Acta Materialia 166 (2019) 424e434

Contents lists available at ScienceDirect

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

Full length article

The effect of randomness on the strength of high-entropy alloys

Luchan Zhang a, b, Yang Xiang a, *, Jian Han b, David J. Srolovitz c, b, d, **
a
Department of Mathematics, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong
b
Department of Materials Science and Engineering, University of Pennsylvania, Philadelphia, PA 19004, USA
c
Department of Materials Science and Engineering, City University of Hong Kong, Kowloon, Hong Kong
d
Department of Mechanical Engineering and Applied Mechanics, University of Pennsylvania, Philadelphia, PA 19004, USA

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

Article history: High-entropy alloys (HEAs), i.e., single-phase, (nearly) equiatomic multicomponent, metallic materials,
Received 14 July 2018 are associated with novel mechanical properties, such as high strength, fracture resistance etc. In this
Received in revised form paper, a stochastic Peierls-Nabarro (PN) model is proposed to understand how random site occupancy
14 December 2018
affects intrinsic strength. The stochastic PN model accounts for the randomness in the composition,
Accepted 17 December 2018
Available online 27 December 2018
characterized by both the standard deviation of the perturbation in the interplanar potential and the
correlation length within the spatial compositional distribution. The model presented includes the ef-
fects of non-uniform compositional distribution both in the direction of dislocation glide and along a
Keywords:
High-entropy alloys
dislocation line to predict overall dislocation glide resistance. The model predicts the intrinsic strength of
Dislocations HEAs as a function of the standard deviation and the correlation length of the randomness. We find that,
Materials strength in most of the parameter space, the compositional randomness in an HEA gives rise to an intrinsic
Peierls-nabarro model strength that far exceeds that of any of the pure metals from which the HEA is composed. This approach
provides a fundamental explanation to the origin of the high strength of HEAs.
© 2018 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction on the classical ideas of solute solution strengthening; e.g., the

Labusch model [8]. The original Labusch model asserts that the
High-entropy alloys (HEAs) are often described as single-phase, critical shear stress for dislocation glide may be described as
(nearly) random solid solutions of four or more elements of (nearly) sc
f 4=3 c2=3 , where c is the concentration of solute atoms and f is
equal composition [1e3]. The crystal structures of such HEAs are the interaction force between a solute atom and a dislocation.
often very simple; e.g., face-centered or body-centered cubic While the original Labusch model is directly applicable for cases
crystals. HEAs exhibit useful and novel mechanical properties [4], where there is a distinction between solute and solvent atoms
fracture resistance [5], corrosion resistance [6], and radiation (unlike in HEAs), extensions to the HEA case have focused on how
damage resistance [7]. The mechanical property profile is charac- to combine contributions from each component to the strength by
terized by a combination of high strength, high ductility and high defining fn and cn for the nth element [9e11]. For example, Toda-
fracture toughness [2,5]; hence, HEAs have attracted considerable Caraballo et al. [10] adopted an averaging procedure in which
P 2=3
interest for potential applications as structural materials. HEA sc
ð n f 2n cn Þ . Varvenne et al. [12,13] explicitly considered the
design (composition and microstructure) will rely on establishing interaction energy between a solute atom and a dislocation in a
the link between structure and mechanical performance. In this matrix that was described as an effective medium composed of all
paper, we examine such a linkage at a fundamental level; in of the other types of atoms; such solute strengthening models were
particular, what are the implications of atomic-level randomness parameterized consistent with atomistic simulation results [14].
on dislocation glide in HEAs. Their model leads to a Labusch-type of expression for the strength:
P 2=3
Many of the existing models for the strength of HEAs are based sc
½ r;n f 2n ðrÞcn  , where r denotes the position of a solute atom
in the effective matrix relative to a dislocation.
In our view, there are two main issues with classical solid so-
lution approaches. First, the interaction between a solute atom and
* Corresponding author.
a dislocation based on elastic interactions associated with atomic
** Corresponding author. Department of Materials Science and Engineering, Uni-
versity of Pennsylvania, Philadelphia, PA 19004 USA.
size misfit and modulus misfit of the solute atom (with respect to
E-mail addresses: [email protected] (Y. Xiang), [email protected] the matrix) does not explicitly account for the interaction between
(D.J. Srolovitz). solute atoms and the dislocation core. Such effects are particularly

https://doi.org/10.1016/j.actamat.2018.12.032
1359-6454/© 2018 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
 L. Zhang et al. / Acta Materialia 166 (2019) 424e434 425

important, because in concentrated solutions the solute concen- parallel to the x-axis, and x-z is the slip plane. The (flat) slip plane
tration near the core may be very large and these interactions are separates two linear elastic, continuum half-spaces (y > 0 and
sensitive to the nature of the atomic bonding in the core. The y < 0). The disregistry across the slip plane (i.e., slip in the x-di-
importance of local bonding details in dislocation cores plays a rection) may be described by the function fðxÞ; fðx/  ∞Þ/0 and
major role in the mechanical response in a wide-range of materials fðx/ þ ∞Þ/b, where b is the magnitude of the Burgers vector. The
and crystal structures [15]. Because of the atomic-scale randomness Burgers vector distribution is rðxÞ ¼ f0 ðxÞ; this distribution char-
associated with HEAs, the local distortion around each atom is very acterizes the dislocation core in this model.
localized. Hence, we suggest that the local bonding effects at the The total energy of a dislocation in the PN model may be written
dislocation core may, in many cases, be at least as important as the as
elastic interactions.
Second, while HEAs are solid solutions, we expect that the dis- Etotal ¼ Eelastic þ Emisfit ; (1)
tribution of elements is not completely random; i.e., there is
chemical short-range order (SRO). While SRO is difficult to measure where Eelastic is the elastic energy in the upper and lower half
experimentally, the existence of SRO in HEAs was reported based spaces delimited by the slip plane and Emisfit is the misfit energy
on direct atomistic simulations [16e20]. SRO is known to increase associated with the nonlinear (atomic) interactions across the slip
alloy strength. Indeed, motion of dislocations through a solute so- plane. The elastic energy is a function of the disregistry
lution with SRO disturbs the local order and produces additional distribution:
resistance for dislocation glide, as suggested by Fisher [21] and
confirmed experimentally [22,23]. SRO may also affect the tem- ð þ∞
1
perature dependence of the yield stress. Unlike in pure face- Eelastic ¼ sxy ðxÞfðxÞdx; (2)
centered cubic (FCC) metals, FCC HEAs exhibit strongly tempera-
2 ∞

ture dependent yield stresses [4,24]. Otto et al. [24] attributed this
where the shear stress across the slip plane may be written in terms
to thermally activated dislocation motion associated with short-
m R þ∞ f ðx1 Þ
0

