Journal of Materials Processing Tech.

259 (2018) 169–179

Contents lists available at ScienceDirect

Journal of Materials Processing Tech.

journal homepage: www.elsevier.com/locate/jmatprotec

Influence of gibbosity on recrystallization behavior of single crystal blade T

casting
⁎
Runnan Wang, Qingyan Xu , Xianglin Su, Zhonglin Li, Baicheng Liu
Key Laboratory for Advanced Materials Processing Technology (MOE), School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China

A R T I C LE I N FO A B S T R A C T

Keywords: Recrystallization (RX) defects formed in the gibbosities induced by platinum chaplet pins were studied by ex-
Single crystal periments and simulations for single crystal (SX) blade castings. Recrystallization occurs in 39 of the 80 gibb-
Recrystallization osities and can penetrate more than 250 μm into the SX blade. The RX occurrence ratios for the gibbosities in the
Platinum chaplet pin upper, middle and lower positions of the blade are 83%, 50%, and 8% respectively, indicating a strong sensi-
Gibbosity
tivity of RX to the location height of the gibbosity. The microstructure is SX without the presence of micro
FEM simulation
Superalloy
misoriented grains in the cross-section of gibbosity of as-cast sample. Secondary carbide particles are found at
the interface between the RX and SX grains. A composite image of backscattered electron (BE) and electron
backscatter diffraction (EBSD) shows that the migration of the RX interface is hindered by coarse eutectic
particles. A thermal elastoplastic model that considers the anisotropic mechanical properties of SX superalloys
and is coupled with a finite-element method (FEM) was utilized to predict the deformation behavior during
directional solidification. The FEM results reveal that the equivalent plastic strain (PEEQ) tends to concentrate in
the gibbosities, and this gives a reasonable and reliable explanation for RX formation.

1. Introduction
Recrystallization (RX) defects can significantly influence the service
life of single crystal (SX) turbine blades and vanes in aero-engines. The
new grain boundaries introduced by RX decrease the creep rupture
resistance of Ni-based SX superalloy blades due to the lack of elements
that can strengthen the grain boundaries (Mathur et al., 2017). Plastic
strain is usually considered as the main reason for the occurrence of RX
in blade castings, and can originate during many processes such as
solidification contraction, shell removal and sand blasting (Jo et al.,
2003). In addition, the difference in thermal expansion coefficients
between superalloys and ceramic moulds is another important reason
why plastic strains can be induced during the cooling process. The
nucleation of RX grains is usually driven by plastic strain along with the
deformation energy stored during the manufacturing process; this type
of RX is called static RX (Reyes et al., 2015).
As the shapes of hollow airfoils with cooling channels become
progressively more complex, some specific technologies (e.g., chaplets
comprising platinum pins) are often used to meet the needs of pro-
duction (Salkeld et al., 1995; Ning et al., 2013; Jago, 1996). During the
directional solidification (DS) process, the core is easily driven to shift
by the flowing molten alloy around it; therefore platinum chaplet pins
are used to keep them stable. As shown in Fig. 1(a) and (b), 12 chaplets
are arranged on either side of a SX hollow blade to fix the core. Fig. 1(c)
presents the schematic diagram of Pt chaplet pin and the induced
gibbosity. The chaplet extends from the core, across the SX blade, to the
inside of ceramic mould, and the gibbosity is formed on the surface of
the blade (red dashed line in Fig. 1(c)). This technique can significantly
reduce the relevant defects induced by the core, such as core break-
through and exposure (black circle in Fig. 1(a), the core penetrates the
thin-wall of blade airfoil), and the rejection rate caused by these defects
decreases from 90% to less than 5%. However, the gibbosity formed in
the position of chaplet on the SX blade surface can be hindered by the
ceramic mould during DS, thereby leading to the formation of shear
stress. This significantly increases the probability of RX, which can
decrease the mechanical properties of the SX blade (Ning et al., 2013).
Porter and Ralph (1981) reported that RX nucleation was prone to
occur by subgrain coalescence, and the subsequent growth was attrib-
uted to strain-induced boundary migration of the grain boundary at
which RX nucleated. Zhuo et al. (2015) pointed out that the subgrain
boundaries formed in the SX matrix were transformed from the dis-
location walls that formed with the migration of dislocations during the
annealing process. Mathur et al. (2017) reported another surface nu-
cleation mode, in which the micro-grains of γ’ phase in surface eutectics
could coarsen during solution heat treatment. Panwisawas et al. (2013)
and Li et al. (2015b) proposed isotropic and anisotropic mechanical

⁎
Corresponding author.
E-mail address: scjxqy@tsinghua.edu.cn (Q. Xu).

