You are on page 1of 14

Acta Materialia 190 (2020) 2942

Contents lists available at ScienceDirect

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

Full length article

Chemical short range order strengthening in a model FCC high entropy


alloy
E. Antillon*,a,b, C. Woodwardb, S.I. Raoa, B. Akdima, T.A. Parthasarathya
a
UES Inc. Dayton, Ohio 45433, USA
b
Materials and Manufacturing Directorate, Air Force Research Laboratory, Wright Patterson Air Force Base, Dayton OH 45433-7817, USA

A R T I C L E I N F O A B S T R A C T

Article History: In order to understand the role of chemical short-range order on deformation mechanisms in FCC composition-
Received 13 September 2019 ally complex alloys, a random model alloy (Co30-Fe16.67-Ni36.67-Ti16.67) is annealed at various temperatures
Revised 13 January 2020 using Hybrid Molecular-dynamics/Monte-Carlo simulations. The simulations produce significant chemical
Accepted 17 February 2020
short-range order (CSRO) that increases with decreasing annealing temperature. Annealing tends to homoge-
Available online 18 March 2020
nize regions of high enthalpy due to: (1) chemical species redistributing into more compact configurations,
and (2) pairs of atoms forming chemical bonds that lower the overall energy of the system; the composition
Keywords:
explored here shows significant amount of ordering in Ti-Fe pairs with respect to random distributions as
Chemical short-range order
High-entropy alloys
described by pairwise (EAM) potentials due to Johnson and Zhou. An energy topology approach is used to
Chemically complex alloys assess the local strengthening behavior in random solid solutions and annealed systems, where an interesting
Solid solution hardening interplay is observed between misfit components and chemical short-range order affecting the overall critical
resolved shear stress. The role of short-range order on the critical yield stress is quantified and compared with
current solid solution models. Finally, we propose and validate an extension to the Labusch-Varvenne class of
high-concentration solid-solution analytic models that incorporates the effects of chemical short range order.
© 2020 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Motivation temperatures below the expected use temperatures. This achieves


some degree of homogenization, which improves structural proper-
Over the last decade there has been increasing interest in alloys ties as well as phase stability in the application. In a solid solution
with high concentrations of multiple elements. Initially studies alloy this produces diffusion of chemical species from regions of high
focused on 5 or more elements of equi-atomic concentrations that enthalpy, which produce chemical short range order (CSRO) and seg-
form a solid solution, the so called High Entropy Alloys. Recently, regation, thereby lowering the overall free energy of the system. Any
research has focused on fewer chemical species (e.g. Medium Entropy practical assessment, or prediction, of material strength at use tem-
Alloys, Compositionally Complex Alloys) and the possibility of taking peratures should allow for this consistent with the anticipated proc-
advantage of second phase precipitation hardening. Calphad surveys essing regime (e.g. annealing temperature). While it is generally
over a wide range of chemistries suggest that the number of stable 4, expected that chemical disorder will increase with temperature, a
5 and higher component HEA’s is quite limited [1]. While CalPhad completely random elemental distribution is consistent with a
approaches must rely on extrapolation to predict phase stability in quench from the solid-liquid phase boundary where the random
the more complex equi-atomic alloys, the results reflect the limita- chemical distribution is frozen in. In atomistic methods this approxi-
tions in available elements consistent with the Hume-Rothery rules. mation should be employed with some caution as it introduces high
Current theory and experiments suggest that ternary and quaternary enthalpy regions which are unstable under annealing. Such regions
alloys may provide the best test bed for studies requiring single- are likely to produce strong interactions with passing dislocations as
phase solid-solution alloys [2]. Further, recent advances in modeling these regions are energetically unstable and cutting them will sub-
Compositionally Complex Alloys (CCA) suggest that one element hav- stantially reduce the energy of the underlying lattice. This manuscript
ing a deviation from the average atomic size is enough to produce a explores how the evolution of local-chemical-structure affects the
significant strengthening effect [3]. solution strengthening in chemically complex solid solution alloys.
Standard metallurgical practice typically involves reasonable cool- Recent findings provide evidence that the most stable thermody-
ing rates from the melt and annealing for extended periods of time at namic configurations at a given temperature in CCAs are not strictly
random. The archetypal multicomponent solid solution alloy, discov-
* Corresponding author. ered by Cantor [4], has been found to be the exception rather than
E-mail address: edwin.antillon@nrl.navy.mil (E. Antillon). the rule. E. George has investigated variants of the Cantor alloy and

https://doi.org/10.1016/j.actamat.2020.02.041
1359-6454/© 2020 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
30 E. Antillon et al. / Acta Materialia 190 (2020) 2942

shown conclusively that replacing most elements (i.e. Ti $ Co) leads Section 4, describes a topological method to capture strengthening
to significant reductions in enthalpy consistent with CSRO at anneal- parameters directly from the simulation under stress. Section 5 com-
ing temperature of 1273 K [5], while at lower annealing temperatures pares the results to analytical solute-solution strengthening models,
(T=700 K), this system is unstable and decomposes into separate and Section 6 summarizes the work presented and draws conclusions.
phases [6]. Similarly, Ding and co-workers have recently used Density
Functional Theory Monte-Carlo calculations to predict CSRO in 2. Computational methods
CrCoNi at 500K [7], while Li and co-workers have used atomistic
interatomic-potential on the same ternary system and concluded 2.1. Simulation cell and dislocation methods
that the presence of CSRO tends to enhance the mechanical yield
strength of the ternary alloy [8]. Such an approach offers a pathway The simulations performed in this work use the large-scale molec-
for ab initio predictions of key metrics for CSRO in chemically com- ular dynamics massively parallel simulator (LAMMPS) [18] and the
plex alloys, such as the Warren-Cowley parameters [9] and diffuse four element, Co-Fe-Ni-Ti, interatomic potential developed by Zhou
anti-phase boundary energies [10]. et al [19]. The simulation dimensions consist of a cell (see Fig. 1) in an
In order to move forward in predicting solution hardening in FCC lattice using an orthogonal coordinate system x=[110] (Lx =
  
these alloys a better understanding of the scale of the effects of CSRO 1236 A) y=[11-2] (Ly = 1141 A), z=[111] (Lz = 202 A).
is required in fcc, bcc as well as hcp systems. In general, several other Periodic boundary conditions are enforced in both the x=[1 -1 0]
paths can be explored including phase transformation, emergent and y=[1 1-2] directions, while the cell boundaries along the z=[1 1 1 ]
ordering (Long-Range and Short-Range), phase separation, and/or direction have free surface boundary conditions. Perfect edge and
precipitation. This paper focuses on understanding the role of chemi- screw dislocations are inserted in between two [111] planes by dis-
cal short-range order (CSRO) in dislocation mobility on a model FCC placing the atoms according to anisotropic elastic Volterra solution
random alloy (Co30 Fe16.67 Ni36.67 Ti16.67) at various temperatures. This [20]. Dislocations are introduced into the annealed samples after the
specific (random) composition has been explored previously by Rao annealing procedure has been performed (described below). The pro-
and co-workers using molecular dynamics simulations [11]. It has cedure to enforce periodic boundary condition for a screw dislocation
been shown to be elastically stable against shear deformations, and is discussed by Rodney [21]. After a dislocation (edge or screw) has
has provided reasonable agreement with experimental yield been inserted into the lattice, and the system is fully relaxed using lat-
strengths and stacking fault energies in analogous FCC composition- tice statics, the system is then strained into a state consistent with con-
ally complex solid-solution alloys. stant shear stress t xz. The strained crystal is then relaxed to this new
In the work presented here, an energy topology of the dislocation configuration by holding the top and bottom Z-layer (a thin slab) fixed,
moving on the glide plane is used to assess the mechanisms control- while the rest of the system is minimized using molecular statics. Once
ling solute strengthening on the aforementioned model alloy. The the minimization procedure is complete, forces are added to the top
approach treats different classes of solution hardening (misfits, statis- and bottom Z-layers in the positive and negative x-direction according
tical distribution and CSRO) on equal footing, and thus provides a to F § ¼ § t A=N; where A and N are the area and number of atoms in
means for assessing there relative contribution[12]. The results of the top(bottom) layers, and t is the applied stress. The initial condi-
this study can provide insights into understanding the effect of CSRO tions for the MD simulations start from the molecular statics relaxed
on dislocation mobility. Historically the sources of strengthening core under the applied stress of interest. Before switching to molecular
mechanisms have been classified as random pinning due to size, dynamics, the system was scaled by a lattice expansion factor to main-
modulus, and chemical misfit interactions from solutes (Fleischer) or tain the pressure near zero while the velocities of the atoms are
solute clusters [13,14], or in alloys with SRO into a bond-breaking assigned according to a Boltzmann distribution at the temperature of
chemical contribution due to ordering [10,15,16]. Currently, models interest. A Nose-Hoover thermostat is used to evolve the atoms at a
of high concentration solution strengthening ignore the role of chem- constant temperature in the simulation. The procedure described
ical short range order. This is consistent with the origins of such mod- above, which is representative of standard practice for this type of sim-
els which were developed for dilute solute concentrations, where ulation, gives fluctuation in the resolved shear stress on the order of
solute-solute interactions are assumed to be negligible. dt xz  10MPa at T=5 K.
The inclusion of chemical ordering in solution hardening models is In order to measure the critical stress to unpin a dislocation, a
important for several reasons. First, when the alloys are annealed, stress-ramping approach is applied. To start, a constant resolved
chemical ordering produces strengthening because the dislocation must shear stress is applied for 50 ps and the position of the partials is
expend additional work when cutting the bonds of the ordered regions monitored as a function of time to determine its motion. If the center

