Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 177 (2017) 263 – 270

XXI International Polish-Slovak Conference “Machine Modeling and Simulations 2016

Modeling of a sintering process at various scales

Jerzy Rojek,*, Szymon Nosewicz, Marcin Maździarz, Piotr Kowalczyk,
Krzysztof Wawrzyk, Dmytro Lumelskyj
Institute of Fundamental Technological Research, Polish Academy of Sciences, Pawińskiego 5B, 02-106 Warsaw, Poland

Abstract

This paper presents modeling of a sintering process at various scales. Sintering is a powder metallurgy process consisting in
consolidation of powder materials at elevated temperature but below the melting point. Sintering models at the atomistic,
microscopic and macroscopic scales have been presented. Sintering is a process governed by diffusion therefore the atomistic
modeling using the molecular dynamics has been focused on investigation of the diffusion process. The micromechanical model
has been developed within the framework of the discrete element method. It allows us to consider microstructure and its changes
during sintering. The macroscopic model is based on the continuum phenomenological approach. It combines elastic, thermal and
viscous creep deformation. The methodology to determine macroscopic quantities: stress, strains and constitutive viscous
properties from the discrete element simulations has been presented. Possibilities of the developed models have been
demonstrated by applying them to simulation of sintering of the intermetallic NiAl powder. Own experimental results have been
used to calibrate and validate numerical models.
©
© 2017
2017The TheAuthors. Published
Authors. by Elsevier
Published Ltd. This
by Elsevier Ltd. is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of MMS 2016.
Peer-review under responsibility of the organizing committee of MMS 2016
Keywords: sintering; modeling; discrete element method; diffusion; molecular dynamics; macroscopic model;

1. Introduction

Sintering is a technique of powder metallurgy in which solid parts are manufactured from metal or ceramic
powder mixtures. Sintering consists in consolidation of loose or weakly bonded powders at elevated temperatures,
close to the melting temperature with or without additional pressure. During sintering a particulate material is
converted into a solid compact body. At the macroscopic level, one can observe a change of the mechanical

* Corresponding author. Tel.: (+48) 22 826 12 81 147.

E-mail address: jrojek@ippt.pan.pl

1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of MMS 2016
doi:10.1016/j.proeng.2017.02.210
264 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270

properties (a material acquires a certain mechanical strength), a change of geometry and volume (shrinkage) and an
increase of bulk density (due to shrinkage).
The macroscopic changes result from processes undergoing at the microscopic level. Figure 1 shows a
microstructure of the intermetallic NiAl at an early stage of sintering. The microstructure during sintering undergoes
an evolution characterized by grain rearrangement and increase of grain compaction. In the initial stage, cohesive
bonds (necks) are formed between powder particles. When the sintering process is continued the necks between
particle grow due to mass transport. Surface and grain boundary diffusion are normally dominant mechanisms of
mass transport in sintering. The stresses in the neck and the surface tension induce particle attraction, which leads to
shrinkage of the system. The neck growth and shrinkage conduct to the reduction and sometimes to the practical
elimination of material porosity.

Fig. 1. Microstructure of NiAl at an early stage of sintering.

Sintering is a complex process and the quality of the sintered components depends on many factors, cf. [1].
Numerical modeling can predict evolution of a sintering process and possible defects of sintered parts, thus helping
to design and optimize the process. Sintering phenomena can be modelled at various scales. Different sintering
models are reviewed in [2–5].
The present paper will present numerical models of sintering phenomena at three scales: atomistic, microscopic
and macroscopic ones. Connections and transitions between the models at different scales will be described.
Analysis of a real sintering process of the intermetallic NiAl powder will demonstrate capabilities of the developed
models and formulations.

