Available online at www.sciencedirect.

com

ScienceDirect
Acta Materialia 65 (2014) 8597
www.elsevier.com/locate/actamat

Size-dependent plasticity in an Nb25Mo25Ta25W25 refractory

high-entropy alloy
Yu Zou a,1, Soumyadipta Maiti b,1, Walter Steurer b,, Ralph Spolenak a,
a

Laboratory for Nanometallurgy, Department of Materials, ETH Zurich, Wolfgang-Pauli-Strasse 10, CH-8093 Zurich, Switzerland
Laboratory of Crystallography, Department of Materials, ETH Zurich, Wolfgang-Pauli-Strasse 10, CH-8093 Zurich, Switzerland

Received 8 October 2013; received in revised form 17 November 2013; accepted 18 November 2013
Available online 21 December 2013

Abstract
High-entropy alloys (HEAs) are evolving multi-component intermetallic systems, wherein multiple principal elements tend to form
single solid-solution-like phases with a strong tendency to solid solution strengthening. In this study, an Nb25Mo25Ta25W25 refractory
HEA was synthesized by arc melting and well homogenized at 1800 C. Single-crystalline HEA pillars in two orientations ([0 0 1] and
[3 1 6]) and with diameters ranging from 2 lm to 200 nm were produced by focused ion beam milling and compressed using a at-punch
tip in a nanoindenter. The HEA pillar samples can reach extraordinarily high strength levels of 44.5 GPa, which is 33.5 times
higher than that of the bulk HEA; meanwhile the ductility is signicantly improved. Compared to pure Nb, Mo, Ta and W pillars,
the HEA pillars exhibit higher strengths than any of them in both absolute and normalized values, and the HEA pillars also show relatively low compressive size eects, as evaluated by the loglog slope of strength vs. pillar diameter. The higher strength levels and lower
size dependence for the HEA could be attributed to the increased lattice resistance caused by localized distortion at atomic length scales.
The correlation between normalized strengths, length scales and temperatures for body-centered cubic structured pillars is illustrated,
and the relevance of a size-eect slope as well as the additivity of strengthening mechanisms is critically discussed.
2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Microcompression; High-entropy alloy; Body-centered cubic; Size eects; Solid solution hardening

1. Introduction
High-entropy alloys (HEAs), conceptualized by Yeh et al.
in 2004 [1], are usually made of ve or more metallic elements
with equimolar or near-equimolar ratios, where their congurational entropy, Sconf, increases with the number of elements n as DSconf = R ln(n), with R the universal gas
constant. The high entropy stabilizes solid-solution phases
at elevated temperatures and single-phase HEAs prevent
the formation of possible intermetallics in these compositions [13]. HEAs may have interesting applications due to

Corresponding authors. Tel.: +41 44 632 66 50 (W. Steurer), tel.: +41

44 632 25 90 (R. Spolenak).
E-mail addresses: steurer@mat.ethz.ch (W. Steurer), ralph.spolenak@mat.ethz.ch (R. Spolenak).
1
These authors contributed equally to this work.

their simple average structure (body-centered cubic, bcc, or

face-centered cubic, fcc), their distorted lattice and low diusion rate in a multi-component system [15]. The main
potential applications of HEAs are in the development of
high-strength and high-temperature sustaining alloys [6,7],
wear-resistant materials [2] and diusion barriers [8]. The
conventional HEAs based on Al, Co, Cr, Cu, Fe and Ni
have reached strengths and workability comparable to
those of steels [6,7]. To achieve higher strengths in the hightemperature regime above 1100 C, the use of refractory
metals in HEAs was implemented by Senkov et al. [9,10],
which is particularly relevant to the aerospace industry, but
their low room-temperature ductility might be a limitation
for further processing steps. Furthermore, refractory metals
or alloys have also been proposed to be employed for electrical
resistors, medical implants and micro- and nanoelectromechanical systems (MEMS and NEMS) [11,12]. It is thus also

1359-6454/$36.00 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.actamat.2013.11.049

86

Y. Zou et al. / Acta Materialia 65 (2014) 8597

of great interest to apply refractory HEAs in the fabrication

of micro- or nanodevices. However, to the authors knowledge, so far all the investigations on refractory HEAs have
been limited to bulk samples, and no study on the mechanical properties of refractory HEAs at the submicron or nanometer scale has been reported.
Small-sized (extrinsic size rather than intrinsic size)
metallic specimens, such as thin lms, wires and pillars,
exhibit size-related strengths in a range from several
microns down to a few nanometers [1315]. Over the last
few years, great advances have been made to understand
the mechanical behavior of materials in micron and submicron regimes by applying the microcompression technique to ion-milled pillars (see reviews in Refs. [16,17]). It
has been widely found that the yield or ow strength (r)
of the pillar can be strongly increased when its dimension
(D) is decreased, commonly expressed by a relationship
of r / Dm [18,19], where m is the size-eect exponent. fcc
metals (e.g. Ni, Au, Al and Cu) exhibit a pronounced
and constant size dependence of plasticity with m in the
range between 0.6 and 0.9 [2022]. bcc metals (e.g.
Nb, Mo, V, Ta and W) have a much more complex sizerelated behavior with various m values ranging from
0.2 to 0.9, reported by Schneider et al. [23,24], Kim
and Greer [25], Kim et al. [25,26] and Han et al. [27].
Schneider and his co-workers [23] noticed that various m
values in bcc metals could be correlated with dierent critical temperatures (Tc), above which ow stress becomes
insensitive to test temperature, and equivalently residual
Peierls potentials: the higher Tc, the lower size dependence.
The popular interpretation of this correlation is that dierent non-planar dislocation cores in bcc metals play important roles in the mobility of screw dislocations, which
inuences the size dependence levels of bcc pillars
[17,24,28,29]. Both the simulation [30] and experiment
[31] suggest that bcc and fcc metal pillars dier in the controlling mechanisms of the size eect. However, the exact
mechanism which determines the size-dependent plasticity
in bcc metals remains under debate.
HEAs are essentially solid solutions with a simple fcc
or bcc structure. Now, two questions arise: what are
the strength and ductility of single-crystalline HEAs at
micron and submicron scales, compared with their bulk
forms? And what is the size-dependent behavior of bcc
refractory HEAs, compared with that of pure bcc metals? In this study, the mechanical properties of Nb25Mo25Ta25W25 HEA pillars with diameters ranging
from 2 lm to 200 nm were investigated using the
microcompression method. This work aims to answer
the two questions above and attempts to shed light
on potential applications of HEAs in micro/nanodevice
design.
2. Materials and methods
Compacted pellets of an approximately equimolar mixture of pure Nb, Mo, Ta and W powders were arc-melted