[10,15,16]. Chemical segregation can also provide a complementary of mass of the partial dislocations moves by less than 5A , the stress is
source of strengthening since atoms that prefer not to be first nearest increased by an additional 7 MPa. This ramping criteria is followed
neighbors can provide resistance to the passage of a dislocation. Even
when disordered, in some systems solid-solution hardening requires a
careful consideration of dislocation geometry and the interaction of the
dislocation with certain clusters of solute atoms [17]. Second, annealing
replaces high-enthalpy atomic configurations with lower enthalpy con-
figurations which influences the distribution of potentially potent solu-
tion hardening volumes. Reshuffling atomic species tends to reduce the
average size misfit interaction with neighboring atoms, effectively
reducing the elastic contribution to hardening. For these reasons, realis-
tic modeling of solid-solution hardening necessitates a careful inclusion
of chemical ordering effects. The main focus on the study presented, is to
understand the interplay between chemical ordering and misfit interaction
on dislocation mobility on complex concentrated alloys.
The paper is organized as follows. Section 2 describes the computa-
Fig. 1. Geometry and coordinate system used for this study. The X and Y dimension are
tional methods used to anneal and apply stress to the simulation cell. periodic, while the Z boundaries have free surface boundary conditions. A perfect edge
Section 3, discusses the effect of chemical short range order on both (screw) dislocation is inserted into the middle of the cell and stress state is applied as
  

bulk properties and on the mobility of edge and screw dislocation. discussed in the text. The cell size is LX=1236A , LY=1141 A, and LZ=202A.
E. Antillon et al. / Acta Materialia 190 (2020) 2942 31

velocities are scaled by the ratio of the masses of the swapped


atom types in order to ensure that the kinetic energy of each
atom is the same after the swap as it was before the swap.
Pfinal
2. Metropolis Criteria: Accept the swap if u < Pinitial , where u is a ran-
dom number 2 [0, 1] and Pi is the Boltzmann distribution Pi ¼
ebUi ; where Ui is the potential energy of the system and b is the
inverse of kBT.
3. MC sample: Repeat steps (1) and (2) 100 times for all pair-types.
4. MD relaxation: Decorrelate the system for 50 molecular dynamic
time steps under constant pressure (NPT) conditions. This step
relaxes any local residual stress that could build up due to the
swaps.
5. Steps (1) through (4) are repeated until reasonable convergence
Fig. 2. Resolved shear stress (left label) and position of the trailing partial (right label) vs
simulation time. The critical unpinning state is denoted by the dashed line around t=300 ps. is obtained as described below

until the partials start to flow continuously. Fig. 2 shows an example


of this procedure where the applied stress starts initially at t xz ¼ To assess convergence in the annealing procedure, the potential
180 MPa, and it is incremented additionally by 7 MPa about four energy (per atom) was tracked as a function of the number of
times before the partials start to flow at around t=300 ps when the attempted swaps as shown in Fig. 3(a). Annealing a sample on the
stress reaches a value around t xz = 210 MPa. Using this approach, the order of 6 Million atoms using the scheme presented above is a com-
critical stress to unpin a dislocation can be measured, with an uncer- putationally demanding task. Therefore, the annealing process was
tainty which is in the order of twice that of the fluctuations of stress halted when a reasonable convergence was observed after about a
state, ie dt xz ~ 20 MPa. million and a half swaps are attempted (for every distinct pair) on
the whole system, which corresponds about 2 MC steps; a MC step is
2.2. Annealing using a hybrid Monte-Carlo/Molecular-Dynamics defined as the number of swap attempts / number of atoms in the lat-
approach tice. Using smaller cells (108,000 atoms), it was determined that

Solute diffusion in this system is mimicked by allowing for an


exchange of any pair of distinct species inside the lattice. This compu-
tational approach ignores the kinematics of vacancy mediated mecha-
nisms that are are thought to dominate solute-substitution diffusion in
FCC metals [22]. We are interested in solution hardening produced
when a dislocation moves through new annealed material, therefore
the hybrid Monte-Carlo method is applied to bulk material without
the presence of the dislocation line. Future studies may consider solute
segregation to the dislocation, forming a Cottrell atmosphere, but that
is beyond the scope of this work. Hence, the approach we follow emu-
lates chemically homogeneous stable regions that are likely to be pres-
ent throughout the lattice under realistic heat processing conditions.
The simulation cell sizes used in the annealing calculations are the
same as the cells with random elemental distributions [Lx, Ly, Lz] and

are built up using the following procedure. An FCC cell (a=3.64 A)

with dimensions [Lx, Ly, 50.4A ] (5,898,240 atoms) is created with an
initially random distribution of elements at the desired composition
using a Park-Miller random number generator [18]. The Annealed
samples are equilibrated using a hybrid Monte-Carlo/Molecular-
Dynamics approach. The Monte Carlo scheme mimics diffusion by
allowing for an exchange of any pair of distinct species (a swap)
inside the lattice, while maintaining the overall concentration (Co30-
Fe16.67-Ni36.67-Ti16.67) constant throughout. This approach is dif-
ferent from a Semi-Grand Canonical Monte-Carlo scheme where
atomic species are swapped with an infinite reservoir which can
result in an overall variation in chemical composition [23]. By com-
bining a Monte-Carlo inter-species moves with Molecular Dynamics,
a representative structure at a fixed concentration isobaric-isotherm
ensemble is obtained. After annealing, the slabs are then replicated
along the z ([111]) direction to create a simulations cell of the same
size used for the random cell calculations. The hybrid molecular-
dynamics/Monte-Carlo algorithm is shown below

Hybrid MD/MC algorithm:

Fig. 3. (a) Potential energy (per atom) with respect to the random case vs MC steps at
1. MC-swap: swap the position of two atoms of different chemical T=700, 1100K, and T=1500K. (b) The acceptance rate for each attempted pair vs MC
composition, chosen randomly within the simulation cell. The steps at T=1100K.
32 E. Antillon et al. / Acta Materialia 190 (2020) 2942

about 10 MC steps are needed to achieve full convergence using the Table 1
Hybrid MC/MD approach for all the temperatures explored. Comparison of lattice properties for annealed and random sys-
tems, where T is the annealing temperature. Lattice constant:
The final energies for annealing temperatures T=700 K and
a, elastic constants, the potential energy difference with
T=1100 K of the larger cells are in reasonable agreement to the con- respect to random case: DU from simulations and DU* (from
verged energies of the smaller cells (within 90%), yet the resultant Eq. (3)), intrinsic stacking fault energies, and configurational
energies produced for T=1500 K annealing lies within 60% of the entropy (from Eq. (4)) .
smaller cell values. This indicates that the T=1500 K anneal likely Property Random Temperature [K]
requires far more annealing time than was at our disposal to obtain
full convergence simulations using this method. Nevertheless, the 1500 1100 700

various annealed samples provide a systematic survey of strengthen- a [A] 3.6407 3.6401 3.6389 3.6387
ing properties with various degrees of CSRO for the model alloy. C11 [GPa] 166.6 171.0 171.8 172.8
Other parallel algorithm schemes have been developed to accelerate C12 [GPa] 131.9 132.3 132.3 132.3
C44 [GPa] 84.2 86.0 86.3 86.7
convergence of large systems using a multi-variable optimization m [GPa] 38.3 40.8 41.3 41.6
problem in term of moments of concentration [24,25]. Yet determi- n [GPa] 0.48 0.47 0.47 0.46
nation of the Lagrange multipliers (chemical potentials) for a multidi- DU [meV/at] 0.0 -6.2 -12.9 -16.6
mensional system is not a trivial task. In the approach taken here, DU* [meV/at] 0.0 -2.9 -12.4 -16.1
g SFE [mJ/m2] 20.0 20.8 24.4 26.7
although slower for large systems, does not require solving equations
Sconf [kB/at] 1.327 1.325 1.300 1.279
of constraint since atoms are simply rearranged in the lattice.
The acceptance rate for all the pairs is shown in Fig. 3(b). Given
that the atomic radius of Titanium is almost 20% larger than the rest
of the atom [19,26], it follows that the acceptance rate of a Ti atom T=700K, 1100K, and 1500 K using the MD/MC hybrid approach
swap with any other species is the lowest. On the other hand, the described in the previous section. These temperatures correspond to
swap Ni$Co is accepted the most frequently since Ni and Co are about 40%, 65% and 85% of the average melting temperatures of the
expected to be soluble in each other given their similar atomic radius constituent elements. Table 1 shows the variation of material proper-
and electronegativity values[27]. A metric to quantify the amount ties as a function of annealing temperature.
CSRO in the annealed systems is discussed next. There is a slight decrease in the equilibrium lattice constant as the
system is annealed at lower temperatures. This results in a slight
2.3. Estimating the degree of chemical short range order increases of the elastic constants: C11 and C44 by less than 3% and C12
by less than half a percent. The anisotropic elastic constants relevant
The Warren-Cowley (WC) parameter [9] is used to characterize for glide of a 12 h 110 i dislocation on the (111) plane are computed
the degree of chemical short range order at the various annealed using the definition due to Bacon and Scatergood [28], where the
temperatures. This parameter can be defined starting from the devia- shear modulus is given by m ¼ Ks ; while the Poisson’s ratio is given
tion of two-point correlation functions on occupation sites (i,j,k) and n ¼ 1Ks =Ke ; and Ks and Ke are the anisotropic elastic a/2⟨110⟩ dislo-
(p,q,r) in the lattice cation line energy per unit length for the screw and edge orienta-
tions. As shown in Table 1, the shear modulus increases by at most 8%
h cijk
n
i h cpqr
m
i  h cijk
n m
cpqr i
anm ¼ ; ð1Þ for the lowest annealed temperature, while the Poisson ratio remains
cn ¢ cm almost unchanged with respect to the annealing temperature.
n As expected, annealing shows a decrease in the interatomic
where cijk denotes the local occupation of species n at lattice site (i,j,
k), and brackets denote averaging over various lattice sites, and cn is potential energy with respect to the fully random configuration (i.e.
the global concentration of the entire system. In particular, lattice DU). Cowley approximates this configurational energy change using
sites (i,j,k) and (p,q,r) that are in close proximity are of importance to a broken-bond model as [29]
quantify the amount of CSRO in the lattice. In this manuscript, unless X
DU  ¼ Z cn cm anm Unm ; ð3Þ
otherwise noted, only the nearest neighbors are used to estimate val- n;m
ues in the WC parameters.
The above expression can be simplified given that the two-point cor- where n and m are the atomic species, Z is the number of nearest-neigh-
relation is expected to be symmetric in its species, that is h cijk
n m
cpqr i ¼ cn bors, cn is the average concentration of type n, Unm is the binding energy
Pnm ¼ cm Pmn ; where Pnm describes the conditional probability for an of pair (n, m) and anm are the WC parameters. Binding energies between
atom of type m being adjacent to an atom of type n. Moreover, noting two solutes are not easily defined in complex concentrated alloys, since
that the local concentration will be on average equal on to the global con- this value can be highly dependent in the local chemical environment.
centration values, it follows that the WC parameter can be written as: Nevertheless, this value can be formally defined with the change in the
energy required to bring two atomic species from far way to the nearest
Pnm
anm ¼ 1 : ð2Þ neighbor position (see Appendix A.1). Also, when comparing energies of
cm the above expression to those obtained from the simulation, care must
A value of the WC parameter of zero would describe a purely ran- be taken to avoid counting the energy contribution from atoms belong-
dom solid solution, whereas negative values represent structures ing to the same pair twice. For those systems that have converged or
showing ordering, and positive values reflect preferences for segrega- are close to the converged values, it was found that the configurational
tion. The values for the WC parameters are obtained by averaging the energy obtained from the broken-bond model (Eq. (3)) are in excellent
estimates of Eq. (2) for all the atoms in the simulation, after the MD/ agreement with the change in the potential energies measured from
MC simulation has converged. the simulation; all systems that ran for 20 MC steps (108 000 atoms)
show excellent agreement with the broken-bond model. Interestingly,
3. Effect of annealing on material properties only the large system annealed at T=1500 K shows a discrepancy with
the broken-bond model, suggesting that this simulation requires further
3.1. Bulk properties annealing in order to fully converge to the target equilibrium configura-
tion. Nevertheless, this annealing result can be a useful test case to dis-
In this section, ordering tendencies on the bulk alloy are investi- cern the minimal effect of ordering on strengthening properties from
gated by annealing a fully random system at various temperatures: those of a purely random alloy.
E. Antillon et al. / Acta Materialia 190 (2020) 2942 33

