Sunteți pe pagina 1din 13

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/280313302

Multiscale modeling of the mechanical


behavior of IN718 superalloy based on
micropillar compression and computational...

Article in Acta Materialia October 2015


DOI: 10.1016/j.actamat.2015.07.006

CITATIONS READS

3 91

9 authors, including:

Bin Gan Koldo Ostolaza


Northwestern Polytechnical University Industria de Turbopropulsores
19 PUBLICATIONS 93 CITATIONS 25 PUBLICATIONS 127 CITATIONS

SEE PROFILE SEE PROFILE

Jon Molina-Aldareguia Javier Segurado


Madrid Institute for Advanced Studies Universidad Politcnica de Madrid
126 PUBLICATIONS 1,976 CITATIONS 76 PUBLICATIONS 2,312 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

MICROMECH (Microstructure based material mechanical models for superalloys) View project

ExoMet: Physical processing of molten light alloys under the influence of external fields (EC FP7
Large-Scale Collaborative Project) View project

All content following this page was uploaded by Javier Segurado on 11 July 2016.

The user has requested enhancement of the downloaded file.


Acta Materialia 98 (2015) 242253

Contents lists available at ScienceDirect

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

Multiscale modeling of the mechanical behavior of IN718 superalloy


based on micropillar compression and computational homogenization
A. Cruzado a,1, B. Gan a,1, M. Jimnez a, D. Barba a, K. Ostolaza b, A. Linaza b, J.M. Molina-Aldareguia a,
J. Llorca a,c,, J. Segurado a,c
a
IMDEA Materials Institute, C/Eric Kandel 2, 28906 Getafe, Madrid, Spain
b
Industria de Turbo Propulsores, 48170 Zamudio, Bizkaia, Spain
c
Department of Materials Science, Polytechnic University of Madrid, E.T.S. de Ingenieros de Caminos, 28040 Madrid, Spain

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

Article history: A multiscale modeling strategy is presented to determine the effective mechanical properties of polycrys-
Received 8 June 2015 talline Ni-based superalloys. They are obtained by computational homogenization of a representative
Revised 2 July 2015 volume element of the microstructure which was built from the grain size, shape and orientation
Accepted 2 July 2015
distributions of the material. The mechanical behavior of each grain was simulated by means of a crystal
Available online 23 July 2015
plasticity model, and the model parameters that dictate the evolution of the critical resolved shear stress
in each slip system (including viscoplastic effects as well as self and latent hardening) were obtained
Keywords:
from compression tests in micropillars milled from grains of the polycrystal in different orientations
Multiscale modeling
Ni-based superalloys
suited for single, double (coplanar and non coplanar) and multiple slip. The multiscale model predictions
Crystal plasticity of the compressive strength of wrought IN718 were in good agreement with the experimental results.
Micropillar 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Computational homogenization

1. Introduction Polycrystalline IN718 presents a strength well above 1000 MPa,


large ductility (>30%), a marked tension-compression strength
Polycrystalline IN718 is a NiFe based superalloy strengthened differential [9], softening under cyclic loading conditions [10] and
with c0 and c00 precipitates, which is widely used for structural signicant losses of strength above 650 C. These properties are con-
applications up to 650700 C because of its good castability and trolled by the microstructural features of the material (grain size,
weldability, high mechanical properties and corrosion resistance precipitate size and volume fraction, etc.) but physically-based
[1]. The microstructure of IN718 is made up by a c phase (Ni FCC models capable of establishing a quantitative relationship between
solid solution) which contains a dispersion of nm-sized the microstructure and the mechanical properties are scarce. This is
c0 -Ni3(Al,Ti) and c00 -Ni3Nb coherent precipitates within the grains important from the fundamental viewpoint as well as from perspec-
together with lm-sized metal carbides and d phase (Ni3Nb) parti- tive of including the microstructural features into the design of
cles at grain boundaries [24]. The c0 phase presents an ordered structural components.
cubic (FCC) L12 structure, while the c00 phase shows a body centered Most of the models for the mechanical behavior of polycrys-
tetragonal (BCT) D022 structure [5] and forms thin ellipsoidal talline IN718 are phenomenological [9,10]. More recently, Fisk
disc-shape precipitates. The c00 phase (which provides most of the and Lundbckl [11] proposed a physically-based ow stress model
strengthening [6]) is a metastable form of Ni3Nb, which tends to to evaluate the residual stresses obtained during welding and heat
stabilize to the orthorhombic d phase with the D0a structure treatment of IN718. The parameters of the dislocation-based model
preferentially at grain boundaries [7]. The volume fractions of were obtained from the stressstrain curves of the polycrystalline
c0 and c00 phases are in the range 35% and 1020% respectively, material at different temperatures, taking into account the
depending on the bulk alloy composition, the heat-treatment and strengthening effects of precipitates. Following this work, Fisk
the degree of element segregation. In general the weight fraction et al. [12] developed a model that includes the evolution
of (c00 + c0 + d) is around 20% [8]. (nucleation, growth and coarsening) of precipitate size and volume
fraction in order to simulate the aging process of IN718. These
models [11,12] were a rst attempt to link the microstructure with
Corresponding author at: IMDEA Materials Institute, Spain. the macroscopic mechanical but the connection of the single crys-
E-mail address: javier.llorca@imdea.org (J. Llorca). tal deformation with the macroscopic ow was carried out through
1
These authors contributed equally to this work.