in argon atmosphere. A Ti getter was used to consume

any trace of oxygen in the argon atmosphere. The button-shaped cast was ipped upside down and re-melted
four times. The cast alloy was then sealed inside a Ta
ampoule and homogenized at 1800 C (65% of the calculated melting temperature [9]) for 7 days. The phase purity
of the homogenized HEA sample was determined by powder X-ray diraction (XRD) (PANalytical XPert PRO diffraction system) using Cu Ka1 monochromatic radiation in
a 2h (diraction angle) range from 20 to 120. In order to
determine the atomic displacement parameters (ADP), a
piece of the HEA crystal smaller than 40 lm was extracted
to collect a single-crystal XRD dataset. A single-crystal diffractometer with Mo Ka radiation source and a CCD
detector (Oxford Diraction, Xcalibur) was used for the
data collection. For microstructure and composition analysis, a SU-70 Hitachi scanning electron microscope (SEM)
combined with energy-dispersive X-ray spectroscopy
(EDX) (X-MAX, Oxford Instruments) was employed. A
FEI Tecnai F30 high-resolution transmission electron
microscope (HRTEM) was used to investigate the structures on sub-micron and nanometer levels. For the TEM
investigation, the HEA sample was embedded inside a copper tube, cut into thin discs, ne polished with sand papers,
dimple-polished on both sides with 3 lm diamond suspension and nally thinned by Ar jet milling.
The hardness of the HEA samples (as-cast and asannealed for 2, 4 and 7 days) and the pure elemental bcc
samples (Nb, Mo, Ta and W as-annealed for 7 days) was
measured in 15 dierent positions using a Vickers microhardness indenter with a load of 200 g and dwell time of
10 s. The HEA sample which was homogenized for 7 days
was used for the microcompression tests. The orientations
of the grains were determined by electron back-scatter diffraction (EBSD) using a FEI Quanta 200 FEG SEM.
Before EBSD characterization, the HEA bulk sample was
cross-sectioned using an alumina cut-o wheel (Struers
50A13), polished using 3 lm diamond paste and nally
polished using 60 nm SiO2 particle suspension. After EBSD
characterization, two orientations were selected to produce
pillars using a focused ion beam (FIB) system (Helios
Nanolab 600i, FEI): [3 1 6] orientation (tolerance angle
<6) with single-slip systems and [0 0 1] orientation (tolerance angle <4) with multiple-slip systems, as circled in
Fig. 1, respectively. The latter orientation can be used to
make a direct comparison with the pure bcc pillars which
have been reported in the literature [23,26].
A two-step milling method was employed: 2.5 nA for
coarse milling and 1040 pA for ne milling. The
FIB-milled pillars have diameters of 2 lm, 1 lm, 500 nm
and 250 nm and aspect ratios of 2.55. A taper of 23
was generally observed in those pillars, and the top diameters were chosen to calculate engineering stresses. At least
four pillars of each size and each orientation were
compressed using a nanoindenter (Triboindenter, Hysitron
Inc., USA) with a diamond at-punch tip (5 lm in
diameter, Synton-MDP, Switzerland) in the displacement

Y. Zou et al. / Acta Materialia 65 (2014) 8597

87

300 m

200 m

Fig. 1. (a) EBSD inverse pole gure map of the cross-section of the HEA bulk specimen before FIB milling. Two grains which were selected to mill pillars
are indicated by circles: the left one is [3 1 6]-oriented (tolerance angle <6) and the right one is [0 0 1]-oriented (tolerance angle <4). The orientations of the
two selected grains are also indicated in the inverse pole gure at the bottom. (b) A typical SEM image (SE mode) of the HEA specimen after FIB milling.
The two selected grains are indicated by circles as well.

control mode by feedback mechanism. A strain rate of

2  103 s1 was used for all the compression tests. The
displacement and loading time were changed according to
the pillar height in order to keep the strain rate constant.
The morphologies of the pillars were characterized using a
high-resolution SEM (MAGELLAN, FEI) before and after
the compressions.
3. Results
3.1. Microstructure and phase analysis of bulk specimens
The powder XRD pattern of the homogenized HEA
indicates a single-phase bcc structure with all the peaks
indexed, as shown in Fig. 2. The experimental lattice
, which is close to the preparameter is 3.222 0.001 A

dicted value of 3.229 A, according to Vegards law [32].

