Computational Materials Science 117 (2016) 397405

Contents lists available at ScienceDirect

Computational Materials Science

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

An atomistic study of the correlation between the migration of planar

and curved grain boundaries
A.T. Wicaksono , C.W. Sinclair, M. Militzer
Department of Materials Engineering, The University of British Columbia, 309-6350 Stores Road, Vancouver, BC V6T 1Z4, Canada

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

Article history: Molecular dynamics simulations were performed to investigate the migration of curved and planar
Received 17 December 2015 boundaries. The reduced mobilities of capillarity driven U-shaped half-loop twin boundaries in b.c.c iron
Received in revised form 9 February 2016 (Fe) were computed between 800 and 1200 K. To rationalize these results, simulations were also per-
Accepted 10 February 2016
formed for planar twin boundaries of different inclination to determine their absolute mobilities and
grain boundary energies. The variation of these properties with inclination was integrated into a contin-
uum model which was found to produce the steady-state shapes of curved boundaries consistent with
Keywords:
those from simulations. A further extension of the continuum model enabled estimations of the reduced
Grain boundary migration
Molecular dynamics simulations
mobility that were in good agreement with simulation results. It was also identified that atomistic events
Iron governing the migration of curved and planar boundaries shared a number of similarities. Overall, the
analyses of the shapes, mobilities and atomic-scale migration mechanisms of curved and planar bound-
aries presented here provide a correlation between the migration of these types of twin boundaries.
Crown Copyright 2016 Published by Elsevier B.V. All rights reserved.

1. Introduction misorientation. A recent experimental study evaluated the migra-

A key property controlling the microstructural evolution of
materials is grain boundary mobility [1]. The mobility is typically
assumed to be independent of the magnitude and type of driving
pressure P, provided that P
kB T=X where kB T is the product of
Boltzmann constant and absolute temperature and X is the bulk
atomic volume [24]. In this regime, the boundary velocity is pro-
portional to the driving pressure, as confirmed with experiments
where the driving pressure is applied, for example, via external
magnetic field [57] or capillarity pressure [810]. The type of
reported mobilities depend on the boundary geometry. Planar
boundaries (e.g. driven by magnetic field [6]) are associated with
absolute mobilities M, while curved boundaries are associated with
reduced mobilities M  , a product of M and stiffness C [11]. Exper-
imentally, both types of mobilities have been reported to typically
follow an Arrhenius behavior, each being associated with a differ-
ent activation energy [610]. There is also evidence from recent
atomistic studies that a number of grain boundaries has non-
Arrhenius mobilities [12].
For the thermally activated migration of grain boundaries in a
material, it is not clear if there is any correlation between the
mobilities of planar and curved boundaries having the same
Corresponding author.
E-mail address: tegar@alumni.ubc.ca (A.T. Wicaksono).
tion of magnetically driven planar boundaries in Zn and the capil-
larity driven migration of quarter loop boundaries of similar
misorientations [7]. Different migration energies were found for
both types of boundaries and interpreted as an indication that a
curved boundary may migrate under a different mechanism than
a planar boundary [7]. This interpretation, also made in [8], would
suggest that the migration mechanism depends on the type of driv-
ing pressure and boundary geometry [1,13].
Experimental measurements of mobilities are challenging. A
minor level of impurities can, for example, significantly affect mea-
sured mobilities [14]. Molecular dynamics (MD) simulations, prove
to be a valuable tool in providing insights to the problems of
boundary migration [15] because they allow for all variables to
be carefully controlled, even those that are difficult to control
experimentally. Several MD techniques for extracting the mobility
have been developed, e.g. the capillarity technique [1618], the
artificial driving force (ADF) technique [1921], the stress-driven
technique [2224], and the random-walk (RW) technique
[2528]. Recent reviews have surveyed their applications and lim-
itations [29,30].
The capillarity technique offers arguably the most direct analy-
sis compared to other driven techniques since it involves no artifi-
cial assumptions about the manner by which driving pressures are
applied. Reduced mobilities of curved boundaries have been
extracted using this technique [17,18]. The shape evolution of

http://dx.doi.org/10.1016/j.commatsci.2016.02.016
0927-0256/Crown Copyright 2016 Published by Elsevier B.V. All rights reserved.
398 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405