http://dx.doi.org/10.1016/j.actamat.2015.07.006
1359-6454/ 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
A. Cruzado et al. / Acta Materialia 98 (2015) 242253 243

the Taylor factor and this simplication limited the scope of the behavior is size independent and representative of the single crystal
approach as microscopic stress and strain elds within each grain behavior within the polycrystalline material. This hypothesis seems
were not considered. reasonable in the case of IN718 because the internal length scale
Polycrystalline homogenization [13] offers an alternative controlling the strength, i.e., the c00 + c0 precipitate spacing, is much
approach for developing models of the mechanical behavior based smaller than the micropillar dimensions.
on the microstructure. Within this framework, the grain structure The multiscale modeling approach based on micropillar com-
is considered through the denition of a representative volume pression tests and computational homogenization is developed
element (RVE), while the single-crystal behavior is accounted for and validated on a precipitation-hardened coarse-grain IN718
by a crystal plasticity model. In particular, computational homog- wrought alloy. The room temperature single-crystal behavior was
enization models [14,15] (based on the resolution of a boundary experimentally measured by compressing single-crystal micropil-
volume problem on an RVE using either nite elements or fast lars machined by focused ion beam (FIB) out of individual grains
Fourier transform) can account for the inuence of the grain size, with different crystallographic orientations. The results of these
shape and orientation distributions on the mechanical properties. tests were used to calibrate a CP model of the c-Ni phase that
The accuracy of these models is strongly dependent on the accounts for the effect of the c00 + c0 precipitation hardening.
single-crystal behavior, which can be represented by a Finally, the polycrystal behavior was obtained by computational
physically-based constitutive model [1619] or by simpler phe- homogenization of an RVE of the microstructure that accounted
nomenological crystal plasticity (CP) model [2022]. In any case, for the actual grain size, shape and orientation distributions. The
the determination of the single-crystal behavior still remains the multiscale model predictions were in good agreement with the
main issue in polycrystalline homogenization. Different macroscopic response of the polycrystal.
approaches can be used to obtain the single crystal properties:
from a multiscale bottom-up approach ranging from the heteroge-
neous subgrain microstructure to the grain mesoscale [18], to an 2. Experimental methodology
inverse analysis strategy in which the single crystal behavior is
chosen to reproduce the results of a set of mechanical tests in poly- 2.1. Microstructural characterization
crystals [1923] or of a set of indentation tests in single crystals
[24,25]. Neither strategy is fully convincing as multiscale modeling The microstructure of the IN718 wrought alloy was character-
strategies are not yet mature enough and the extrapolation of the ized by optical microscopy, scanning electron microscopy (SEM),
parameters obtained by inverse analysis strategies to other scenar- using a FEI Helios Nanolab 600i instrument, and electron backscat-
ios (different from the one used in the inverse analysis) can be ter diffraction (EBSD) in an Oxford AZTEC system. The microstruc-
questioned. ture was mainly composed of c phase grains, with negligible
In this investigation, an alternative approach is proposed to amounts of dispersed metal carbides (<1 vol.%), Fig. 1a. The average
obtain the single-crystal behavior based on the use of micropillar grain size was 95 lm and the grain orientations presented a ran-
compression tests within individual grains from the polycrystalline dom crystallographic texture, as shown in the inverse pole gure
microstructure [2628]. Micropillar compression allows the direct (IPF) map of Fig. 1b. Negligible amounts of d phase particles
determination of the single-crystal stressstrain behavior without (<1 vol.%) were found, all decorating the grain boundaries
the need of complex inverse analysis, as in the case of inverse anal- (Fig. 1c). Transmission electron microscopy analysis, performed
ysis strategies based on indentation. However, it is well known that using a JEOL 3000F instrument, revealed the presence of profuse
the mechanical behavior of micropillars might be affected by their amounts of c0 and c00 precipitates in the interior of the grains with
size [29], and this approach would only be reliable if the micropillar sizes below 50 nm (Fig. 1d).

(a) (b)

Carbides

250 m 125 m
(c) (d)
phases

phases

phases
20 m 20 nm

Fig. 1. Microstructure of wrought IN718 superalloy. (a) Optical image showing the homogeneous distribution of metal carbides throughout the sample. (b) Inverse pole gure
showing the grain orientation distribution. (c) SEM micrograph showing the decoration of grain boundary by d phase. (d) TEM micrograph showing the distribution of c0 and
c00 precipitates within the grain.
244 A. Cruzado et al. / Acta Materialia 98 (2015) 242253

Based on the microstructural observations, the metal carbide The micropillar aspect ratio (length/diameter) was 2.4 to avoid
and d phase content was rendered negligible and the material buckling during compression [30] and the micropillar diameter
was modeled as a polycrystalline aggregate of c grains, containing was varied between 1 and 18 lm to determine the range of
a ne dispersion of c0 and c00 precipitates. Thus, the micropillar micropillar sizes that rendered a size independent response. One
compression tests in single crystals provided the mechanical of the as-milled micropillars, with 5 lm in diameter and 12 lm
response of the c0 and c00 strengthened c-Ni phase. in length, is depicted in Fig. 3a. The annular milling parameters
used resulted in a minimum taper (<1.5) of the pillars, as shown
in the FIB cross-section of Fig. 3b.
2.2. Mechanical characterization
Most of the pillars were compressed using a circular diamond
at punch of 10 lm in diameter inside an instrumented nanoin-
The compressive stressstrain behavior of the wrought IN718
dentation system (Hysitron TI950). Tests were carried out in dis-
alloys was determined using a universal testing machine. Tests
placement control at three different strain rates (104, 103 and
were conducted at room temperature and a constant strain rate
102 s1). Only pillars with diameter <7 lm (which constitute the
of 5.0  104 s1 on cylindrical specimens 6 mm in diameter and
majority of the results included in this paper) could be compressed
9 mm in length.
in this system because of the at punch diameter. Larger pillars (up
to 18 lm in diameter) were tested in a Micromaterials Nanotest
2.3. Micropillar compression tests system (maximum load of 500 mN) with a circular diamond at
punch of 30 lm in diameter. The experimental loaddisplacement
The CP model of the single crystal was calibrated from the curves were corrected for the extra compliance associated with the
results of micropillar compression tests machined out of individual elastic deection of the matrix at the base of the pillar using the
grains within the polycrystal. The micropillars were milled in the Sneddon correction [31]. They were transformed in either engi-
center of selected grains (Fig. 2a) by means of FIB (FEI Helios neering stressstrain curves or engineering resolved shear stress
Nanolab 600i) following an annular milling strategy, with a nal strain curves from the initial length and upper diameter of the pil-
polishing step using a current of 230 pA to minimize FIB lar and the initial Schmid factor calculated from the crystallo-
surface damage. Special care was taken to ensure that the resulting graphic orientation of the pillar.
pillars were completely embedded in individual grains to avoid
grain boundary effects. The crystallographic orientation of the 3. Multiscale modeling strategy
pillars was determined by EBSD (Fig. 2b). A wide range of crystal-
lographic orientations were tested, and they are shown in the 3.1. Crystal plasticity model
reduced inverse pole gure (IPF) of Fig. 2b. Plastic deformation in
FCC c-Ni phase occurs on the octahedral {1 1 1} h1 1 0i slip A detailed description of the CP constitutive model and its
systems and the micropillar orientations were selected to promote implementation is provided in [14]. It is briey outlined here for
single slip, double (coplanar and non-coplanar) and multiple slip. completeness. The CP model is based on the multiplicative

Fig. 2. (a) SEM micrograph of polycrystalline IN718 with the micropillars machined out in the center of various grains. (b) IPF showing the orientation of the micropillars.