After homogenization for 7 days, the grain size of the
HEA is larger than 200 lm. The compositional homogeneity of the sample was measured by EDX with line scans
inside a grain and across a grain boundary, respectively,
as shown in Fig. 3. The back-scattered electron (BSE)

images show no signicant contrast either within a grain

(Fig. 3a) or between two adjacent grains (Fig. 3c) and no
trace of a dendritic segregation was found as that in the literature [9,10]. The corresponding elemental analyses
obtained by EDX are plotted in Fig. 3b and d. The elemental composition varies by 12% and the overall atomic
compositions vary within 3%, compared to 10% variation
in the literature [9]. The average hardness value of the HEA
is between 4.52 and 4.85 GPa (Table 1), close to the value
as reported in the literature (4.46 GPa) [10]. The standard
deviations of the hardness decrease from 387 to 58 MPa
over a homogenization time of 7 days. The HEA shows a
much higher hardness than the pure bcc elements measured
here, even 1 GPa higher than pure W.
The bright-eld HRTEM image (Fig. 4a) was taken with
the zone axis along the [1 0 0] direction. Fig. 4b shows a
Fourier-ltered image of the lattice fringes in Fig. 4a.
Two boxed regions in Fig. 4b are enlarged at the bottom
and traced for their lattice fringes, as shown in Fig. 4cf.
The HRTEM images (Fig. 4a and b) show the (1 1 0) set
of lattice planes and the fringes continue through the whole
length of the sample. The Fourier-ltered image (Fig. 4b)
shows that there are visible mismatches and local distortions in both the two sets of parallel lattice fringes. These
local lattice distortions induce a slight amount of kinks
and bends in the lattice layers as also observed in bcc CoCrFeNiAl HEA [33].
In order to obtain the ADP, the single-crystal diraction
dataset was rened by the program SHELXL97 [34]. The
average ADP value of the HEA was determined to be
2. The average thermal component of the ADP,
0.0091 A
Ut, calculated from the DebyeWaller factors of the pure
2 [35]. The static component of ADP,
elements is 0.0037 A
Us, is calculated in a similar approach as in Ref. [40] and
has the form:
U s Rci d i  d al 2

Fig. 2. X-ray diraction pattern (Cu Ka1) of the NbMoTaW HEA

annealed at 1800 C for 7 days. The pattern indicates a single-phase bcc
structure.

where ci is the mole fraction of the element i, di is the lattice

parameter of the pure element and dal is the lattice parameter of the HEA. The expected total ADP of the alloy, Ual,
including the thermal and static components are calculated

88

Y. Zou et al. / Acta Materialia 65 (2014) 8597

100 m

100 m

Fig. 3. SEM images (BSE mode) of the cross-sections of the NbMoTaW HEA annealed at 1800 C for 7 days: (a) inside a grain and (c) including a
grain boundary. The arrows indicate the line-scans for EDX analysis. (b and d) The corresponding atomic proportions of the four elements along the lines
in (a) and (c), respectively. The typical EDX resolution limit is 1.0 at.%.

Table 1
The average values and standard deviations (in MPa) of the Vickers microhardness of 15 random indents on the HEAs (as-cast and as-annealed at 1800 C
for 2, 4 and 7 days) and pure Nb, Mo, Ta and W (as-annealed at 1800 C for 7 days).
Materials

Nb 7D

Mo 7D

Ta 7D

W 7D

HEA cast

HEA 2 days

HEA 4 days

HEA 7 days

Average
Standard deviation

1086
16

1898
51

1488
52

3714
96

4853
387

4766
362

4803
234

4515
58

2, which is
to be Ut + Us = 0.0037 + 0.0055 = 0.0092 A
2
.
close to the experimental value of 0.0091 A
3.2. Compression of pillar samples
As shown in the SEM images (Fig. 5a and b), single slips
are observed in 2 lm and 1 lm [3 1 6]-oriented HEA pillars,
which have slip bands traversing along the gauge length of
the samples. The slip bands are oriented at 4070 o the
loading axis. A second slip system could be also activated
when the pillars experienced large strains. Some localized
shear osets along slip planes are observed at the top part
of the pillars (Fig. 5b). For smaller pillars (500 nm and
250 nm in diameter, Fig. 5c and d), multiple slips are usually observed. The multiple slips might be due to a slight
misalignment between the pillar top and the at punch or
due to the inuence of the tolerance angle. The multiple
slips are expected to contribute to strain hardening. As
shown in Fig. 5b, occasionally some degree of bending is

observed. The data from bent pillars are not considered

in the analysis.
Compressed [0 0 1]-oriented HEA pillars are shown in
Fig. 6. Wavy morphologies can be found in both large
and small pillars. This wavy-slip feature may be attributed
to the cross-slip of screw dislocations along h1 1 1i directions, which is commonly observed in deformed bcc metals
[36,37]. This post-deformed morphology of the HEA pillars is similar to that of the W pillars as reported in Refs.
[23,26].
Fig. 7a and b shows engineering stressstrain relationships for the representative [3 1 6]- and [1 0 0]-oriented
HEA pillars, respectively. Although the crystal orientations
are dierent, the two groups of pillars exhibit similar features of both stress magnitude and characters of curves:
the small pillars have higher ow strengths than the big pillars and displacement bursts occurred in both big and small
pillars, showing a similar phenomenon to that observed in
fcc and bcc metal pillars [27,36,37]. The displacement

Y. Zou et al. / Acta Materialia 65 (2014) 8597

89

Fig. 4. High-resolution TEM images of the NbMoTaW HEA homogenized at 1800 C for 7 days: (a) a bright-eld TEM image oriented in [1 0 0] zone
axis and the corresponding electron diraction pattern; (b) an inverse fast Fourier transform image of the area (a); (c) and (e) are enlarged images of the
indicated boxes in (b); (d) and (f) show lattice fringes which are traced for (c) and (f), respectively, to indicate the regions with lattice distortions.

a
1 m

2 m

2 m

500 nm

500 nm

Fig. 5. SEM images (SE mode) of post-compressed [3 1 6]-oriented HEA pillars with approximate diameters of: (a) 2 lm, (b) 1 lm, (c) 500 nm and (d)
250 nm. An enlarged image which presents sharp slip bands is shown in the inset of (a).

bursts may be due to a relatively low feedback rate in displacement mode compared with the burst events, as
explained in Refs. [38,39]. In both orientations, the smaller
pillars exhibit stronger displacement bursts in both magni-

tude and frequency; meanwhile the smaller pillars show a

higher strain-hardening rate than the big pillars, which
might be due to more activated and interacting slip systems
in the small dimension samples. To reduce the inuence of

90

Y. Zou et al. / Acta Materialia 65 (2014) 8597

500 nm

1 m

2 m

400 nm

500 nm

Fig. 6. SEM images (SE mode) of post-compressed [0 0 1]-oriented HEA pillars with approximate diameters of: (a) 2 lm, (b) 1 lm, (c) 500 nm and (d)
250 nm. An enlarge image which presents a wavy morphology is shown in the inset of (a).