2. Atomistic modeling of sintering phenomena

There have been attempts to simulate directly particle sintering using the molecular dynamics, e.g. [6], but such
simulations are limited to nanoparticles. Typical particle diameter in such simulations reaches a few nanometers
while the particle diameter in the considered NiAl powder is three orders of magnitude larger. Atomistic simulations
of sintering of such particles would require enormous computational effort. Therefore atomic and atomistic
simulations of sintering mechanisms are usually aimed at determining the diffusivity of a given material system.
Similarly, the present work is focused at studies of diffusion as the main mechanism of sintering. The main objective
of these studies is to estimate parameters for the micromechanical particle-based model of NiAl sintering.
The solid state sintering is modelled in the present work, thus, the diffusion occurring in solids is studied.
Diffusion in solids is very slow, however, at elevated temperatures, it gets much faster. The sintering temperature is
usually in the range 0.7–0.9 of the melting point, high enough for an intensive diffusion.
In the continuum approach, intensity of a stationary diffusion process is described by the Fick’s law:
۸ ൌ െ‫݊׏ܦ‬ (1)
where J is the diffusion flux, n is the scalar concentration field, ‫ ݊׏‬is the concentration-gradient vector, and D is the
diffusion coefficient of the considered species. The form of Eq. (1) is valid for diffusion in isotropic media. The
 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270 265

diffusion flux is given by the number of particles passing through a unit area per unit time (it has the dimension
[m-2s-1)]), the concentration n is defined as the number of particles per unit volume in [m-3], and consequently, the
diffusion coefficient is given in [m2s-1].
Diffusion in crystalline solids occurs by atomic hops in a lattice, either to an interstitial position (interstitial
diffusion) or to a neighbouring vacancy (vacancy diffusion). The energy necessary for an elementary jump is called
the activation energy Q. The activation energy is an important parameter characterizing diffusion. The diffusion
coefficient D is often expressed by an Arrhenius-type relationship:

െܳ (2)
‫ ܦ‬ൌ  ‫ ’š‡ Ͳܦ‬൬ ൰
 ܶ
where ‫ Ͳܦ‬is the pre-exponential factor, T is the absolute temperature and   is the Boltzmann’s constant. The
relationship given by Eq. (2) predicts that the lower the activation energy and the higher temperature are, the higher
the diffusivity is. The diffusivity enters directly as a parameter in the definition of the micromechanical model of
sintering.
Diffusion data can be obtained experimentally but this is neither easy nor cheap. In some cases, complete
information is almost impossible to get. Therefore, possibility to simulate diffusion process at the atomistic level is
important. Molecular Dynamics (MD) provides a straightforward technique to calculate the diffusivity. Since a
diffusion process is mediated by a certain type of point defects, the defects are created in the simulation block, and
their random walk through the model is enabled by a sufficiently long MD run. The coefficient of diffusion can be
calculated from the Einstein-Smoluchowski relation [7] either in the average form:

ͳ ‫ۃ‬ሾ‫ݎ‬ሺ‫ Ͳ ݐ‬൅‫ݐ‬ሻെ‫ݎ‬ሺ‫ Ͳ ݐ‬ሻሿʹ ‫ۄ‬ (3)

‫ ݃ݒܽܦ‬Ǥ ൌ Ž‹‫ݔ‬՜λ ,
ʹ݀ ‫ݐ‬

or in the instantaneous one:

ͳ ߲‫ۃ‬ሾ‫ݎ‬ሺ‫ Ͳ ݐ‬൅‫ݐ‬ሻെ‫ݎ‬ሺ‫ Ͳ ݐ‬ሻሿʹ ‫ۄ‬ (4)
‫ ݐݏ݊݅ܦ‬Ǥ ൌ Ž‹‫ݔ‬՜λ ,
ʹ݀ ߲‫ݐ‬

where, d is the dimensionality (d = 3 for volume, d = 2 for surface and grain-boundary diffusion); t is the time and
‫ۃ‬ሾ‫ݎ‬ሺ‫  Ͳݐ‬൅ ‫ݐ‬ሻ െ ‫ݎ‬ሺ‫ Ͳݐ‬ሻሿʹ ‫ۄ‬is the ensemble average mean-squared displacement (MSD).