Fig. 3. (a) SEM micrograph of a micropillar machined from the IN718 polycrystalline specimen. (b) Cross-section of the micropillar showing the upper and lower diameter
and the height.
A. Cruzado et al. / Acta Materialia 98 (2015) 242253 245

decomposition of the deformation gradient, F, into its elastic Fe and hardening modulus at large strains and ca is the accumulated shear
plastic part Fp according to strain in all slip systems, which is given by
Z tX
F Fe Fp 1
ca jc_ a jdt 10
The plastic velocity gradient Lp in the intermediate (relaxed) 0 a
conguration is dened as the sum of the shear rates, c_ a , on all
the slip systems a according to 3.2. Single-crystal CP constitutive model
N slip
1 X
Lp F_ p Fp c_ a sa0  ma0 2 The CP model was implemented in a user subroutine UMAT for
a1 the c0 and c00 strengthened c-Ni phase. Slip trace analysis in poly-
where sa0 and ma0 are the unit vectors in the slip direction and the crystalline IN718 [33], as well as the micropillar compression tests
normal to the slip plane in the reference conguration respectively. that will be presented below, indicate that plastic deformation
The elastic strain is dened using the Green-Lagrange strain only took place in octahedral {1 1 1} h1 1 0i slip systems. This is con-
tensor, Ee, is expressed as sistent with the FCC structure of the c phase and it is presumably
due to the low volume fraction (<20%) of c0 and c00 small precipi-
1 eT e tates, as opposed to the high volume fraction (>50%) of large c0 pre-
Ee F F  I 3
2 cipitates observed in other Ni-based superalloys where
where I stands for the second order identity tensor. The symmetric pseudo-cubic systems operate [34]. Therefore, only the 12 octahe-
second Piola-kirchhoff stress tensor in the intermediate congura- dral {1 1 1} h1 1 0i slip systems were implemented in the CP model.
tion, S, is related with the Green-Lagrange strain tensor according to The single crystal elastic properties were obtained from literature
[21] and can be found in Table 1. The parameters that control the
S CEe 4 plastic deformation of each slip system s0 , ss , h0 , hs and qab were
where C stands for the fourth order elastic stiffness tensor on the obtained by comparison of the experimental stress-strain curves
single crystal. Thus, the resolved shear stress sa on the slip plane of micropillars of 5 lm in diameter in different orientations with
a is obtained as the projection of the Kirchoff stress on the slip sys- the nite element simulations of the micropillar compression tests
tem according to presented below.
 
1 eT e
sa C F F  I : sa0  ma0 5 3.3. Finite element model of micropillar compression
2
The crystal is assumed to behave as an elasto-viscoplastic solid The FE model used for the simulation of the micropillar com-
in which the plastic slip rate for a given slip plane system follows a pression tests is shown in Fig. 4. The model represents the geome-
power law dependency try of the micropillars with 5 lm in diameter, including the radius
 a 1=m of curvature of the llet (approximately r = 1.5 lm) and the taper
js j angle b  1.5, as measured in Fig. 3b. Including the taper is crucial
c_ a c_ 0 sgnsa 6
sac because it is has been shown that neglecting taper might result in
an overestimation of the elastic modulus and of strain hardening
where c_ 0 stands for the reference strain rate, sac is the critical
and an increase of the apparent yield stress [35].
resolved shear stress (CRSS) of ath slip system at the reference
The micropillar and the supporting material were modeled as a
strain rate and m the rate sensitivity parameter.
single crystal. Eight node lineal brick elements (C3D8 with full
The evolution of the CRSS of a given slip system a, sac , is
integration in Abaqus [36]) were used for discretization. The total
expressed as,
number of elements in the model was 15359 elements, following
X
s_ ac hab jc_ b j 7 a mesh convergence analysis. The at punch was modeled as a
b rigid body, with a lateral stiffness of 10 lN/nm, which approxi-
mately corresponds to the lateral stiffness of the indenter used in
where b stands for any slip system. hab is the strain hardening mod-
the experiment [37]. The contact between the at punch and the
ulus due to self and latent hardening, which can be expressed in a micropillar head was modeled according to a Coulomb friction
simplied manner according to
hab qab h 8

where qab are the interaction coefcients that represent the inu-
ence of the hardening between different slip systems, and h corre-
sponds to the self hardening modulus. The evolution of the self
hardening was described according to the Voce hardening model
proposed in [32],
 
h0 hs ca
hca hs h0  hs exph0 ca =ss s0 9
ss  s0
where h0 is the initial hardening modulus, s0 the initial yield shear
stress, ss the saturation yields shear stress, hs the saturation

Table 1
Elastic constants of the cubic IN718 single crystal [21].

C11 (GPa) C12 (GPa) C44 (GPa)


259.6 179 109.6
Fig. 4. Finite element model of the micropillar compression tests.
246 A. Cruzado et al. / Acta Materialia 98 (2015) 242253