The WC parameters after annealing at various temperatures are configurational entropy for systems with high solute concentrations
shown in Fig 4(a) for all the unique pairs in the system (numerical val- and non-zero CSRO using a cluster variation method (CVM) in the
ues are shown in Appendix A.2). For random configurations the WC pair approximation [32]:
parameters are found to be zero within the standard deviation: "  #
P ZX Z
s ðaij Þ ¼ 0:001. At the highest annealing temperature (Tanneal ¼ 1500K), SConf ¼ kB ðZ 1Þ n ðcn lncn cn Þ
2 n;m
ð h cn cm i ln h cn cm i  h cn cm i Þ þ
2
1 ;
the largest positive values correspond to Ti-Ti pairs, whereas the pairs
Ti-X (X=Co,Fe,Ni) show similar negative values. This can be understood ð4Þ
as pure size effect for an alloy composed of large and small atoms, where Z is the number of nearest-neighbors, cn is the concentration of
where the neighbors of a small atom will correspond to larger atoms species n, and ⟨cncm⟩ describes the probability of finding an n-m pair.
more often than in a random alloy, while the neighbors of a large atom The above expression reduces to the ideal Bragg-Williams approxima-
will tend to be small ones. tion [33] when the correlation between the species become indepen-
As the annealing temperature in the Monte-Carlo sampling is low- dent as expected for random systems, i.e. h cn cm i ¼ > h cn i h cm i ;
ered the contribution due to entropic effects decreases accordingly, P
SConf ¼ > kB n ðcn lncn cn Þ. The estimated values of SConf for the
and it is expected that certain atomic pair interactions can lower their model FCC alloy in random and annealed structures are displayed in
energy by forming bonds when these are brought close together. Table 1. As expected SConf decreases relative to the random configura-
Because of the strong attractive Fe-Ti interactions, Fe ends up avoid- tion as the annealing temperature is reduced.
ing all other species, even itself. Interestingly the Nickel-Cobalt sub- In summary, as the annealing temperature is decreased, CSRO
space is largely unaffected by the annealing procedure as these starts to build up inside the crystal. This has several consequences:
atoms are expected to be completely soluble in each other. The mag- (1) large misfits will tend to avoid one another thereby reducing the
nitude of WC parameters as a function of the nearest neighbor shell average misfit interactions inside the lattice, and (2) atomic bonds
is shown in Fig 4(b) for Tanneal ¼ 1100K. It is important to note that that are energetically favored will start to form, thereby lowering the
the lack of CSRO beyond the 3rd shell for all pair-probabilities, is an overall energy of the system, and (3) the SFE becomes more positive
indication that the system is not forming a characteristic wavelength as the annealing temperature decreases. The role of annealing on dis-
associated with long-range order. location mobility is discussed next.
The composition of our model FCC alloy (Co30Fe16.6Ni36.6Ti16.6) was
chosen to produce stacking fault energies close to experimental values 3.2. Resolved critical shear stress
of medium-to-high entropy alloys g SFE ~ 20mJ/m2[30]. The large
annealed cells are used to find the intrinsic stacking fault energies (SFE) The critical stress to unpin edge and screw dislocations as a func-
by displacing half of the cell the stable intrinsic stacking fault vector tion of simulation temperature, for the random and annealed cells,
along < 112 > . Sampling the SFE value over several (111) planes are shown in Fig. 5. For clarity the t (T) results for the different
showed less than 0.5 mJ/m2 variability. As shown in Table 1, annealing annealing temperatures are plotted using different symbols and col-
shows a stabilization of the stacking fault energy with decreasing ors. The annealed cells approximate the equilibrium chemical distri-
annealing temperature. Ding et al have shown that the stacking fault bution at each temperature and can be compared to experimental
energies in CrCoNi systems correlates strongly to the amount of local samples annealed for long times at the same temperature. The critical
chemical short range order in the lattice [31]. In particular, the SFE val- stress values obtained using this method for the random configura-
ues becomes more positive as the annealing temperature decreases [8]. tions are in good agreement with those obtained Rao et al. [11] using
In general this effect is expected in chemically complex alloys contain- identical (random) composition and system dimensions.
ing significant concentrations of a (HCP) stabilizers. In classical solute-dislocation hardening models solute interaction
In the dilute limit the configurational entropy of a solid solution is with the dislocation are assumed to be uncorrelated. Assuming that
often approximated using the Bragg-Williams approach. This approx- this interaction is dominated by misfits, the critical stress needed to
imation assumes that the different atoms mix randomly on the lat- move a dislocation with an edge component is expected to be 1=ð1 nÞ
tice, or a subset of the Wyckoff positions for a crystal with a non greater that the screw counterpart. Using the anisotropic value of Pois-
trivial basis. For systems with high solute concentrations ( >  7%) son’s ratio from Table 1, the interaction energies of the edge component
or exhibiting some degree of CSRO the Bragg Williams approximation are expected to be about 90% greater than the screw, yet the measured
is inadequate. Caro et al. have suggested an expression for the value is only about 30% larger. A fully predictive theory for the edge

Fig. 4. (a) WC parameter for every pair at various annealing temperatures (first shell only). (b) WC parameter as a function of the nearest neighbor shells used to define the pair
probability for a fixed annealing temperature (T=1100 K).
34 E. Antillon et al. / Acta Materialia 190 (2020) 2942

Interestingly, the difference in the critical stress between


1100700 K annealed alloys and the random alloys, is maintained
throughout a large temperature range. This is consistent with the fact
that the strength due to ordering terms is athermal, assuming the
timescales of the dislocation motion are smaller than diffusional
transport [15]. On the other hand, the strength due to random
obstacles is a thermally activated process. This indicates that the
thermal drop-off in the critical stress for this system is still largely
dominated by randomly distributed obstacles.
As a dislocation traverses the simulation cell variations in chemis-
try produce different stacking fault energies. In the random simula-
tion cell the screw dislocation samples volumes that produce large
fluctuations in the stacking fault energy, consistent with the presence
of high-enthalpy atomic-configurations. In these volumes the partial
dislocations are effectively constricted and the dislocation segment
has a finite probability of cross slipping onto another (111) plane. The
partials are observed to pinch off and pin a short length of the dislo-
cation line. The pinning is overcome as the dislocation moves forward
and the pinned dislocation segment cross-slips back onto the original
glide plane. Double cross slip debris are not observed. Such incipient
cross-slip is expected to harden the screw dislocation in the random
simulation environment, but appears to be small as reflected in the
low temperature strength for the screw dislocation in Fig. 5, of the
random and 1500K annealed alloys. Fig. 6 shows a histogram of the
stacking fault distance as the system glides in the random and
annealed systems. It is clear that annealing tends to homogenize the
Shockley partial splitting distance. This is particularly important for
the screw dislocation since it reduces the amount of incipient cross-
slip in screw dislocation of the annealed samples, but also lowers the
probability of having a large partial splitting distance of both edge
and screw random systems.
An estimate of the partial splitting distance (req) can be obtained
from the balance of forces between the partials and knowledge of its
Fig. 5. (TOP) Critical Stress vs temperature for (a) an edge dislocation and (BOTTOM) stacking fault energy [34], viz
an screw dislocation. The different labels correspond the annealing temperatures used
prior to inserting the dislocation into the lattice . mb2
req ¼ f ðnÞ; ð5Þ
g ISF 8p
 
where the factor f ðnÞ ¼ 11n  13 for an edge partial dislocations and
component using the Labusch-Leyson-Varvenne (LLV) approach will be h i
f ðnÞ ¼ 1 3ð1
1
nÞ for a screw partial dislocations, and b is the burgers
discussed in Section 5.
In annealed systems, the critical stress now reflects a balance vector. Using the stacking fault energy, the ⟨111⟩ shear modulus, and
between the loss of size misfit strengthening and the hardening pro- anisotropic Poisson ratio from Table 1, the splitting distance of the
duced when the dislocation cuts CSRO regions of the glide plane. edge and screw partials on the random systems are expected to be
 
edge
Assuming that the size misfit dominates the solute-dislocation hard- rrandom  80A and rrandom
screw
 18A respectively. On the other hand for
ening we expect that the critical stress for the edge dislocation would the T=700 K anneal, these values are expected to decrease by about
decrease and the screw would be unaffected at this anneal tempera- edge
20% and 15% respectively, i.e rannealed