range obstacles (barriers). The short-range obstacles may corre- of the disregistry sxy ðxÞ ¼ 2pð1nÞ ∞ xx1 dx1 (m and n are the shear
spond to local regions of large SRO. Although SRO influences the modulus and Poisson ratio). The misfit energy may also be written
mechanical properties of HEAs, SRO is not included in the HEA in terms of the disregistry:
solute strengthening models.
Our paper has two main goals: (1) to understand the effects of ð þ∞
distributed solutes on dislocation cores and how they move, and (2) Emisfit ¼ gðfðxÞÞdx; (3)
to consider the effect of the random nature of this solute distri- ∞
bution (and its spatial correlations). We examine these effects
where gðfÞ is the periodic, interplanar potential (generalized
generically, so that they can be used in a wide variety of contexts
stacking-fault energy). In the classical PN model, gðfÞ is approxi-
and as additional effects to those already well-known models. We
mated by the Frenkel sinusoidal potential [27,28],
emphasize that this paper is not a complete description of all effects
that solutes have on dislocations nor is it focused on the specific  
distribution of solutes in any particular concentrated alloy. mb2 2pf
gðfÞ ¼ 1  cos ; (4)
More specifically, we present a model for the influence of the 4p2 d b
randomness in the spatial distribution of composition on the
strength of HEAs that naturally describes dislocation core effects in where d is the atomic interplanar spacing perpendicular to the slip
a generic manner; i.e., based upon a Peierls-Nabarro (PN) disloca- plane.
tion model [25,26]. While such a model is not an accurate Equilibrium is obtained by minimizing the total energy with
description of the true complexity of dislocation cores, it does respect to the disregistry distribution: dEtotal =df ¼ 0. The equilib-
provide a heuristic description of dislocation core effects that may rium disregistry function is
be generalized to describe the influence of SRO. In the PN model,
 
the disregistry across the glide plane is explicitly considered, and an b x b
fðxÞ ¼ tan1 þ ; (5)
interplanar potential is used to model the local effect of chemical p z0 2
bonding in the dislocation core. The distribution of chemical
components can also be included by making the interplanar po- where z0 is the dislocation core width:
tential spatially varying in a purely random manner or with SRO.
We first introduce the classical PN model and then extend it to form d
a random, stochastic PN model. Next, we consider the effect of SRO z0 ¼ : (6)
2ð1  nÞ
by making this random variation in composition spatially variable
with a well-defined correlation length. Finally, we extend our There is energetic barrier that must be overcome to move the
approach beyond the one-dimensional PN model to describe dislocation along the slip plane. The energy barrier and the critical
dislocation motion on a two-dimensional slip plane and add some stress for doing so are known as Peierls barrier and Peierls stress,
simplistic applications to HEAs. respectively. In the PN model, the Peierls barrier and stress for the
Frenkel interplanar potential are E0p ¼ mb2 =½pð1  nÞexpð2pz0 =bÞ
2. The classical Peierls-Nabarro model and

 
The Peierls-Nabarro (PN) model is a continuum model that m 2pz0
s0p ¼ exp  ; (7)
combines a long-range description of the strain field of a disloca- 1n b
tion and an atomic-level description of its core. In its classical form,
it describes a straight edge dislocation [25e27] with its core spread both are functions of the ratio of dislocation core width to Burgers
over a finite region along the slip plane. We assume that the vector z0 =b. In our numerical calculations below, we will assume
dislocation line is located along the z-axis, its Burgers vector is d ¼ b and n ¼ 0:347.
426 L. Zhang et al. / Acta Materialia 166 (2019) 424e434

3. Stochastic Peierls-Nabarro model: uniform randomness We consider that the random variable u is described by a
normal/Gaussian distribution with mean 1 and standard deviation
In a high-entropy alloy (HEA), the local environment varies with D, i.e., its probability function is
location in the lattice because the material is a solid solution of
many elements. However, unlike a random dilute binary alloy,
2
1 ðu1Þ
where most of the lattice sites are occupied by atoms of a single Pu ðu; 1; DÞ ¼ pffiffiffiffiffiffie 2D2 : (12)
type (only a minority of sites are occupied by atoms of another D 2p
type), in an HEA, the local site occupancies vary from site to site on
an atomic scale. Hence, a dilute binary alloy model is inappropriate In the mathematical literature, this normal distribution is often
for describing HEAs. Here, we extend the PN to describe the written as u
N ð1; DÞ; a notation we adopt below. The mean value
randomness in the elemental site occupancies of an HEA as a 1 implies that the averaged value of the interlayer potential at each
random variation in the amplitude of the interplanar potential. site on the slip plane of the HEA equals that of the lattice with a
This, in turn, leads to stochastic variations in the dislocation core single-type atom, as in Eq. (4). The probability distribution of the
width, Peierls stress, and Peierls energy and, as we demonstrate (normalized) dislocation core size ~
z (using Eq. (11) with u following
later, significantly alters the yield strength of the material. this normal distribution) for several values of D is shown in Fig. 1.
Specifically, we incorporate randomness into the PN model to The mean core size and its standard deviation are tabulated in the
describe an HEA, by scaling the interlayer potential (Eq. (4)) with a inset to that figure for the value of D. These results demonstrate
random variable u; i.e., that, when D ¼ 0, the probability density of ~ z is a delta function,
which corresponds to the classical PN model core width. As D in-
ð þ∞ ð þ∞ creases, the most probable core width (peak of the distribution)
Emisfit ¼ uðxÞgðfðxÞÞdxzu gðfðxÞÞdx: (8) gets smaller and the core width distribution broadens. However,
the mean value 〈~
∞ ∞
z〉 and standard deviation D~z of the core width
Since this randomness is allowed to vary from position to po- both increase with increasing D. This means that, as the standard
sition along the slip plane uðxÞ, the first expression can account for deviation of the randomness of the interplanar potential amplitude
random variation on any scale. In other words, uðxÞ may account for increases, the most probable dislocation core width decreases,
any short-range order that may occur in the HEA (i.e., spatial, whereas the mean value of the dislocation core width increases.
chemical short-range correlations). In the second expression, A small aside regarding the assumption that the randomness is
however, we replace the actual variation along the slip plane by characterized by a normal distribution (Eq. (12)) is in order. This
considering an ensemble of models, each with a spatially uniform distribution was chosen for its flexibility, that it is well-known, and
interlayer potential whose amplitude determined by the random x- because it is simpler to use than many other possible choices.
independent variable u (i.e., infinite spatial correlation). This is the However, it is an undesirable characteristic that this distribution
uniform randomness case which is the focus of this section. We has a finite probability for u < 0; this is unphysical. We have
return to consider the first (i.e., spatial short-range correlation) case examined the effect that this has on several dislocation properties
in the next section. For each case, we examine the dislocation core by comparing predictions using the normal distribution to that
width, the Peierls stress, and the distance a dislocation may travel obtained using the Gamma distribution with the same mean and
under a fixed, applied stress in the HEA. standard deviation (this distribution is finite only for u > 0). The
We assume that the random variable u is described by a prob- results are shown in Supplemental Material. Comparisons show
ability density function Pu ðuÞ (in this notation, u refers to a that the results using the normal distribution are good approxi-
particular value of the random variable u). The cumulative proba- mations to those using the Gamma distribution.
Ru
bility density function associated with Pu is Fu ðuÞ ¼ 0 Pu ðvÞdv. The randomness in the interlayer potential also makes the
Incorporating the randomness in the interlayer potential leaves Peierls stress a distributed quantity. Using the misfit energy with
the elastic energy unchanged except for a rescaling of the misfit the random interplanar potential in Eq. (8) and the disregistry
energy by a factor of u. Following the solution of the equilibrium function in Eq. (9) and following the standard procedure of lattice
disregistry function in Eq. (5), we assume that, with the random- summation of the misfit energy in the classical PN model, we
amplitude interlayer potential, the equilibrium disregistry func- determine the Peierls stress: sp ¼ um=ð1  nÞexpð  2pz=bÞ. Un-
tion that describes the dislocation core still takes the form fortunately, the probability density function of sp cannot be solved
  analytically from that of the random variable u. Instead, we look for
b x b
fðxÞ ¼ tan1 þ : (9)
p z 2
The equilibrium dislocation core width z is determined from
dEtotal =dz ¼  mb2 =½4pð1  nÞz þ mb2 u=2pd ¼ 0:

d z
z¼ ¼ 0; (10)
2ð1  nÞu u

where z0 is the core width in the classical PN model (Eq. (6)). The
core width, relative to that in the uniform system, is ~z ¼ z=z0 ¼ 1=u
(see Eq. (10)). The probability distribution for the core width P~z ðuÞ is
related to the probability distribution Pu of u as
 2  
1 1
P~z ðuÞ ¼ Pu : (11)
u u
Fig. 1. The (normalized) dislocation core width ~z distribution for several values of D for
Accordingly, the cumulative distribution for the core width F~z ðuÞ is normally distributed u, i.e., N ð1; DÞ. The inset to the figures shows the mean 〈~z〉 and
related to that for the random variable u as F~z ðuÞ ¼ 1  Fu ð1=uÞ. standard deviation D~z of the core width.
 L. Zhang et al. / Acta Materialia 166 (2019) 424e434 427

a reasonable approximation for the Peierls stress that yields

analytical forms of the probability density function. Noting that the
Peierls stress sp is sensitive to the dislocation core width z, we keep
the exponential term in sp and approximate u in its factor by its
mean value 1. As shown in the Supplemental Material, this is a good
approximation for the physically reasonable case where the stan-
dard deviation of u is small compared with its mean value 1. This
leads to the following approximate formula of the Peierls stress in
the HEA: sp ¼ m=ð1  nÞexpð  2pz=bÞ, or the reduced Peierls stress:
.   
s~p ¼ sp s0p ¼ exp b1 1  u1 ; (13)

where b ¼ b=2pz0 .
The probability distribution for the normalized Peierls stress s ~p
is Ps~p ðuÞ ¼ Pu ð1=ð1  b log uÞÞðb=uÞð1  b log uÞ2 . This probability Fig. 2. The probability distribution function of the Peierls stress s
~p for several values of
D, where u is normally distributed. The inset shows its mean hs~p i and standard de-
density function is shown in Fig. 2 for several values of D, along viations Ds~p for each D.
with its mean hs ~p i and standard deviation Ds~ . For D ¼ 0, the
p
probability distribution of s ~p is a delta function and we recover the
classical Peierls stress. With increasing D, the most probable Peierls any finite temperature, the enthalpy of mixing is non-zero. This
stress (peak in the probability density function) shifts towards implies that some degree of short-range order (SRO) will always be
smaller s ~p and the distribution widens. For sufficiently large D, present and any particular atom type will have either an enhanced
however, the probability distribution of s ~p becomes monotonically or reduced probability for neighboring atoms to have particular
decreasing (narrowing with increasing s ~p ). The mean value hs ~p i atom types. While this is a thermodynamic effect, kinetics associ-
and standard deviation Ds~p of the Peierls stress, on the other hand, ated with synthesis or processing may alter the degree of SRO that
increase with increasing D for all D. This means that as the standard occurs.
deviation of the randomness of the interplanar potential amplitude
increases, the most probable Peierls stress decreases, whereas the
mean value of the Peierls stress increases.
When a stress is applied to the HEA sa (s ~a ≡sa =s0p ), a dislocation
will glide until the local Peierls stress sp is too large to overcome at
the applied stress. The probability that a dislocation will be able to
move past any particular atomic site is Pðs ~p < s ~a Þ ¼ Fs~ ðs
p
~a Þ ¼ Fu ð1=
½1  b logðs ~a ÞÞ. Hence, the probability that the dislocation will be
able to travel a distance D ¼ Nb before it meets a barrier that it is
unable to overcome is PD ðN; s ~a Þ ¼ P N ðs ~p < s~a ÞPðs~p > s
~a Þ ¼
FN
u ð1=½1  b logð s
~ a ÞÞð1  Fu ð1=½1  b logð s
~ a ÞÞÞ. This implies that
the mean distance D a dislocation moves under an applied stress is
X X
D¼ NbPD ðN; s
~a Þ ¼ b NF N
u ðqÞ½1  Fu ðqÞ; (14)
N N

where q≡1=½1  b logðs ~a Þ. This is valid in the physically meaningful

regime where sa < m=ð1  nÞ (or s ~a < e1=b ).
Fig. 3(a) shows the mean distance a dislocation moves under an
applied stress for several values of D. When D ¼ 0, the mean dis-
tance a dislocation can travel is either zero (sa < s0p ) or infinite
(sa  s0p ). Importantly, when the width of the distribution of u is
finite (D > 0), the dislocation will only be able to travel a finite
distance for any applied stress. Fig. 3(b) shows the distribution of
distances that a dislocation will travel for several applied stresses.
This is not surprising, since a moving dislocation has a finite
probability of encountering a local Peierls stress that is above the
applied stress. The mean distance a dislocation can move increases
exponentially with applied stress. Fig. 3(a) can also be examined in
terms of what stress must be applied in order to move a dislocation
the same distance. Clearly, the stress required to move a dislocation
any non-trivial distance increases rapidly with increasing
randomness D. These results imply that the randomness that
characterizes HEAs will have a profound effect on the yield
behavior of the material (see below). Fig. 3. (a) The mean distance D a dislocation will move in an HEA as a function of the
applied stress s ~a for several values of D. (b) The probability distribution for this dis-
4. Stochastic Peierls-Nabarro model: short-range correlations tance PD ðN; s
~a Þ for several applied stresses. Note that when the applied stress is large
(s
~a ¼ 10), the dislocation traveling a long distance is (almost) equally probable as
traveling a short distance. As a result, the probability for the dislocation to stop at each
A high-entropy alloy is never completely random. The fact that specific place (at any distance) is small (this is a result of the normalization of the
atomic interactions depend on atom type (element) implies that, at probability).
428 L. Zhang et al. / Acta Materialia 166 (2019) 424e434

The existence of SRO in HEAs suggests that a moving dislocation width z. As in the previous section, we assume that the equilibrium
will not see a truly random distribution of neighboring environ- disregistry function fðxÞ that describes the dislocation core still
ments. In this section, we propose an approach to incorporate takes the form in Eq. (9). Recall that the equilibrium dislocation
spatial correlation of environments on dislocation motion. To this core width z is that which minimizes the total energy dEtotal =dz ¼
end, we return to the general form of the misfit energy in the 0. The misfit energy is (see Eqs. (8), (9) and (16))
Peierls-Nabarro model; i.e., the first expression in Eq. (8) for the
rffiffiffi ð þ∞ ð x  
interplanar potential. That is, we replace the classical, sinusoidal mb2 z mb2 2 xs 2z
2
Frenkel interplanar potential gðfðxÞÞ (Eq. (4)) with uðxÞgðfðxÞÞ, Emisfit ¼ þ D exp  dWs dx;
2pd 4p2 d l ∞ ∞ l x2 þ z
2
where uðxÞ is a stochastic coefficient function with short-range
spatial correlation. Formally, we describe uðxÞ as a stationary (18)
Gauss-Markov process [29,30]. In particular, this is done through
application of the Ornstein-Uhlenbeck process [29e31] which is a and the elastic energy is as per Eq. (2). For D≪1, the dislocation core
widely applied, nontrivial form of this process and which satisfies width ~
z ¼ z=z0 can be expressed as a function of l as
the stochastic differential equation rffiffiffi ð þ∞ ð ∞  
1 2z 2 xs x2
rffiffiffi ¼1þ 0 D exp    dxdWs :
1  uðxÞ 2 ~
z p l ∞ s l 2 2
duðxÞ ¼ dx þ DdWx ; (15) x2 þ z0
l l
(19)
where l is the spatial range of the correlation or the correlation
length of the random perturbation uðxÞ over the space, D is the 1=~
z follows a normal distribution with mean 1 and standard
standard deviation of uðxÞ at each lattice site (as before), and Wx is a deviation
continuous-time stochastic process that leads to Brownian motion
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 3
(i.e., the Wiener process [32]). The expectation of uðxÞ remains rffiffiffi u
uð þ∞ ð þ∞   2
〈uðxÞ〉 ¼ 1, such that, in the absence of the stochastic effect, the 2z0 2 u 6 x  s x 2
7
misfit energy in Eq. (8) reduces to the deterministic formula in Eq.
D¼ Du
t 4 exp    dx 5 ds;
p l ∞ s l 2 2
(3). x2 þ z0
The stochastic coefficient function uðxÞ can be determined from (20)
Eq. (15) as