for the tessellation, so the grain size distribution in the RVE fol-
lowed the experimental grain size distribution (Fig. 5). The details
of the algorithm developed can be found in Appendix A.
The Voronoi tessellation was carried out with Neper [40] from
an extended cloud of points obtained by a periodic copy in the
three directions of space of the point distribution obtained by the
Monte-Carlo method in order to preserve the periodicity of the
microstructure in the RVE. The nal shape of the RVE was not cubic
because grains intersecting the cube faces were not cut and copied
into the opposite face but were maintained in their positions. This
strategy avoided meshing problems related with the development
of very small grains near the cube surfaces. The periodic RVE was
nally meshed with the open source program Gmsh [41].
10-node quadratic tetrahedral elements (C3D10 with full integra-
tion in Abaqus [36]) were used for the discretization and the mesh
quality was very high (only 0.26% of distorted elements). Because
the texture in the material was negligible, the grains orientations
were randomly generated in the rotation group SO3. Finite element
models of RVEs containing from 20 to 472 grains and 615 to 26007
elements per grain were created to analyze the inuence of ele-
ment size and number of grains in the numerical results. One of
Fig. 5. Log-normal grain size distribution of polycrystalline IN718 (solid line) and them, which includes 210 grains, is depicted in Fig. 6.
the corresponding grain size distribution in the RVE of the microstructure. The generated RVEs are periodic along the axes of the original
cube of length L although they do not have a cubic shape because
grains are not limited by the faces of the original cube to avoid dis-
model, with a friction coefcient of 0.1 [37]. The base of the sup- torted elements. Nodes A and B in the same location of opposite
porting material was fully constrained, while a at punch was surfaces can be grouped in pairs of periodic nodes and the initial
moved in the vertical direction at a constant speed. distance between them is given by xB  xA LAB , where
LAB = (L,0,0), (0,L,0) or (0,0,L) depending of the surface orientation.
3.4. Computational homogenization framework Uniaxial compression of the RVE was simulated by applying peri-
odic boundary conditions to each node pair according to
The effective behavior of polycrystalline IN718 was obtained by   ILAB
uB  uA F 11
computational homogenization of an RVE of the microstructure.
where F  stands for prescribed macroscopic deformation gradient
The grain size distribution in the polycrystal was measured from
a cross-section that included approximately 300 grains. The exper- and uA and uB the displacement vectors of the paired nodes. If some
imental 2D grain size distribution was transformed to a 3D distri- components of the far-eld deformation gradient are not known a
bution assuming spherical grains. The open source software priori (mixed boundary conditions, as in uniaxial compression),
StripStar [38] was used to this purpose. The 3D grain radii distribu- the corresponding components of the effective stresses are applied
tion was then approximated by a lognormal function (r = 0.70862, instead following the strategy presented in [23].
l = 3.65) by means of the LevenbergMarquadt algorithm (correla-
tion coefcient of 0.93) and is shown in Fig. 5. The mechanical 4. Results
behavior of each crystal in the RVE was given by the CP model pre-
sented in Sections 3.1 and 3.2. The cubic RVEs were generated from 4.1. Effect of pillar size
the Voronoi tessellation of a set of points, which divided the initial
volume in a number of polyhedral [39]. A Monte Carlo algorithm The effect of pillar size on the experimental stressstrain
was developed to generate the position of the set of points used response was assessed by performing tests on pillars with

Fig. 6. Finite element model with 130000 elements of an RVE containing 210 grains following the lognormal distribution in Fig. 5. A cross-section of the RVE parallel to one of
the cube faces is shown to the right, showing the periodicity of the microstructure.
A. Cruzado et al. / Acta Materialia 98 (2015) 242253 247

Fig. 7. CRSS vs. strain curves of micropillars of different diameter oriented for single slip tested in compression. (a) Diameter range from 1 to 7.5 lm. (b) Diameters of 6 and
18 lm.

diameters in the range 118 lm at an average strain rate of parameters were obtained from the behavior of micropillars of
103 s1. To this end, micropillars with different size were milled 5 lm in diameter.
in grains with the same or similar crystallographic orientation
(close to either h2 4 5i or h2 3 5i) to rule out any plastic anisotropy
4.2. Calibration of the CP model
effects. They were favorably oriented for single slip with Schmid
factors (SF) of 0.452 and 0.445 for the orientations h2 4 5i and
As shown in Fig. 7, the slope of the initial loading decreases with
h2 3 5i respectively. The critical resolved shear stress (CRSS) vs.
micropillar diameter, leading to a much more compliant response
strain curves of 4 micropillars with diameters between 1 and
than that found upon unloading. While the unloading slope yields
7.5 lm are plotted in Fig. 7a and the corresponding SEM
the correct elastic modulus, it is well known that the initial loading
micrographs of the deformed micropillars are shown in Fig. 8ad,
slope in a micropillar tests is strongly affected by asperities of the
which conrmed that plastic deformation took place by single slip.
surface and/or incorrect alignment between the at punch and the
The micropillar of 1 lm in diameter presented a stiffer
head of the pillar [37]. This misalignment induces a stress concen-
initial response and a higher CRSS while the CRSSstrain behav-
tration at the initial contact point of the micropillar, which leads to
ior of the micropillars with diameters between 3 lm and 7.5 lm
early yielding and to the apparent reduction of the initial loading
were practically identical, except for a slight reduction in the slope
stiffness while the unloading slope is free for this artifact once full
of the initial loading slope with pillar diameter, that will be
plastic contact has been established between the at punch and
discussed in more detail in the next section. In all cases, the
the pillar. Unfortunately, this experimental uncertainty makes it
CRSSstrain curves displayed unloading events as a result of
difcult to establish the onset of plastic yielding from the stress
dislocation bursts, upon which the high frequency feedback loop
strain curves, and this information is necessary to calibrate the
of the system reacts relaxing the load to maintain the prescribed
parameters of the CP model. This limitation can be overcome fol-
displacement rate.
lowing the strategy proposed by Kupka et al. [26] to obtain the
In addition to these tests, two new pillars with 6 and 18 lm in
evolution of the stress with the plastic strain in micropillar com-
diameter were milled in grains with h3 2 6i (SF = 0.467) orientation.
pression tests. To this effect, the experimental strain er was rst
The CRSSstrain curves are plotted in Fig. 7b and the SEM micro-
corrected by subtracting the elastic strain, which was obtained for
graphs of the deformed micropillars are depicted in Fig. 8e and f,
each stress level from the unloading slope Sunload according to
showing single slip. These micropillars were tested in a different
nanoindentation system, as described in Section 2.2, due to the r
large load required for the 18 lm pillar. The dynamic response of
e1 r er  12
Sunload
this system was not fast enough to unload the pillar in reaction
The original and the corrected stress-strain curves are shown in
to the dislocation bursts, resulting in repeated pop-in events,
Fig. 9a for the case of a micropillar of 3 lm in diameter oriented for
instead of load relaxations. Except for this, the CRSSstrain curves
single slip in the [3 2 5] direction. After this correction, the unload-
measured in these micropillars were indistinguishable from those
ing segment of the stressstrain curve appears as a vertical line.
measured in micropillars with diameter between 3 and 7.5 lm.
Afterwards, the initial yield point (ry , e1y ) was determined as the
These results conrm the hypothesis that the micropillar response
point at which the maximum slope of the loading curve decreased
can be regarded as independent of the micropillar diameter, and
by 20% with respect to the initial slope, and the corrected plastic
representative of the single-crystal bulk behavior, at least for
strain free of the initial loading artifacts-was given by
micropillars with diameters above 3 lm in the case of IN718.
This behavior presumably occurs because the internal length scale ep r e1 r  e1y 13
controlling the strength, i.e. the c00 + c0 precipitate spacing, is much
smaller (of the order of 50 nm) than the micropillar dimensions. and the corresponding stressplastic strain curve can be found in
However, the CRSSstrain curves in Fig. 7a also showed a clear size Fig. 9b.
effect of the type the smaller the stronger in micropillars of 1 lm The CRSSplastic strain curves corresponding to the micropillar
in diameter. The origin of this behavior is intriguing but is beyond compression tests were obtained using this procedure and they
the scope of this paper. Based on these results, the CP model were used to determine the parameters of the CP model as follows.
248 A. Cruzado et al. / Acta Materialia 98 (2015) 242253