 63A and rannealed
screw
 16A. The


ture. Some of these aspects are reflected in Fig. 5 as follows.


measured values of the equilibrium distance of the edge partials
First, the critical stress to move an edge dislocation through a vol-
shows a decrease on the mean values of the distribution of about
ume that has been annealed at T=1500 K shows initial softening on edge 
edge 

consistent with a decrease in the misfit contribution, while the 10%: h rrandom i  56A vs h rannealed i  51A. Whereas the measured
strengthening contribution due to chemical bond formation is negli-
gible. As described in the previous section, at this annealing tempera-
ture ordering is expected to be dominated by the segregation of
Ti-pairs only, while all other WC-parameters are significantly smaller.
On the other hand, the screw dislocation has a weak coupling to the
(size) misfit, so that any decrease in misfit with annealing tempera-
ture is expected to have a small effect on the critical stress. Second,
the derived critical stresses for annealing temperature below 1100 K
implies that significant bond formation has occurred in the Fe-Ti
pairs which results in additional strengthening mechanism kicking
in. Thus, while the edge dislocation exhibits an initial softening with
decreasing anneal temperature, significant solution hardening is
recovered at annealing temperatures at T=700 K. In contrast, the
screw dislocation exhibits a continuous surge in solution hardening
from 1500 K to 700 K annealing, consistent with the increasing CSRO Fig. 6. Stacking fault distribution for edge/screw dislocations, for random and
on the glide plane. annealed cases.
E. Antillon et al. / Acta Materialia 190 (2020) 2942 35

value for the screw components remains quite similar: h rrandom


screw
i  into two length scales that describe the dislocation bowing required to
 
25 A vs h rannealed
screw
i  24 A. reduce its total energy in the presence of the solutes [35]. According to
An increase in SFE due to annealing results in a more uniform distri- Labusch [13], near the critical unpinning point, it is the smallest scales
bution of the equilibrium distance. This manifests in a reduction of large of bowing that control the unpinning state.
tails on the distribution of the partial splitting for both the edge and Leyson and Varvenne have extended the Labusch model by deriv-
screw components. Similarly, annealing also reduces the amount of ing the length scales self-consistently, using the energetics of the sys-
cross-slip events, which particularly affects screw dislocation since it tems analytically [3,36,37]. In their approach, they envision a local
reduces the amount of incipient cross-slip in the annealed samples, and sinusoidal energy landscape with a periodic distance wc, such that a
in turn increases the mean of the distribution. Fewer cross-slip events minimum amount of work (t bξ cx) is needed to move a dislocation
can be attributed to the reduction of high-enthalpy volumes which segment by a distance x out of its local minimum.
  
correlate to clustering of Ti-rich configurations in the lattice. Fig. 7 DEb px
shows representative snapshots for the screw partials where the color- Hðt ; xÞ ¼ 1 cos  t bξ c x; ð6Þ
2 wc
ing represents the degree of short-range order on the Ti-Ti pairs.
On summary, the results of this section suggest that annealing tends -where x is the coordinate describing the dislocation glide, DEb is
to homogenizes local high-enthalpy volumes and results in a more uni- a characteristic energy associated with a segment length ξ c paral-
form Shockley partial splitting distance. The introduction of chemical lel to the dislocation line describing typical pinning distances,
short range order into the lattice also results in an overall increase of and the roughness wc describes the range of bowing perpendicu-
the resolved critical shear stress with respect to the random case, but lar to the dislocation line. Explicit knowledge about energy land-
affects the screw and edge components in different ways. In order to scape can provide information about the strengthening
understand this interplay and develop a predictive theory, a method parameters governing dislocation flow. While a sinusoidal land-
that measures characteristic energies and length scales near the critical scape is a convenient construct in analytical models, it may not
unpinning under stress is presented next. The results of this approach be appropriate in crystal lattices with large solute-concentrations
are then used to incorporate the effect of CSRO on the prediction of or those involving large distortions [35,38]. This approximation
strengthening properties from solid solution models. can be relaxed by performing large scale molecular dynamic sim-
ulations on chemically complex systems.
Previously [12], the authors demonstrated a technique that
4. Measuring strengthening parameters using an energy topology includes all the strengthening mechanisms based on the energy land-
approach
scape sampled by the gliding dislocation. This also offers a direct
means for assessing the strengthening parameters associated with
A general energy landscape that captures the interplay of local pin-
local deformation using atomistic, molecular-dynamics simulations.
ning forces and the externally applied load provides a useful theoretical The approach has been validated in a system exhibiting strong chem-
construct in describing athermal and thermally activated deformation
ical effects [39] in a binary alloy. Here the technique is applied to the
processes. The Labusch theory of solid-solution-strengthening maps the current multicomponent alloy in order to directly measure the con-
interaction of a dislocation with a random distribution of solute sites
tribution of solution hardening due to CSRO. Then, this information is
used to derive the strengthening parameters used in a Labusch
description of solution hardening as discussed below.
Consider a generic energy profile E(x) obtained as a dislocation
moves in the random alloy. This profile is obtained by computing the
work done in moving dislocation over obstacles in the lattice and
subtracting the expected (analytical) work for an equivalent straight
dislocation moving through a defect free glide plane. In this case the
work done on a straight dislocation can be readily estimated by the
! ! !
work done by the Peach-Koehler force, viz F =L ¼ ðt ¢ bÞ  dl where
!
b is the burgers vector of the dislocation, t is the local stress tensor,
!
and dl is the unit vector tangent to the dislocation line [12].
A two-point auto-correlation function can be used to extract char-
acteristic energy and length scales from a complex landscape. As an
application example, take the analytical energy landscape (Eq. (6)),
apply the auto-correlation function,
Z L
Cðxo Þ ¼ hðxÞhðx þ xo Þdx ð7Þ
0

 
L ¢ DEb2 px o
¼ cos ð8Þ
8 wc
where hðxÞ  Hð0; xÞ h Hð0; xÞ i ; here the mean value has been sub-
tracted in order to make the fluctuation symmetric. The values xo
where the above function vanishes identifies the smallest lengths
scale at which the energy landscape becomes uncorrelated. For a
sinusoidal function this occurs when xo ¼ wc =2. Moreover, the above
Fig. 7. Snapshots for screw partials for (TOP) annealed system (T=1100 K Annealing) self-correlation function can also be used to estimate a characteristic
and random system (BOTTOM). The partials (black lines) move from left to right. The energy barrier, viz
colors in the lattice indicate the sign of the WC parameter for Ti-Ti pairs: Blue (nega- pffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tive), Green (zero), and Red (positive) (For interpretation of the references to color in DEb ¼ 2 2 Cðxo ¼ 0Þ=ðLÞ ð9Þ
this figure legend, the reader is referred to the web version of this article.).
36 E. Antillon et al. / Acta Materialia 190 (2020) 2942

Clearly, this approach is not needed if the energy landscapes are ana- until the segments become unpinned. This process can repeat itself
lytical. Its usefulness lies in estimating characteristic energy and in other neighboring regions, until the dislocation experiences
length scales from landscapes that are rugged and poorly represented enough stress and acquires significant inertia to overcome the rest of
using a simple trigonometric function. This approach is now applied the obstacles in the lattice. While MD strain rates are inherently large,
to the energy landscapes obtained directly from the simulation. this effect can be minimized by focusing on regimes where the dislo-
Energy landscapes generated as an edge dislocation moves through cation overcomes obstacles in a quasi-static manner. The velocity of
different sections of a random alloy are shown on Fig. 8 (a)(c). Recall the trailing partial is shown in Fig. 8 (d)(f), where a regime can be

that we are interested in quantifying the smallest scale of bowing near seen where the partials moves by  1A =ps ¼ 100m=s  c=60 (where
the unpinning state under stress and their corresponding energy fluctu- c is the speed of sound). This region, indicated between the two verti-
ations. Hence, the chosen temperature is 5K to reduce thermal noise cal dashed lines will be considered quasistatic, whereas outside of
and the stress is ramped from below the critical stress as described in these regions the partials can be seen acquire larger velocity, and as a
the methods section, until the unpinning state occurs. In these low tem- result the energy fluctuations (due to the dislocation interacting with
perature simulations we explicitly ignore zero point energy effects and the lattice) grow accordingly. Fig. 8 (g)(i) illustrates snapshots of
issues related to low temperature thermal transport that could in princi- the partials taken every picosecond during the dislocation motion. As
ple effect low temperature (cryogenic) strength [40,41]. the partials come to rest momentarily, these lines collapse onto the
Unpinning occurs once the dislocation defeats local obstacles same shape before venturing onto the next favorable position. An
inhibiting its motion. As the applied load is increased, the line length example of where this has occurred within the quasistatic regime has
of the dislocation increases by bowing (increasing its total energy), been highlighted with a red line.