rffiffiffi ð x i.e.,
 
2 xs
uðxÞ ¼ 1 þ D exp  dWs : (16) 1  
l ∞ l
N 1; D : (21)
~
z
The values of uðxÞ at two different lattice sites, x1 and x2 , are
correlated as More details are provided in the Supplemental Material.
Fig. 5 shows the dislocation core width ~ z distribution for
   
jx  x2 j different values of l and D. These results are also tabulated in
ðuðx1 Þ  1Þðuðx2 Þ  1Þ ¼ D2 exp  1 : (17)
l Table 1. As D/0 and/or l/0, we recover the classical Peierls-
Nabarro prediction for the dislocation core width. With the in-
This means that l is the correlation length of the random pertur- crease of D at fixed l, ~
z distribution broadens and the peak shifts to
bation over the space. In other words, spatial correlations in the smaller core size. More interestingly, increasing correlation length l
composition decay exponentially with separation between atomic at fixed D leads to broadening of the core width ~ z distribution. As
sites with a decay distance of l. Fig. 4 shows uðxÞ for several l/∞, we recover the uniform randomness results from the pre-
different values of l at fixed D. When the correlation length l/0, h vious section.
ðuðx1 Þ  1Þðuðx2 Þ  1Þi/0, which indicates that the randomness at We now examine the effect of spatial correlations in the
x1 and x2 are independent of each other. When l/∞, the variance composition (uðxÞ) in the HEA on the Peierls stress. As discussed in
h½uðx1 Þ  uðx2 Þ2 i/0, and therefore uðx1 Þ ¼ uðx2 Þ; this is the case Section 3, it is reasonable to approximate the Peierls stress distri-
of uniform randomness considered in Section 3. bution based on the assumption that it is controlled by the
We now examine the effect of spatially correlated randomness randomness in the dislocation core width following Eq. (7). This
on dislocation properties, beginning with the dislocation core yields

Fig. 4. The random coefficient function uðxÞ defined by the Ornstein-Uhlenbeck process for several different values of the spatial correlation length l at D ¼ 0:1. Although the mean
value of uðxÞ appears to be greater than one in (d), over a very large distance (x[l) the average will be unity.
 L. Zhang et al. / Acta Materialia 166 (2019) 424e434 429

Fig. 5. The dislocation core width ~z distribution with spatially correlated random co- Fig. 6. The Peierls stress s
~p distributions for the case of spatially correlated random
efficient function uðxÞ (a) for different values of D at l=z0 ¼ 1 and (b) for different coefficient function uðxÞ are shown in (a) for several values of D at l=z0 ¼ 1:0 and in (b)
values of l=z0 at D ¼ 0:1. for several values of l=z0 at D ¼ 0:1.

h  i
distribution. Increasing the spatial correlation length l at fixed D
s~p ¼ exp b1 1  ~z ; (22)
leads to increasing mean Peierls stress, small shifts in the most
probably Peierls barrier to lower values and a broadening of the
where dislocation core width ~ z is as per Eq. (19), and 1=~z is Peierls stress distribution. Consistent with the dislocation core
described by a normal distribution given by Eqs. (20) and (21). width ~z results, as D/0 and/or l/0, the Peierls stress goes to its
Fig. 6 shows the probability density function of the Peierls stress classical Peierls-Nabarro value and as l/∞ the Peierls stress dis-
s~p for several values of D and l=z0 . The associated mean hs~p i and tribution goes to the uniform randomness limit, as discussed above.
standard deviation Ds~p of the Peierls stress s ~p distribution are
We now consider how a dislocation moves in an HEA with
summarized in Table 2. Increasing D at fixed l leads to increasing spatially correlated atomic site occupancies under an applied stress
mean Peierls stress, a shift in the most probably Peierls barrier to s~a ¼ sa =s0p , following the same approach as in the previous section.
lower values and an increase in the width of the Peierls stress The probability that a dislocation can move a distance D ¼ Nb

Table 1
Table 2
The mean value 〈~
z〉 and standard deviation D~z of the core width ~z for several values The mean value hs~p i and standard deviation Ds~ of the Peierls stress s
~p for several
of D and l=z0 . p
values of D and l=z0 .
l=z0 D 0 0.10 0.14 0.32 0.45
l=z0 D 0 0.10 0.14 0.32 0.45
0.2 〈~
z〉 1.0 1.0 1.0 1.01 1.01
0.2 hs
~p i 1.0 1.0 1.01 1.03 1.07
D~z 0 0.02 0.03 0.07 0.10 Ds~p 0 0.11 0.15 0.35 0.50
1 〈~
z〉 1.0 1.0 1.0 1.02 1.03 1.0 hs
~p i 1.0 1.01 1.02 1.10 1.19
D~z 0 0.04 0.05 0.13 0.20 Ds~p 0 0.19 0.26 0.61 0.88
5 〈~
z〉 1.0 1.0 1.01 1.05 1.12 5 hs
~p i 1.0 1.03 1.06 1.28 1.52
D~z 0 0.07 0.10 0.27 0.53 Ds~p 0 0.32 0.45 1.10 1.59
10 〈~
z〉 1.0 1.01 1.01 1.07 1.17 10 hs
~p i 1.0 1.04 1.08 1.37 1.64
D~z 0 0.08 0.11 0.36 0.69 Ds~p 0 0.37 0.53 1.29 1.81
100 〈~
z〉 1.0 1.01 1.02 1.11 1.21 100 hs
~p i 1.0 1.06 1.11 1.53 1.80
D~z 0 0.09 0.13 0.47 0.84 Ds~p 0 0.46 0.66 1.60 2.09
∞ (uniform) hs
~p i 1.0 1.06 1.13 1.58 1.83
∞ (uniform) 〈~
z〉 1.0 1.01 1.02 1.14 1.24
Ds~p 0 0.49 0.70 1.70 2.15
D~z 0 0.10 0.15 0.60 0.97
430 L. Zhang et al. / Acta Materialia 166 (2019) 424e434