curved boundaries has also been used to deduce the absolute

mobilities of planar boundaries of the same misorientation [16].
The ADF [19] and the stress-driven techniques [22] were
designed to examine the migration of planar boundaries. Both
techniques have been applied to several f.c.c metals, e.g. Al
[19,28] and Ni [24,20,21], and have further been found to produce
consistent results in terms of the absolute mobilities [29]. Recent
studies pointed to a need for considerable care in analyzing the
absolute mobilities determined from the ADF technique [28,31].
For example, difficulties have been experienced in analyzing the
mobility of low misorientation angle boundaries [28], thus neces-
sitating a post-simulation correction to the mobility analysis [31].
In contrast to the driven boundary techniques (i.e. the capillary,
the ADF and the stress-driven techniques), the RW technique [26],
or equivalently the boundary fluctuation technique [32], computes
the absolute mobility of a planar grain boundary in a non-driven
bicrystal. Above the roughening temperature [33], boundaries
undergo uncorrelated spatial fluctuations indicative of a random
walk process. The mean-squared displacement of such a boundary
can be recorded over time to extract the absolute mobility [26].
While the absolute mobility computed via this technique may
not be derived in a conventional manner, i.e. via the migration of
a grain boundary, the RW technique provides the benefit of being
in the limit of zero driving pressure. One strong limitation to this
technique is that the mobility computation is only possible when
the temperature is higher than the roughening transition temper-
ature [30].
Previous studies have shown that the same absolute mobilities
were found (within the simulation error) from the ADF, the stress-
driven and the RW techniques [28,29]. There are, however, no sys-
tematic MD simulations that investigate whether the mobilities of Fig. 1. Two types of bicrystal cells prepared in this study, (a) type-1 cell containing
a U-shaped half-loop boundary, and (b) type-2 cell containing two planar
planar and curved boundaries, for the same boundary misorienta-
boundaries.
tion, can be related to one another. In this study, a direct correla-
tion is established between the mobilities of curved and planar
boundaries for a specific boundary misorientation in a b.c.c crystal. Table 1
D E
The migration of 1 1 0 curved twin boundaries in pure b.c.c Fe Crystallographic axes of bicrystals containing a U-shaped grain boundary (the type-1
cell) investigated in this study, see Fig. 1(a) for symbol definition.
is first examined in terms of boundary shape and reduced mobility.
Orthogonal axes Designation
Planar inclined segments that make up the curved boundaries are
then simulated, to quantify their energies and absolute mobilities. Coherent loop Incoherent loop
D E
Finally, the properties of planar boundaries are incorporated into XA ; XB 110
continuum models that allow for the prediction of the shape and D E D E
YA 112 111
mobility of a curved boundary having the same misorientation.
ZA h1 1 1i h1 1 2i
The findings are discussed in terms of atomistic mechanisms D E D E
YB 112 111
underlying the migration of both types of boundaries. D E D E
ZB 111 112

2. Simulation methodology

Simulations were carried out using LAMMPS [34] with a time D E

step of 1 fs. Visuals were produced using AtomEye [35] and OVITO
[36]. The Ackland-04 EAM potential was used to model the interac-
tion among Fe atoms [37], since it has been extensively bench-
marked with MD simulations of b.c.c Fe.
Two types of simulation cells were prepared, i.e. the type-1 cell,
a bicrystal whose grains are separated by a U-shaped half-loop
grain boundary (Fig. 1(a)) and the type-2 cell, a bicrystal with
two planar boundaries (Fig. 1(b)).
The type-1 cell was constructed by first creating a single b.c.c
crystal with orthogonal axes of (XA ; YA ; ZA ). Atoms in the dark-
shaded region in Fig. 1(a) were removed and replaced by atoms
occupying b.c.c lattice sites of different orientation, i.e. (XB ; YB ,
ZB ). The resulting grain boundary is a U-shaped half-loop bound-
ary, consisting of a curved end cap and planar sides, i.e. the h?
and hjj sides, see Fig. 1(a). In this work, two different type-1 cells
were prepared, their axes being defined in Table 1. The planar sides
of both cells are symmetric coherent and incoherent twin bound-
aries. The twofold symmetry of 1 1 0 -axis renders the planar
sides complementary to each other, e.g. if the h? side is the coher-
ent twin, the hjj side will be the incoherent twin. To simplify the
nomenclature, the curved portion in a type-1 cell will be referred
to based on the segment at the tip, e.g. the coherent loop is a
curved boundary with the symmetric coherent twin at its tip.
Periodic boundary conditions were applied to walls normal to
the X- and Z-axes. A shrink-wrapped boundary condition was
imposed to walls normal to the Y-axis such that two free surfaces
were created. Atomic planes within a distance of 1 nm from the
free surface were fixed to prevent grain rotation [18]. Molecular
statics were then performed to the cells while the cell volume
was modified (using the box/relax technique on LAMMPS [34])
in order to find the minimum energy configuration at zero
pressure.
The type-2 cell was constructed by populating the light- and
dark-shaded region in Fig. 1(b) with atoms occupying b.c.c lattice
 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405 399

sites of orientation (XA ; YA , ZA ) and (XB ; YB ; ZB ), respectively. Table 2
provides a list of type-2 bicrystals that have been investigated in
this study, each generating an inclined planar boundary, i.e. the
boundary that forms an inclination angle u with the planar coher-
ent twin. This is equivalent to an inclination of 90  u when
measured with respect to the planar incoherent twin. Periodic
boundary conditions were applied in all directions, thus creating
two planar boundaries of the same inclination, see Fig. 1(b). The
equilibrium structure of type-2 cells at 0 K was obtained following
the procedure previously employed for planar boundaries [38].
Both types of bicrystal cells were brought to a finite constant
temperature by expanding their dimensions uniformly based on
the appropriate lattice constant [39] and initializing the atomic
velocities. Simulations using the type-1 cells were run under a con-
stant volume condition for 12.5 ns at a given temperature. For the
type-2 cells, simulations were first performed under a zero-
pressure condition for 1 ns to obtain the equilibrium structure
for a given temperature. The equilibriated cells were then simu-
lated for up to 6 ns under a constant volume condition.
The type-1 cells were simulated to extract the reduced mobili-
ties of curved boundaries. During the migration of the half-loop
boundary with diameter DU , see Fig. 1(a), the driving pressure is
given by P Cp=DU [18] where C is the boundary stiffness [11],
C c d2 c=du2
To identify atoms in the shrinking grain, each atom was
assigned an order parameter g, a unitless property determined
from the atoms position relative to their nearest neighbors
[19,20]. The existing LAMMPS sub-routine for determining g has
been modified to make it compatible with b.c.c crystals. We imple-
mented the following criteria for identifying bulk or boundary
atoms, respectively, by setting cut-off values glo ; ghi . Using the
bulk atom of grain A as a reference, an atom is considered to belong
to grain A if g < glo , to grain B if g > ghi , and to the grain boundary
otherwise. The mobility results were found to be independent of
the choice of the reference grain and the cut-off values when
0 6 glo 6 0.3 and 0.7 6 ghi 6 1.
The simulation results were found to be independent of cell size
when the length LU was at least 10 atomic planes, the thickness LX
was at least 10 atomic planes and the loop diameter DU was at least
25 nm and 35 nm for the coherent and incoherent loop, respec-
tively. In summary, the minimum dimensions of the cell were
LX 4:1 nm, LY 58 nm and LZ 40 nm, corresponding to at least
750,000 Fe atoms.
To explain the reduced mobilities of curved boundaries (type-1
cells), independent sets of simulations were carried out to deter-
mine the properties of planar inclined boundaries (type-2 cells),
namely their energies c and absolute mobilities M. The energy of
planar inclined boundaries at a finite temperature, cT, was com-
1 puted using the excess energy technique [40], i.e.