Fig. 8. Topological energy landscape of an edge dislocation moving through a lattice of random distributed elements at 5 K. Figs (a)(c) show the landscape due to the movement of
the partials vs position of the trailing partial starting at 3 different position in the lattice. Figs (d)(f) show the corresponding velocity profiles, and Figs (g)(i) show snapshots of the
trailing partials at various locations in the lattice taken every picosecond. The region between the dashed lines illustrate the domains where the dislocation moves quasi-statically as
described in the text.
E. Antillon et al. / Acta Materialia 190 (2020) 2942 37

Characteristic energies scales can now be extracted from the rugged method is demonstrated next when the method is applied to systems
energy profile near the unpinning sections using the auto-correlation exhibiting various degrees of short-range ordering.
formalism mentioned above. Only regions where the dislocation motion
is quasi-stationary, can be considered useful in order to extract charac- 4.0.1. Topological landscape for systems with chemical short range
teristic scales relevant to the critical unpinning state. Using the portion order
of the landscape inside the dashed vertical lines in Fig. 8, yields charac- Dislocations moving inside an annealed alloy will now encounter
teristic roughening scale (starting from left to right) on the order of an additional strengthening mechanism due to the energy needed to

wc ¼ f10:0; 12:7; 13:4g A with corresponding estimate on the energy destroy the chemical order represented by atoms adjacent to the
scale DEb ¼ f2:4; 3:30; 3:22g eV.
glide plane. Upon shearing the chemically ordered and segregated
The energies obtained from the landscape apply to the whole dislo- regions, it is expected that the randomness in the lattice also
cation line (length L), hence to obtain the energy at the smallest

scale of
increases the overall energy [15]. In contrast, any energy spend in
bowing (2ξ c), the total energy was scaled by a factor 2Lξ c [12]. To esti- breaking local-bonds in purely random alloys will be gained by
mate the critical length scale along the dislocation line (ξ c), a similar
another segment forming similar bonds, such as the randomness of
approach used on the energy landscape can be used, except that the dis- the solutes is maintained. Thus, the change in energy (due to segment
location shape Y(X) at the critical points is used instead to extract the
of length L moving distance X) is expected to equal to the work done
smallest bowing length scales. Applying the auto-correlation function by the applied stress, that is DE t bLX ¼ 0.
on the dislocation shape yðxÞ  YðxÞ h YðxÞ i ; where X ¼ h 112 i and
 For systems, with some degree of ordering, Fisher [15] proposes
Y ¼ h 110 i ; from Fig. 8(g)(i) yields values of ξ C ¼ f98; 122; 102g A, that the energy increase (per unit area) can serve as an estimate for
respectively.
chemical short range order; this energy is often referred to as a dif-
In a similar vein, one can also estimate the bowing length (w) by fuse anti-phase boundaries energy (DAPB) since this is the amount of
measuring the height fluctuations of the dislocation in the ⟨110⟩
energy needed to disrupt the localized order of the nearby atoms.
direction. Assuming a
that the dislocation line follows a sinusoidal Deviations from the expression above serve to quantify the energy
form yðxÞ ¼ wc0 sin p2ξx ; use of the auto-correlation function gives
per unit area produced by the diffuse anti-phase interface, viz
Z L  
px DE=ðLXÞ t b ¼ g DAPB ; ð13Þ
C 0 ð0Þ ¼ wc02 sin ð10Þ
0 2ξ
DAPB
where g is the energy per unit area produced by the disordered
!
wc02 L ξ sin p2ξx interface. Measuring this energy explicitly from the simulation can
¼ 1 ; ð11Þ be done in a straightforward manner provided the applied stress t
2 Lp
and the change in energy DE due to the movement of a dislocation
whence in the limit L  ξ , an estimate for ffiffi height
fluctuation of
the
pffiffiffip
can be measured reliably [12]
the dislocation is given by wc0  2 ðC 0 ð0Þ=L . Using the As shown in the previous section, in random system the changes in
highlighted red lines shown in Fig. 8(g)(i) as before, values of wc0 energy tend to fluctuate around a mean value expected from the

¼ f17:8; 19:1; 12:9g A are obtained. applied stress. On the other hand, system with non-vanishing short-
The estimates for the bowing length using the energy landscape range order will show a systematic deviation in the quantity DE t LbX;
(wc) tend to be smaller than estimates obtained using the dislocation due to the additional work needed to overcome ordered regions. Such
shape near the critical stress (wc0 ). This is in part because on the latter deviations are shown in Fig. 9 for system with various degrees of short-
a specific form of the dislocation has been assumed to quantify its range order as a function of the trailing partial. The measured slopes
height fluctuation, for p example
ffiffiffi assuming insteadpffiffiffi a saw-tooth shape (per unit length) on the above graph provide estimates for the diffuse
2
edge ¼ 5, 14, 17 mJ/m for
would yield a factor of 3 instead of the value 2 above. On the other anti-phase boundary (via Eq. (13)), yielding g DAPB
hand using the energy landscape, no explicit shape has been Tanneal ¼ 1500, 1100, and 700 K, respectively for the edge dislocations,
assumed, rather the scale wc is obtained by finding the smallest value 2
screw ¼ 4, 13, 21 mJ/m for the screw components. Note that for
and g DAPB
(xo) where the convolution of the energy landscape as a function of
the random alloys, the deviation in energy with respect to work due to
x¼ h 110 i vanishes (within the quasi-static regime). A more detailed
the Peach-Koehler force is less than 0.25mJ/m2, which is the same order
analysis on the dislocation shape can be envisioned to introduce a
of magnitude for the standard deviation of the fits.
hierarchy of length scales that filter out larger harmonics not relevant
The ordering energy can also be estimated by reversing the
to dislocation unpinning. In principle this approach should bring
applied load and driving the dislocation back through the already cut
down the estimate of wc0 . Moreover, during free-flow and/or at larger
glide plane. As described in the methods section, the dislocations are
working temperatures, larger bowing scales are expected to follow
introduced into the lattice using the anisotropic elastic solution for
self-similar scaling [35,42,43]. Yet, these effects are not relevant to
the atomic displacement field of the dislocation. Effectively the cell is
the main study of this manuscript, which is to study the role of CSRO
sheared (edge or screw) in the glide plane, placing the dislocation in
on dislocation mobility, and can be explored in a future publication.
the cell interior. This operation shears possibly chemically ordered
Assuming sinusoidal energy landscape Eq. (6), it follows that the
regions. If the load is reversed, the moving dislocation recreates the
critical stress can be derived when the enthalpy term vanishes under
CSRO present in the original annealed crystal. This produces a nega-
stress, @Hð@xt;xÞ ¼ 0; yielding: tive DE t bLX, with the same slope as shown in Fig. 9. The difference
p DEb in the critical stress needed to move a dislocation in opposite direc-
t o ðwc Þ ¼ : ð12Þ tions was found to be Dt  100.0 MPa for Tanneal ¼ 1100 K. Since con-
2 bξ c wc
tributions due to random obstacles are expected to be the same in
the critical unpinning stress estimated using this equation is tp 
c ffiffiffi
either direction, the difference in stress can be used to estimate the
amount of CSRO, viz g DAPB ¼ Dt b=2  12½mJ=m2 ; which agrees with
 
230 MPa, using DEb  3.0, wc  12 A, ξ ¼ 105A , and bedge ¼ a= 2.
This value is in reasonable agreement with the values obtained on the values obtained above.
the critical resolved shear stress by the simulation of an edge disloca- The method presented in this section has been shown to capture
tion on a random solid solution. This suggest that the method can be the energetics of different obstacle-dislocation interactions on a
used at low temperatures to quantify the energy barriers for model alloy. These effects have then been quantified in terms of
obstacles irrespective of their nature. Moreover, the versatility of this strengthening parameters that can be used to inform strengthening
38 E. Antillon et al. / Acta Materialia 190 (2020) 2942

In general, the standard deviation in potential energy between a


set of atoms in lattice sites (xi, yj, zk) and (xp, yq, zr) interacting with
the dislocation at the origin can written as,
X Xh i
s 2DUtot ¼ pqr i  h nijk i h npqr i DUijk DUpqr
h nnijk nm n m n m
ð15Þ
i;j;k;p;q;r n;m

In the original description by Varvenne et al. [3], ignores the inter-


action between different atomic rows (xi, yj) 6¼ (xp, yq) since such sites
are assumed to be uncorrelated, and thus, the standard deviation is
done only for terms in the same atomic row (xi, yj). In this case, the
energy fluctuation along zk can be easily obtained assuming that the
occupancy of the solutes in the lattice follows random numbers
according to the binomial distribution. The moments of this distribu-
tion can be readily generated and used to estimate the overall stan-
dard deviation of the total energy for uncorrelated solutes interacting
with the dislocation. The results can be shown to be linear in concen-
Fig. 9. Energy landscape comparison for various degrees of short-range order as a
tration of the solutes (see A.11 in [3]):
function of the trailing partial position. Note that systems with some degree of order-
XX
ing show a systematic deviation which can be used to estimate the energy contribution
s 2DUsolute ¼ Ns cn DUi;j
n
ðwÞ2 ; ð16Þ
due to CSRO. The slope of the curves corresponds to the average ordering energy (per tot
i;j n
unit length) encountered as the dislocation shears the lattice.
where Ns denotes the number of atomic sites along z direction for a
models. Let us now consider analytical models which can describe dislocation segment of length z, confined to the atomic row (xi, yj). In
the interplay between chemical short-range ordering and that due to principle this analysis can also describe possible local variations in
random misfit interaction in complex alloys. the solute/dislocation interaction energy due to the local atomic envi-
ronment around any given solute. This effect appears via an addi-
tional sigma term: s 2DU . In random alloys this effect is dismissed as
i;j
5. Incorporating CSRO into solid solution models being negligible [3], while for system with some degree of ordering,
it is not clear how this term arises.
A general model for high concentration solid solution hardening We propose an extension of Labusch-Leyson-Varvenne solution
should account for the effect of local misfit defects as well as chemical hardening models that includes the hardening effects of CSRO with-
short-range order (CSRO). Labusch-type models have been successful at out assuming that solutes in different atomic rows are uncorrelated.
quantifying overall contribution of random obstacles affecting disloca- The contribution of these solute-solute terms to the total fluctuation
tion glide. This approach attributes the strength of alloys to the magni- in energy can be included directly from Eq. (15) by counting the like-
tude of the fluctuations in various interaction energy terms with the lihood that the various nearest neighbors (xp, yq, zr) occur relative to
dislocation [13]. Models that consider CSRO attribute the strength (xi, yj, zk), viz
instead to preferential bond-formation in the crystal [17], and the energy XX
s 2DUsolute pair ¼ Z cn cm anm DUijk
n
DUim0 j0 k0 ; ð17Þ
needed to break these bonds is what inhibits dislocation flow [16]. tot
i;j;k n;m
In this work, we follow closely the interplay between these com-
n,m
peting schools of thought with the intention of combining them into where a is the WC short range order coefficients for first neighbors
a general framework for modeling the effects of chemical short range (see Eq. (1)), and the sum over the second nearest neighbors sites (p,
order on the critical stress of multicomponent concentrated alloys. q,r) has been replaced by the coordination number Z of the lattice,
One requirement of the analytic approach is that any necessary and DUim0 j0 k0 is the energy change at a representative nearest-neighbor
parameters can be calculated using atomistics or ab initio methods, site of solute type m.
these include the WC parameters and the diffuse anti-phase bound- In what follows, the energy changes after the dislocation moves a
ary energy (g DAPB). To show proof of concept, the FCC model alloy distance w is split into two classes: (1) those interacting defects
investigated in the previous section is analyzed next. whose paired contribution is assumed to be just the sum of individual
The interaction energy, Un(xi, yj, zk), between a solute and a disloca- defect separately, (2) and those that do not. Misfit volumes interact-
tion segment can be written in terms of the position of the solute (xi, yj) ing with the pressure field of the dislocation can be described by the
relative to the center of the dislocation, and its position zk along the dis- former class, whereas energy changes due to atomic bond formation
location directions ⟨112⟩. The method introduced by Varvenne et al uses represent the latter type of interactions.
an effective medium to describe the effect of a multiple solutes on a con- Focusing on the first class, the interaction energy due to the dislo-
centrated solid solution. The potential energy change as the dislocation cation pressure field (p) and misfit volume DVn of type n can be writ-
glides a distance w along the x direction ⟨110⟩ is given by [3] ten as Unmisfit ¼ p  DV n . In the recent work of Varvenne the relevant
XX solute-dislocation interactions are expressed as DVn2 weighted by the
DUtot ðz; wÞ ¼ ij DUi;j ðwÞ;
nm m
ð14Þ
solute concentration cn, whereas the cross-terms are assumed to van-
i;j m
ish. While this assumption can be justified in purely random alloys, in
where nm is the occupation of solutes in atomic row (xi, yj) of type m alloys with some degree of ordering this approximation needs to be
ij P
along the dislocation line zk. The sum m nm ij ¼ 1; since at least one revisited. In order to include a correction for CSRO into solution hard-
type occupies any given point in the lattice. Of particular importance ening models of the Labusch type we assume that regions with a
in this model, are standard deviations on the interactions energies non-zero WC parameter (an,m) are statistically best represented by
arising both from fluctuations in solute concentration and their vari- the density defect pairs (n,m) interacting with the dislocation. A
ous interaction with the dislocation.
pffiffi Unlike the interaction energies defect pair DVnm acts as the sum of the volume changes due to each
which fall as ~ 1/r, where r ¼ ðx2i þ y2j Þ; the energy fluctuations fall individual defect: DVnm ¼ DVn þ DVm ; direct atomistic simulations
off at a faster rate ( ~ 1/r2) so that only a finite regions around the dis- were used to verify this hypothesis. Thus the correction due to paired
location needs to be considered [44]. misfit volumes can be written using Eq. (17), with nearest neighbor
E. Antillon et al. / Acta Materialia 190 (2020) 2942 39