https://doi.org/10.1016/j.jmatprotec.2018.04.036
Received 17 July 2017; Received in revised form 20 April 2018; Accepted 21 April 2018
Available online 22 April 2018
0924-0136/ © 2018 Elsevier B.V. All rights reserved.
R. Wang et al.
Fig. 1. SX turbine blade with platinum chaplet pins and gibbosities: (a) front
and (b) back sides, (c) schematic diagram of a Pt chaplet pin and an induced
gibbosity.
models to simulate the scale of plastic strain accumulated in SX castings
during the DS process, respectively. Zambaldi et al. (2007) simulated
the RX evolution around the indentation in SX superalloy by combining
the cellular automaton method and finite-element method (FEM) based
on crystal plasticity theory. Rettberg and Pollock (2014) studied the
influence of localized RX on macroscopic creep rates and developed a
model for the prediction of RX-accelerated tertiary creep. Jo and Kim
(2003) found that the surface RX in the SX CMSX-2 superalloy did not
reduce creep rupture life at 982 °C, 240 MPa owing to the strengthening
effect of a surface oxide layer. The early initiation of surface cracks on
grain boundaries normal to the applied stress implemented harmful
effects to the SX specimens, in accordance with the investigation of
Zhang et al. (2012). However, the influence of surface gibbosity on RX
behavior and the prediction of deformation in the gibbosity of blade
casting were seldom reported.
The present research focuses on the investigation of RX behavior in
blades with gibbosities on their surfaces. In this regard, a blade casting
with several gibbosities on its surface was designed, directionally soli-
dified and heat-treated. The recrystallized gibbosities at various heights
were counted and characterized. The nucleation mechanism of the RX
grains was discussed. To further study the driving force for nucleation
and growth, a thermal elastoplastic model considering the anisotropic
mechanical properties of SX superalloys was utilized to predict the
deformation behavior of blade castings during DS.
2. Methods
2.1. Experimental procedures
A model of a blade casting with five platinum chaplet pins and
gibbosities was designed for a length of 99 mm that approximately
equals the length of an aero-engine turbine blade, as shown in Fig. 2.
Wax patterns were fabricated by an additive manufacturing technology,
and spiral grain selectors were added to the very base of them. Chaplets
of platinum wire were inserted into the wax blades, and the gibbosities
formed around them. Four wax blade patterns were arranged in one
cluster. In this study, three types of mould clusters with angles of 0°, 45°
and 90° were fabricated (Fig. 3). The ceramic shells were produced by
ceramic slurries and particles. The superalloy used in this study was a
second-generation SX superalloy, DD6, whose nominal chemical com-
position is presented in Table 1. The SX blade castings were prepared in
an ALD industrial vacuum Bridgman furnace. The pouring temperature
was 1550 °C. After preheating, pouring and heat preservation, the chill,
ceramic shell and alloy were withdrawn from the heating zone into a
170
Journal of Materials Processing Tech. 259 (2018) 169–179
Fig. 2. Positions of Pt chaplet pins and induced gibbosities on the surface of the
designed blade casting: (a) front and (b) back sides.
cooling zone through a baffle at the constant rate of 5 mm/min. After
casting, the blades were subjected to the standard heat treatment (SHT)
consisting of the solution and aging processes (Table 2), which is si-
milar to the process employed in commercial production.
To inspect the surface RX grains, a mixed chemical solution (50%
hydrochloric acid and 50% hydrogen peroxide) was used for etching the
blade castings. After macro-etching, the macro RX areas can be easily
detected on a sample surface by optical inspection. When such a sample
is viewed under a directional light source, the RX areas are seen as local
chromatic aberrations, with different reflectivities indicating that they
differ in crystallographic orientation from the SX matrix. The number of
gibbosities (na, subscript a denotes gibbosity #a) that are recrystallized
at each gibbosity position was counted to calculate the RX occurrence
rate at this position (na/12, 12 is the total number of gibbosities at each
gibbosity position). In order to observe the RX microstructure, the
gibbosity positions were carefully cut down from the blade casting
using electrical discharge machining (EDM) normal to the surface
across the gibbosity peak. Subsequently, mechanical grinding and pol-
ishing were performed. The as-polished samples were etched using
Marble’s reagent to enhance the contrast between the precipitate and
matrix, then observed by optical microscopy (OM, Zeiss AM10 OM) and
scanning electron microscopy (SEM, Merlin FEG-SEM). Composition
measurements were performed using energy-dispersive spectroscopy
(EDS). The samples for crystal orientation measurements were elec-
trochemically polished, and the images were captured in backscattered
electron (BEI) and electron backscatter diffraction (EBSD) mode.
2.2. Mathematical model
In classical theory, plastic strain, as the main driving force, has been
widely used to explain the nucleation and growth of RX grains.
However, the exact temperature and time at which plastic strain occurs
is difficult to measure experimentally. Therefore, it is necessary to use
FEM and a constitutive model that considers the anisotropic mechanical
behavior of SX superalloys to calculate the equivalent plastic strain
(PEEQ) in blade castings during the DS process. Because the elastic and
plastic strains are mainly caused by the difference in coefficients of
thermal expansion between the castings and the shells/cores during the
solidification process, the accuracy of PEEQ is dependent on that of
temperature distribution. This distribution is extracted from the simu-
lation results of ProCAST and then imported to Abaqus as a pre-defined
temperature field. Both ProCAST and Abaqus are commercially avail-
able FEM software packages. To ensure accuracy of the temperature
field calculation, experimentally measured temperature/time-
R. Wang et al. Journal of Materials Processing Tech. 259 (2018) 169–179

Fig. 3. Schematic of three mould clusters. Four blade patterns are arranged in one cluster at (a) 0°, (b) 45° and (c) 90° angles with respect to the radial direction.