and u is the inclination, see Fig. 1(a). Assuming that the velocity V is
a product of absolute mobility M and driving pressure P; V can be
expressed as [18]
V M p=DU
 
where M MC is the reduced mobility. In simulations, M was
extracted by monitoring the rate at which v U , the shrinking grain
volume, decreases during the steady-state migration, i.e.
dv U =dt DU LX dLU =dt DU LX V. Substituting Eq. (2) yields
M  pLX 1 dv U =dt XpLX 1 dNU =dt
where X is the bulk atomic volume and NU is the number of atoms
in the shrinking grain.
cT EGB =2A Etot  NFe eT =2A 4
where EGB is the boundary excess energy, which is also referred to
as the boundary enthalpy [18], Etot is the potential energy of the
2 bicrystal, N Fe is the number of Fe atoms, eT is the potential energy
of an Fe atom in a b.c.c crystal at temperature T; A is the boundary
area and the factor 2 is to account for periodic boundary conditions,
see Fig. 1(b). To obtain eT , separate simulations using a b.c.c single
crystal cell with dimensions of 203 unit cells were performed at a
zero-pressure condition for 3 ns. The ratio between the average
3 total potential energy and the number of atoms in the single-
crystal cell yields eT . It is important to recognize that the first equal
sign in Eq. (4) employs an approximation where the boundary
energy, c, is assumed to be equal to the boundary excess energy

Table 2 D E
Geometrical setup of type-2 bicrystals used for determining the absolute mobility of planar boundaries, their common XA =XB axis being 1 1 0 with a dimension LX of 12.2 nm,
see Fig. 1(b).

Inclination Orthogonal axes Dimensions [nm] N Fe [atoms]

u [] YA ZA YB ZB LY LZ
D E D E
0.0a h1 1 1i 112 111 h1 1 2i 14.9 33.7 518,400
D E D E
8.1 h3 3 4i 223 11 11 8 h4 4 11i 15.1 28.3 437,040
  D E
15.8 h4 4 7i 778 221 h1 1 4i 13.0 29.3 388,200
D E D E
19.5 h1 1 2i 111 552 h1 1 5i 12.7 29.9 388,080
D E D E
26.8 h3 3 8i 443 19 19 4 h2 2 19i 15.7 44.2 707,280
D E
35.3 h1 1 4i 221 h1 1 0i h0 0 1i 12.2 29.2 376,800
D E D E
44.7 h1 1 8i 441 h17 17 4i 2 2 17 14.0 39.6 568,560
D E D E
54.7 h0 0 1i 110 221 h1 1 4i 13.0 30.1 387,000
D E D E
63.2 h2 2 19i 19 19 4 443 h3 3 8i 11.1 62.5 706,800
  D E
70.5 h1 1 5i 552 111 h1 1 2i 12.0 42.2 516,480
D E D E
74.2 h1 1 4i 221 778 h4 4 7i 14.7 20.6 308,640
D E D E
81.9 4 4 11 h11 11 8i 223 h3 3 4i 14.2 40.2 585,120
b D E D E
90.0 h1 1 2i 111 112 h1 1 1i 14.0 19.7 285,600

a
The planar coherent twin.
b
The planar incoherent twin.
400 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405

per unit area, EGB =2A. Such an approximation has been employed in
previous studies, e.g. [18,40,41], and further rationalized in [41]
where the boundary enthalpy is deemed to be an appropriate mea-
sure of the relative (but not absolute) boundary free energy.
To obtain the absolute mobilities M, the random-walk (RW)
technique [19] was chosen over the ADF technique because the
mobility extracted from the latter depends on the range of driving
pressure being deemed as the range where the velocity-driving
pressure trend is linear. In the RW technique, if the boundary is
rough at the temperature of observation, it performs a random
walk that exhibits diffusive characteristics. The average position
of the boundary was determined every 10 ps by averaging the posi-
tion of boundary atoms. The evolution of the mean-squared dis-
D E
placement, h2 , of the average boundary position yields a
coefficient that is proportional to M, via [26]
D E
h2 2MkB T s=A 5

where s is the time-interval.

In addition to extracting the absolute mobilities of grain bound-
aries, the mechanisms underlying grain boundary migration were
also studied. Because such information is not available from the
RW technique, since no net displacement of the boundary is
involved, the ADF technique was employed instead. In the ADF
technique, atoms in one of the grains receive an extra energy ug
where ug umax 1  cos pg, with umax being the maximum
energy added to an atom [19]. The system lowers its total energy Fig. 2. (a) An example of the grain shrinkage due to curvature-driven boundary
by shrinking the high-energy grain, thus promoting grain bound- migration taken from a 25-nm diameter coherent loop at 900 K, the shrinkage rate
at steady state highlighted and (b) Arrhenius plot of reduced mobilities, dashed
aries to migrate.
lines indicating the Arrhenius fit (see text).
For each planar boundary, simulations to compute their M and c
at a given temperature were performed in four independent runs.
The dimensions of type-2 cells and the number of Fe atoms N Fe ,
having been checked for finite size effects, are given in Table 2.