(a) (425) (b) (325)

500 nm 1
m
m

(c) (235) (d) (245)

1.5 m 2 m
(e) (326) (f) (326)

1.5 m 10 m

Fig. 8. SEM micrographs of micropillars with different size deformed in the single slip condition. (a) 1 lm oriented in [4 2 5]. (b) 3 lm oriented in [3 2 5]. (c) 5 lm oriented in
[2 3 5]. (d) 7.5 lm oriented in [2 4 5]. (e) 6 lm oriented in [3 2 6] and (f) 18 lm oriented in [3 2 6].

Fig. 9. Correction of the stressstrain curves of the micropillar compression tests to obtain the stressplastic strain curve. (a) Original and corrected stressstrain curve
according to Eq. (12). The yield point is indicated in the corrected curve. (b) Stressplastic strain curve obtained from Eq. (13). The results in this gure correspond to a
micropillar of 3 lm in diameter oriented for single slip in the [3 2 5] direction and tested at an average strain rate of 103 s1.

According to Eq. (6), the actual CRSS in a given slip system a, sas , where c_ a0 and sac stand for the reference strain rate and the CRSS in
depends on the shear strain rate c_ a in the form the system a at the reference strain rate. The rate sensitivity expo-
 a nent m was obtained from the CRSSplastic strain curves at average
c_ 1 sa
ln ln j sa j 14 strain rate of 102 s1 in h1 2 3i micropillars and at 103 s1 and
c_ a0 m sc 104 s1 in h2 3 5i micropillars oriented for single slip (SF = 0.467
A. Cruzado et al. / Acta Materialia 98 (2015) 242253 249

Fig. 10. (a) CRSS vs. plastic strain of micropillar oriented for single slip tested at different strain rates at room temperature. (b) Strain rate sensitivity parameter m of IN718 at
room temperature as obtained from micropillar compression tests.

Fig. 11. Experimental and simulated CRSS vs. plastic strain curves of compression tests of micropillars of 5 lm in diameter oriented for single slip. (a) Average strain rate
103 s1, micropillar orientation [2 3 5]. (b) Average strain rates 102 s1 and 103 s4, micropillar orientation [1 2 3] and [2 3 5].

Table 2 regression shows a relative poor correlation index with the experi-
Self hardening and latent hardening parameters of the Voce law. mental data.
s0 (MPa) ss (MPa) h0 (GPa) hs (GPa) qab The second step to calibrate the crystal plasticity model was to
obtain the parameters of the Voce law each slip system, namely h0 ,
465.5 598.5 6.0 0.3 1
s0 , ss and hs . To this end, the nite element model of the micropil-
lar compression test, presented in Section 3.3, was used to simulate
the stressstrain curve of h2 3 5i micropillars, oriented favorably for
and SF = 0.451 for h1 2 3i and h2 3 5i, respectively), Fig. 10a. The slip with a SF = 0.451 at an average strain rate of 103 s1. The
average strain rate in each slip system can be computed as the aver- experimental CRSS vs. plastic curve is plotted in Fig. 11a together
age strain rate of the micropillar divided by the Schmid factor and with the results of the numerical simulation obtained with the
the corresponding CRSS was taken as the one at a plastic strain of parameters shown in Table 2. It is worth noting that the values
0.04. The choice of the plastic strain to determine the CRSS was of the initial and saturation CRSSs (s0 and ss ) obtained from the
not very critical as the CRSSplastic strain curves were practically nite element simulation of the micropillar are slightly below
parallel for plastic strains >0.02. These results (normalized by the the ones that would be directly obtained from the CRSSplastic
shear strain rate and the CRSS for the tests carried out at an average strain curve and the Schmid factor because the numerical model
strain rate of 103 s1) are plotted in Fig. 10b in bilogarithmic coor- accounts for the geometrical strain hardening introduced by the
dinates. The strain rate sensitivity parameter m = 0.017 was tapering of the micropillar as well as for lattice rotation during
obtained from the inverse of the slope, the straight line tted by the test. The numerical model can also take into account the slight
the least squares method to these experimental data, according to inuence of the strain rate on the mechanical response of the
Eq. (14). This result indicates that strain rate sensitivity of IN718 micropillars, as shown in Fig. 11b, in which the results of the
at room temperature is very small and it is not surprising that linear numerical simulations are compared with the CRSS vs. plastic
250 A. Cruzado et al. / Acta Materialia 98 (2015) 242253

Fig. 12. (a) Simulated stress vs. plastic strain curves of micropillars oriented for single slip (h2 3 5i) and double slip (h4 1 4i) obtained with qab = 1 and 2. (b) Experimental and
simulation results of micropillars oriented in h0 1 2i (non co-planar double slip) and in h4 1 4i (co-planar double slip). Simulations were carried out with qab = 1. (c) SEM of a
micropillar deformed along h0 1 2i showing non coplanar double slip. (d) SEM of a micropillar deformed along h4 1 4i showing coplanar double slip. All simulations and
experiments in this gure were carried out in micropillars of 5 lm in diameter at an average strain rate of 103 s1.