Table 1 σ ⎡ D1111 D1122 D1133 D1112 D1113 D1123 ⎤ ε11 ε

⎧ 11 ⎫
⎧ σ11 ⎫ ⎢ D2222 D2233 D2212 D2213 D1113 ⎥ ⎧ ⎫
Nominal chemical composition of the DD6 superalloy (Editorial committee,
⎪ 22 ⎪ ⎢ ⎪ ε22 ⎪ ⎪ ε22 ⎪
2001). ⎪ σ33 ⎪ D3333 D3312 D3313 D3323 ⎥ ⎪ ε33 ⎪ ⎪ ε33 ⎪
=⎢ ⎥ = [D el] γ
Element Cr Co Mo Ta Re Nb Al Hf Ni ⎨ σ12 ⎬ ⎢ D1212 D1213 D1223 ⎥ ⎨ γ12 ⎬ ⎨ 12 ⎬
⎪ σ13 ⎪ ⎢
⎪ σ23 ⎪ ⎢ sym D1313 D1323 ⎥ ⎪ γ13 ⎪ ⎪ γ13 ⎪
⎪ γ23 ⎪ ⎪ γ23 ⎪
wt.% 4.3 9 2 7.5 2 0.5 5.6 0.1 Balance ⎩ ⎭
⎣ D2323 ⎥
⎦ ⎩ ⎭ ⎩ ⎭
(3)
Table 2 where σij is the element of stress vector, [Del] the elastic stiffness matrix,
Standard heat treatment of the DD6 superalloy (Editorial εii and γij the elements of strain vector. If an orthotropic elastic model is
committee, 2001). employed, the number of independent constants in the stiffness matrix
Step Heat treatment conditions decreases from 21 to 9. For a SX FCC material with three identical
principal orientations, the number of variables will further decrease to
1 1290 °C / 1 h 3. As a result, the form of the compliance matrix for a SX can be written
2 1300 °C / 2 h
as:
3 1315 °C / 4 h, air cooling
4 1120 °C / 4 h, air cooling
5 870 °C / 32 h ⎛ 1/ E − μ/ E − μ/ E 0 0 0 ⎞
⎜− μ/ E 1/ E − μ/ E 0 0 0 ⎟
⎜− μ/ E − μ/ E 1/ E 0 0 0 ⎟
[C ]m = [D el]−1 =
dependent boundary conditions and temperature-dependent physical ⎜ 0 0 0 1/ G 0 0 ⎟
⎜ 0 0 0 0 1/ G 0 ⎟
properties of the DD6 superalloy (parameters are from Editorial com- ⎜ ⎟
mittee (2001) and Li et al. (2012)) were used in this study. ⎝ 0 0 0 0 0 1/ G ⎠ (4)
During last two decades, crystal plasticity finite-element methods where E, μ and G denote the Young’s modulus, Poisson’s ratio, and shear
(CPFEM) such as the one introduced by Roters et al. (2010) have gained modulus, respectively, all of which are defined along the < 001 >
great popularity in describing the heterogeneous characteristics. How- orientation. Using the spatial transformation of linear orthotropic
ever, they can hardly describe the stored-energy distribution at the elastic behavior, the variables in the stiffness matrix can be defined as:
entire-blade scale, owing to the limitation of their required computa-
tion. Hence, for the present work, a macroscopic phenomenon-based D1111 = E (1 − μ2 ) ν = D2222 = D3333 (5)
deformation model considering orthotropic mechanical properties was
D1122 = E (μ + μ2 ) ν = D1133 = D2233 (6)
utilized for the SX superalloy. This three dimensional model was es-
tablished using an elasto-plastic material law under the assumption of D1212 = G = D1313 = D2323 (7)
orthotropic elasticity and anisotropic plasticity. The model has been
proved using regression analysis in the study of Li et al. (2015b). As- where v is the constant related to Poisson’s ratio μ:
suming the plasticity is rate-independent, the thermal strain (εth ), elastic 1
ν=
strain (εel ), and plastic strain (εpl ) follow the relations below: 1 − 3μ2 − 2μ3 (8)
εth + εel + εpl = 0 (1)
2.4. Plastic behavior
εth = αΔT = α (T − Tref ) (2)
The Hill yield criterion considering the anisotropic plastic property
where Tref is a reference temperature, such as the room temperature.
of material is utilized, and the general form can be described by Eq. (9).
The variable α denotes the thermal expansion coefficient.
The constants F, G, H, L, M and N are determined by testing the material
in different orientations experimentally in terms of rectangular
2.3. Orthotropic elastic properties Cartesian stress components. Assuming that the three principal axes of
the SX are identical, Eq. (9) can be simplified into the form of Eq. (10),
In comparison with the isotropic elastic property that was used for where K = L/F.
modeling SX materials, the orthotropic elastic property is more suitable.
2 2 2
It can accurately describe the mechanical properties of crystals with f (σ ) = F (σ22 − σ33 )2 + G (σ33 − σ11 )2 + H (σ11 − σ22 )2 + 2Lσ23 + 2Mσ31 + 2Nσ12
three orthotropic planes such as face-centered cubic (FCC) and body- (9)
centered cubic (BCC) lattices, in which three < 100 > orientations are
1 2 2 2
identical. According to the shear stress reciprocal theorem τij = τji , a f (σ ) = (σ22 − σ33 )2 + (σ33 − σ11 )2 + (σ11 − σ22 )2 + 2K (σ23 + σ31 + σ12 )
2
symmetric form of the strain tensor matrix with six independent vari-
ables can be obtained. A fully anisotropic form of the constitutive (10)
equation can be written as: For this function, the flow rule is:

171
R. Wang et al. Journal of Materials Processing Tech. 259 (2018) 169–179

Table 3
Mechanical parameters used in the simulation (Li et al., 2015b).
Parameter 25 °C 700 °C 760 °C 850 °C 980 °C 1070 °C 1100 °C

E < 001 > (GPa) 131.5 107 105.5 98 80.5 69.5 67.5
μ < 001 > 0.344 0.374 0.377 0.383 0.390 0.399 0.413
G < 001 > (GPa) 142 93.6 98.3 71.2 85.4 75.7 66.6
K 6.2 3.47 3.44 2.46 1.87 2.15 2.36
L 3.1 1.74 1.72 1.23 0.94 1.08 1.18

∂f dλ
dε pl = dλ = b
∂σ f (11)
where b can be written as:

⎡− G (σ33 − σ11) + H (σ11 − σ22 ) ⎤

⎢ F (σ22 − σ33) − H (σ11 − σ22) ⎥
⎢ ⎥
− F (σ22 − σ33) + G (σ33 − σ11 ) ⎥
b=⎢
⎢ 2 Nσ 12 ⎥
⎢ 2Mσ31 ⎥ Fig. 5. Number of gibbosities with RX grains at five gibbosity positions in three
⎢ ⎥ mould clusters.
⎣ 2Lσ23 ⎦ (12)
The mechanical parameters used in the simulation are given in than that of #4 (27.5 mm), #5 (27.5 mm), #2 (30.5 mm), and the up-
Table 3. The strain results of nodes are interpolated from integration permost one #1 (45 mm). The RX occurrence rate at #3 is 8% (1/12),
points; hence, the PEEQ are obtained from integration points owing to whereas those of #4, #5 and #2 are 25% (3/12), 50% (6/12) and 75%
their high accuracy. PEEQ (ε pl ) is calculated from the equivalent plastic (9/12), respectively. Surprisingly, the value at gibbosity#1 reaches up
strain rate (ε̇ pl ) by integrating on time (t), as shown in Eq. (13), where σ to 83% (10/12). In summary, the RX occurrence rates for the gibbos-
is the stress tensor, ε̇ pl the plastic strain rate tensor, and σ° the reference ities at the upper, middle and lower positions are 83% (10/12), 50%
yield stress defined for Hill yield potential function. (18/36) and 8% (1/12), respectively, showing significant difference.
t t σ : ε˙ pldt The RX tendency increases evidently with the increase in the gibbosity’s
ε pl = ∫0 ε˙ pldt = ∫0 σ0 (13) location height.

3. Results and discussion 3.2. The depth and microstructure of RX grains

3.1. Macroscale recrystallized gibbosities
After macroscale chemical etching, the RX grains can be detected
from the various reflectivities observed under light irradiation. As
shown in Fig. 4, the samples annealed using SHT show large RX areas
on the gibbosity surface, including some twinning grains. However,
almost no RX was observed at other locations on the blade surface. The
macroscale experimental results reveal a strong sensitivity of RX to the
gibbosity on the surface of a blade casting. The largest RX and twinning
grains almost cover the whole surface of the gibbosity, but cannot cover
its round edge.
The number of gibbosities with RX grains at five gibbosity positions
in three mould clusters is summarized in Fig. 5. There are four identical
blade castings arranged in one mould cluster, such that the maximum
number of recrystallized gibbosities at the same gibbosity position in
one mould cluster is four. The location height of #3 (10 mm) is lower
The RX grain morphology in the cross-section of a gibbosity is
shown in Fig. 6 (a). The grain can be divided into two parts by the red
dashed line that represents the boundary between the SX blade and the
gibbosity: (1) the part inside the gibbosity (marked by the yellow line)
and (2) the part inside the SX blade (marked by the red line). The
former will be removed during the subsequent process of removal of the
gibbosity, while the latter will actually destroy the integrity of the SX
structure and reduce its mechanical properties.
As shown in Fig. 6(a), a one-dimensional coordinate system was
defined to describe the grain depth x. The origin of the coordinate
(0 μm) is on the red dashed line (the boundary between the SX blade
and the gibbosity), and the value above zero represents that RX grains
have grown into the inside of the SX blade casting (left side of the red
dashed line). The minimum and maximum depth of RX grains and the
total number of recrystallized gibbosities at five gibbosity positions are
summarized in Fig. 7. Although RX occurred in nearly half of the

Fig. 4. Macroscale-photographs of blade cast-

ings after SHT and macro-etching: (a) twelve
blade castings; (b) (c) local magnified pictures
of RX at five gibbosities. The gibbosities with
RX grains are marked by red circles (For in-
terpretation of the references to colour in this
figure legend, the reader is referred to the web
version of this article).