3. Results

3.1. Curvature-driven grain boundary migration

The kinetics of a migrating curved boundary is observed by

monitoring the evolution of the shrinking grain volume, see
Fig. 2(a) for an example. Within a given time period, the grain
shrinkage is considered to have occurred at steady state if the
0.1-ns moving linear regression in that period has a slope that var-
ies by less than 20%. The reduced mobility M  was obtained from
the steady-state shrinkage rate at a given temperature and DU ,
highlighted in Fig. 2(a), via Eq. (3).
The reduced mobility determined in temperature range of 800
1200 K is presented in Fig. 2(b). Each data point and its error bar
represent the average and the standard deviation, respectively,
over results taken from three half-loop diameters: DU of 25, 30
and 40 nm for the coherent loop and DU of 35, 42 and 56 nm for
the incoherent loop. The reduced mobilities show an Arrhenius
relationship, i.e.
M  M 0 exp Q m =RT 6

where M0 is the prefactor, Q m is the effective migration energy, and

 
R is the gas constant. Parameters M0 ; Q m are (4.2 0.9  107 m2
s1; 10.5 0.1 kJ mol1) and (4.4 1.1  107 m2 s1; 12.0 0.1 kJ
mol1) for the coherent and incoherent loop boundary, respectively.
The coherent loop has a reduced mobility that is 1020%
higher than that of the incoherent loop in the investigated temper-
ature range. More importantly, both loops developed significantly
different shapes, as illustrated in Fig. 3. Their morphological differ- Fig. 3. The shape of (a) 25-nm diameter coherent loop and (b) 35-nm diameter
ence suggests that the kinetics of migration varies as a function of incoherent loop during their steady-state migration (>1 ns) at 1000 K, atoms
inclination. In order to verify such a hypothesis, the absolute colored based on their potential energy.
 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405 401

mobilities of planar segments composing the curved boundaries

were investigated, starting with the segment at the loop tip, i.e.
the planar symmetric coherent and incoherent twin boundaries,
respectively.

3.2. Absolute mobilities of planar symmetric twin boundaries

The absolute mobilities of boundary segments at the loop tip,

i.e. the planar coherent and incoherent twin boundaries (u 0 ,
90 in Table 2), were computed using the RW technique. Fig. 4(a)
and (b) show the Gaussian displacement distribution for the planar
symmetric coherent and incoherent twin at 1000 K, respectively.
The former has a width that shows little variation with s while
the width of the latter increases as s increases, indicating a diffu-
sive random walk.
The mean-squared displacement (MSD) of both boundaries can
be computed from these distributions as a function of time inter-
val, see Fig. 5(a). The planar incoherent twin has an MSD that
increases with increasing time-intervals, where the deviation from
a perfectly straight line is a signature of the thermal noise from the
displacement distribution shown in Fig. 4(b) [42]. Despite these
undulations, the data is well fit by a linear regression, allowing
the absolute mobility to be calculated by Eq. (5). The planar coher-
ent twin, in contrast, shows an insignificant MSD evolution. This
indicates an absolute mobility that is close to zero.
The same procedure was repeated to extract the absolute
mobilities of the planar incoherent twin as a function of tempera- Fig. 5. (a) The evolution of MSD for planar twin boundaries, the slope of linear fit
ture, shown in Fig. 5(b), where error bars are the standard devia- proportional to M via Eq. (5) and (b) Arrhenius plot of M of the planar incoherent
twin.
tion from four runs. The results were fit to an Arrhenius
relationship M M 0 exp (Q m /RT) where M 0 ; Q m are
(2.7 1.8  105 m4 J1 s1; 51 5.0 kJ mol1). has also been observed in the experimental study on Zn grain
The migration energies of the planar twin boundaries can be boundaries, and was attributed to the possibility of curvature dri-
compared with those of the curved twin boundaries, i.e. Fig. 2(b), ven boundaries migrating under a different mechanism than pla-
since both cases employed boundaries of the same misorientation, nar boundaries [7]. In order to rationalize this observation, we
albeit different geometry. The migration energy of the planar inco- next consider if the kinetics of curved boundary migration can be
herent twin (Q m 51 kJ mol1) is four times larger than that of the predicted from the behavior of their planar constituents.
curved incoherent twin (Q m 12 kJ mol1). On the other hand, the
planar coherent twin is effectively immobile, as indicated in Fig. 5
3.3. Absolute mobilities and energies of planar inclined boundaries
(a), this being in contrast with the mobile curved coherent twin
(Q m 10 kJ mol1). Such a discrepancy in the migration energies
The RW technique was also applied to extract the absolute
mobilities of planar inclined boundaries (u 0 ; 90 in Table 2).
The MSD evolution of the inclined boundaries was monitored at
800, 1000 and 1200 K. Each boundary was found to perform a dif-
fusive random walk, permitting the use of Eq. (5) to compute their
absolute mobilities M. The energy of these planar inclined bound-
aries at each temperature, cT, was also obtained via Eq. (4), the
potential energy per bulk Fe atom eT at 800, 1000 and 1200 K hav-
ing been determined from bulk crystal simulation as 3:90; 3:87
and 3:84 eV, respectively. The variation of grain boundary energy
c and absolute mobilities M with the inclination angle u as
obtained from the simulations are presented in Fig. 6. Each data
point and its error bar represent the average and the standard devi-
ation from four runs, respectively (see Supplementary Materials
(online) for tabulated values).
For a given temperature, the grain boundary energy increases as
the inclination goes from the planar symmetric coherent twin
(u 0 ) to the planar symmetric incoherent twin (u 90 ), see
Fig. 6(a). For the purpose of the continuum modeling to be per-
formed later, the relationship between c and u was further fit to
a curve,

cu cmax  cmin 1  exp B1 u cmin 7

Fig. 4. The displacement distributions of (a) the planar coherent twin and (b) the where fitting parameters B1 ; cmax , and cmin are given in Table 3. Eq.
planar incoherent twin, taken for several s at 1000 K. (7) is chosen as the fitting function because its mathematical form
402 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405