strain curves measured at 102 and 104 s1 in micropillars ori- which is the stiffest and strongest orientation of the crystal.
ented for single slip in orientations h1 2 3i and h2 3 5i, respectively. Deformation along h1 1 1i leads to the formation of a shear band
The previous simulations were carried out under the assump- (Fig. 13c), presumably as a result of the activation of a secondary
tion that qab = 1 (Eq. 8). The choice of qab was irrelevant in the case slip system during deformation, although the exact sequence of
of micropillars oriented for single slip, as shown in the simulations events leading to this behavior is not known and is currently under
depicted in Fig. 12a for a micropillar of 5 lm in diameter oriented study. Finally, simulation of the deformation along h0 0 1i was very
along h2 3 5i for single slip. Nevertheless, the contribution of latent accurate (Fig. 13b) although this orientation led to multiple slip
hardening was not negligible in the case of double slip and the ow with the activation of 8 simultaneous slip systems (Fig. 13d).
stress of micropillars oriented along h4 1 4i for double slip Finally, the predictive capacity of the CP model can be assessed
depended on qab (Fig. 12a). The actual value of qab was obtained by comparing the SEM of the deformed micropillar after compres-
from the mechanical response of micropillars oriented along sion along the h2 1 2i orientation (where both coplanar slip systems
h4 1 4i and h0 1 2i. h0 1 2i micropillars showed activity in two non are active) (Fig. 14a), with the corresponding predictions of the
coplanar slip systems, {1 1 1} h1 0 1i and {1 1 1}h1 0 1i, with a nite element model (Fig. 14b). The model was able to reproduce
SF = 0.49 (Fig. 12c), while the micropillar oriented in h4 1 4i accurately the shape of the deformed micropillar as well as the
deformed in two coplanar slip systems, {1 1 1}h0 1 1i and location of the slip band.
{1 1 1}h1 1 0i, with SF = 0.43 (Fig. 12d). The simulations carried
out with qab = 1 provided the best approximation to the experi- 4.3. Computational homogenization
mental data (Fig. 12b) indicating that IN718 follows an isotropic
hardening model. The CP model calibrated in the previous section was able to
The CP model of IN718 was validated by comparing the results reproduce the mechanical behavior of IN718 single crystals in dif-
of the micropillar compression tests in different orientations with ferent orientations, which include single, double (coplanar and non
actual experimental results. They are shown in Fig. 13a for coplanar) and multiple slip. This model was used as the constitu-
micropillars of 5 lm in diameter tested at an average strain rate tive equation of the single crystals within a computational homog-
of 103 s1 along h1 2 6i (SF = 0.488) and h4 5 6i (SF = 371) orienta- enization framework to obtain the effective properties of the
tions. Numerical simulations were very close to the experimental polycrystal.
results in h1 2 6i while the simulations in h4 5 6i orientation overes- A sensitivity analysis was rst carried out to assess the inu-
timated the experimental ow stress. This difference may be due ence the mesh size and of the number of grains in the RVE on
to the proximity of the h4 5 6i direction to the h1 1 1i orientation, the predictions of the effective response of the polycrystal.
A. Cruzado et al. / Acta Materialia 98 (2015) 242253 251

(111) (001)
(c) (d)

Fig. 13. (a) Experimental and simulation results of micropillars oriented in h5 4 6i and h2 1 6i orientations. (b) Experimental and simulation results of micropillars oriented in
h0 0 1i orientation (c) SEM micrograph of a square micropillar deformed along h1 1 1i showing the formation of a shear band. (d) SEM micrograph of a micropillar deformed
along h0 0 1i showing multiple slip. All simulations and experiments in this gure were carried out in micropillars of 5 lm in diameter at an average strain rate of 103 s1.

Fig. 14. (a) SEM micrograph of micropillar of 5 lm in diameter deformed up to 0.2 in h2 1 2i orientation at an average strain rate of 103 s1. (b) Contour plot of the
accumulated plastic strain superposed to the deformed mesh obtained from the nite element model.

Numerical simulations were performed with three different mesh log-normal distribution measured experimentally and numerical
sizes with 615, 2466 and 26007 nite element per grain in RVEs simulations were carried out with four different realizations of
containing 21 and 199 grains. The differences in the stress-strain the random grain orientation distribution. The differences in the
curve between the coarse and the ne meshes were always below mechanical response calculated with different realizations were
1% up to an applied strain of 20%. Similarly, simulations were car- below 1.3%, which indicates the error in the effective response of
ried out with RVEs containing 20, 210 and 472 grains and 610 nite the polycrystalline model for a given set of CP parameters.
elements per grain. The differences in the stressstrain curves The effective response obtained by computational homogeniza-
between the models were also very small (<1%). tion is compared with the experimental behavior in Fig. 15 in
Thus, the model with 210 grains and 610 nite elements pre terms of the true stresslogarithmic strain curves. The agreement
grain was selected to predict the mechanical behavior of the between them was fairly good: the maximum difference in the
IN718 polycrystal. The grain size distribution followed the compressive ow stress was below 4% and the strain hardening
252 A. Cruzado et al. / Acta Materialia 98 (2015) 242253

The mechanical behavior in compression of the polycrystal pre-


dicted with this multiscale strategy was in good agreement with
the experimental results and validated the whole strategy. The
methodology presented in this paper, together with the recent
advances in the characterization of the mechanical properties of
lm-sized specimen milled from grains and grain boundaries in
the polycrystal, presents a large potential to carry out virtual tests
of polycrystalline Ni-based superalloys to predict the inuence of
the microstructure on key macroscopic features. This includes,
among others, the strength differential, the effect of temperature
on the strength, the creep and the fatigue resistance, etc.

Acknowledgments

This investigation was supported by the MICROMECH, funded


by the European Union under the Clean Sky Joint Undertaking,
7th Framework Programme (CS-GA-2013-620078). TEM was car-
ried out in the National Center for Electron Microscopy of the
Complutense University of Madrid.