172
R. Wang et al.
Fig. 6. Structure morphology in the cross-section of gibbosity. (a) OM graphic. (b) Twinning RX grains by EBSD detection. The inset is a local magnified composite
image of backscattered electron (BE) and EBSD, showing the eutectic retarding the migration of the RX interface. (c) Eutectic particles at grain boundary.
Fig. 7. Minimum and maximum depth of RX grains and the number of re-
crystallized gibbosities at five gibbosity positions.
gibbosities, the front interface of most RX grains stopped moving in the
gibbosities and did not actually penetrate the SX blade casting. How-
ever, there still exist some RX grains that penetrate as deep as ap-
proximately 300 μm. We can find that the integrity of the SX structure
in several blades has been damaged because of the deep RX grains at
gibbosity #1, #2 and #5. Meanwhile, the RX grains of the only re-
crystallized gibbosity in the position of gibbosity#3 is some distance
away from the real SX blade.
It is well known that many small RX nuclei initially appear in the
dendritic arms (DAs) from the onset of heat treatment (Li et al., 2015c).
Subsequently, by undergoing long-time SHT at high temperature
(solvus and sub-solvus), the small grains will overgrow the DAs and
173
Journal of Materials Processing Tech. 259 (2018) 169–179
enter into the interdendritic regions (IDRs). The RX behavior of the
deformed matrix at this stage is caused by the release of stored de-
formation energy. At the next stage, the RX grain growth is controlled
by the difference of interfacial energy (IFE) between small and large RX
grains. The number of grain boundaries and IFE can be reduced by
growing and merging of RX grains, so this is a spontaneous process.
Thus, there exist only large RX grains in gibbosities instead of finer
grains after the SHT process (Figs. 4 and 6). Owing to the formation of
stacking faults during the migration of the RX interface, annealing twin
grains structure can usually be found in RX regions, appearing as
strictly parallel grain boundaries, which have 60° misorientation angles
in EBSD figures, as shown in Fig. 6(b). Apart from large RX grains in-
duced by the deformed matrix, eutectic phases are observed by OM
after the corrosion of Marble’s reagent. As shown in Fig. 6(b) and (c),
the migration of the RX interface can be retarded by some coarse par-
ticles, which are determined to be eutectic phases by the composite
image of BE and EBSD detection, with the same orientation as SX ma-
trix. These phases mainly exist in IDRs and are hard to completely
dissolve during heat treatment (Wang et al., 2006). Otherwise, without
the existence of these eutectic particles, the RX grains probably would
have grown deeper into the SX blade casting.
It is commonly accepted that the recovery behavior is relatively
weak for the studied Ni-based superalloy with low stacking fault energy
(Liu et al., 2015). Therefore, the recovery effect can be ignored during
the RX process in this study, and the onset temperature of RX increases
with the decrease in RX capability. Coarse γ’ and eutectic particles at
grain boundaries are considered to form during solidification, and they
strongly limit the migration of RX grain interfaces owing to high so-
lubility and diffusivity in the RX interface (Porter and Ralph, 1981).
Besides these particles, some irregular new precipitated phases (NPP)
that do not influence the interface morphology are found at the grain
R. Wang et al.
Fig. 9. Elemental segregation in region near the grain boundary between RX and SX matrix of as-heat-treated samples.
boundaries between SX matrix grain and mis-orientated RX grains
(Fig. 8(b)). Area scanning results of EDS measurement (Fig. 9) shows
that the contents of elements W, Mo and Re are extremely high in NPPs
compared to the SX and RX regions. Thermodynamic calculation was
conducted with Pandat software and PanNi database which is based on
the experimental measurement. According to the calculation results,
the phases γ, γ’, M6C, M23C6 and μ can form in DD6 superalloy with
nominal chemical composition, but the atom activity gradually
174
Journal of Materials Processing Tech. 259 (2018) 169–179
Fig. 8. Microstructure of DD6 superalloy after
SHT: (a) cubic γ' in RX and SX bulk; (b) new
precipitated carbides at interface. (c)Size dis-
tribution of cubic γ' strengthening phases at
both sides of grain boundary.
decreases with the decrease of temperature. Hence, phases formed at
elevated temperature are likely to remain. The elements such as C and
Hf are likely to accumulate in grain boundaries and play a role in grain
boundary strengthening. Besides, the chemical composition close to the
formation position of precipitated phase dose not conform to the
nominal chemical composition of DD6 superalloy, so the formation
condition of these phases does not completely conform to the calcula-
tion results. There are difference in chemical composition between M6C
R. Wang et al. Journal of Materials Processing Tech. 259 (2018) 169–179

Fig. 10. SEM and EDS spectroscopy of the precipitated particles in grain boundary: (a) particle morphology, the particle for point EDS is marked at the upper left
corner; (b) the EDS spectroscopy of the marked precipitated particle.

Table 4 significantly occurred, was considered in this study.

The quantitative chemical composition of precipitated particle (Fig. 10(a)) In the fixed blade coordinate system, to check the influence of
measured by point EDS. crystal orientation on simulated PEEQ, three [100] orientations ([100],
Element C W Mo Cr Re Ni Co [100]* and [100]** in the inset of Fig. 11) with an interval of 30° were
set for the casting. Simulations were carried out using identical para-
wt.% 14.22 23.51 8.04 4.46 23.21 14.99 7.84 meters except for the set crystal orientations. The simulated PEEQ of
ten points, from the lower to upper part of the blade, were chosen for
comparison, and the corresponding integration points and elements
and M23C6 carbides. M6C is rich in W and Mo, while M23C6 is rich in Cr.
were identical in the three simulation cases. As shown in Fig. 11, the
The EDS spectroscopy of the precipitated particle in Fig. 10 (a) is shown
simulation results of three cases are very close, with most relative de-
in (b), and corresponding quantitative chemical composition is shown
viation less than 5%. The maximum relative deviation of approximately
in Table 4. It indicates that the particles are enriched in C, W and Mo,
20% occurs at point #6, but the value of PEEQ is small, which means
which are much higher than the nominal chemical composition of DD6
the relative deviation can increase under very slight random error.
superalloy, but the content of Cr almost does not show any difference.
Because the variation of simulation results with crystal orientations is
Hence the precipitated particles are determined to be M6C carbide,
small, for convenience, the orientation is defined to be 0° in the fol-
which agrees with the observation of Xiong et al. (2009).
lowing simulations.
It is well known that the cubic γ' phases re-precipitated after solu-
In addition to crystal orientation, the blade castings can be arranged
tion and aging heat treatment play the most important role in the
in one cluster with any angle, which can also change the temperature
strengthening mechanism of Ni-based superalloys. As shown in Fig. 8,
distribution, solidification sequence and strain distribution of the blade
the γ' precipitates in the RX regions have an orientation different from
during the DS process. Three angles (0°, 45° and 90° with respect to the
those in the SX matrix. The size distributions of the cubic γ' strength-
radial direction of the cluster) were tested in this study, with the cor-
ening phases in both sides of the grain boundary were obtained by an
responding mould clusters shown in Fig. 3. As shown in Fig. 12(a) and
image processing method, as shown in Fig. 8c. The cubic nature of the γ'
(b), the temperature distribution near gibbosity #1 varies with mould
phases in the original SX matrix conforms well with the observations by
clusters, and the inclination angle of solid-liquid interface of the blade
Xiong et al. (2015) and Shi et al. (2015). The cuboidal γ' precipitates
in clusters of 0° and 90° can change by approximately 20°. This is as-
with a volume fraction of 70% are distributed uniformly in the γ matrix
cribed to the difference in the solidification sequence induced by a
of the original region along the [001] direction, which is parallel to the
shadow effect (Ma et al., 2012) in the blade castings. Decreased tem-
horizontal direction. The similarities in the elemental components and
perature makes the castings gradually contract, but how can the soli-
microstructures of the RX region and SX matrix reveal that the reduc-
dification sequence influence the PEEQ? This can be explained by a
tion in mechanical properties is due to the formation of high-angle
comparison of the process of plastic strain accumulation. When the Hill
grain boundaries.
potential function exceeds the yield strength at the same the