Fig. 7. Representative engineering stress strain curves for (a) [3 1 6]-oriented and (b) [0 0 1]-oriented single crystalline HEA pillars with the diameters
ranging from 2 lm to 200 nm.

the displacement bursts on analysis and make a direct comparison with pure bcc metal pillars in the references [23,26],
the highest ow stress values measured below 5% strain
and 8% strain, which are dened as r0.05 and r0.08, respectively, are used to compare the strengths for dierent pillar
dimensions.
The changes of r0.05 and r0.08 due to dierent pillar
diameters for both [3 1 6] and [0 0 1] orientations are plotted
in Fig. 8: for r0.05, [3 1 6] and [0 0 1]-orientated HEA pillars
have size-eect exponents (m) of 0.30 0.02 and
0.33 0.02, respectively; for r0.08, [3 1 6] and [0 0 1]-orientated HEA pillars have m values of 0.32 0.02 and

0.36 0.02, respectively. The absolute strength levels

for the pillars in two orientations are close to each other.
[3 1 6]-oriented pillars have a slightly lower size dependence
than [0 0 1]-oriented pillars.
4. Discussion
4.1. Solid solution eect
Some physical properties of Nb, Mo, Ta, W and the
HEA are listed in Table 2. The atomic sizes of Nb and
Ta are 5% larger than Mo and W. This atomic size mist

Y. Zou et al. / Acta Materialia 65 (2014) 8597

Fig. 8. The relationship between engineering stresses at 5% strain and 8%

strain (r0.05 and r0.08) and pillar diameters for [3 1 6]- and [0 0 1]-oriented
HEA pillars.

of the constituent elements in the HEA can cause a highly

distorted lattice (Fig. 4) with localized strains throughout
the whole sample. The lattice distortion in the HEA specimen appears as splits in the atomic positions and localized
shearing of several adjacent lattice fringes. The calculated
ADP was 2.5 times higher than the expected thermal
ADP, also suggesting a local lattice distortion due to the
dierence in atomic sizes. The ADP values have been used
as a measure of the local lattice distortion in a bcc ZrNbHf
alloy [43], where it was observed that the average ADP of
the alloy was many times higher than the expected thermal
ADP. In addition, the thermal and static components of
the modeled ADPs, if added up, match the experimental
rened ADP of the HEA closely. This might validate the
simple model of calculating static ADPs, and the average
. Morestatic displacement of the atoms could be 0.074 A
over, the modulus mist between the constituent elements
has a large range: 4% between W and Mo; 70% between
W and Nb. Compared with the pure bcc elements, the
binding forces around dierent solute atoms in the HEA
could vary depending on surrounding elements. This nonuniform bonding feature at atomic length level may lead

91

to extremely inhomogeneous stress elds throughout the

HEA specimen.
In bcc metals, edge dislocations move much more easily
than screw dislocations. Thus, the latter ones mainly control plastic ows, by a processes of kink nucleation and
motion [44]. The eect of lattice distortion on plastic
strength may be twofold [45,46]: on the one hand, the lattice distortion, by its stress eld, may facilitate the nucleation of new kink pairs, leading to solid-solution
softening; On the other hand, the propagation of screw dislocations is retarded by the stress elds of localized solute
obstacles along the slip planes, resulting in solid-solution
hardening. These two mechanisms compete in bcc solid
solutions during plastic deformation. However, the phenomenon of solid-solution softening is mostly observed at
intermediate low temperatures (100250 K) and low concentrations (less than 5 at.%). In our study, due to the
room-temperature measurement and high-concentration
alloying, the solid-solution hardening eect should be dominant in the HEA.
4.2. HEA pillar vs. HEA bulk: ductility and strength
4.2.1. Ductility
Since the concept of HEA design was introduced [1],
HEAs have been considered as potential high-performance
structural materials. However, a big limitation to using
refractory HEAs is their low ductility and toughness at
room temperature. Senkov et al. [10] found that Nb25Mo25Ta25W25 and Nb20Mo20Ta20V20W20 HEAs fractured along
grain boundaries at 2% compressive strain at room temperature. In Fig. 9a, a fracture surface along the grain
boundaries of the bulk HEA can be seen. Two HEA pillars
with and without a grain boundary are compared, as
shown in Fig. 9b and c. It is found that a crack that propagated along the grain boundary caused the failure of a
3 lm bicrystal pillar (Fig. 9b). The pillar top area and the
corresponding forcedisplacement curve are shown in the
insets. Fig. 9c exhibits a post-deformed [0 0 1]-oriented single-crystalline pillar, which can even bear large-strain bending (75) without any fracture or crack on the surface,
corresponding to a tensile strain larger than 20% on one

Table 2
Physical properties of pure Nb, Mo, Ta, W and the HEA.
)
Metal
a (A
q (g cm3)
s0 (MPa)
G (GPa)

s0 (103)

Tm (K)

Tc (K)

m for r0.05

m for r0.08

Nb
Mo
Ta
W
HEA calc.
HEA exp.