Mu B2 up exp B3 uq 9

where fitting parameters B2 ; B3 ; p and q are given in Table 3. The
extent of scatter between the data and the model varies with u.
The M of the 44.7 -inclination from simulations, for example, coin-
cides well with the fitting curves for 800 and 1000 K, but is about
twice as high as the predicted value at 1200 K. Such scatter empha-
sizes that Eq. (9) is only a first order approximation to describe the
absolute mobility from simulations. This must be kept in mind
when the results of the continuum models that make use of these
analytical expressions are compared to the simulation results in
the discussion below.

4. Discussion

As a starting point for predicting the migration of the coherent

and incoherent loop, an approach similar to that used in [16] has
been employed to predict the steady-state loop shapes via a con-
tinuum model [3,45]. The model asserts that the boundary shape
at steady state, x; y, can be obtained via [3]
Z /
x/ v 1 M/C/ sin /d/
0
Z /
y/ v 1 M /C/ cos /d/ 10
0

where / is the inclination relative to the segment at the loop tip, i.e.
Fig. 6. (a) Grain boundary energy and (b) absolute mobility as a function of / u for the coherent loop and / 90  u for the incoherent
inclination at several temperatures, dashed lines in (a) and (b) drawn using Eqs. (7) loop, and v is a scaling factor with a unit of velocity determined
and (9), respectively. from imposing the boundary conditions, i.e. y/ 90
DU =2.
An isotropic half-loop boundary, i.e. M; C f /, will maintain its
initial shape throughout its migration. The analytical fits to C/
and M/ from Eqs. (8) and (9) were substituted into Eq. (10). The
Table 3
Fitting parameters in Eqs. (7) and (9) to draw the approximation of cu and M u in steady-state shapes x; y were then solved for 90 6 /  90 .
Fig. 6. The predicted shapes are in reasonable agreement with the simula-
Parameters 800 K 1000 K 1200 K
Eq. (7) cmin =cmax [J m2] 0.25/1.36 0.20/1.32 0.19/1.30
B1 [deg1] 4:04  102 3:53  102 2:26  102
Eq. (9) B2 [m4 J1 s1 degp] 4:46  10 4
5:32  10 0
1:49  101
B3 [degq] 83.6 78.6 69.5
p/q 6.0/0.8 4.0/1.0 4.0/0.9

provides a compromise between precision and the number of fitting

parameters. The relationship between stiffness and inclination is
obtained by substituting Eq. (7) into Eq. (1), yielding

Cu cmax  1 B21 cmax  cmin expB1 u 8

The boundary energy also varies with temperature T, i.e. c decreases

as temperature increases. The temperature effect on c is second
order and more pronounced in the inclined boundaries compared
to the symmetric twin boundaries. A similar trend of c with T has
been reported, from experiments [43] and simulations [44,41].
The opposite c-T trend, i.e. where c increases as temperature
increases, has also been reported, e.g. for f.c.c Ni [40]. Ref. [41] pos-
tulated that such a different c-T trend may be due to grain bound-
aries undergoing a structural transformation upon heating.
Fig. 6(b) shows the variation of absolute mobilities with the
inclination angle as obtained from the RW technique, including
the earlier results for the planar coherent and incoherent twin
boundaries. The numerical values of M for these boundaries, as Fig. 7. Superimposition between the predicted shapes of curved boundaries (solid
well as their Arrhenius fit, are available in Supplementary Materi- red lines, see Eq. (10)) and the shapes observed from simulations (i.e. Fig. 3) for (a)
als (online). The boundaries having inclinations between 15 and the coherent loop and (b) the incoherent loop at 1000 K. The C/ and M/ used in
the continuum model are shown on the right side of the figure. Note that the C/
55 have the highest mobilities. While there is no clear monotonic and M/ data set are the same for both boundaries, but shifted along the /-axis by
trend between the absolute mobilities M and the inclination angle 90 . (For interpretation of the references to colour in this figure legend, the reader is
u, the dependence of M on u for a given temperature was fit to referred to the web version of this article.)
 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405 403

tion results, as shown in Fig. 7. Some discrepancies, primarily Table 4

observed in the case of the coherent loop (Fig. 7(a)), can be attribu- Reduced mobilities extracted from simulations (Fig. 2(b)) and a continuum model (Eq.
(11)) where D = (M mod  M  )/M  .
ted to the fact that the simplified Eqs. (7) and (9) do not capture all
of the details of the c and M dependencies on / from simulations, T [K] Coherent loop Incoherent loop
see Fig. 6(a) and (b). M M mod D [%] M M mod D [%]
This model can be extended to estimate the reduced mobility of
[108 m2 s1] [108 m2 s1]
the curved boundaries. Consider the half-loop boundary migrating
at steady state, shown in Fig. 8. The maximum inclination of the 800 8.64 8.56 1 7.37 5.30 28
1000 12.0 14.3 19 10.3 10.4 1
curved boundary, /max , may be less than 90 . Beyond /max , the 1200 14.4 17.0 18 13.3 15.8 19
curved portion is terminated. A segment ds at the curved portion
is associated with its properties, M and C. The segment ds, which
lies at an inclination / from the tip, has a velocity of
!
nature of the underlying atomistic mechanisms controlling both
V / j V / j M/ C/ j/ where j/ is the local curvature of processes. To investigate these mechanisms, the following analysis
segment ds. Each segment migrates at a velocity of was performed.
!
j V x j V / cos / along a direction that forms an angle / with the A layer of atoms in the 1 1 0-plane was extracted from a num-
! ber of migrating boundaries, both curved and planar. The initial
segments normal. At steady state, the magnitude of j V x j is uni- and final position of each atom were monitored for a given time
form for all segments, and is equal to the velocity V in Eq. (2). interval. Vectors representing each atoms displacement were then
! !
The curvature is defined as j/ dj T / j/ds where T / is the unit mapped.
tangent vector for segment ds [11]. By substituting Eq. (2) to V / , In the migration of the incoherent loop, Fig. 9(a), atoms located
the reduced mobility, M mod , in the continuum model is given by around the loop tip underwent a random shuffling, see the dashed
box in inset (a1). The shuffling event is different from the event
R
DU M /C/ cos /dT /
M mod R 11
p ds