3.3. Crystal orientation and mould cluster

According to the thermal-mechanical model in this study, the or-

thotropic elastic property and anisotropic intensity strongly depend on
the crystal orientation of the microstructure in the castings. Generally,
the crystal orientation employed in the simulation did not conform to
reality. In addition, the angle at which the blades are arranged in one
mould cluster, can influence the temperature distribution in the cast-
ings during solidification and then influence the simulated PEEQ.
Hence the effect of crystal orientation and mould cluster on PEEQ
should be tested and analyzed. The simulation results are extracted
from the blade castings at 800 °C. The reason why 800 °C is chosen in
this study is that the predicted PEEQ is usually larger than the real
value when the temperature of the solid casting decreases to room
temperature. At lower temperature, the strength of the solid superalloy
will be high enough to tense the ceramic shell to fracture, especially
when the ceramic has become fragile after undergoing high tempera-
ture, but this phenomenon was ignored in the simulation. According to
the investigation of Li et al. (2015a), the sensitive RX temperature of
DD6 superalloy usually exceeds 800 °C. Therefore, a temperature range Fig. 11. Simulated PEEQ of different positions in the blade castings with three
of 800–1400 °C, in which the fracture of the ceramic shell has not [100] crystal orientations at 800 °C.

175
R. Wang et al.
Fig. 12. Difference between temperature distribution of blade castings in (a) 0° and (b) 90° clusters. (c) The comparison of PEEQ evolution at gibbosity#1 in 0° and
90° clusters during solidification.
temperature, plastic strain will accumulate. This process is described in
Fig. 12(c); the point chosen for comparison is the same position of
gibbosity #1 in the 0° and 90° clusters. As can be seen, there is a minor
difference in the evolution curves of the Hill potential function, leading
to different yield behaviors, but the variation tendencies are similar.
Once the PEEQ accumulates, it is irreversible, and hence it is related to
the stress status of every moment. Different temperature distributions
lead to the differences in stress status in the three mould clusters. As a
result, the PEEQ is influenced by the mould clusters. Fig. 13 indicates
that although the simulated PEEQ varies with clusters, the stress status
and variation trend inside the blades are similar. For the semi-quanti-
tative judgment of RX, it is completely acceptable as long as the trends
and values are correct in a single cluster. Therefore, the 0° cluster, the
results of which are closest to the average values of the three clusters, is
chosen for the following simulation.
176
Journal of Materials Processing Tech. 259 (2018) 169–179
3.4. Evolution of PEEQ
PEEQ represents the extent of plastic deformation and crystal slip,
which is essentially ascribed to the formation of dislocation. Zhuo et al.
(2015) have observed that the dislocation can migrate to form the
dislocation walls during annealing, and the SX matrix can be separated
by these dislocation walls to provide the RX nucleation sites, which
conforms with the traditional RX nucleation mode of dislocation in-
duction. On the macro scale, PEEQ can well represent the extent of
plastic deformation and slip. Owing to heterogeneity of the finite-ele-
ment mesh, it is important to select reasonable mesh size to ensure
calculation accuracy. In this study, coarse and fine meshes were allo-
cated to the simple blade body and complex geometric features, re-
spectively, to achieve an appropriate balance between computational
efficiency and accuracy, as shown in Fig. 14. When the temperature of
R. Wang et al.
Fig. 13. Simulated PEEQ of different positions in the blade castings of three
mould clusters at 800 °C.
gibbosity #1 decreases to 800 °C, the PEEQ distribution on both sides of
the blade casting is shown in Fig. 14. Most PEEQ accumulations appear
near the gibbosities, with a little near the chaplet hole. In Fig. 14a, the
PEEQ of gibbosity#1 appears to be less than those of #2 and #5. When
the temperature of gibbosity#1 decreases to 800 °C, the temperatures of
#2 to #5 are all less than 800 °C, and hence they have accumulated
more PEEQ. The PEEQ mainly exists in the top and bottom regions of
the gibbosities. The evolution of PEEQ distribution in the cross-section
of gibbosity#1 is demonstrated in Fig. 15, with the temperature ranging
from 1342 °C (solidus) to 800 °C.
The variations in PEEQ with temperature at the top and bottom
parts of five gibbosities are shown in Fig. 16. At the top of a gibbosity,
the PEEQ tends to accumulate at temperatures above 1250 °C and in the
range of 1000–1100 °C. The PEEQ in the bottom part of a gibbosity is
mostly introduced at the early stage of solidification. According to
previous research, the deformation temperature has been proved to be
crucial to the RX behavior (Li et al., 2015a). The RX sensitivity sig-
nificantly depends on the deformation temperature; generally, elevated
deformation temperature can facilitate RX formation. Compared to the
top part of a gibbosity, the bottom part has larger deformed regions and
higher deformation temperature; hence, PEEQ at this place can exert a
greater impact on RX formation. In this study, the average PEEQ (A-
PEEQ) and the PEEQ at the bottom part of gibbosity (B-PEEQ) are paid
Fig. 14. PEEQ distributions in blade casting when the temperature of gibbosity#1 (the uppermost gibbosity) decreases to 800 °C: (a) front side; (b) back side.
177
Journal of Materials Processing Tech. 259 (2018) 169–179
more attention than that at the top part (T-PEEQ).