3. Micromechanical model of sintering

A number of sintering models have been developed within the discrete element method (DEM), e.g. [8–12]. The
discrete element models consider a powder during sintering as a collection of spherical particles (discrete elements)
interacting with one another. This enables consideration of a microstructure of a powder material and its changes
during sintering.
This work employs the authors’ own original viscoelastic model implemented in the in-house DEM code [13].
The particle interaction model in this model is an extension of the particle sintering model developed originally by
[14] and [15]. The two-particle model of sintering [14, 15] is derived considering diffusion as the main mechanism
of sintering. By analysis of mass transport and stresses at the grain boundary between two sintered particles the
following equation for the particle interaction during sintering is obtained:

Sa 4 ª § <· <º (5)

Fn vrn  SJ S «4r ¨1  cos ¸  a sin »
8Deff ¬ © 2¹ 2¼

where Fn is the normal force between two particles, vrn – the normal relative velocity, r – the particle radius, a – the
radius of the interparticle boundary, < – the dihedral angle, γs – the surface energy. The geometrical parameters of
the model are defined in Fig. 2 The value of diffusion parameter directly depends on certain diffusion mechanism,
which is considered at the constitutive model of sintering. Assuming that the grain boundary diffusion is a dominant
mechanism in the neck growth and shrinkage of the system the effective grain boundary diffusion coefficient Deff is
given by the following formula:
266 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270

‫ߗߜ „‰ܦ‬ (6)

‫ ˆˆ‡ܦ‬ൌ
݇‫ܶ ܤ‬

where Dgb is the grain boundary diffusion coefficient with the width δ, Ω is the atomic volume, kB is the Boltzmann
constant and T is the temperature. The model takes into account the thermal expansion – the particle radius r is
calculated as follows:

‫ ݎ‬ൌ  ‫ Ͳݎ‬ሺͳ ൅ ߙሺܶ െ ܶͲ ሻሻ (7)

where T0 is the reference temperature, r0 is the particle radius at the reference temperature and α is the linear thermal
expansion coefficient. The rate of radius change due to thermal expansion is taken into account in calculation of the
normal relative velocity vrn.

Fig. 2. Two-particle model of sintering.

High computational cost is one of the main drawbacks of the discrete element method which impedes its wider
use in engineering practice. Therefore it is important to combine the DEM with a more efficient continuous
macroscopic approach. In the present work, the results of the discrete elements simulations will be used to determine
macroscopic viscous properties used in the formulation of the macroscopic model of sintering.

4. Macroscopic model of sintering

The macroscopic model of sintering used in this work combines the elastic, thermal and viscous deformation
induced by external loading and sintering driving stress. The strain rate ઽሶ is decomposed into the elastic, thermal and
viscous parts, ࢿሶ ‡ , ࢿሶ  and ࢿሶ ˜ respectively:

ઽሶ ൌ ࢿሶ ‡ ൅ ࢿሶ  ൅  ࢿሶ ˜ (8)

The thermal strain rate is given for isotropic material by:

ࢿሶ  ൌ ߙܶሶ۷ (9)

where α is the coefficient of linear thermal expansion and I is the identity tensor.
The constitutive equation can be written as follows:

ɐሶ ൌ  ‫ ׷‬ሺࢿሶ ‡ െ ࢿሶ  െ  ࢿሶ ˜ ሻ ൅ ۷ɐሶ •‹– (10)

where σ is the Cauchy stress tensor, C – the elastic stiffness tensor, ߪ •‹– – the sintering driving stress. Employing
the decomposition of the stresses and strain rates in the viscous element into the deviatoric and volumetric parts the
relationship between the stress and viscous strain rate can be written as follows [16]:

ߪ ൌ †‡˜ሺߪሻ  ൅  ߪ ۷ ൌ ʹߟ• †‡˜ሺࢿሶ ˜ ሻ  ൅  ߟ˜ –”ሺࢿሶ ˜ ሻ۷ ൅ ۷ߪ •‹– (11)