atoms with misfit volumes DVn and DVm. Note that in order to avoid our present alloy Co30Fe16.67Ni36.67Ti16.67 which is on the order of 2,
counting the same bond (i,j,k)-(i’,j’,k’) twice, the sum over all lattice gives f1(wc) ~ 0.2 and f2(wc) ~ 5.0 [46]. The values of DVn for Co, Ni, Fe
sites (i,j,k) needs to be corrected. and Ti solutes in the average Co30Fe16.67Ni36.67Ti16.67 alloy have been
We suggest a “misfit” correction that arises for systems with non- calculated using atomistic simulations and is given in reference[11].
vanishing chemical short range order. The overall standard deviation Neglecting the contribution from s DVn ; t y0 for a random Co30Fe16.67-
on the energy interaction due to misfit terms that also includes CSRO Ni36.67Ti16.67 alloy is calculated to be 222 MPa which is very close to
can be written as: the value given by direct atomistic molecular dynamics simulations.
" # In the proposed extension of the Labusch-Leyson-Varvenne model,
P P Zð111Þ X
s 2;misfit
DUtot ¼ Ns i;j p2ij ðwÞ n cn DVn2  cn cm an;m ððDVn þ DVm Þ2  DVn2  DVm
2
Þ ; the effect of short range order on the dislocation core coefficients, f1(wc)
2 n;m
and f2(wc) is assumed to be negligible. Using Eqs. (19) and (20) with
ð18Þ
non-vanishing short range order coefficients given in Table 3, the contri-
where pij ðwÞ  pðxi w; yj Þpðxi ; yj Þ; is the pressure difference due to bution of elastic size misfit to the CRSS of a/2⟨110⟩ edge dislocations at
the dislocation from site ðxi w; yj Þ to site (xi, yj)[45], w is the bowing 0K for various annealing conditions (1500K, 1100K and 700K) can be
scale along the glide direction, and the factor of 12 avoids counting the derived. Note that to first order this correction is not expected to affect
same bond twice when summing over the lattice sites (i,j), and Z(111) the critical strength of the screw components.
is the number of first nearest neighbors on the (111) glide plane. Der- The discussion so far has introduced a correction into the analyti-
ivation of this last equation assumes that the volume change due to a cal framework due to Varvenne et al. which describes the change of
solute pair is the sum of the volume changes due to the individual statistics of the local misfit interactions with the dislocation. How-
solutes, and that their interaction with the (edge) dislocation is addi- ever, this approach neglects to include chemical interaction effects
tive. For solutes lying in the same (111) plane, this can be justified that arise from lowering the energy of the system due to bond forma-
since the change in pressure over a bowing distance w is of the same tion. From this point of view, slip encounters clusters with non-van-
order for both misfit defects, and therefore the interaction energy of ishing CSRO dispersed uniformly throughout the lattice.
the solute pair can be easily written down as additive. The standard Quantification of such effect to the strength of the lattice, dates back
deviation of the interaction energy due to the pair sites then enters to the works of Fisher [15], Flinn [16], and Cohen [10], which ascribe
as DEpair
2
/ pij ðwÞ2  ðDVn þ DVm Þ2 . Moreover, the contribution of strength in alloys to the energy needed to break bonds during slip in
each terms has to be subtracted in order to avoid counting their indi- a lattice with non-vanishing local order.
vidual contribution more than once and provide the right limits as For an fcc lattice, slip on the close-packed plane, (111), consists of
shown in Eq. (18). For example, if either DVn or DVm is zero the con- Z(111) = 6 that lie in this plane, while three lie in the plane above, and
tribution from the solute pairs is zero as required. For solute pairs not three in the plane below. If we displace the plane above or below by
lying in the glide plane but are across the glide plane, the term pij(w) one Burgers vector (b = l/2 [110]), only two of the three neighbors in
will vary at different yj sites, and therefore the addition argument this plane will change bonds [16]. Unlike the misfit components
does not hold. It is worth pointing out that the correction we propose whose interaction with the dislocation can span several (111) planes,
can take either sign depending on the sign of WC parameter and the the bond-breaking contribution is localized to the slip plane. The
individual misfit volumes, unlike the misfit term s 2DV which is energy change in alloys with non-zero CSRO should increase given
i;j
always positive quantity. For example, species m with a large misfit that a dislocation movement introduces more disorder in the lattice.
volume will have a misfit correction term second order in concentra- Hence, the energy change due to breaking chemical bonds can be
tion that lessens the magnitude of the energy fluctuations of the self- written using Eq. 17 with Z=2,
terms, since on average such types will tend to avoid one another X
DEbondbreak ¼ 2 cn cm an;m U n;m ; ð21Þ
(am,m > 0). The converse being true as well. n;m
The prediction of CRSS for the motion of a/2[110] edge disloca-
n,m
tions at 0K, t o and the energy barrier to escape the energy well, DEb, where U is the binding energy between atoms of type n and m that
can now be written as [46] are nearest neighbors. Using the mean-field potential method due to
  Varvenne et al. [47], these values have been estimated for all pairs as
1 þ n 4=3 shown in Table 2 and have been validated using large statistics with
t y0 ¼ 0:051a1=3 mb4  f1 ðwc Þ
1 n full random lattice as shown in Appendix A.1. For the Fe-Ti case, the
" #2=3 FeTi
binding energy yields a value of Ubond e0:058; where as for Ti-Ti this
X X
 cn ðDVn2 þ s 2DVn ÞZð111Þ cn cm an;m DVn DVm ð19Þ value repulsive Ubond e0:10. An order of magnitude estimate on the
TiTi
n n;m Fe-Ti binding energy is calculated by noting bonds are not formed at
  T=1500 K, but they start to become prevalent at T=1100 K. That is the
1þn energy to break a Ti-Fe bond can be estimated as 32 kB DTe0:05eV
DEb ¼ 0:274a1=3 mb  f2 ðwc Þ
1 n where DT ¼ 1500K 1100K; which is similar to the value found in
" #1=3
X X Table 2.
 cn ðDVn þ s DVn ÞZð111Þ
2 2
cn cm a DVn DVm
n;m
; The above energy change can now
n n;m pffiffi be divided by the area per
atom in the slip plane, which is a2 ð3Þ=4; to yield the total energy
ð20Þ (per atom) released in breaking the ordered bonds due to slip, viz
where m is the shear modulus, n is the Poisson ratio (see Table 1), the 8 X
line tension G is written as amb2 with a ~ 0.123, b is the Burgers vector g DAPB ¼ pffiffiffi cn cm an;m Un;m ; ð22Þ
a2 3 n;m
of the dislocation, wc is the critical amplitude of the bowed out edge dis-
location segment, f1(wc) and f2(wc) are the minimized dislocation core and the critical stress due to the diffuse anti-phase boundary is
coefficients. For large values of the Shockley partial splitting width, the obtained from t DAPB b ¼ g DAPB ; where b is the burgers vector. This
dislocation core coefficients are only dependent on the core width of expression suggests that atomistic results can provide an estimate for
the Shockley partials, s , which is inversely proportional to the unstable the diffuse anti-phase boundary energy (g DAPB) using values for the
stacking fault energy, g uns, as 0:28mb=½4pð1 nÞg uns [3,46]. binding energies (Unm) and WC parameters (anm) for all pairs. Note
For random systems, Eqs..(19) and (20) assume that the WC- that ab initio methods can be used to calculate the diffuse anti-phase
parameters vanish, therefore using the values of s (in units of b) for boundary energy directly [7] or by cluster expansion methods [48].
40 E. Antillon et al. / Acta Materialia 190 (2020) 2942