Fig. 15. Experimental result and numerical simulation obtained by computational Appendix A.
homogenization of an RVE of the true stressstrain curve in compression of IN718.

of the model and of the material was identical. This is remarkable An algorithm has been developed to generate the coordinates of
taking into account the scatter associated with the micropillar a set of N points p = pi that can lead by Voronoi tessellation to an
compression tests was around 5%. Moreover, it should be high- RVE whose grain size distribution follows a prescribed one. To this
lighted the multiscale nature of the approach because all the ingre- end, it is necessary to minimize an error function O(p) dened as
dients in the polycrystalline homogenization strategy were X
N
obtained at lower length scales from the microstructure of the Op jPDF exp V j  PDF p V j j DV A1
polycrystal and the mechanical properties of the single crystals j1

within the polycrystal. where PDF exp and PDF p are the probability density functions of the
An interesting consequence of the accuracy of this prediction is experimental and the target grain size distributions, respectively,
that grain boundaries do not contribute to the hardening of this which indicate the probability for a grain to have a volume in the
coarse grained polycrystal. If this mechanism were relevant here, range V j and V j DV, where DV V max V min =N and V max and
the simulated stressstrain curves (that do not account for slip gra-
V min stand for the maximum and minimum grain sizes, respectively.
dients) should have underestimated the experimental ow stress.
An iterative process based on the Monte Carlo method was used
The physical reason of the negligible effect of grain boundaries is
to minimize O(p). Let p(k) the set of points at iteration k, that leads
probably the small distance between precipitates that controls
to an error Opk higher than a prescribed value. The new set p(k+1)
the plastic response of the material, and renders negligible the
is obtained by moving a random point pi in a random direction a
strengthening contribution of grain boundaries for this grain size.
random distance Dx in the range between 0 and 3.5D, where D is
the average grain size of the experimental distribution. A new
5. Conclusions Voronoi tessellation is carried for the set pk1 , and the error func-
tion Opk1 is computed from the new PDF p . Whether or not the
A multiscale modeling strategy has been developed to predict
set pk1 is chosen as the starting point for the next iteration
the mechanical behavior of polycrystalline Ni-based superalloys.
depends on the parameter prob, which is given by
The effective properties of the polycrystalline material were   
obtained by computational homogenization of an RVE of the Opk1  Opk
prob min 1; exp  A2
microstructure that accounts for the grain size, shape and orienta- A Op k

tion distributions. The mechanical behavior of each grain is given


by a crystal plasticity model and the model parameters that dictate where A is a parameter, chosen equal to 0.02 in this case. If the error
the evolution of the CRSS in each slip system (including viscoplas- is reduced, Opk1 6 Opk , prob = 1 and the set pk1 is chosen as
tic effects as well self and latent hardening) were obtained from the starting point for the next iteration. On the contrary,
compression tests in micropillars milled from grains of the poly- prob 2 0; 1 if the error increases and a random number is gener-
crystal in different orientations suited for single, double (coplanar ated, r 2 0; 1. The set pk1 is chosen as the starting point for
and non coplanar) and multiple slip. The ability of the crystal plas- the next iteration if r < prob; otherwise, the old set pk is again
ticity model to reproduce the mechanical behavior of micropillars used as the starting point. This algorithm, that is able to accept dis-
was validated by the accurate prediction of micropillar compres- tributions that slightly increase the error, leads to an improvement
sion tests in different orientations and at various strain rates. In of the efciency of the Monte Carlo method. The process is repeated
addition, the model was able to capture the dominant deformation until the error function Op is smaller than a prescribed tolerance.
mechanisms in the micropillars, including the formation of the slip This algorithm has been programmed in Matlab.
bands observed. It was demonstrated that mechanical behavior of
micropillars was size independent (and, thus representative of the References
behavior of the material within the grains) for micropillars with a
diameter above 3 lm because the ow stress in this material was [1] H.J. Wagner, A.M. Hall, Physical metallurgy of alloy 718. DMIC Report 217,
Battelle Memorial Institute, Columbus, June 1, 1965.
controlled by the distance between nm-sized c0 and c00 precipitates [2] Y.S. Song, M.R. Lee, J.T. Kim, Effect of grain size for the tensile strength and the
dispersed in the solid solution c face. low cycle fatigue at elevated temperature of alloy 718 cogged by open die
A. Cruzado et al. / Acta Materialia 98 (2015) 242253 253