under an applied stress s ~a is PD ðN; s ~a Þ ¼ F N

N ð1;DÞ
ð1=ð1  by the normal distribution:
b log s~a ÞÞ½1  FN ð1;DÞ ð1=ð1  b log s~a ÞÞ. From this, we calculate the  
mean distance a dislocation moves D under applied stress s ~a : u
N 1; D ; (25)
X    
1 1
D¼b NF N
1  FN 1;D
: where the standard deviation D is given by Eq. (20), which
N 1;D 1  b logðs
~a Þ 1  b logðs
~a Þ
N explicitly depends on both l and D.
(23) In fact, by using this homogenized model in Eq. (24), the
dislocation core width ~ z is given by Eq. (10), i.e., 1=~z ¼ u. We can
Note that in the correlated randomness case, 1=~ z is described by the see that it is the same as the full stochastic model result in Eq. (21) if
normal distribution N ð1; DÞ, instead of N ð1; DÞ as in the uniform Eq. (25) holds. That is, the random dislocation core widths in the
randomness case (previous section). two models following the same distribution. As a result, the Peierls
Fig. 7 shows the mean distance a dislocation moves D under stresses and the distance a dislocation moves calculated from the
applied stress s~a in the spatially correlated random composition dislocation core widths are also the same using the two models. In
(uðxÞ) case for several values of D and l. Dðs ~a Þ exhibits a similar this sense, we conclude that the model in Eqs. (24) and (25)
variation with D for the spatially correlated and uniform random- represent a homogenized version of the fully stochastic model
ness case, Fig. 3(a); i.e., for any finite D, the dislocation travels a presented in the previous section.
finite distance at any applied stress, and as the applied stress s ~a The relation between the standard deviation D of the random
increases, D increases exponentially (when D ¼ 0 the dislocation variable u and the spatial correlation length l of uðxÞ is shown in
will either not move or move infinitely far). The stress that must be Fig. 8. We see that the standard deviation D of the random variable
applied to move a dislocation the same distance increases with u scales in a nonlinear but monotonic manner with the spatial
increasing D. More interestingly for the present discussion, correlation length l. When l/0, the standard deviation D/0,
increasing the spatial correlation length l implies either that the which corresponds to the classical Peierls-Nabarro model; when
dislocation travels a shorter distance for a fixed applied stress or l/∞, the standard deviation D/D, which corresponds to the
that greater applied stresses are required to move the dislocation a uniform randomness case in Section 3. This is consistent with the
fixed distance. In the limit that the correlation length l=z0 /∞, the discussion at the beginning of Section 4. The standard deviation D
results converge to the uniform random case, Fig. 3(a), as expected. of the random variable u is linearly dependent on the parameter D,
In general, increasing correlation length l interpolates the which is the standard deviation of uðxÞ at each location x.
dislocation properties between that of the non-random case
(classical Peierls-Nabarro prediction) to the uniform randomness
case (of the previous section). 6. Peierls stress in two dimensions

5. Homogenized stochastic model The model presented is one-dimensional and should be viewed
in the same manner as the classical Peierls-Nabarro model. One
While the stochastic model with correlated randomness provide limitation of focusing on one dimension is that dislocation motion
a useful approach to understanding the effects of randomness on can be completely stopped by a Peierls barrier at one atomic site. Of
dislocation motion in HEAs, it would be convenient to recast the course, in a two-dimensional slip plane, a dislocation can bow
model in such a way that u is no longer a function of position. That around an obstacle that it cannot cross. Here, we make use of the
is, instead of making the approximation that led to the second one-dimensional model to provide insight into the strength of local
expression in Eq. (8), we would like to have a homogenized model barriers and use it as input to a simple two-dimensional model.
that accounts for spatial correlation in the randomness without Consider a dislocation line in two dimensions. Along the dislo-
assuming that the randomness is uniform. To this end, we rewrite cation line, the Peierls stress at each lattice site may be approxi-
the misfit energy as mated by Eq. (22) following the one dimensional stochastic model
in Section 4. This stochastic model may be represented by a random
ð þ∞ variable u as discussed in Section 5, and the associated Peierls
Emisfit ¼ u gðfðxÞÞdx: (24) stress may be approximated by s ~p ¼ exp½b1 ð1  u1 Þ, where u

∞
N ð1; DÞ and the standard deviation D is as per Eq. (20).
We can do this by choosing a u as a random variable described

Fig. 7. Under applied stress s

~ a , the mean value D of the dislocation traveled distance in
a HEA with spatially correlated random coefficient function uðxÞ with several different Fig. 8. The standard deviation D of the random variable u as a function of the spatial
values of D and l. correlation length l of uðxÞ.
 L. Zhang et al. / Acta Materialia 166 (2019) 424e434 431

We now consider the ability of a dislocation to bow around

obstacles, in the Orowan-looping sense [33]. The critical (Orowan)
stress for a dislocation to bypass a set of “pinning sites” is

sc ¼ mb=L; (26)

where L is the average spacing between the pinning sites. Along the
dislocation line, the ratio of the number of pinning lattice sites to
that of the total sites is approximately Pðs
~p > s
~c Þ, that is,

P s~p > s
~ c ¼ a0 =L; (27)

where s ~c ≡sc =s0p and a0 is the lattice constant (bza0 ). The critical
stress, therefore, must satisfy the condition (using Eqs. (26) and
(27)):
Fig. 9. The critical stress sec determined using the analytical approximation in Eq. (31)
s0p

s~ ¼ P s~p > s~c : (28) (solid lines) and by numerical solution of the full equation in Eq. (29) (dashed lines), as
m c functions of D for two different correlation lengths, l=z0 .

This probability can be written as Pðs ~p > s ~cÞ ¼ 1  F  ð1 

N ð1;DÞ solute strengthening models consider some atoms as solvents and
b log s~c Þ. Further using 1  FN ð1;DÞ ðuÞ ¼ 12 erfc p1ffiffiffi ðu1  1Þ , the others as solute, such distinctions are unclear in HEAs where there
2D
equation for the critical stress Eq. (28) becomes is no nominal solvent since most HEAs consist of many elements in
   near equiatomic concentrations. Varvenne et al. [13] proposed us-
2s0p 1 1 ing an “average alloy” reference with properties (e.g., elastic con-
s~ ¼ erfc pffiffiffi 1 : (29)
m c 2D 1  b log s
~c stants) chosen/fit to match the HEA. In a concentrated alloy, local
interactions between solute atoms screens the long-range elastic
Now, we simplify Eq. (29) for the case where the critical stress sc
interactions between any particular solute and the dislocation core.
is close to the Peierls stress s0p (i.e., s
~c
1). When s~c ¼ 1, the left-
Hence, short range solute/dislocation core interactions are more
hand-side of Eq. (29) is 2s0p =m. Let z0 be the value that satisfies 2s0p =
important in HEAs than in dilute solid solutions. At small length
m ¼ erfcðz0 Þ. Approximating the error function as
d erfcðz Þðz  z Þ scales, it is not possible to separate these solute/dislocation elastic
erfcðzÞzerfcðz0 Þ þ dz 0 0 in Eq. (29) for
1 interactions from dislocation core effects. A complete description
z0 ≡erfc ð2sp =mÞ, we find that
0
should, of course, fully account for the details of the dislocation