 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270 267

where ߪ is the mean stress

–”ሺߪሻ (12)
ߪ ൌ 
͵
and ߟ• and ߟ˜ are the shear and volumetric viscous moduli, respectively. The inverse relationship can be easily
obtained in the following form:
–”ሺࢿሶ ˜ ሻ †‡˜ሺߪሻ ߪ െ ߪ •‹– (13)
ࢿሶ ˜ ൌ †‡˜ሺࢿሶ ˜ ሻ ൅ ۷ൌ ൅
͵ ʹߟ• ͵ߟ˜
In the phenomenological modeling, constitutive model parameters are obtained by fitting experimental data. The
present work will show an alternative way to determine the macroscopic properties of a sintered material based on
the simulation of sintering process at a lower, microscopic, scale.
First, macroscopic strains and stresses will be obtained by averaging over representative volume elements [17,
18]. Then, assuming that the constitutive relationships (11) and (13) are valid for the effective averaged stresses and
strains the macroscopic moduli can be evaluated as follows:
ԡ†‡˜ሺߪሻԡ ߪ െ ߪ •‹– (14)
ߟ• ൌ ǡ ߟ˜ ൌ
ʹԡ†‡˜ሺࢿሶ ˜ ሻԡ –”ሺࢿሶ ˜ ሻ
where the symbol ԡǤ ԡ denotes the norm of a second order tensor – for the deviatoric parts of the stress and viscous
strain tensors the norms are defined as follows:
ͳ ͳ
ԡ†‡˜ሺࢿሶ ˜ ሻԡ ൌ ሺ†‡˜ሺࢿሶ ˜ ሻǣ †‡˜ሺࢿሶ ˜ ሻሻʹ . (15)
ԡ†‡˜ሺߪሻԡ ൌ ሺ†‡˜ሺߪሻǣ †‡˜ሺߪሻሻʹ ǡ

5. Numerical results

Numerical models described above have been applied to simulation of a sintering process of the NiAl powder.
Sintering of the NiAl at temperature of 1673 K under uniaxial pressure of 30 MPa has been investigated in our
laboratory (see Fig. 1). Diffusivity parameters estimated by molecular dynamics simulations have been used in the
discrete element simulation of sintering as initial data. The final values of micromechanical parameters have been
established by fitting the numerical evolution of the relative density with the experimental results. Finally, discrete
element simulations have been used to determine macroscopic constitutive properties.

5.1. Estimation of diffusivity for NiAl by molecular dynamics simulations

Diffusion process in the NiAl in the temperature range from 1573 to 1673 K (corresponding to the sintering
temperature in laboratory tests) has been studied using the MD program LAMMPS (Large-scale Atomic/Molecular
Massively Parallel Simulator) [19]. The embedded-atom method potential EAM2009 [20] has been used in the
atomistic model. The quality of this potential has been assessed checking the melting point temperature predicted by
the potential and a set of parameters such as the lattice constant, cohesive energy, elastic constants, bulk modulus,
surface energy, defect energies, migration energies calculated for NiAl by molecular statics approach [21, 22].
The grain boundary diffusion has been investigated using the models representing bicrystals of NiAl with
different crystal orientations shown in Fig. 3. Grain boundary diffusivity for different bicrystals are shown in the
form of the Arrhenius plot in Fig. 4. The numerical values of diffusion parameters are given in Table 1. The results
confirm that the grain boundary diffusivity depends on the relative monocrystal orientation.

Table 1. Grain-boundary diffusivity results.

Bicrystal type D0b [m2/s] Q [kJ/mol]
-7
[100 - 110] 3.66·10 208
[100 - 111] 1.94·10-5 214
[110 - 111] 1.73·10-5 218
268 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270

a) b)

c)
Fig. 3. Atomistic models for simulation of grain boundary diffusion in NiAl bicrystal.

Fig. 4. Diffusivity Arrhenius plot.