where the integration is performed over the entire loop, i.e. /max
6 / 6 /max . The magnitude of ds and dT / can be expressed in terms
1=2 2 2 1=2
of / as ds = x0 2 y0 2  d/ and dT / T 0x T 0y  d/, the 0 -
sign denoting the first-order derivative with respect to /.
Using the steady-state shapes produced by Eq. (10) and the fits
to C/ and M/ from Eqs. (8), (9), (11) was used to estimate the
reduced mobilities of the coherent and incoherent loop. Table 4
compares the mobilities computed from this continuum model
M mod to those found from simulations M  , see Fig. 2(b), showing
that good agreement is found. The largest deviation (28%) belongs
to the incoherent loop at 800 K, which is most likely due to the
continuous fits to M and c being a first order approximation.
Additionally, Mmod values were fit to an Arrhenius relationship.
The obtained Q m of the coherent loop is 14 kJ mol1, in good agree-
ment with the simulation result (10 kJ mol1 in Fig. 2(b)). On the
other hand, the obtained Q m of the incoherent loop is 22 kJ mol1,
which is about a factor of two larger than the simulation result
(12 kJ mol1 in Fig. 2(b)). The discrepancy can be attributed to
the monotonic temperature trend of the mobility deviations, D,
see Table 4.
The continuum model discussed above aims to explain the
shape and reduced mobilities of curved boundaries using the prop-
erties of planar boundaries. While it is evident from Fig. 7 and
Table 4 that the migration of planar and curved boundaries are cor-
related, it is important to substantiate this correlation with the

Fig. 8. Geometrical features of a curved boundary at steady state as a visual for Eq.
(11).
Fig. 9. Snapshots taken from the migration of (a) the incoherent loop, (b) the
coherent loop, (c) the planar boundary of 27 inclination, (d) the planar symmetric
incoherent twin. Arrows indicate displacement vectors of each atom between 0.5
and 0.9 ns after the steady-state migration started. The planar boundaries in (c) and
(d) were driven using the ADF technique. Dashed boxes indicate the random
shuffling of atoms, while dotted ovals indicate the cooperative atomic motion (see
text).
404 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405
occuring near the planar portion of the boundary, see the dotted
oval in inset (a2), where a cooperative motion of atoms along the
h1 1 1i direction is operational. This cooperative motion can also
be identified in the coherent loop migration, see the dotted oval
in Fig. 9(b), where it appears to be the predominant mechanism
throughout the curved portion of the boundary. Additionally, the
random shuffling is apparent in the planar portion of the boundary.
Both atomistic events, i.e. the cooperative motion and the ran-
dom shuffling, are also present in the migration of planar bound-
aries, see Fig. 9(c) and (d). These planar boundaries were driven
using the ADF technique at 1000 K with a driving pressure of
80 MPa. As noted earlier, the ADF technique was used instead of
the RW technique as the atomic mechanisms underlying the
migration cannot be analyzed from the stationary boundary used
in the RW analysis. It is found, however, that both techniques give
the same values for the absolute mobility of the boundaries studied
here, in agreement with recent studies [28,29].
D E
In Fig. 9(c), the net boundary migration occurs in the 2 2 19 //
 