  cores and the subtlety of bonding in the highly distorted core re-
2s0p 2s0p 2exp z20 1
s~c z  pffiffiffi pffiffiffi bðs
~c  1Þ  z0 (30) gion. Models based on standard solute strengthening approaches
m m p 2D cannot account for such core effects. Our approach focuses on
dislocation core effects within the idealized (and relatively simple)
or
framework of the Peierls-Nabarro model which includes explicit
2z0 expðz20 Þ pffiffiffi core effects (including non-linear elastic deformation in the core).
pffiffiffi
p 2 2pz0 z0 We estimate the core strengthening effect in an HEA based on
s~c ¼ 1 þ pffiffiffi z1 þ D: (31)
2s0p 2b expðz20 Þ b the PN model, by applying Eq. (31). As a specific case, we focus on
m þ pffiffiffi
pD the body-centered cubic (BCC) HEA, Mo0.2Nb0.2Ta0.2V0.2W0.2. We
assume that z0 zb and set several of the fundamental alloy prop-
This approximation is valid for small D. Recall that D is the standard erties to averages over those in the component pure metals, i.e., the
deviation of the effective random variable u, Eq. (20), and depends average Peierls stress is s0p ¼ 0:586 GPa and the average shear
on the correlation length l and the standard deviation D at each modulus is m ¼ 90:24 GPa [35]. Fig. 10 shows the critical stress sc as
location of the random coefficient function uðxÞ.
The critical stress s ~c determined by using the analytical
approximation in Eq. (31) is shown in Fig. 9 along with those ob-
tained by numerically solving the full equation in Eq. (29). Clearly,
the approximation (Eq. (31)) provides a good approximation to the
critical stress s
~c for small D. These results demonstrate that the
dimensionless critical stress s ~c increases linearly with the standard
deviation of the randomness D at each lattice, and increases
monotonically (but nonlinearly) with the correlation length l (see
Fig. 8).
The critical stress sc is the yield strength increment associated
with randomness in the spatial distribution of the elements in the
HEA. This strengthening arises from the interaction between the
dislocation core and the locally varying atomic-level environment.

7. Discussion
Fig. 10. Critical stress sc (Eq. (31)) as a function of the standard deviation D and the
correlation length l, which is estimated based on the parameters for BCC Mo0.2N-
Our approach to describing strengthening in high-entropy alloys b0.2Ta0.2W0.2V0.2 HEA. The contour lines for several sc values are denoted by white
is fundamentally different from strengthening models based upon dashed lines. The contour line for sc ¼ 0:9 GPa corresponds to the Peierls stress of pure
Fleischer [34] or Labusch [8] types of solute strengthening. While W.
432 L. Zhang et al. / Acta Materialia 166 (2019) 424e434

a function of the standard deviation of the amplitude of the effect of randomness in HEAs; it clearly does not account for
generalized stacking-fault energy in the stochastic PN model D (see changes in alloy chemistry that change the energy difference be-
Eq. (31)) and its correlation length l for this HEA system. In general, tween FCC and HCP (which controls stacking fault energy and
sc increases with increasing D and l. Fig. 10 shows that for most of partial spacing); in fact we note that in some FCC HEAs, the HCP
the D-l parameter space, the critical stress for the HEA is larger phase may be associated with even lower energy than the FCC
than the Peierls stress of any of the pure metals from which it is phase which suggests a negative stacking fault energy (if the sys-
formed. This implies that the strength of HEAs will, in general, be tem were equilibrated). Clearly, the effect of changes in the average
higher than those of any of its constituents. Returning to the spe- magnitude of the stacking fault energy is outside the realm of this
cific example of Mo0.2Nb0.2Ta0.2V0.2W0.2, we assume that the cor- model.
relation length for the site occupancy is twice the dislocation core The stochastic PN model discussed here is based upon several
size (l ¼ 2z0 ) and we set D to the average Peierls barriers of the five assumptions; we outline the major assumptions here. Many of
constituent elemental solids (D ¼ 1:5), we obtain sc ¼ 7:27 GPa. If these assumption could be relaxed via the same approaches
we set D ¼ 1, we obtain sc ¼ 5:04 GPa. These values should be employed for the generalization of the elemental PN model. First,
compared with the average Peierls stress of the five constituent our stochastic PN approach is built upon the classical PN model
elemental solids, 0.59 GPa, or that of the elemental solid with the where a nonlinear potential energy across the slip plane is
largest Peierls stress (i.e., W, 0.9 GPa). This comparison shows that described in terms of a simple cosine function. Of course, more
randomness easily provides very large strength increments accurate nonlinear potential energy surfaces (the generalized
compared with the HEA constituents. Of course, this is a crude stacking fault energy or g-surface [43]) may be incorporated (e.g.,
estimate, but makes a strong, clear case. directly from density functional theory calculations [44e49]).
The finite-temperature, finite-strain-rate yield stress can be Available techniques for PN models with general g-surfaces may be
estimated from the energy barrier for dislocation glide DEb applied in the stochastic PN framework; these include, e.g., repre-
following common practice [36] as, for example, applied by Varv- senting the g-surfaces by a finite-term Fourier series that reflects
enne [13]: the lattice symmetry [47e53], approximate analytical methods
[48e50], and solutions based on discrete lattice point models
"  23 # [45,52,54].
kB T ε_ 0 Second, unlike our one-dimensional description (where dislo-
sðTÞ ¼ sc 1 ln ; (32)
DEb ε_ cations are edges), dislocations in real materials are not straight and
therefore not uniformly edge or screw. Many such curved disloca-
where sc is the critical stress obtained by Eq. (31) and ε_ 0 is a tion PN models have been proposed, e.g., Refs. [51e58]. The dif-
reference strain rate. For most metals, sc is usually assumed to be a ference between edges and screws are important for BCC metals
constant (temperature independent), such that the entire temper- (including HEAs), where there are a predominance of screws.
ature dependence is associated with the term in the square Generalizations of the classical PN model to screw dislocations have
brackets. On the other hand, in HEAs sc depends on the composi- been proposed (with partial dissociation in multiple slip planes in
tional correlation length l. The correlation length depends on the BCC lattices) [59,60]; our stochastic PN model can similarly be
short range order in the material; i.e., l decreases with increasing T generalized to include such screw dislocations (with nonplanar
in HEAs [37]. This implies that sc ðlðTÞÞ is nearly independent of T at cores [61]). We have extended our one-dimensional results to two
low temperature and very sensitive to T at high temperature; see dimensional slip planes by application of an Orowan pinning
the dependence of D (sc is proportional to D) with respect to l in model. This, of course, is oversimplified, but may be extended more
Fig. 8. Hence, in HEAs, the temperature dependence of the yield effectively in two dimensions by considering a two-dimensional
stress is controlled by the term in square brackets in Eq. (32) at low inter-planar potential.
T, but by sc ðTÞ at high temperature. This implies that the temper-
ature dependence of the yield stress is “normal” at low and inter- 8. Conclusions
mediate T, but will fall rapidly with increasing T at high T. The is
consistent with experimental measurements of the temperature In this paper, we developed a model to predict the intrinsic yield
dependence of the yield strength of BCC HEAs [38] (in these ex- strength of high-entropy alloys (HEAs) within the framework of the
periments, the sample were annealed at the test temperature prior Peierls-Nabarro (PN) model. We assume that the intrinsic yield
to loading). strength of HEAs is associated with the stress required for dislo-
In this paper, we incorporated the effects of randomness in HEAs cation glide; i.e., the Peierls (friction) stress. In the classical PN
based on the classical PN model [25,26] for undissociated perfect model, the Peierls stress results from the periodic interplanar po-
dislocations. For HEAs in which the dislocations do not dissociate tential for dislocation glide in a discrete lattice. In the case of HEAs,
into partials, e.g. BCC HEAs, this stochastic PN model applies we relate the spatial randomness of the composition distribution as
directly. For FCC HEAs [39,40] in which the two partials are well stochastic perturbations to the interplanar potential. Accounting
separated the stochastic PN model can be applied using the for these stochastic perturbations in the PN framework leads to the
approach employed previously in the classical PN model for pure stochastic PN model developed here.
FCC metals [41,42]. While the effects of randomness on the Peierls We developed two versions of the stochastic PN model. The
stress in FCC materials follows the same general description as that simple version of the stochastic PN model accounts for the
for BCC metals discussed here, it is important to keep in mind that randomness in the composition in terms of a distribution of misfit
the magnitude of the Peierls stress on the strength of FCC metals is energies, from which we estimate the probability distributions of
much smaller than in BCC metals (the Peierls stress in most pure the dislocation core size, Peierls stress and the distance a disloca-
FCC metals is very small). In FCC metals, the strength is more tion can travel under a particular applied stress. This simple model
commonly dominated by dislocation-dislocation interactions and provides an estimate of the variation of the Peierls stress with
extrinsic strengthening mechanisms; hence, while randomness respect to the amplitude of the HEA randomness.
plays the same role in FCC and BCC alloys its contribution to the Of course, multicomponent alloys must have finite short-range
overall strength will be significantly smaller in FCC than BCC alloys. order and hence are not completely random. Our more complete
We recall that the main point of this paper is to examine the version of the stochastic PN model explicitly considers spatial
 L. Zhang et al. / Acta Materialia 166 (2019) 424e434 433