Table 2 higher annealing temperatures, which results in an CRSS being almost


The first neighbor solute-solute interaction ener- identical to the fully random material. However, at lower annealing tem-
gies are obtained by using the mean-field potential
peratures (T = 700K) the misfit term saturates, and as a consequence the
due to Varvenne [47].
bond-breaking contribution continues to increase accordingly, such that
Uij [eV] Ni Co Fe Ti the overall critical stress becomes larger than the purely random term.
Ni -0.002 0.001 0.008 -0.010 On the other hand, for the screw dislocation the misfit terms of the dis-
Co 0.002 0.009 -0.015 sociated partial are not expected to affect the yield strength. The model
Fe 0.022 -0.058 misfit strength on the screw components in Fig. 10 is taken as the value
Ti 0.100
measured using molecular dynamics for the random system. We see
that the contribution due to random misfit obstacles appears constant
throughout for all the annealing temperatures, although there is a small
Finally, the relative contribution of the various terms affecting drop in the annealed systems compared to the random case which could
CRSS discussed above can be compared. Fig. 10 breaks down the vari- be due to misfit terms on the dissociated partials. Because the misfit
ous contribution to the CRSS in terms of analytical models (AM) and term appears constant throughout all annealing temperatures, the
the values calculated (MD) in the previous section for both an edge strength of the screw components increases with the amount of chemi-
and a screw dislocation. The solid bars correspond to the direct mea- cal short-range order. Interestingly, at T=700 K annealing, the overall
surement from the molecular dynamic simulations (at T=5K), which critical stress in the screw and edge values yield similar values and this
have been broken down into contributions from the fluctuating trend persists over a wide range of working temperatures (see Fig. 5).
chemical field (dark-red) and CSRO (cyan). The latter values are taken
from analysis of Fig. 9. The dashed bars correspond to predictions 6. Summary
based on the modified-misfit model which accounts for CSRO (Eq.
(19)) and the bond-breaking model (Eq. (22)) as described above. The The effects of chemical short range order on the deformation of an
results due to the modified analytical model are in fairly good agree- fcc Compositionally Complex Alloy (Co30-Fe16.67-Ni36.67-Ti16.67)
ment with direct atomistic simulations, given that the uncertainty on were studied using an atomistic approach. Embedded Atom Method
the critical stress is of the order of dt ~ 20MPa (see methods section). potentials of the Johnson form produce a stable fcc solid solution at

For an edge dislocation, the misfit term and bond-breaking terms this composition. A large simulation cell (1236 x 1141 x 202 A) was
contributing to the overall strength appear to be anti-correlated at constructed in the dislocation coordinate system, with the elements
distributed randomly throughout the lattice. A hybrid Molecular

Dynamics Monte Carlo method was then used to anneal a 50 Aslab of
(  6 million) atoms at three temperatures (700, 1100 and 1500K)
which was then replicated to the dimensions of the larger simulation
cell. Molecular dynamic simulations were used to explore the effects
of chemical short range order on the strength of the random and
annealed cells with incremental increases in the applied load for
edge and screw dislocations in the designated glide plane
The degree of CSRO produced by annealing was assessed using the
Warren-Cowley metric for local chemical composition. The atomistic
representation of this CCA produces strong chemical ordering
between Fe and Ti, while Ti segregates from Ti atoms. Annealing also
produces a systematic redistribution of the atomic sites with the larg-
est local hydrostatic strain (i.e. size misfit), moving to smaller values
such that the overall distribution becomes more compact at lower
annealing temperatures. As expected, these trends increase with
decreasing annealing temperature.
Key observation of this study is that the average misfit interaction
inside the crystal decreases as the system becomes more ordered,
while at the same time, atomic bonds that are energetically favored
begin to form lowering the overall energy of the system. It was found
that for a dislocation with an edge character, these two mechanisms
have an opposite effect on the critical stress needed to unpin a dislo-
cation, whereas for a screw dislocation, the effect of the misfit term is
negligible and the critical stress rises with increasing CSRO. Thus, the
overall strength is essentially a competition between these two
effects which is likely to be system dependent. The methodology
developed in this study is intended to decouple these effects and pro-
vide an approach for incorporating the contribution due to chemical
short-range order into current solid solution models.
Correctly measuring the interplay of dislocation with misfit inter-
actions and chemical effects (bond formation and chemical segrega-
tion) poses various challenges for multicomponent alloys. The
approach taken in this study was to evaluate solid solution strength-
ening effects by calculating the energy topology of the dislocation
Fig. 10. Misfit (red) and short-range ordered (cyan) contribution of the critical stress at
moving through the complex field of atoms as compared to the ana-
T= 5 K measured from MD. The dashed bars correspond to the contribution due to the
analytical models (AM) using bond-breaking terms and change in misfit terms as lytic work required to move the dislocation through a unit area. The
described in the text. (For interpretation of the references to colour in this figure leg- measurement from this methodology allowed us to separate the con-
end, the reader is referred to the web version of this article.) tributions due to chemical ordering and other misfit solution
E. Antillon et al. / Acta Materialia 190 (2020) 2942 41

hardening mechanisms. Moreover, this approach provides quantita-


tive information on the sources of solution hardening that can be
used to inform solid solution models. A simple modification to the
statistics of random solid solution model due to Varvenne et al is pro-
posed which accounts for the correlation of neighboring solutes on
the misfit interaction with the dislocation. Additionally, a supple-
mental strengthening mechanism is added which captures the ener-
getics involved in breaking favorable bonds during slip. The
inclusion of both of these contribution captures the observed trends
in the yield stress due to the increasing degree of ordering in the lat-
tice.
While the approach was demonstrated on an model alloy sys-
tem composition we expect that the methodology developed
here is quite general and can be applied to a wide variety of mul-
ticomponent alloys. The required parameters for the extended
analytic model, Warren-Cowley estimates of CSRO, the diffuse
anti-phase boundary energy, or the change in the configurational
energy to relate these parameters, are readily accessible from
current DFT methods. It is expected that chemically complex
commercial alloys as well as the general class of High Entropy
Alloys will exhibit various degrees of chemical ordering which
will influence strength over a wide range of temperatures.

7. Acknowledgments

The authors (EA, CW, SR, and BA) dedicate this work to the
memory of Dr. Triplicane Parthasarathy, our friend and colleague Fig. 11. Distribution of the binding energies for Ti-Ti pairs (TOP) and Fe-Ti pairs
of many years. EA would like to gratefully acknowledge an insight- (BOTTOM) using Eq. (23) for random alloy (solid lines) and the average alloy
ful discussion with W. Curtin on the role of short-range order on (dashed lines). The mean values of the distribution and the mean field approach
are shown in the legend.
solid solution hardening models. This research was supported in
part by a grant from the Air Force Research Laboratory, Materials
and Manufacturing Directorate (F33615-01-C-5214), and com-
puter time from the DoD High Performance Computing Moderniza-
tion Program at the Air Force Research Laboratory DoD changed to species A and B, respectively, providing an estimate of
Supercomputing Research Center (AFRL-DSRC). U AþB . Therefore, an estimate of the binding energy UA,B can be readily
obtained via Eq. (23).
The variability of the estimate on the binding energy for species A
Appendix A and B is obtained by repeating this approach on various nearest
neighbors and different independent configurations of the composi-
A1. Binding energies in a random lattice tion Ni36.6Co30.0Fe16.6Ti16.6 with a system size of 9216 atoms. Energies
were estimated using conjugate gradient minimization. A total of 500
Formally the binding energy is defined as the change in the different random configuration were used to evaluate each pair, and
energy required to bring two atomic species from far way to the near- for each configuration, a single central atom was selected at random,
est neighbor position. In practice, this energy can be defined as the while all of its 12 neighbors neighbors were used to collect statistics;
difference between a system with the pair minus the energy of the this amounts to a total of 6000 entries per pair. The distribution of Ti-
individual parts Ti and Fe-Ti pairs is shown in histograms from on Fig 11. The dashed
U A;B ¼ U AþB U A U B : ð23Þ (gray) lines correspond to the energy obtained using the average lat-
tice approach due to Varvenne et al. Note that the distributions from
In the average alloy obtained using a mean-field approach on the the random alloy shows a peak at zero which corresponds to pairs
interatomic potentials [47], evaluation of these binding energies is that happened to already be species A and B. This contribution needs
straight forward. The results are shown in Table 2. In real concen- to be counted as well, and in doing so, the mean value tends to
trated alloys, the binding energy is not easily defined since this value slightly shift away from other non-zero modes present in the distri-
can be highly dependent in the local chemical environment. Never- bution. Nevertheless, the mean values of the random distributions
theless, this value can be obtained by evaluating the above expres- and the estimate of the mean field approach are in very close agree-
sion at various sites in the lattice. ment, showing that the mean-field approach is able to capture the
To evaluate the binding energy of atoms A and B, an atom was energetics of complex concentrated alloys.
selected at random (central atom) along with one of its nearest
neighbors (neighbor atom). The species of the central atom is
changed to species A; this provides an estimate of energies UA. Next, A2. WC parameters
the species of the neighbor atom is changed to species B indepen-
dently of the central atom; this provides an estimate of energies UB. Numerical values for the Warrren-Cowley parameters as obtained
Next, both the species of the central and the neighboring atom are by the Hybrid MC/MD simulations are shown in Table 3
42 E. Antillon et al. / Acta Materialia 190 (2020) 2942