4 4 3 direction but the underlying atomic event is the cooperative
motion along the h1 1 1i direction, e.g. see the dotted oval. The ran-
dom shuffling is also apparent in Fig. 9(c), e.g. see the dashed box,
but not as prevalent as the cooperative motion. Fig. 9(d) shows the
opposite situation where the migration of the planar incoherent
twin occurs predominantly by the random shuffling event.
The above analyses demonstrate that the migration of curved
and planar boundaries share a number of similarities in terms of
their mechanisms, both involving a combination of the random
shuffling and the cooperative atomic motion. It can be inferred
from the mobility values shown in Figs. 2 and 6(b) that boundaries
migrating predominantly via cooperative motion, e.g. boundaries
in Fig. 9(a) and (c), are more mobile than those advancing primarily
via random shuffling, e.g. boundaries in Fig. 9(b) and (d). This sug-
gests that the cooperative motion event is faster than the random
shuffling event.
These analyses can further be applied to reanalyze recent work
published in the literature. An MD study on the migration of
curved twin boundaries in FeCr alloys [16] determined that the
curved coherent twin was highly mobile due to the fast coopera-
tive motion being responsible for its migration. Based on this, the
authors suggested that the planar coherent twin must be highly
mobile too [16]. It was acknowledged that this deduction differed
from the classical view which postulates that the planar coherent
twin has a low mobility due to the absence of line defects [4]. In
this work, we have shown (Section 3.2) that the planar coherent
twin has a mobility close to zero. We propose that the high mobil-
ity of the curved coherent twin is attributed to the fast cooperative
motion along the h1 1 1i direction, see Fig. 9(b). The cooperative
motion is operational only if triggered by the random shuffling that
takes place along the planar segments that connect to the end-
points of the curved portion. The random shuffling is available on
these segments because they consist of the planar incoherent twin,
see Fig. 9(d). On the other hand, no random shuffling is present to
trigger the cooperative motion in the planar coherent twin due to
the entire boundary being coherent. As a result, the planar coher-
ent twin has zero mobility.
5. Conclusions
The migration of curved and planar grain boundaries has been
analyzed in terms of boundary shape, mobility and underlying
atomic-scale migration mechanisms. The shape that curved
boundaries assume during their migration can be explained by a
continuum model that incorporates the stiffness and absolute
mobility of individual boundary segments that make up the curved
boundaries. The model was extended to provide a quantitative
estimation of the reduced mobilities of curved boundaries, which
is in good agreement with the simulation results. Thus, for the
investigated twin boundaries, the kinetics of curved boundary
migration can be obtained from the kinetics of planar boundary
migration. Finally, analyses at the atomic-scale have determined
that, while the detailed mechanisms for the migration of both
curved and planar boundaries may not be exactly the same, their
similarities provide further evidence of the correlation between
their migration behavior. This further implies that the mechanism
of boundary migration is independent of the type of driving pres-
sure, a seemingly intuitive concept that has recently been called
into question.
The conclusion above was derived using an approach that inte-
grated perspectives from atomistic simulations and continuum
models. Another potential application of such an approach is
related to the implementation of experimentally derived mobilities
into models of microstructure evolution. A large number of exper-
imental studies employed the capillarity technique to measure
grain boundary mobilities, e.g. via half-loop or quarter-loop bicrys-
tals as well as grain growth. Since these reported values are essen-
tially reduced mobilities, they may not be immediately accessible
by microstructural evolution models which often require absolute
mobilities as their input parameters. The approach presented here
can therefore be applied in reverse, i.e. to estimate the absolute
mobilities given the information on the migration of curvature-
driven grain boundaries.
Acknowledgments
The authors gratefully acknowledge the financial support from
Natural Sciences and Engineering Research Council of Canada and
the computing resources provided by WestGrid, a division of Com-
pute Canada. ATW thanks Dr. J.J. Hoyt and H. Song for useful dis-
cussions and their warm hospitalities during his visit to
McMaster University.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.commatsci.2016.
02.016.
References
[1] F.J. Humphreys, M. Hatherly, Recrystallization and Related Annealing
Phenomena, Elsevier, 2004. doi:http://dx.doi.org/10.1016/B978-008044164-
1/50005-0.
[2] J. Burke, D. Turnbull, Prog. Met. Phys. 3 (1952) 220292, http://dx.doi.org/
10.1016/0502-8205(52)90009-9.
[3] G. Gottstein, L. Shvindlerman, Grain Boundary Migration in Metals, CRC Press,
2010. doi:http://dx.doi.org/10.1201/9781420054361.
[4] A.P. Sutton, R.W. Balluffi, Interfaces in Crystalline Materials, OUP, 2007.