5.2. Discrete element simulation of sintering

Laboratory tests of sintering of the NiAl powder have been simulated using the discrete element model.
Experimental tests have been performed under pressure of 30 MPa and at temperature of 1673 K. The simulations
have been performed using the DEM model formed by 1751 particles filling a cylinder with fixed walls confining
the powder subject to uniaxial compression by a plate representing the punch. The initial configuration of the
particle assembly before pressing is shown in Fig. 5a.
The model has been created maintaining the original size (mean particle radius 3.97 μm) and size distribution of
powder particles and taking a reduced diameter (200 μm) of a cylinder confining the powder. It has been assumed
that such a reduced geometric model represents correctly sintering process in a real specimen with diameter of 120
mm. This assumption is justified provided the material behavior is uniform in the real specimen volume.
The specimen geometry at different stages of simulation can be seen in Fig. 5. Figure 5b shows the geometry of
the specimen under pressure of 30 MPa before heating. The final geometry of the sintered specimen is presented in
Fig. 5c. A significant height reduction in the process of sintering can be observed. The volumetric shrinkage is
associated with the changes of the bulk density.
Initial analysis of sintering has been performed assuming the average diffusivity data from the MD simulations:
D0b = 1.24‫ڄ‬10-5 m2/s, Q = 213 kJ/mol. Figure 6 shows the numerical density evolution curve in comparison with
experimental data. It can be seen that the numerical simulation with initial diffusivity parameters has predicted a
slower densification. The experimental data have been matched taking the pre-exponential factor of the grain
boundary diffusion D0b as the fitting parameter. A good agreement with experimental results has been obtained
taking D0b = 9‫ڄ‬10-5 m2/s, Q = 213 kJ/mol.
 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270 269

a) b) c)
Fig. 5. DEM simulation of sintering: a) specimen before pressing, b) specimen after pressing and before sintering,
c) specimen after sintering.

Fig. 6. Evolution of the relative density during the process – Fig. 7. Evolution of the relative density for different pressures.
comparison of numerical and experimental results.

Fig. 8. Bulk viscous modulus as a function of the relative density. Fig. 9. Shear viscous modulus as a function of the relative density.

5.3. Determination of macroscopic viscous moduli

The DEM model with diffusivity parameters determined in the previous section has been used in simulations
specially conceived to give macroscopic viscous moduli of the sintered material. For this purpose, the simulations of
NiAl sintering have been carried out for the temperature 1673 and three values of uniaxial pressure: 5, 15 and 30
MPa. The evolution of the relative density during the whole process is plotted in Fig. 7. Using the methodology
described in Sec. 4 the macroscopic viscous moduli have been determined. The bulk and shear viscous moduli
obtained for different pressure levels have been plotted as functions of the relative density in Figs. 8 and 9,
respectively. It can be easily noticed that the curves corresponding to different pressures coincide very well. This
confirms correctness of our simulations. The micromechanical model does not assume any pressure dependence of
the constitutive parameters and this is reflected in the macroscopic constitutive properties. It can be seen from Figs.
270 Jerzy Rojek et al. / Procedia Engineering 177 (2017) 263 – 270

8 and 9 that the numerical calculations predict an increase of the viscous moduli with increasing density. This
increase becomes faster when the relative density tends to unity.
6. Conclusions
Possibilities of modeling of sintering phenomena have been presented. Modeling at the atomistic scale using the
molecular dynamics enables estimation of the diffusivity parameters which can be used in the micromechanical
particle model of sintering. The diffusivity parameters require certain fitting to match numerical and experimental
results. Micromechanical discrete element simulations can be used to determine macroscopic constitutive properties.
The presented models and inter-scale transitions form a framework for multiscale modeling of sintering.
Acknowledgements
This project has been financed from the funds of Polish National Science Centre (NCN) awarded by decision
numbers DEC-2013/11/B/ST8/03287.