forging press, in: E.A. Loria (Ed.), Superalloys 718, 625, 706 and Derivatives [23] V. Herrera-Solaz, J. Llorca, E. Dogan, I. Karaman, J. Segurado, An inverse
2005, TMS, 2005, pp. 539549. optimization strategy to determine single crystal mechanical behavior from
[3] C. Ruiz, A. Obabueki, K. Gillespie, Evaluation of the microstructure and polycrystal tests: application to AZ31 Mg Alloy, Int. J. Plast. 57 (2014) 115.
mechanical properties of delta processed alloy 718, in: S.D. Antolovich (Ed.), [24] R. Snchez-Martn, M.T. Prez-Prado, J. Segurado, J. Bohlen, I. Gutirrez-
Superalloys 1992, TMS, 1992, pp. 3342. Urrutia, J. Llorca, J.M. Molina-Aldareguia, Measuring the critical resolved shear
[4] J.F. Radavich, The physical metallurgy of cast and wrought alloy 718, in: E.A. Loria stress in Mg alloys by instrumented nanoindentation, Acta Mater. 71 (2014)
(Ed.), Superalloy 718-Metallurgy and Applications, TMS, 1989, pp. 229240. 283292.
[5] H.E. Jianjong, c00 precipitate in Inconel 718, J. Mater. Sci. Technol. 10 (1994) [25] B. Eidel, Crystal plasticity nite-element analysis versus experimental results
293303. of pyramidal indentation into (0 0 1) fcc single crystal, Acta Mater. 59 (2011)
[6] S.J. Hong, W.P. Chen, T.W. Wang, A diffraction study of the c00 phase in Incocel 17611771.
718 superalloy, Metall. Mater. Trans. A 32 (2001) 18871901. [26] D. Kupka, N. Huber, E.T. Lilleodden, A combined experimental-numerical
[7] J. Dong, X. Xie, Z. Xu, S. Zhang, M. Chen, J.F. Radavich, TEM study on approach for elasto-plastic fracture of individual grain boundaries, J. Mech.
microstructure behaviour of alloy 718 after long time exposure at high Phys. Solids 64 (2014) 455467.
temperatures, in: E.A. Loria (Ed.), Superalloys 718, 625, 706 and Various [27] R. Soler, J.M. Molina-Aldareguia, J. Segurado, J. Llorca, R.I. Merino, V.M. Orera,
Derivatives 1994, TMS, 1994, pp. 649658. Micropillar compression of LiF [1 1 1] single crystals: effect of size, ion
[8] Y.C. Fayman, Microstructural characterization and element partitioning in a irradiation and misorientation, Int. J. Plast. 36 (2012) 5063.
direct-aged superalloy (DA718), Mater. Sci. Eng. 92 (1987) 159171. [28] M. Kuroda, Higher-order gradient effects in micropillar compression, Acta
[9] S.K. Iyer, C.J. Lissenden, Multiaxial constitutive model accounting for the Mater. 61 (2013) 22832297.
strength-differential in Inconel 718, Int. J. Plast. 19 (2003) 20552081. [29] R. Soler, J.M. Wheeler, H.-J. Chang, J. Segurado, J. Michler, J. Llorca, J.M. Molina-
[10] D. Gustafsson, J.J. Moverare, K. Simonsson, S. Sjstrm, Modeling of the Aldareguia, Understanding size effects on the strength of single crystals
constitutive behaviour of Inconel 718 at intermediate temperatures, J. Eng. Gas through high temperature micropillar compression, Acta Mater. 81 (2014)
Turb. Power 133 (2011) 14. 5057.
[11] M. Fisk, A. Lundbck, Simulation and validation of repair welding and heat [30] D. Raabe, D. Ma, F. Roters, Effects of initial orientation, sample geometry and
treatment of an alloy 718 plate, Finite Elem. Anal. Des. 58 (2012) 6673. friction on anisotropy and crystallographic orientation changes in single
[12] M. Fisk, J.C. Ion, L.-E. Lindgren, Flow stress model for IN718 accounting for crystal microcompression deformation: a crystal plasticity nite element
evolution of strengthening precipitates during thermal treatment, Comput. study, Acta Mater. 55 (2007) 45674583.
Mater. Sci. 82 (2014) 531539. [31] I. Sneddon, The relation between load and penetration in the axisymmetric
[13] R.A. Lebensohn, C.N. Tome, A selfconsistent approach for the simulation of Boussinesq problem for a punch of arbitrary prole, Int. J. Eng. Sci. 3 (1965)
plastic deformation and texture development of polycrystals: application to 4757.
Zirconium alloys, Acta Metall. Mater. 41 (1993) 26112624. [32] C. Tome, G.R. Canova, U.F. Kocks, N. Christodoulou, J.J. Jonas, The relation
[14] J. Segurado, J. Llorca, Simulation of the deformation of polycrystalline between macroscopic and microscopic strain hardening in FCC polycrystals,
nanostructures Ti by computational homogenization, Comput. Mater. Sci. 76 Acta Metall. 32 (1984) 16371653.
(2013) 311. [33] C.J. Boehlert, H. Li, L. Wang, Slip system characterization of IN 718 using in-situ
[15] R.A. Lebensohn, N-site modelling of a 3D viscoplastic polycrystal using Fast scanning electron microscopy, Adv. Mater. Processes 168 (2010) 4145.
Fourier Transform, Acta Mater. 49 (2001) 27232737. [34] A. Vattre, B. Devincre, A. Roos, Orientation dependence of plastic deformation
[16] F. Roters, P. Eisenlohr, L. Hantcherli, D.D. Tjahjanto, T.R. Bieler, D. Raabe, in nickel-based single crystal superalloys: discretecontinuous model
Overview of constitutive laws, kinematics, homogenization and multiscale simulations, Acta Mater. 58 (2010) 19381951.
methods in crystal plasticity nite-element modelling: theory, experiments, [35] H. Zhang, B.E. Schuster, Q. Wei, K.T. Ramesh, The design of accurate micro-
applications, Acta Mater. 58 (2010) 11521211. compression experiments, Scr. Mater. 54 (2006) 181186.
[17] A. Ma, F. Roters, A constitutive model for fcc single crystals based on [36] ABAQUS. Standard Users Manual Version 6.10, Hibbitt, Karlsson, and Sorensen
dislocation densities and its application to uniaxial compression of aluminium Inc., Pawtucket, Rhode Island, USA, 2010.
single crystals, Acta Mater. 52 (2004) 36033612. [37] R. Soler, J.M. Molina-Aldaregua, J. Segurado, J. Llorca, Effect of misorientation
[18] S. Keshavarz, S. Ghosh, Multi-scale plasticity nite element model approach to on the compression of highly anisotropic single-crystal micropillars, Adv. Eng.
modelling nickel-based superalloys, Acta Mater. 61 (2013) 65496561. Mater. 14 (2012) 10041008.
[19] M. Shenoy, J. Zhang, D.L. McDowell, Estimating fatigue sensitivity to [38] R. Heilbronner, D. Bruhn, The inuence of three-dimensional grain size
polycrystalline Ni-base superalloy microstructures using a computational distributions on the rheology of polyphase rocks, J. Struct. Geol. 20 (1998)
approach, Fatigue Fract. Eng. Mater. Struct. 30 (2007) 889904. 695707.
[20] R.J. Asaro, A. Needleman, Overview no. 42 texture development and strain [39] R. Quey, P.R. Dawson, F. Barbe, Large-scale 3d random polycrystals for the
hardening in rate dependent polycrystals, Acta Metall. 33 (1985) 923953. nite element method: generation, meshing and remeshing, Comput. Methods
[21] G. Martin, N. Ochoa, K. Sa, E. Herv-Luanco, G. Cailletaud, A multiscale model Appl. Mech. Eng. 200 (2011) 17291745.
for the elastoviscoplastic behavior of Directionally Solidied alloys: application [40] Romain Quey, Neper Reference Manual, 2014.
to FE structural computations, Int. J. Solids Struct. 51 (2014) 11751187. [41] C. Geuzaine, J.-F. Remacle, Gmsh: a three-dimensional nite element mesh
[22] C.A. Sweeney, P.E. McHugh, J.P. McGarry, S.B. Leen, Micromechanical generator with built-in pre- and post-processing facilities, Int. J. Numer. Meth.
methodology for fatigue in cardiovascular stents, Int. J. Fatigue 44 (2012) Eng. 79 (2009) 13091331.
212216.

View publication stats

S-ar putea să vă placă și