[5] W.W. Mullins, Acta Metall. 4 (1956) 421432, http://dx.doi.org/10.1016/0001-
6160(56)90033-5.
[6] D.A. Molodov, G. Gottstein, F. Heringhaus, L.S. Shvindlerman, Acta Mater. 46
(1998) 56275632, http://dx.doi.org/10.1016/S1359-6454(98)00258-4.
[7] C. Gnster, D.A. Molodov, G. Gottstein, Acta Mater. 61 (2013) 23632375,
http://dx.doi.org/10.1016/j.actamat.2013.01.007.
[8] M. Winning, G. Gottstein, L.V. Shvindlerman, Acta Mater. 50 (2002) 353363,
http://dx.doi.org/10.1016/S1359-6454(01)00343-3.
[9] V.G. Sursaeva, B.B. Straumal, A.S. Gornakova, L.S. Shvindlerman, G. Gottstein,
Acta Mater. 56 (2008) 27282734, http://dx.doi.org/10.1016/j.
actamat.2008.02.014.
[10] J.E. Brandenburg, L.A. Barrales-Mora, D.A. Molodov, G. Gottstein, Acta Mater.
61 (2013) 55185524, http://dx.doi.org/10.1016/j.actamat.2013.05.043.
[11] C. Herring, The Physics of Powder Metallurgy, McGraw-Hill, 1951 (Chapter 8.
Surface tension as a motivation for sintering), pp. 143179. doi:http://dx.doi.
org/10.1007/978-3-642-59938-5_2..
[12] E.R. Homer, E.A. Holm, S.M. Foiles, D.L. Olmsted, JOM 66 (2014) 114120,
http://dx.doi.org/10.1007/s11837-013-0801-2.
[13] M.L. Taheri, D.A. Molodov, G. Gottstein, A.D. Rollett, Z. Metallkd 96 (2005)
11661170, http://dx.doi.org/10.3139/146.101157.
 A.T. Wicaksono et al. / Computational Materials Science 117 (2016) 397405
[14] K.T. Aust, J.W. Rutter, Trans. AIME 215 (1959) 119127.
[15] Y. Mishin, M. Asta, J. Li, Acta Mater. 58 (2010) 11171151, http://dx.doi.org/
10.1016/j.actamat.2009.10.049.
[16] I. Toda-Caraballo, P. Bristowe, C. Capdevila, Acta Mater. 60 (2012) 11161128,
http://dx.doi.org/10.1016/j.actamat.2011.11.021.
[17] M. Upmanyu, R.W. Smith, D.J. Srolovitz, Interf. Sci. 6 (1998) 4158, http://dx.
doi.org/10.1023/A:1008608418845.
[18] H. Zhang, M. Upmanyu, D.J. Srolovitz, Acta Mater. 53 (2005) 7986, http://dx.
doi.org/10.1016/j.actamat.2004.09.004.
[19] K.G.F. Janssens, D. Olmsted, E.A. Holm, S.M. Foiles, S.J. Plimpton, P.M. Derlet,
Nat. Mater. 5 (2006) 124127, http://dx.doi.org/10.1038/nmat1559.
[20] D.L. Olmsted, E.A. Holm, S.M. Foiles, Acta Mater. 57 (13) (2009) 37043713,
http://dx.doi.org/10.1016/j.actamat.2009.04.015.
[21] C. Deng, C.A. Schuh, Phys. Rev. B 84 (2011) 214102, http://dx.doi.org/10.1103/
PhysRevB.84.214102.
[22] B. Schnfelder, D. Wolf, S.R. Phillpot, M. Furtkamp, Interf. Sci. 5 (1995) 245
262, http://dx.doi.org/10.1023/A:1008663804495.
[23] B. Schnfelder, G. Gottstein, L.S. Shvindlerman, Metall. Mater. Trans. A 37
(2006) 1757, http://dx.doi.org/10.1007/s11661-006-0118-7.
[24] H. Zhang, M.I. Mendelev, D.J. Srolovitz, Acta Mater. 52 (2004) 25692576,
http://dx.doi.org/10.1016/j.actamat.2004.02.005.
[25] J.M. Rickman, S.R. Phillpot, D. Wolf, D.L. Woodraska, S. Yip, J. Mater. Res. 6
(1991) 22912304, http://dx.doi.org/10.1557/JMR.1991.2291.
[26] Z.T. Trautt, M. Upmanyu, A. Karma, Science 314 (2006) 632635, http://dx.doi.
org/10.1126/science.1131988.
[27] J.J. Hoyt, Z.T. Trautt, M. Upmanyu, Math. Comput. Simul. 80 (2010) 1382,
http://dx.doi.org/10.1016/j.matcom.2009.03.012.
[28] M.J. Rahman, H.S. Zurob, J.J. Hoyt, Acta Mater. 74 (2014) 3948, http://dx.doi.
org/10.1016/j.actamat.2014.03.063.
[29] M.I. Mendelev, C. Deng, C.A. Schuh, D.J. Srolovitz, Modell. Simul. Mater. Sci.
Eng. 21 (2013) 045107, http://dx.doi.org/10.1088/0965-0393/21/4/045017.
[30] J.J. Hoyt, Modell. Simul. Mater. Sci. Eng. 22 (2014) 033001, http://dx.doi.org/
10.1088/0965-0393/22/3/033001.
405
[31] F. Ulomek, C.J. OBrien, S.M. Foiles, V. Mohles, Modell. Simul. Mater. Sci. Eng. 23
(2015) 025007, http://dx.doi.org/10.1088/0965-0393/23/2/025007.
[32] S.M. Foiles, J.J. Hoyt, Acta Mater. 54 (2006) 33513357, http://dx.doi.org/
10.1016/j.actamat.2006.03.037.
[33] D.L. Olmsted, S.M. Foiles, E.A. Holm, Scr. Mater. 57 (2007) 11611164, http://
dx.doi.org/10.1016/j.scriptamat.2007.07.045.
[34] S.J. Plimpton, J. Comput. Phys. 117 (1995) 119, http://dx.doi.org/10.1006/
jcph.1995.1039.
[35] J. Li, Modell. Simul. Mater. Sci. Eng. 11 (2003) 173177, http://dx.doi.org/
10.1088/0965-0393/11/2/305.
[36] A. Stukowski, Modell. Simul. Mater. Sci. Eng. 18 (2010) 015012, http://dx.doi.
org/10.1088/0965-0393/18/1/015012.
[37] G.J. Ackland, M.I. Mendelev, D.J. Srolovitz, S. Han, A.V. Barashev, J. Phys.:
Condens. Matter 16 (27) (2004) S2629, http://dx.doi.org/10.1088/0953-8984/
16/27/003.
[38] D.L. Olmsted, E.A. Holm, S.M. Foiles, Acta Mater. 57 (2009) 36943703, http://
dx.doi.org/10.1016/j.actamat.2009.04.007.
[39] M.I. Mendelev, Y. Mishin, Phys. Rev. B 80 (2009) 144111, http://dx.doi.org/
10.1103/PhysRevB.80.144111.
[40] M.I. Mendelev, H. Zhang, D.J. Srolovitz, J. Mater. Res. 20 (2005) 11461153,
http://dx.doi.org/10.1557/JMR.2005.0177.
[41] S. Ratanaphan, D.L. Olmsted, V.V. Bulatov, E.A. Holm, A.D. Rollett, G.S. Rohrer,
Acta Mater. 88 (2015) 346354, http://dx.doi.org/10.1016/j.
actamat.2015.01.069.
[42] R. LeSar, Introduction to Computational Materials Science: Fundamentals to
Applications, Cambridge University Press, 2013.
[43] G.S. Rohrer, E.A. Holm, A.D. Rollett, S.M. Foiles, J. Li, D.L. Olmsted, Acta Mater.
58 (2010) 50635069, http://dx.doi.org/10.1016/j.actamat.2010.05.042.
[44] S.M. Foiles, Scr. Mater. 62 (2010) 231234, http://dx.doi.org/10.1016/j.
scriptamat.2009.11.003.
[45] V.E. Fradkov, L.S. Shvindlerman, Phys. Chem. Mech. Surf. 1 (1982) 180.