References

[1] M. Chmielewski, D. Kaliński, K. Pietrzak, and W. Włosiński. Relationship between mixing conditions and properties of
sintered 20AlN/80Cu composite materials. Archives of Metallurgy and Materials, 55 (2010) 579–585.
[2] J.-M. Missiaen. Modelling of sintering: recent developments and perspectives. Rev. Met. Paris, 12 (2002) 1009–1019.
[3] J. Pan. Modelling sintering at different length scales. International Materials Reviews, 2,17 (2003) 69–85.
[4] H.E. Exner and T. Kraft. Review on computer simulations of sintering processes. In Powder Metallurgy World Congress 2,
1998, pp. 278–283, EPMA, Shrewsbury, U.K.
[5] D. Gendron. Edtude numerique et experimentale du frittage a lechelle du grain. PhD thesis, L’Universite Bordeaux I, 2001.
[6] Brian J. Henz, Takumi Hawa, and Michael Zachariah. Molecular dynamics simulation of the kinetic sintering of Ni and Al
nanoparticles. Molecular Simulation, 35 (2009) 804–811.
[7] H. Mehrer. Diffusion in Solids - Fundamentals, Methods, Materials, Diffusion-Controlled Processes. Springer, 2007.
[8] F. Parhami and R.M. McMeeking. A network model for initial stage sintering. Mechanics of Materials, 27 (1998) 111–124.
[9] C.L. Martin, L.C.R. Schneider, L. Olmos, and D. Bouvard. Discrete element modeling of metallic powder sintering. Scripta
Materialia, 55 (2006) 425–428.
[10] B. Henrich, A.Wonisch, T. Kraft, M. Moseler, and H. Riedel. Simulations of the influence of rearrangement during
sintering. Acta Materialia, 55 (2007) 753–762.
[11] A. Wonisch, T. Kraft, M. Moseler, and H. Riedel. Effect of different particle size distributions on solid-state sintering: A
microscopic simulation approach. J. Am. Ceram. Soc., 92 (2009) 1428–1434.
[12] S. Martin, M. Guessasma, J. Lechelle, J. Fortin, K. Saleh, and F. Adenot. Simulation of sintering using a Non Smooth
Discrete Element Method. Application to the study of rearrangement. Computational Materials Science, 84 (2014) 31–39.
[13] S. Nosewicz, J. Rojek, K. Pietrzak, and M. Chmielewski. Viscoelastic discrete element model of powder sintering. Powder
Technology, 246 (2013) 157–168.
[14] R.L. Coble. Sintering of Crystalline Solids. Intermediate and Final State Diffusion Models. J.Appl.Phys., 32(1961) 787–792.
[15] D.L. Johnson. New method of obtaining volume, grain boundary, and surface diffusion coefficients from sintering data. J.
Appl. Phys., 40 (1969) 192–200.
[16] V.V. Skorokhod, E.A. Olevsky, and M.B. Shtern. Continuum Theory for Sintering of Porous Bodies: Model and
Application. Science of Sintering, 32 (1991) 79–91.
[17] K. Bagi. Analysis of microstructural strain tensors for granular assemblies. Int. J. Solids & Structures, 43 (2006) 3166–3184.
[18] S. Luding. Micro-macro transition for anisotropic, frictional granular packings. Int. J. Solids & Structures, 41, (2004) 5821–
5836.
[19] S. Plimpton. Fast Parallel Algorithms for Short-Range Molecular Dynamics. J. Comp. Physics, 117 (1995) 1–19.
[20] G.P. Purja Pun and Y. Mishin. Development of an interatomic potential for the Ni-Al system. Philosophical Magazine, 89
(2009) 3245–3267.
[21] M. Maździarz, T.D. Young, and G. Jurczak. A study of the affect of prerelaxation on the nanoindentation process of
crystalline copper. Archives of Mechanics, 63, 2011, pp. 533.
[22] M. Maździarz, T.D. Young, P. Dłużewski, T. Wejrzanowski, and K.J. Kurzydlowski. Computer modelling of
nanoindentation in the limits of a coupled molecular–statics and elastic scheme. J. Comp. Theor. Nanosci., 7, 2010, pp. 1172.