Table 3 [15] J. Fisher, On the strength of solid solution alloys, Acta Metall. 2 (1) (1954) 9–10.
Warren-Cowley parameters (in units of 103 ) for an FCC alloy [16] P. Flinn, Solute hardening of close-packed solid solutions, Acta Metall. 6 (10)
Co30 Fe16.67 Ni36.67 Ti16.67 in a completely random system and (1958) 631–635.
annealed system at various temperatures. The error in paren- [17] L. Proville, S. Patinet, Atomic-scale models for hardening in fcc solid solutions,
thesis correspond to the error on the last digit shown . Phys. Rev. B 82 (5) (2010) 054115.
[18] S. Plimpton, Fast parallel algorithms for short-range molecular dynamics, J. Com-
Ni Co Fe Ti put. Phys. 117 (1) (1995) 1–19.
RANDOM [19] X.W. Zhou, R.A. Johnson, H.N.G. Wadley, Misfit-energy-increasing dislocations in
vapor-deposited cofe/nife multilayers, Phys. Rev. B 69 (2004) 144113, doi:
Ni +0.2(2) + 0.2(3) - 0.2(3) - 0.6(4) 10.1103/PhysRevB.69.144113.
Co + 0.2(3) + 0.2(3) - 0.6(5) - 0.2(5) [20] A. Stroh, Dislocations and cracks in anisotropic elasticity, Philos. Mag. 3 (30)
Fe - 0.2(4) - 0.7(5) + 0.5(7) + 1.2(7) (1958) 625–646.
Ti - 0.5(4) - 0.2(5) + 1.1(7) + 0.5(7) [21] D. Rodney, Molecular dynamics simulation of screw dislocations interacting with
T=1500 K Ni Co Fe Ti interstitial frank loops in a model fcc crystal, Acta Mater 52 (3) (2004) 607–614.
ANNEAL [22] J.W. Christian, The Theory of Transformations and Metals and Alloys, Pergamon
Press, 1965.
Ni + 4(1) + 4(1) + 5(1) - 21(1)
[23] D. Frenkel, B. Smit, Understanding molecular simulation: from algorithms to
Co + 5(1) + 4(1) + 5(1) - 23(1)
applications, Vol. 1. Elsevier, 2001.
Fe + 3(1) + 6(1) + 6(1) - 25(1) [24] B. Sadigh, P. Erhart, A. Stukowski, A. Caro, E. Martinez, L. Zepeda-Ruiz, Scalable
Ti - 20(1) - 23(1) - 27(1) + 115(1) parallel monte carlo algorithm for atomistic simulations of precipitation in alloys,
T=1100 K Ni Co Fe Ti Phys. Rev. B 85 (2012) 184203, doi: 10.1103/PhysRevB.85.184203.
ANNEAL [25] L. Koch, F. Granberg, T. Brink, D. Utt, K. Albe, F. Djurabekova, K. Nordlund, Local
Ni - 14(1) - 2(1) + 45(1) - 11(1) segregation versus irradiation effects in high-entropy alloys: steady-state condi-
Co - 2(1) - 10(1) + 58(1) - 36(1) tions in a driven system, J. Appl. Phys. 122 (10) (2017) 105106.
Fe + 47(1) + 58(1) + 58(1) - 264(1) [26] O. Senkov, D. Miracle, Effect of the atomic size distribution on glass forming abil-
Ti - 12(2) - 37(1) - 262(1) + 354(1) ity of amorphous metallic alloys, Mater Res. Bull. 36 (12) (2001) 2183–2198.
T=700 K Ni Co Fe Ti [27] W. Hume-Rothery, R.W. Smallman, C.W. Haworth, The Structure of Metals and
ANNEAL Alloys, The Institute of Metals, 1969.
Ni - 33(1) - 17(1) + 94 (1) + 9(1) [28] D. Bacon, D. Barnett, R.O. Scattergood, Anisotropic continuum theory of lattice
defects, Prog. Mater Sci. 23 (1980) 51–262.
Co - 17(1) - 9(1) + 82 (1) - 28(1)
[29] J. Cowley, An approximate theory of order in alloys, Phys. Rev. 77 (5) (1950) 669.
Fe + 95(1) + 82(1) + 39 (1) - 395(2)
[30] S. Liu, Y. Wu, H. Wang, J. He, J. Liu, C. Chen, X. Liu, H. Wang, Z. Lu, Stacking fault
Ti + 10(1) - 30(1) - 393(1) + 426(1)
energy of face-centered-cubic high entropy alloys, Intermetallics 93 (2018) 269–
273.
[31] J. Ding, Q. Yu, M. Asta, R.O. Ritchie, Tunable stacking fault energies by tailoring
local chemical order in crconi medium-entropy alloys, Proc. Natl. Acad. Sci. 115
References
(36) (2018) 8919–8924, doi: 10.1073/pnas.1808660115.
[32] A. Tamm, A. Aabloo, M. Klintenberg, M. Stocks, A. Caro, Atomic-scale properties of
[1] D. Miracle, J. Miller, O. Senkov, C. Woodward, M. Uchic, J. Tiley, Exploration and ni-based fcc ternary, and quaternary alloys, Acta Mater. 99 (2015) 307–312.
development of high entropy alloys for exploration and development of high [33] R. DeHoff, Thermodynamics in Materials Science, CRC Press, 2006.
entropy alloys for structural applications, Entropy 16 (2014) 494–525. [34] J.L. J. Hirth, P. Anderson, Theory of Dislocations, Cambridge University Press, 2017.
[2] D. Miracle, O. Senkov, A critical review of high entropy alloys and related con- [35] M. Zaiser, Dislocation motion in a random solid solution, Philos. Mag. A 82 (15)
cepts, Acta Mater 122 (2017) 448–511, doi: 10.1016/j.actamat.2016.08.081. (2002) 2869–2883.
[3] C. Varvenne, A. Luque, W.A. Curtin, Theory of strengthening in fcc high entropy [36] G.P.M. Leyson, W.A. Curtin, L.G. Hector Jr, C.F. Woodward, Quantitative pre-
alloys, Acta Mater 118 (2016) 164–176. diction of solute strengthening in aluminium alloys, Nat. Mater 9 (9) (2010)
[4] B. Cantor, I. Chang, P. Knight, A. Vincent, Microstructural development in equia- 750–755.
tomic multicomponent alloys, Mater. Sci. Eng. A 375 (2004) 213–218. [37] G. Leyson, L. Hector, W. Curtin, Solute strengthening from first principles and
[5] F. Otto, A. Dlouhy0 , C. Somsen, H. Bei, G. Eggeler, E.P. George, The influences of application to aluminum alloys, Acta Mater. 60 (9) (2012) 3873–3884.
temperature and microstructure on the tensile properties of a cocrfemnni high- [38] L. Zhang, Y. Xiang, J. Han, D.J. Srolovitz, The effect of randomness on the strength
entropy alloy, Acta Mater 61 (15) (2013) 5743–5755. of high-entropy alloys, Acta Mater. 166 (2019) 424–434.
[6] F. Otto, Y. Yang, H. Bei, E.P. George, Relative effects of enthalpy and entropy on the [39] E. Rodary, D. Rodney, L. Proville, Y. Bre chet, G. Martin, Dislocation glide in model
phase stability of equiatomic high-entropy alloys, Acta Mater 61 (7) (2013) 2628–2638. ni (al) solid solutions by molecular dynamics, Phys. Rev. B 70 (5) (2004) 054111.
[7] J. Ding, Q. Yu, M. Asta, R.O. Ritchie, Tunable stacking fault energies by tailoring [40] Z.S. Basinski, The instability of plastic flow of metals at very low temperatures,
local chemical order in crconi medium-entropy alloys, Proc. Natl. Acad. Sci. 115 Proc. R. Soc. Lond. Ser. A. Math. Phys. Sci. 240 (1221) (1957) 229–242.
(36) (2018) 8919–8924, doi: 10.1073/pnas.1808660115. [41] L. Proville, D. Rodney, M.-C. Marinica, Quantum effect on thermally activated glide
[8] Q.-J. Li, H. Sheng, E. Ma, Strengthening in multi-principal element alloys with local- of dislocations, Nat Mater. 11 (10) (2012) 845.
chemical-order roughened dislocation pathways, Nat. Commun. 10 (2019) 3563. [42] G. Leyson, W. Curtin, Solute strengthening at high temperatures, Modell. Simul.
[9] D. de Fontaine, The number of independent pair-correlation functions in multi- Mater. Sci. Eng. 24 (6) (2016) 065005.
component systems, J. Appl. Crystallogr. 4 (1) (1971) 15–19, doi: 10.1107/ [43] S. Patinet, D. Bonamy, L. Proville, Atomic-scale avalanche along a dislocation in a
S0021889871006174. random alloy, Phys. Rev. B 84 (17) (2011) 174101.
[10] J. Cohen, M.E. Fine., Some aspects of short-range order, J. Phys. Radium 23 (10) [44] C. Varvenne, G.P.M. Leyson, M. Ghazisaeidi, W.A. Curtin, Solute strengthening in
(1962) 749–762. random alloys, Acta Mater. 124 (2017) 660–683.
[11] S.I. Rao, C. Woodward, T.A. Parthasarathy, O. Senkov, Atomistic simulations of dis- [45] R. Johnson, Elastic interaction between a point defect and an edge dislocation, J.
location behavior in a model fcc multicomponent concentrated solid solution Appl. Phys. 50 (3) (1979) 1263–1266.
alloy, Acta Mater 134 (2017) 188–194. [46] C.R. LaRosa, M. Shih, C. Varvenne, M. Ghazisaeidi, Solid solution strengthening
[12] E. Antillon, C. Woodward, S. Rao, B. Akdim, T. Parthasarathy, A molecular dynam- theories of high-entropy alloys, Mater. Charact. 151 (2019) 310–317.
ics technique for determining energy landscapes as a dislocation percolates [47] C. Varvenne, A. Luque, W.G. No € hring, W.A. Curtin, Average-atom interatomic
through a field of solutes, Acta Mater 166 (2019) 658–676, doi: 10.1016/j.acta- potential for random alloys, Phys. Rev. B 93 (2016) 104201.
mat.2018.12.037. [48] A. Van De Walle, M. Asta, First-principles investigation of perfect and diffuse
[13] R. Labusch, Cooperative effects in alloy hardening, Czech. J. Phys. B 38 (1988) 474–481. antiphase boundaries in hcp-based ti-al alloys, Metall. Mat Trans. A 33
[14] H. Suzuki, et al., Chemical interaction of solute atoms with dislocations, 4 (1952) (2002) 735.
455–463.

You might also like