The simulated A-PEEQ, B-PEEQ, T-PEEQ and PEEQ distributions on
the surfaces and cross-sections of five gibbosities at 800 °C are shown in
Fig. 17. The simulation results were extracted from the integration
points with maximum PEEQ at the top or bottom of each gibbosity. The
deformation-affected area can penetrate the airfoil. The amounts of A-
PEEQ and B-PEEQ in different gibbosities vary with their location
heights, with the minimum value appearing at gibbosity #3, which has
the minimum location height among the five gibbosities. This agrees
well with the minimum RX occurrence ratios at gibbosity #3 in the
experimental results. From qualitative perspective, the relation of linear
thermal expansion/contraction vs temperature variation follows
ΔL = α (T − T0) , where α and T0 are thermal expansion coefficient and
reference temperature, respectively. It indicates that the gibbosity with
larger location height is likely to accumulate more deformation due to
the longer solidification part below it, and this agrees with the ob-
servation in industrial blade manufacturing process such as the occur-
rence position of crack and RX. In previous studies, Panwisawas et al.
(2013) performed experiments and simulations of bobbin castings
during DS process, and found the critical plastic strain for RX occur-
rence is in the range of 1.5%–2.5%. Li et al. (2016) used Gleeble test
machine to obtain the specimens with different uniaxial deformation
amount at various temperature, and then annealed them at different
temperature to examine the RX occurrence. He found the critical plastic
strain for RX occurrence is 2–3%. The A-PEEQ in most gibbosities
reaches 4–5%, which already exceeds the threshold strain range of
2–3% to propagate RX. Moreover, for the SHT process that includes the
maximum solution temperature of 1315 °C for the DD6 superalloy, RX is
more likely to occur compared to the normal sub-solvus isothermal heat
treatment. However, for gibbosity#3, B-PEEQ at the bottom portion is
so small that there is not enough deformed area that can give rise to RX.
The simulated B-PEEQ at gibbosity #2 and #5 are larger than those for
the others due to the presence of the chaplet hole near them. This
agrees with the experimental results that recrystallized gibbosities arise
more at gibbosity #2 and #5 than #4, though they have nearly the
same location height.
The PEEQ profile along the red line (upper-right inset) on the cross-
section of gibbosity#1 is shown in Fig. 18. From the trend of the curve,
it can be seen that the maximum PEEQ concentration is at the corner
and decreases progressively from the outside to the inside. Combining
the experimental and simulated results, the nucleation mechanism of
RX grains is discussed. For SX components, there are two typical modes
of RX nucleation: one is through the dislocations in γ and stacking faults
in γ’, which are induced from the PEEQ (Zhuo et al., 2015); the other is
R. Wang et al.
Fig. 15. Evolution of PEEQ distribution in the cross-section of gibbosity#1 at (a) solidus 1342 °C, (b) 1200 °C, (c) 1000 °C and (d) 800 °C.
Fig. 16. Variation of simulated PEEQ with the decrease in temperature at the
(a) top and (b) bottom parts of five gibbosities during solidification.
through the micro-grains of γ’ in the surface eutectic, concurrently in
the presence of moderate casting deformation (Mathur et al., 2017).
The micro-grains in the latter nucleation mode resemble the freckle-
chain defects of DS components, and it should be noted that significant
elemental segregation remains even after heat treatment, as demon-
strated in the experimental results of Mathur et al. (2017). For the blade
castings with gibbosities, the EDS maps of different elements in the
grain boundary between RX and SX grains are presented in Fig. 9. The
secondary carbide particles, enriched in Ta, W, Re and Co, appeared at
the grain boundary; however, almost no difference in the elemental
distribution was measured between the RX and SX regions. In addition,
no misoriented grains were found in the surface and cross-sections of
as-cast gibbosities by OM and EBSD observations. Neither of these re-
sults conform to the characteristics of the latter nucleation mode.
Moreover, the FEM results reveal that the Pt pin and gibbosity here act
178
Journal of Materials Processing Tech. 259 (2018) 169–179
Fig. 17. Simulated A-PEEQ, B-PEEQ, T-PEEQ and PEEQ distributions on the (a)
(c)(e)(g)(i) surfaces and (b)(d)(f)(h)(j) cross-sections of five gibbosities at
800 °C.
Fig. 18. PEEQ profile along the red line (upper-right inset) in the cross-section
of gibbosity#1. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article).
as concentration features that induce large mechanical deformation,
and the induced PEEQ is the driving force of RX nucleation and growth.
4. Conclusions
Experiments and simulations were conducted to investigate the ef-
fect of gibbosities on RX formation in SX blade castings. Based on the
results obtained above, the following conclusions can be drawn:
1) 48% (29/60) of gibbosities exhibited RX grains after SHT (solution
and aging). Most RX grains exist inside the gibbosities, but some still
R. Wang et al.
penetrate the blades to a depth of 300 μm, damaging the integrity of
the SX structure.
2) The location height of a gibbosity has a great influence on RX grain
formation and depth. RX occurrence rates of the gibbosities at the
upper, middle and lower positions are 83%, 50% and 8%, respec-
tively. The RX grain at gibbosity#3, which has the lowest location
height, is very shallow, with a distance of 1342 μm from the SX
blade. Although the chaplets and gibbosities must be used in the