correlations within the spatial composition distribution (described [11] Z. Wu, Y. Gao, H. Bei, Thermal activation mechanisms and Labusch-type
strengthening analysis for a family of high-entropy and equiatomic solid-
here by a Ornstein-Uhlenbeck process); this introduces a spatial
solution alloys, Acta Mater. 120 (2016) 108e119.
correlation parameter. This model gives a Peierls stress that is a [12] C. Varvenne, A. Luque, W.A. Curtin, Theory of strengthening in fcc high en-
function of both the amplitude of the HEA randomness and the tropy alloys, Acta Mater. 118 (2016) 164e176.
correlations within it. From the viewpoint, the simple model should [13] C. Varvenne, G.P.M. Leyson, M. Ghazisaeidi, W.A. Curtin, Solute strengthening
in random alloys, Acta Mater. 124 (2017) 660e683.
be viewed as one with an infinite correlation length. [14] G.P.M. Leyson, W.A. Curtin, Solute strengthening at high temperatures, Model.
The PN model is one dimensional (1D) and can only be directly Simulat. Mater. Sci. Eng. 24 (2016), 065005.
applied to the case where every point along the dislocation line is in [15] M.S. Duesbery, V. Vitek, Plastic anisotropy in bcc transition metals, Acta Mater.
46 (1998) 1481e1492.
exactly the same (compositional) environment. Of course, a dislo- [16] M.F. Del Grosso, G. Bozzolo, H.O. Mosca, Modeling of high entropy alloys of
cation in an HEA will experience random variations in the refractory elements, Physica B 407 (2012) 3285e3287.
composition both as it moves as well as along its length. If there is [17] P. Singh, A.V. Smirnov, D.D. Johnson, Atomic short-range order and incipient
long-range order in high-entropy alloys, Phys. Rev. B 91 (2015) 224204.
indeed random composition variations along the dislocation line, [18] A. Tamm, A. Aabloo, M. Klintenberg, M. Stocks, A. Caro, Atomic-scale prop-
the dislocation will not be straight. We extend the 1D stochastic PN erties of Ni-based FCC ternary, and quaternary alloys, Acta Mater. 99 (2015)
model to dislocation glide along its 2D slip plane based on an 307e312.
[19] A. Sharma, P. Singh, D.D. Johnson, P.K. Liaw, G. Balasubramanian, Atomistic
Orowan approach, where the average spacing between the pinning clustering-ordering and high-strain deformation of an Al0.1CrCoFeNi high-
sites is the probability that an applied stress is larger than the local entropy alloy, Sci. Rep. 6 (2016) 31028.
Peierls stress. [20] F. Ko € rmann, A.V. Ruban, M.H.F. Sluiter, Long-ranged interactions in bcc
NbMoTaW high-entropy alloys, Mater. Res. Lett. 5 (2017) 35e40.
Based upon the stochastic PN model, we see that randomness in
[21] J.C. Fisher, On the strength of solid solution alloys, Acta Metall. 2 (1954) 9e10.
the spatial distribution of the composition in HEAs gives rise to a ment, A. Coujou, Dynamic friction stresses in the g
[22] M. Jouiad, F. Pettinari, N. Cle
distribution of local Peierls stresses. This distribution of local Peierls phase of a nickel-based superalloy, Phil. Mag. A 79 (1999) 2591e2602.
[23] F. Pettinari-Sturmel, M. Jouiad, H.O.K. Kirchner, N. Cle ment, A. Coujou, Local
stresses implies that dislocations will move by bowing between
disordering and reordering phenomena induced by mobile dislocations in
pinning points (where the local Peierls stress exceeds the applied short-range-ordered solid solutions, Phil. Mag. A 82 (2002) 3045e3054.
stress). The net result is that the underlying randomness of an HEA [24] F. Otto, A. Dlouhỳ, C. Somsen, H. Bei, G. Eggeler, E.P. George, The influences of
naturally gives rise to an intrinsic yield strength that exceeds that of temperature and microstructure on the tensile properties of a CoCrFeMnNi
high-entropy alloy, Acta Mater. 61 (2013) 5743e5755.
its constituents. Examples were provided that show that the [25] R. E. Peierls, “The size of a dislocation,” Proc. Phys. Soc. 52, 34.
effective Peierls stress for an HEA exceeds that of any of the con- [26] F. R. N. Nabarro, “Dislocations in a simple cubic lattice,” Proc. Phys. Soc. 59,
stituent single-component metals. The extent of the short range 256.
[27] J.P. Hirth, J. Lothe, Theory of Dislocations, second ed., John Wiley, New York,
order within the HEA has a strong influence on both its intrinsic 1982.
yield strength and its temperature dependence. These predictions [28] J. Frenkel, Zur theorie der elastizita€tsgrenze und der festigkeit kristallinischer
are consistent with a range of experimental data in BCC HEAs. ko€rper, Z. Phys. 37 (1926) 572e609.
[29] G.E. Uhlenbeck, L.S. Ornstein, On the theory of the brownian motion, Phys.
Rev. 36 (1930) 823e841.
Acknowledgements [30] M.C. Wang, G.E. Uhlenbeck, On the theory of the brownian motion ii, Rev.
Mod. Phys. 17 (1945) 323e342.
[31] J.L. Doob, The brownian movement and stochastic equations, Ann. Math. 43
The authors gratefully acknowledge useful discussions with (1942) 351e369.
Prof. Abba Krieger of the Department of Statistics, The University of [32] I. Karatzas, S.E. Shreve, Brownian Motion and Stochastic Calculus, Springer-
Pennsylvania. YX acknowledges support of the Hong Kong Research Verlag, 1988.
[33] E. Orowan, Symposium on internal stress in metals and alloys, Metals and
Grants Council General Research Fund 16302818.
Alloys (1948) 451.
[34] R.L. Fleischer, Substitutional solution hardening, Acta Metall. 11 (1963)
Appendix A. Supplementary data 203e209.
[35] L. Dezerald, L. Ventelon, E. Clouet, C. Denoual, D. Rodney, F. Willaime, Ab initio
modeling of the two-dimensional energy landscape of screw dislocations in
Supplementary data to this article can be found online at bcc transition metals, Phys. Rev. B 89 (2014), 024104.
https://doi.org/10.1016/j.actamat.2018.12.032. [36] A. Argon, Strengthening Mechanisms in Crystal Plasticity, Oxford University
Press, 2008.
[37] A. Ferna ndez-Caballero, J.S. Wro bel, P.M. Mummery, D. Nguyen-Manh, Short-
References range order in high entropy alloys: theoretical formulation and application to
Mo-Nb-Ta-V-W system, J. Phase Equilibria Diffusion 38 (2017) 391e403.
[1] J.-W. Yeh, S.-K. Chen, S.-J. Lin, J.-Y. Gan, T.-S. Chin, T.-T. Shun, C.-H. Tsau, S.- [38] O.N. Senkov, G.B. Wilks, J.M. Scott, D.B. Miracle, Mechanical properties of
Y. Chang, Nanostructured high-entropy alloys with multiple principal ele- Nb25Mo25Ta25W25 and V20Nb20Mo20Ta20W20 refractory high entropy alloys,
ments: novel alloy design concepts and outcomes, Adv. Eng. Mater. 6 (2004) Intermetallics 19 (2011) 698e706.
299e303. [39] A.J. Zaddach, C. Niu, C.C. Koch, D.L. Irving, Mechanical properties and stacking
[2] Y. Zhang, T.T. Zuo, Z. Tang, M. Gao C, K. Dahmen A, P.K. Liaw, Z.P. Lu, Micro- fault energies of NiFeCrCoMn high-entropy alloy, J. Occup. Med. 65 (2013)
structures and properties of high-entropy alloys, Prog. Mater. Sci. 61 (2014) 1780e1789.
1e93. [40] S. Huang, W. Li, S. Lu, F. Tian, J. Shen, E. Holmstro € m, Temperature dependent
[3] D.B. Miracle, O.N. Senkov, A critical review of high entropy alloys and related stacking fault energy of FeCrCoNiMn high entropy alloy, Scripta Mater. 108
concepts, Acta Mater. 122 (2017) 448e511. (2015) 44e47.
[4] A. Gali, E.P. George, Tensile properties of high-and medium-entropy alloys, [41] F.R.N. Nabarro, Theoretical and experimental estimates of the peierls stress,
Intermetallics 39 (2013) 74e78. Philos. Mag. A 75 (1997) 703e711.
[5] B. Gludovatz, A. Hohenwarter, D. Catoor, E.H. Chang, E.P. George, R.O. Ritchie, [42] G. Schoeck, The peierls stress and the flow stress in fcc metals, Scripta Metall.
A fracture-resistant high-entropy alloy for cryogenic applications, Science 345 Mater. 30 (1994) 611e613.
(2014) 1153e1158. [43] V. Vitek, Intrinsic stacking faults in body-centred cubic crystals, Philos. Mag.
[6] Y. Shi, B. Yang, P.K. Liaw, Corrosion-resistant high-entropy alloys: a review, 18 (1968) 773e786.
Metals 7 (2017) 43. [44] E. Kaxiras, M.S. Duesbery, Free energies of generalized stacking faults in Si and
[7] N.A.P.K. Kumar, C. Li, K.J. Leonard, H. Bei, S.J. Zinkle, Microstructural stability implications for the brittle-ductile transition, Phys. Rev. Lett. 70 (1993)
and mechanical behavior of FeNiMnCr high entropy alloy under ion irradia- 3752e3755.
tion, Acta Mater. 113 (2016) 230e244. [45] V.V. Bulatov, E. Kaxiras, Semidiscrete variational Peierls framework for
[8] R. Labusch, A statistical theory of solid solution hardening, Phys. Status Solidi dislocation core properties, Phys. Rev. Lett. 78 (1997) 4221e4223.
B 41 (1970) 659e669. [46] G. Lu, N. Kioussis, V.V. Bulatov, E. Kaxiras, Generalized stacking fault energy
[9] O.N. Senkov, J.M. Scott, S.V. Senkova, D.B. Miracle, C.F. Woodward, Micro- surface and dislocation properties of aluminum, Phys. Rev. B 62 (2000)
structure and room temperature properties of a high-entropy TaNbHfZrTi 3099e3108.
alloy, J. Alloy. Comp. 509 (2011) 6043e6048. [47] S. Zhou, J. Han, S. Dai, J. Sun, D.J. Srolovitz, van der Waals bilayer energetics:
[10] I. Toda-Caraballo, P.E.J. Rivera-Díaz del Castillo, Modelling solid solution generalized stacking-fault energy of graphene, boron nitride, and graphene/
hardening in high entropy alloys, Acta Mater. 85 (2015) 14e23. boron nitride bilayers, Phys. Rev. B 92 (2015) 155438.
434 L. Zhang et al. / Acta Materialia 166 (2019) 424e434