8.7
4.6
5.4
1.72.1

2750
2896
3290
3695
3158

350 [23], 290 [26]

480 [23], 465 [26]
450 [23], 440 [26]
800 [23], 760 [26]
520 [23], 489 [26]
9001200 [10]

0.48 [23]
0.38 [23]
0.41[23]
0.21 [23]
0.37
0.33

0.93
0.44
0.43
0.44
0.56
0.36

3.301
3.147
3.303
3.165
3.229
3.222

8.57
10.28
16.65
19.25
13.69

415
730
340
280350

47.2 [41]
158 [24]
62.8 [41]
164 [42]
114

[26]
[26]
[26]
[26]

The crystal lattice parameter, a, density, q, and melting temperature, Tm, of pure Nb, Mo, Ta and W are adapted from Table 2 of Ref. [9]; Peierls stress, s0 ,
is collected from Ref. [42], the corresponding shear modulus, G, is chosen for the active slip systems of {1 1 2}h1 1 1i [24,41,42]; the values of critical
temperature, Tc, and size-eect exponent, m, are chosen from Refs. [23,26]. To give an estimation of Tc in the bulk HEA [10], an intersection point between
the high-temperature linear part and the low-temperature linear part in the measured stresstemperature curve is chosen as the value of Tc. HEA calc. is
calculated due to the rule of mixtures. HEA exp. is the value measured in this study.

92

Y. Zou et al. / Acta Materialia 65 (2014) 8597

2m

400 m

10 m

5 m

Fig. 9. Typical SEM images (SE mode) of: (a) a fracture surface in the HEA bulk sample, showing the fracture occurred along grain boundaries; (b) a
post-deformed HEA pillar that contains a grain boundary (indicated by arrows), where the fracture occurred. The corresponding forcedisplacement
curve and the enlarged area of the fracture region are shown in the insets; (c) a 2 lm [0 0 1]-oriented single-crystalline HEA pillar, which was severely bent
without any fracture or crack after deformation.

side of the pillar. This comparison suggests that the elimination of grain boundaries and decrease of sample size
could signicantly increase the ductility of HEAs.
4.2.2. Strength
Submicron-sized HEA pillars exhibit extraordinarily
high strengths compared to the bulk HEA (44.5 vs.
1 GPa [10]). The origin of the higher strength for the
HEA pillars could be the same as the size-eect phenomena
for the other metal pillars. It is generally believed that when
the sample dimension is reduced into micron and submicron
regimes, the strength is increased due to the decreased average size of dislocation sources. Weinberger and Cai [30] suggest, for bcc pillars, that a single dislocation can also
multiply itself repeatedly and forms dislocation segments
and hard junctions, contributing to the increased strength.
4.3. HEA bcc pillar vs. pure elemental bcc pillars: size eects
Size-dependent strengths have been commonly and
empirically characterized by a power-law relation, either
in a non-normalized form, r = A(D)m, or in a normalized
form, s/G = A(D/b)m, where s is the resolved shear stress
on primary slip planes, A is a constant, D is the pillar

diameter and b is the Burgers vector [16,17,19,47,48]. For

the calculation of s in this study, {1 1 2} slip planes are chosen [24,41], and the Schmid factors for [3 1 6] and [0 0 1] orientations are 0.41 and 0.47, respectively. Here, we apply
the power-law ts to the HEA pillars (this study) and to
[0 0 1]-oriented Nb, Ta, Mo and W (literature data from
Refs. [23,26]). In both non-normalized and normalized tting lines (Fig. 10ad), the HEA pillars exhibit higher
strength levels than the pure bcc metal pillars. For example, the 200 nm HEA pillars have higher ow strengths
than pure bcc metal pillars by a factor of 24. In addition
to higher strength levels, the HEA pillars also exhibit a
reduced size eect (a smaller absolute value of m) compared to the pure bcc pillars here. The only exception to
this trend is that the size dependence of the HEA is slightly
larger than that of W reported by Schneider et al. [23], but
it is smaller than that of W reported by Kim et al. [26]
(compare Table 2).
Fig. 10e gives a schematic illustration of the normalized strengthdiameter relationship of fcc and bcc pillars
summarized from Refs. [17,23,26] and the HEA pillars
in this study. In the size range of a few microns, bcc
pillars have higher normalized strength levels than fcc
pillars, but in the submicron regime the strength levels

Y. Zou et al. / Acta Materialia 65 (2014) 8597

93

Fig. 10. Size-dependent strengths for the [3 1 6]- and [0 0 1]-oriented HEA pillars in this study and pure Nb, Mo, Ta and W as reported by Schneider et al.
[23] and Kim et al. [26]. (a and b) r0.05 and r0.08 vs. pillar diameters (D); (c and d) resolved ow strengths normalized by corresponding shear modulus (s/
G) vs. pillar diameters normalized by Burgers vector (D/b). (e) Schematic illustration of size-dependent strengths for dierent metallic systems: FIB-milled
pure fcc and bcc pillars (data summarized from Refs. [17,23,26]) and the HEA bcc pillars in this study, and the range of each group is indicated by a
colored solid ellipse. The HEA bcc pillars exhibit both higher absolute and normalized strength levels than any other bcc metals but a relatively low size
dependence of strengths.

of bcc pillars converge to those of fcc pillars. The HEA

bcc pillars in this study show extraordinarily high strength
compared to the pure bcc elements. In order to understand the dierent size eects for pure bcc and HEA

bcc pillars, we propose a simple analysis on the resolved

ow stress of a pillar sample. The applied resolved shear
stress, s, is traditionally expected to be a sum of lattice
friction, s*, elastic interactions between dislocations

94

Y. Zou et al. / Acta Materialia 65 (2014) 8597

(i.e., Taylor hardening), sG, and

strength, ssource, expressed as [48,49]:

source-controlled

s s sG ssource

In Eq. (2), s is the stress required to overcome the Peierls

potential and arises as a consequence of the forcedistance
relation between individual atoms in a periodic lattice
structure. s is temperature-dependent and can be expressed as [50]:


Tt 

s 1
3
s with T t < T c ;
Tc 0
s  0

with T t P T c

where Tt is test temperature, usually room temperature and

s0 is the Peierls stress at 0 K. Above Tc, there is sucient
thermal energy to overcome the Peierls barriers by thermal
activation. The second term in Eq. (2), sG, is an athermal
component, which arises from the resistance to dislocation
motion due to long-range elastic interactions, such as the
interactions between dislocations. Here, sG may be simply
approximated by using the Taylor-hardening relation as:
p
5
sG aGb q0 qD
where a is a constant falling in the range 0.1 to 1.0, q0 is the
initial dislocation density before pillar compression and qD
is the increased dislocation density due to the compression.
For a small amount of strain, the dislocation density is in
the order of 10121013 m2 for most metals. Unlike in bulk
metals, dislocation storage in small-scale pillar specimens
could be in a lower level, because new generated dislocations may move out of a pillar more easily due to a small
conned dimension. Another contribution in Eq. (2), ssource,
is the minimal stress required to operate a dislocation
source. In bulk samples dislocation segments in the length
104b can act as a FrankRead source [50]. However, in pilk, is limited, and it is
lar samples the average source length, 
proportional to the pillar dimension. Single-ended sources
could be dominant in the size range of 0.520 lm
(102105b) [51]. The single-end source has also been seen
in aluminum pillars using in situ TEM [52]. The activation
stress of a dislocation source in a pillar sample has been estimated by three-dimensional (3-D) discrete dislocation
dynamics simulations [51] and could be expressed as:
k=b
ln
ssource KG 
k=b

4.3.1. Pure elemental bcc pillars

In order to illustrate how the three mechanisms underlying the three terms in Eq. (7) inuence the size dependence
of the strength in bcc metals, the parameters for Mo, which
have been most investigated for pillar compression, are cho , (q0 + qD)  5.0  1012 m2,
sen, as: a  0.5, b  2.728 A

K  0.5 and k  D [48,50,51]. It should be noted that K
is dependent on Poissons ratio, dislocation type and
anisotropy of dislocation line tension, and k is also inuenced by dislocation densities and their distribution [51].
Here, we use the above values to give an estimation of pillar strengths. Fig. 11 shows a 3-D graph of normalized
strength (s/G) vs. normalized length (D/b) and normalized
temperature (Tt/Tc) for Mo pillars. The top surface with
contour lines represents a sum of all the three mechanisms
in Eq. (7). Each mechanism is also plotted in a single color
below separately: s (blue), sG (red) and ssource (green). The
graph clearly shows how the local slope, m, increases in
magnitude with an increase of normalized temperature.
According to this graph, we could make the following predictions: at 0 K, if the sample size is smaller than 1000b,
the strength is source-controlled, having strong a size
dependence; if the size is larger than 1000b, the strength
is controlled by the Peierls potential, showing nearly no size
eect. However, when Tt equals Tc (480 K for Mo), if the
size is smaller than 20,000b, the strength is controlled by
source action; but if the size is larger than 20,000b, dislocation interactions play an important role in reducing the
size eect. At room temperature (i.e. Tt of 300 K), this
0.0300
0.0121
0.0051
0.0014
0.0005
0.0002

0.0001

source
G

where K is the source-strengthening constant in the order

of 0.1. Although dierent controlling mechanisms could
operate for bigger pillars (>20 lm in diameter) and even
smaller pillars (<100 nm in diameter), the single-end
source model applies to the size range discussed in this
study. Because Tt is smaller than Tc in most bcc metals,
merging Eqs. (3), (5), and (6) into Eq. (2), we obtain:


p
k=b
s
T t s0
ln
1
ab q0 qD K 
7
G
Tc G
k=b

Fig. 11. A 3-D illustration for the size and temperature dependence of the
strengths in Mo pillars according to Eq. (7): normalized strength (s/G) vs.
(normalized length scale (D/b), normalized temperature (Tt/Tc)) in a size
range of 102105b and a temperature range of 0Tc. In order to give the
best estimation of the strength, the parameters are chosen as: a  0.5,
, (q0 + qD)  5.0  1012 m2, K  0.5 and k  D, [48,50,51].
b  2.728 A
The top rainbow-colored surface with contours is a sum of all the
strengthening mechanisms in Eq. (7). The contributions from the lattice
resistance (s*), the Taylor hardening (sG) and the source strength (ssource)
are plotted in blue, red and green, respectively. (For interpretation of the
references to color in this gure legend, the reader is referred to the web
version of this article.)

Y. Zou et al. / Acta Materialia 65 (2014) 8597

calculated strengthsize curve according to Eq. (7) is compared to the experimental data of [0 0 1]-Mo pillars [23,26]
and [1 1 1]-Mo pillars [53], as shown in Fig. 12. The dashed
lines present the individual contributions from the lattice
friction (s*, blue), the Taylor hardening (sG, red) and the
source strengthening (ssource, green) to the overall strength
(s, black), respectively. In the case of Mo, the experimental
data points are in a good agreement with the calculated
strength curve at a reasonable scatter level, especially for
the sample size larger than 100 nm. Surface image stresses
may have a large eect at the length smaller than 100 nm.
In the region in which two or three mechanisms are similar
in magnitude, the normalized strength will change by up to
a factor of three, compared to a scenario where only the
strongest strengthening mechanism is relevant. This is the
only scenario where an understanding of which strengthening mechanisms are additive and which are not becomes
important. If only one mechanism dominates, the distinction is secondary.
4.3.2. HEA bcc pillar
Here, we attempt to predict a strengthsize curve for the
HEA pillars using Eq. (7). Although there is no experimental data of the Peierls stress and shear modulus for the
HEA, the values of s0 =G are available for Nb, Mo, Ta
and W [42] (Table 2), which are between 103 and 102.
Wang [54] also calculates s0 =G theoretically and estimates
that the values of s0 =G are 103 for bcc edge dislocations
and 102 for bcc screws. Here, the maximum s0 =G value

