation of ξ with deformation temperature and
∂ln(m / m + 1) 2c + 6d logε˙
ξ (ε˙ ) = +m= +m strain rate at a true strain of 0.5 for the SPSed sample. Dark shaded
∂lnε˙ m (m + 1) ln10 (12) regions represent the flow instability domains with negative ξ values.
In temperature and strain rate regimes where ξ has negative values For the SPSed sample, the instability domain is located at the de-
flow instabilities occur. formation conditions of T = 1150–1215 °C/ ε̇ = 10−2-10−1 s−1. The

Fig. 5. 3D variation of strain rate sensitivity, m, with strain rate and deformation temperature at a true strain of 0.5 for SPSed (a) and conventional extruded (b) TNB-
V5.


M. Bambach et al. ,QWHUPHWDOOLFV  ²

diffusion, as already reported for a similar high Nb containing TiAl

alloy with (α2+γ) microstructure as well [57]. Moreover, the value of
Qapp is nearly equal to the activation energy of the dislocation climb
(386 kJ mol−1 [61]) for TiAl alloys. We, therefore , can conclude that
the dislocation climb in the γ lattice is the major rate-controlling me-
chanism governing the superplastic deformation of the studied TNB-V5
alloy at elevated temperatures [57].
The BSE-SEM images of the specimens compressed at 1250 °C at
different strain rates demonstrate that noticeable grain growth occurred
at 1250 °C/10−3 s−1. However, at higher strain rates of 10−1 s−1, the
grains are significantly smaller than the grains observed at lower strain
rates of 1.3 × 10−3 s−1. Therefore, the higher strain rate domain
(10−2-10−1 s−1) denoted as D2 in Fig. 8b with a peak power efficiency
of 60% represents the dynamic recrystallization of α-grains. This cor-
responds to a strain rate sensitivity value of ≈0.45. The occurrence of
DRX at higher temperature and strain rate is consistent with the flow
stress curves showing a more pronounced peak of flow stress at a strain
rate of 10−1 s−1.
The processing map also exhibits a domain of flow instability
shaded in Fig. 8b. Our microstructural analysis reveals the formation of
Fig. 7. Flow instability maps of the SPSed TNB-V5 alloy at a true strain of 0.5. a considerable amount of porosities in the specimens hot compressed at
The shaded region corresponds to flow instability.
temperatures below 1100 °C. Thus, it can be concluded that the me-
chanism of flow instability observed for SPSed TNB-V5 alloys in this
values of strain rate sensitivity and efficiency of power dissipation study might be the void formation and/or cracking. This conclusion is
under these conditions vary from 0.05 to 0.40 and 10% to 60%, re- in a good agreement with Ref. [40] reporting that the flow instability
spectively. This domain may represent cracking of the sample as al- region expands over a wide temperature range and at strain rates higher
ready reported for the similar Ti-(45–47)Al alloys containing 2–7 at.% than 0.01s−1 for Nb containing Ti-Al alloys.
Nb [40,56]. In summary, dynamic recrystallization and superplasticity were
It is worth noting that the region of flow instability observed in found to be the main softening and deformation mechanisms at ele-
Fig. 7 is identified as a stable region by power efficiency and strain rate vated temperature observed for the studied TNB-V5 alloy. However,
sensitivity maps. This proves that the domains with high η or m values void formation and cracking were identified as the major instability
do not always coincide with those characterized as safe working re- mechanisms occurring during deformation at lower temperatures and
gions. A comprehensive consideration of strain rate sensitivity, power higher strain rates.
dissipation efficiency, and instability criterion is required for identi-
fying the stable or damage working regions. 3.4. Characterization of initial powder

3.3.4. Processing map and determination of deformation mechanisms Fig. 9 reveals the microstructure of the gas atomized powder
A processing map is built by superimposition of the power dissipa- showing a β-dendrite microstructure with residual segregation as re-
tion map and the instability map, which reveals the domains or win- ported in the literature [46,62]. The average concentration of the
dows representing various microstructural mechanisms and the con- constituting elements of powder particles obtained by EDX analysis is
straints of flow instabilities. The optimum intrinsic workability of the shown in the table.
materials can be achieved by using the processing windows that are the Fig. 10 displays the microstructure of the initial powder particles
safe processing domains with the most efficient power dissipation. after SPS processing at minimal punch pressure. Such a treatment al-
Fig. 8 shows 3D processing maps and the corresponding iso-contour lows for achieving an equilibrium microstructure by transformation of
maps of the SPSed TNB-V5 alloy at a true strain of 0.5. The shaded the solidification microstructure to an equilibrium (α+γ)-two-phase
region in the iso-contour map represents the instability domain with microstructure that will then transform to (α2+γ) at room temperature
negative ξ values. The processing map reveals two safe processing [63]. As shown in Fig. 10, the microstructure consists of two phases
windows for the studied alloy. The lower strain rate domain (10−3- with different contrasts of dark and bright gray. The bright contrast
10−2 s−1) denoted as D1 in Fig. 8b with a peak power efficiency of 90% phase is mainly observed in near-surface regions of the powder parti-
and a strain rate sensitivity value of ≈0.8 represents superplasticity of cles. Phase identification studies by EBSD mapping clearly indicate that
the (α+γ)-phase containing microstructure. The superplastic de- the bright phase surrounding powder particles is the α2-Ti3Al phase and
formation of high Nb containing TiAl alloys with fine and uniform the dark phase is γ-TiAl. The α2-Ti3Al grains mainly appear as a net-
(α2+γ) microstructure (present at room temperature) was extensively work at the necks between powder particles and a few α2-grains are
reported [57–60]. The presence of a homogeneous microstructure also observed within the powder particles. Analysis of the chemical
consisting of fine equiaxed α- and γ-grains leads to a high degree of composition by EDX line scan shown in Fig. 10b reveals that the near
superplasticity in the studied alloy at elevated temperatures. Grain surface regions are rich in Ti and poor in Al compared to the inner
boundary, GB, sliding significantly contributes to the superplastic de- regions.
formation. To avoid premature failure of the specimens due to stress The reason why near-surface regions of powder particles are de-
concentration on the triple junctions of the grains, the GB sliding must pleted in Al (≈42%) compared to inner regions (≈46%) and why α2-
be accommodated by several mechanisms such as diffusion and dis- Ti3Al phase forms predominantly as a network at near surface regions
location slip. The apparent activation energy of the hot deformation, calls for further investigations. Trzaska et al. [63] and Jabbar et al. [64]
Qapp, calculated in the present study equals to 380.30 kJ mol−1, which studied spark plasma sintering mechanisms at the necks between TiAl
agrees well with Qapp values reported for various microcrystalline powder particles and did not find evidence of localized overheating,
(α2+γ) TiAl alloys. The value of Qapp = 380.30 kJ mol−1 is quite close plasma formation or other electrically induced effects for the presence
to the Al self-diffusion energy in the γ-TiAl lattice implying that the of the α2-phase. These findings support the idea that oxygen take-up on
superplastic deformation of the studied alloy is controlled by lattice the surface of the powder particles might account for the observed