[48] Z. Pei, M. Eisenbach, Acceleration of the Particle Swarm Optimization for [56] M. Koslowski, A.M. Cuitino, M. Ortiz, A phase-field theory of dislocation dy-
Peierls-Nabarro modeling of dislocations in conventional and high-entropy namics, strain hardening and hysteresis in ductile single crystals, J. Mech.
alloys, Comput. Phys. Commun. 215 (2017) 7e12. Phys. Solid. 50 (2002) 2597e2635.
[49] Z. Pei, G.M. Stocks, Origin of the sensitivity in modeling the glide behaviour of [57] C. Shen, Y. Wang, Incorporation of g-surface to phase field model of disloca-
dislocations, Int. J. Plast. 106 (2018) 48e56. tions: simulating dislocation dissociation in fcc crystals, Acta Mater. 52 (2004)
[50] G. Schoeck, The Peierls model: progress and limitations, Mat. Sci. Eng. A 683e691.
400e401 (2005) 7e17. [58] J.R. Mianroodi, B. Svendsen, Atomistically determined phase-field modeling of
[51] Y. Xiang, H. Wei, P.B. Ming, W. E, A generalized PeierlseNabarro model for dislocation dissociation, stacking fault formation, dislocation slip, and re-
curved dislocations and core structures of dislocation loops in Al and Cu, Acta actions in fcc systems, J. Mech. Phys. Solid. 77 (2015) 109e122.
Mater. 56 (2008) 1447e1460. [59] M.S. Duesbery, V. Vtek, D.K. Bowen, The effect of shear stress on the screw
[52] H. Wei, Y. Xiang, “A generalized PeierlseNabarro model for kinked disloca- dislocation core structure in body-centred cubiclattices, Proc. Roy. Soc. Lond.
tions, Philos. Mag. 89 (2009) 2333e2354. A 332 (1973) 85e111.
[53] S. Dai, Y. Xiang, D.J. Srolovitz, Structure and energy of (111) low-angle twist [60] A.H.W. Ngan, A generalized Peierls-Nabarro model for nonplanar screw
boundaries in Al, Cu and Ni, Acta Mater. 61 (2013) 1327e1337. dislocation cores, J. Mech. Phys. Solid. 45 (1997) 903e921.
[54] C. Shen, J. Li, Y. Wang, Predicting structure and energy of dislocations and [61] S.I. Rao, C. Varvenne, C. Woodward, T.A. Parthasarathy, D. Miracle,
grain boundaries, Acta Mater. 74 (2014) 125e131. O.N. Senkov, W.A. Curtin, Atomistic simulations of dislocations in a model bcc
[55] G. Xu, A.S. Argon, Homogeneous nucleation of dislocation loops under stress multicomponent concentrated solid solution alloy, Acta Mater. 125 (2017)
in perfect crystals, Phil. Mag. Lett. 80 (2000) 605e611. 311e320.