95

among the bcc metals, 8.7  103, is chosen to give an esti , and Tc
mation for the HEA pillars as well as b of 2.799 A
of 1050 K (Table 2). As can be seen in Fig. 12, the experimental data points are higher than the predicted curve by a
factor of 2. The reason might be that unlike pure bcc elements the solute atoms in the HEA have dierent atomic
dimensions which can induce signicant localized lattice
distortion (Fig. 4). While the rule of mixtures may be
appropriate for determining the shear modulus, the severe
lattice distortions in the HEA are expected to result in a
signicantly higher Peierls potential than for each of the
constituents. Moreover, the non-uniform stress elds
throughout the HEA sample might cause an increased
dynamic drag eect and the following phenomena might
occur [46,55,56]: the emission of elastic waves during the
deceleration and acceleration of dislocation sliding along
a distorted lattice, the excitation of local vibrations of solute atoms and the radiation of phonons by dislocation
vibration like a string. Dierent from pure and lightly
alloyed metals with a relative ideal lattice, the dynamic
drag eect could be prominent in the HEA, and therefore
the lattice friction could be signicantly increased, leading
to strong strengthening. However, to make a convincing
conclusion, a precise experimental evaluation of Tc and
s0 =G as well as detailed microstructural analyses and
atomic simulations of the non-planar dislocation core
structure in HEAs will be a subject for future investigation.
As we have shown in Fig. 11, the apparent size eect
exponent m depends not only on the superposition of
strengthening mechanisms but also on the experimentally
accessible size range. Nevertheless it is instructive to correlate m to the normalized temperature, if the analyzed size
ranges and dislocation densities are comparable. Here, we
adapted the method used by Schneider et al. [23] to correlate m and Tt/Tc for pure bcc and HEA bcc pillars. According to Eq. (7), the value of m can be expressed as:

*
source
Nb

W
Mo Ta

Fig. 12. The calculated normalized strength vs. normalized length for Mo
pillars at room temperature (300 K) according to Eq. (7). The solid black
line is a sum of all the mechanisms for Mo and the contribution of each
mechanism is plotted separately in a dashed color line: s* (blue), sG (red)
and ssource (green). The black points are the experimental data of [0 0 1] Mo
pillars [23,26] and [1 1 1] Mo pillars [53]. The predicted curve for the HEA
using Eq. (7) and the experimental data of [0 0 1] HEA in this study are
also plotted. In order to calculate the strength levels of the HEA, the
, Tc  1050 K (Table 2) and
following parameters are chosen: b  2.799 A
the maximum s0 =G value among the bcc metals, 8.7  103. (For
interpretation of the references to color in this gure legend, the reader
is referred to the web version of this article.)

the HEA

Fig. 13. The absolute values of m vs. Tt/Tc: the correlation between the
size eects and the critical temperatures of pure Nb, Mo, Ta and W pillars
[23,26] and the HEA pillars in this study.

96

m ln

Y. Zou et al. / Acta Materialia 65 (2014) 8597

  
 
b
s T t
ln B  0
D
GA T c

where B is a material-independent constant. Because s0 =G

is nearly constant for most bcc metals and in the order of
103 to 102, the m value could be mostly inuenced by
Tt/Tc. Using the values in Table 2, Fig. 13 indicates that
the material that has higher Tt/Tc is expected to have a lower size eect. The HEA from this study is in reasonable
agreement with this pattern.
5. Summary
In this work, an Nb25Mo25Ta25W25 refractory HEA
was synthesized by arc melting and well homogenized at
1800 C for 7 days. The HEA shows a simple bcc structure with the average lattice parameter of Nb, Mo, Ta
and W, while localized lattice distortions at atomic length
scales were observed using HRTEM. The mechanical
properties of [3 1 6]- and [0 0 1]-oriented HEA pillars have
been measured using the microcompression technique.
Orientation change has a minor inuence on the size
dependence of strengths. In both orientations, the HEA
pillars exhibit higher strength levels than pure Nb, Mo,
Ta and W pillars by a factor of 25, as well as a relatively low size eect (loglog slope of strength vs. pillar
diameter of 0.3). Both the increased strength levels
and reduced size dependence in the HEA could be attributed to a higher lattice friction in the HEA than that in
pure bcc metals. In this paper, we also illustrate how
the normalized strength correlates to the normalized
length scale and the normalized temperature for bcc structures, and elucidates the contributions to the size-dependent strength from lattice resistance, dislocation
interactions and source strength, respectively. Additionally, towards the application, both the strength and ductility of single-crystalline HEA pillar samples are
signicantly improved compared to polycrystalline bulk
forms. These ndings indicate that refractory HEAs show
promise as potential structural materials in micro/nanodevice design.
Acknowledgements
The authors would like to thank P. Gasser, Dr. K.
Kunze and Dr. Fabian Gramm (EMEZ, ETH Zurich)
for their help in the sample preparation using FIB and
HRTEM; Huan Ma and Matthias Schamel (LNM, ETH
Zurich) for their help in SEM and EBSD characterization;
and Claudia Muller (LNM, ETH Zurich) for proof-reading
the manuscript. The authors also gratefully acknowledge
nancial support through SNF Grants (200021_143633
and 200020_144430).
References
[1] Yeh JW, Chen SK, Lin SJ, Gan JY, Chin TS, Shun TT, et al. Adv
Eng Mater 2004;6:299.