manufacturing process for technological reasons, in order to de-
crease the RX tendency, it is highly recommended to design and
assemble chaplets at lower positions.
3) The scale of A-PEEQ and B-PEEQ in the different gibbosities vary
with their location heights, and the minimum value occurs at
gibbosity #3 which has the minimum location height among the five
gibbosities. The A-PEEQ in most gibbosities reach 4–5%, which al-
ready exceeds the threshold strain range of 2–3% required to pro-
pagate RX. Such modeling can aid in the assessment of casting
strains to predict RX grain formation prior to casting and determine
the position of Pt chaplets to minimize PEEQ.
Acknowledgments
This research was funded by the National Key R&D Program of
China (2017YFB0701503) and National Basic Research Program of
China (2011CB706801). The authors are grateful to Sun Changbo and
Xiong Jichun from Aero Engine Corporation of China (AECC) for their
support with the experimental equipment.
References
Editorial committee, 2001. Engineering Materials Handbook. Standards Press of China,
Beijing (In Chinese).
Jago, M., Investment casting. US005505250A. Apr. 9, 1996. [Patent].
Jo, C.Y., Cho, H.Y., Kim, H.M., 2003. Effect of recrystallisation on microstructural evo-
lution and mechanical properties of single crystal nickel base superalloy CMSX-2 part
1 - microstructural evolution during recrystallisation of single crystal. Mater. Sci.
Technol. 19, 1665–1670.
Jo, C.Y., Kim, H.M., 2003. Effect of recrystallisation on microstructural evolution and
mechanical properties of single crystal nickel based superalloy CMSX-2 part 2 - creep
behaviour of surface recrystallised single crystal. Mater. Sci. Technol. 19, 1671–1676.
Li, J., Xiong, J., Tang, D., 2012. Advanced High Temperature Structural Materials and
Technology. National Defense Industry Press, Beijing (In Chinese).
Li, Z., Fan, X., Xu, Q., Liu, B., 2015a. Influence of deformation temperature on re-
crystallization in a Ni-based single crystal superalloy. Mater. Lett. 160, 318–322.
179
Journal of Materials Processing Tech. 259 (2018) 169–179
Li, Z., Xiong, J., Xu, Q., Li, J., Liu, B., 2015b. Deformation and recrystallization of single
crystal nickel-based superalloys during investment casting. J. Mater. Process.
Technol. 217, 1–12.
Li, Z., Xu, Q., Liu, B., 2015c. Microstructure simulation on recrystallization of an as-cast
nickel based single crystal superalloy. Comp. Mater. Sci. 107, 122–133.
Li, Z., Xu, Q., Liu, B., 2016. Experimental investigation on recrystallization mechanism of
a Ni-base single crystal superalloy. J. Alloys Compd. 672, 457–469.
Liu, Y., Lin, Y.C., Li, H., Wen, D., Chen, X., Chen, M., 2015. Study of dynamic re-
crystallization in a Ni-based superalloy by experiments and cellular automaton
model. Mater. Sci. Eng. A 626, 432–440.
Ma, D., Wu, Q., Andreas, B.P., 2012. Some new observations on freckle formation in
directionally solidified superalloy components. Metall. Mater. Trans. B 43, 344–357.
Mathur, H.N., Panwisawas, C., Jones, C.N., Reed, R.C., Rae, C.M.F., 2017. Nucleation of
recrystallisation in castings of single crystal Ni-based superalloys. Acta Mater. 129,
112–123.
Ning, Y., Wang, Y., Li, H., Ding, S., Liu, J., Method for enhancing strength of core in single
crystal blade by metal chaplet. CN103056313A. Apr. 24, 2013. [Patent].
Panwisawas, C., Mathur, H., Gebelin, J., Putman, D., Rae, C.M.F., Reed, R.C., 2013.
Prediction of recrystallization in investment cast single-crystal superalloys. Acta
Mater. 61 (1), 51–66.
Porter, A., Ralph, B., 1981. The recrystallization of nickel-base superalloys. J. Mater. Sci.
16, 707–713.
Rettberg, L.H., Pollock, T.M., 2014. Localized recrystallization during creep in nickel-
based superalloys GTD444 and René N5. Acta Mater. 73, 287–297.
Reyes, L.A., Páramo, P., Zamarripa, A., Garza, M., Guerrero Mata, M.P., 2015. Grain size
modeling of a Ni-base superalloy using cellular automata algorithm. Mater. Des. 83,
301–307.
Roters, F., Eisenlohr, P., Hantcherli, L., Tjahjanto, D.D., Bieler, T.R., Raabe, D., 2010.
Overview of constitutive laws, kinematics, homogenization and multiscale methods
in crystal plasticity finite-element modeling: theory, experiments, applications. Acta
Mater. 58, 1152–1211.
Salkeld, R., Field, T., Ault, E., Preparation of single crystal superalloys for post-casting
heat treatment. US005413648A. May 9, 1995. [Patent].
Shi, Z., Liu, S., Yu, J., Li, J., 2015. Tensile behavior of the second generation single crystal
superalloy DD6. J. Iron Steel Res. Int. 22, 738–742.
Wang, L., Xie, G., Zhang, J., Lou, L.H., 2006. On the role of carbides during the re-
crystallization of a directionally solidified nickel-base superalloy. Scripta Mater. 55,
457–460.
Xiong, J., Li, J., Zhao, J., Liu, S., Dong, J., 2009. Characteristic and formation mechanism
of precipitates at recrystallization grain boundaries of single crystal superalloy DD6.
Acta Metall. Sin. 10, 1232–1236.
Xiong, X., Quan, D., Dai, P., Wang, Z., Zhang, Q., Yue, Z., 2015. Tensile behavior of
nickel-base single-crystal superalloy DD6. Mater. Sci. Eng. A 636, 608–612.
Zambaldi, C., Roters, F., Raabe, D., Glatzel, U., 2007. Modeling and experiments on the
indentation deformation and recrystallization of a single-crystal nickel-base super-
alloy. Mater. Sci. Eng. A 454, 433–440.
Zhang, B., Lu, X., Liu, D., Tao, C., 2012. Influence of recrystallization on high-temperature
stress rupture property and fracture behavior of single crystal superalloy. Mater. Sci.
Eng. A 551, 149–153.
Zhuo, L., Xu, T., Wang, F., Xiong, J., Zhu, J., 2015. Microstructural evolution on the
initiation of sub-solvus recrystallization of a grit-blasted single-crystal superalloy.
Mater. Lett. 148, 159–162.