[2] Huang PK, Yeh JW, Shun TT, Chen SK. Adv Eng Mater 2004;6:74.
[3] Yeh JW, Chen YL, Lin SJ, Chen SK. Mater Sci Forum 2007;560:1.
[4] Senkov ON, Senkova SV, Woodward C, Miracle DB. Acta Mater
2013;61:1545.
[5] Zhu C, Lu ZP, Nieh TG. Acta Mater 2013;61:2993.
[6] Tong CJ, Chen MR, Chen SK. Metall Mater Trans A 2005;36A:1263.
[7] Tsai CW, Tsai MH, Yeh JW, Yang CC. J Alloy Compd 2010;490:160.
[8] Tsai MH, Yeh JW, Gan JY. Thin Solid Films 2008;516:5527.
[9] Senkov ON, Wilks GB, Miracle DB, Chuang CP, Liaw PK.
Intermetallics 2010;18:1758.
[10] Senkov ON, Wilks GB, Scott JM, Miracle DB. Intermetallics
2011;19:698.
[11] Kaufmann D, Monig R, Volkert CA, Kraft O. Int J Plast
2011;27:470.
[12] Voyiadjis GZ, Almasri AH, Park T. Mech Res Commun 2010;37:307.
[13] Arzt E. Acta Mater 1998;46:5611.
[14] Kraft O, Gruber PA, Monig R, Weygand D. Annu Rev Mater Res
2010;40:293.
[15] Dehm G. Prog Mater Sci 2009;54:664.
[16] Uchic MD, Shade PA, Dimiduk DM. Annu Rev Mater Res
2009;39:361.
[17] Greer JR, De Hosson JTM. Prog Mater Sci 2011;56:654.
[18] Dimiduk DM, Uchic MD, Parthasarathy TA. Acta Mater
2005;53:4065.
[19] Dou R, Derby B. Scripta Mater 2009;61:524.
[20] Uchic MD, Dimiduk DM, Florando JN, Nix WD. Science
2004;305:986.
[21] Greer JR, Oliver WC, Nix WD. Acta Mater 2005;53:1821.
[22] Ng KS, Ngan AHW. Acta Mater 2008;56:1712.
[23] Schneider AS, Kaufmann D, Clark BG, Frick CP, Gruber PA, Monig
R, et al. Phys Rev Lett 2009:103.
[24] Schneider AS, Frick CP, Clark BG, Gruber PA, Arzt E. Mater Sci
Eng A Struct 2011;528:1540.
[25] Kim JY, Greer JR. Acta Mater 2009;57:5245.
[26] Kim J-Y, Jang D, Greer JR. Acta Mater 2010;58:2355.
[27] Han SM, Bozorg-Grayeli T, Groves JR, Nix WD. Scripta Mater
2010;63:1153.
[28] Malygin GA. Phys Solid State 2012;54:1220.
[29] Han SM, Feng G, Jung JY, Jung HJ, Groves JR, Nix WD, et al. Appl
Phys Lett 2013;102:041910.
[30] Weinberger CR, Cai W. Proc Natl Acad Sci 2008;105:14304.
[31] Greer JR, Weinberger CR, Cai W. Mater Sci Eng A Struct
2008;493:21.
[32] Denton AR, Ashcroft NW. Phys Rev A 1991;43:3161.
[33] Zhou YJ, Zhang Y, Wang FJ, Chen GL. Appl Phys Lett 2008:92.
[34] Sheldrick GM. Acta Crystallogr A 2008;64:112.
[35] Peng LM, Ren G, Dudarev SL, Whelan MJ. Acta Crystallogr A
1996;52:456.
[36] Spolenak R, Dietiker M, Buzzi S, Pigozzi G, Loer JF. Acta Mater
2011;59:2180.
[37] Loer JF, Buzzi S, Dietiker M, Kunze K, Spolenak R. Philos Mag
2009;89:869.
[38] Dimiduk DM, Woodward C, LeSar R, Uchic MD. Science
2006;312:1188.
[39] Dimiduk DM, Nadgorny EM, Woodward C, Uchic MD, Shade PA.
Philos Mag 2010;90:3621.
[40] Yeh JW, Chang SY, Hong YD, Chen SK, Lin SJ. Mater Chem Phys
2007;103:41.
[41] Duesbery MS, Vitek V. Acta Mater 1998;46:1481.
[42] Suzuki T, Kamimura Y, Kirchner HOK. Philos Mag A 1999;79:1629.
[43] Guo W, Dmowski W, Noh JY, Rack P, Liaw PK, Egami T. Metall
Mater Trans A 2013;44A:1994.
[44] Vitek V. Prog Mater Sci 1992;36:1.
[45] Butt MZ, Feltham P. J Mater Sci 1993;28:2557.
[46] Neuhauser H, Schwink C. Materials science and technology. London: Wiley-VCH; 2006.
[47] Korte S, Clegg WJ. Philos Mag 2011;91:1150.
[48] Lee SW, Nix WD. Philos Mag 2012;92:1238.

Y. Zou et al. / Acta Materialia 65 (2014) 8597

[49] Parthasarathy TA, Rao SI, Dimiduk DM, Uchic MD, Trinkle DR.
Scripta Mater 2007;56:313.
[50] Hull D, Beacon DJ. Introduction to dislocations. Oxford: Pergamon;
2001. p. 193232.
[51] Rao SI, Dimiduk DM, Tang M, Parthasarathy TA, Uchic MD,
Woodward C. Philos Mag 2007;87:4777.
[52] Oh SH, Legros M, Kiener D, Dehm G. Nat Mater 2009;8:95.

97

[53] Huang L, Li Q-J, Shan Z-W, Li J, Sun J, Ma E. Nat Commun

2011;2:547.
[54] Wang JN. Mater Sci Eng A Struct 1996;206:259.
[55] Alshits V, Indenbom V. In: Nabarro FRN, editor. Dislocations in
solids, vol. 7. Amsterdam: Elsevier Science; 1986.
[56] Natsik VD, Chishko KA. Cryst Res Tech 1984;19:763.