J Mater Sci (2020) 55:2661–2681

Review
REVIEW

The Hall–Petch and inverse Hall–Petch relations

and the hardness of nanocrystalline metals
Sneha N. Naik1 and Stephen M. Walley1,*

1
SMF Fracture and Shock Physics Group, Cavendish Laboratory, J.J. Thomson Avenue, Cambridge CB3 0HE, UK

Received: 28 June 2019 ABSTRACT

Accepted: 30 October 2019 We review some of the factors that influence the hardness of polycrystalline
Published online: materials with grain sizes less than 1 lm. The fundamental physical mecha-
14 November 2019 nisms that govern the hardness of nanocrystalline materials are discussed. The
recently proposed dislocation curvature model for grain size-dependent
Ó The Author(s) 2019 strengthening and the 60-year-old Hall–Petch relationship are compared. For
grains less than 30 nm in size, there is evidence for a transition from dislocation-
based plasticity to grain boundary sliding, rotation, or diffusion as the main
mechanism responsible for hardness. The evidence surrounding the inverse
Hall–Petch phenomenon is found to be inconclusive due to processing artefacts,
grain growth effects, and errors associated with the conversion of hardness to
yield strength in nanocrystalline materials.

Introduction the resulting imprint area. The indentation process

has been found to depend strongly on indenter
In this review, we focus on how the hardness of geometry, depth of indentation, and specimen size
metals is affected by the interaction between dislo- [1].
cations and grain boundaries, particularly as the In the 1950s, Hall and Petch demonstrated that the
grain size is reduced to the nanometre scale. Hard- yield stress r (Fig. 1) [11, 12] (and hence the hard-
ness (H) has been related to the compressive flow ness) scales with the inverse square root of grain size,
stress of a material by the following relation: d, in polycrystalline materials according to the fol-
H ¼ 3r0:08 ; ð1Þ lowing relation:

where r0:08 is the compressive flow stress of the tes- r ¼ r0 þ kd1=2 ; ð2Þ
ted material at a strain of 8% [1]. Equation (1) and where k is a measure of the local stress needed to
variants of it [2, 3] have been in use for about a initiate plastic flow at a grain boundary and r0 is the
century [4–8]. Hardness is most commonly measured resistance to dislocation motion in the grain interior
through indentation testing [9, 10], which involves [13]. This relationship has been explained by a dis-
pressing a piece of hard material (the indenter) into a location pile-up model for the stress concentration at
test specimen under a known force and measuring

Address correspondence to E-mail: [email protected]

https://doi.org/10.1007/s10853-019-04160-w
2662 J Mater Sci (2020) 55:2661–2681

reported for nanocrystalline materials having typical

grain sizes less than 30 nm: the so-called inverse
Hall–Petch effect [25, 26]. The inverse Hall–Petch
effect has been observed both experimentally [25] as
well as in molecular dynamics (MD) simulations
[27–30]. The weakening has been attributed to (1)
processing artefacts [18, 26], (2) disordered grain
boundaries [31], and (3) the higher percentage of
material in grain boundaries for nano- as opposed to
micro-grained materials [32, 33] coupled with the
intrinsic relative softness of material in grain
boundaries (Fig. 2) [30]. Other studies, such as the
one reported by Conrad [34], considered a possible
explanation for the effect as being the transition from
dislocation controlled to grain boundary and diffu-
sion-dominated deformation [19].
One important problem at present with MD sim-
ulations is that since the cost of computation time
needed to track the motion of atoms in a solid is very
high [35], the timescale that can be investigated is
Figure 1 The relationship between lower yield point (rLYP) and
grain size, d, in mild steel. The yield stress of the single crystal
was obtained from Ref. [16]. From [11].

the tip of a slip band [13, 14], but recently, the validity
of this relationship has been debated [15].
In the 1980s, Gleiter et al. [17] pioneered research
into polycrystalline materials whose grains are of
nanometre size. It was thought then that these
materials would exhibit superior hardness as well as
superior wear resistance and fracture strength com-
pared with their coarse-grained counterparts due to
the large volume fraction of grain boundaries they
contain as grain boundaries were known to govern
the response of metals to deformation [18, 19]. Since
that time, ultrafine-grained materials have been
defined as having grain sizes in the range
100 nm \ d \ 500 nm, and nanocrystalline materials
as having grain sizes less than 100 nm. There have
been reports of nanocomposite coatings with Vickers
microhardness (Hv) of up to * 40 GPa [20], which is
of the same order of magnitude as diamond (Hv *
70–90 GPa) [21]. This ‘super-hardness’ of nanocrys-
talline materials is of interest to the biomedical [22],
military and electronics industries [23, 24].
Although hardness measurements of some
nanocrystalline samples have been reported to be
consistent with the behaviour expected by the Hall– Figure 2 Molecular dynamics simulations of hardness-depth
Petch law, grain boundary weakening has also been relations for a on a grain boundary and b near a grain boundary
for nanocrystalline pure iron. From [30].
J Mater Sci (2020) 55:2661–2681 2663

very short. The consequence of this is that the strain

rate in the simulations is many orders of magnitude
higher than the experiment [29] lying high up in the
shock-loading regime (1010 s-1) whereas the strain
rate for normal indentation experiments is quasistatic
(typically 10-3 s-1). Recent studies by Gurrutxaga-
Lerma and colleagues have shown that the qua-
sistatic theory of dislocations is not valid for shock
plasticity [36, 37]. This is because a quasistatic anal-
ysis ‘ignores the finite time for elastic signals to travel
in the medium’ so that the ‘stresses created by dis-
locations behind the shock front are felt instanta-
neously by [dislocation] sources ahead of the shock
front’ [36]. The practical outcome of applying a qua-
sistatic analysis is that ‘dislocation sources [are]
activated ahead of the shock front’ (Fig. 3), which
does not happen.
In ‘‘The role of dislocations in the deformation of
nanocrystalline materials’’ section of this review, the
evidence for and against key theories that have been
developed to explain the deformation mechanisms
operating in nanocrystalline materials are discussed
along with recent reports claiming the absence of the
Hall–Petch effect in grain size strengthening. In ‘‘The
inverse Hall–Petch phenomenon’’ section, the inverse
Hall–Petch phenomenon is discussed and the mech-
anisms postulated to explain grain size weakening Figure 3 Quasistatic analysis showing unphysical dislocation
are summarised. In ‘‘Synthesis of ‘super-hard’ nucleation ahead of a shock front at a 0.9 ns and b 2 ns. From
[36].
nanocrystalline materials’’ section, the main methods
used to synthesise nanocrystalline materials are
by Eq. (2), the Hall–Petch relation. Figure 4 shows
summarised and the importance of grain boundary
that the hardness of nanometre-sized nickel follows a
structure on the hardness of metals is discussed.
Hall–Petch dependence as the grain size is reduced,
‘‘Summary and conclusions’’ section presents the
supporting the dislocation pile-up theory.
overall conclusions reached.
However, there exists a large body of evidence
suggesting that the Hall–Petch relation is not uni-
versally valid for nanocrystalline materials. Pande
The role of dislocations in the deformation
et al. [41] argued that Eq. (2) is only valid if there are
of nanocrystalline materials a large number of dislocations in a pile-up (Fig. 5)
and that, as grain size decreases, the pile-up mecha-
Extending the classic dislocation pile-up
nism saturates when the number of dislocations in
mechanism
the pile-up tends to 1 [42, 43]. This limit is discussed
Under an applied stress, many dislocation loops are further in ‘‘Expansion of a single dislocation loop
generated in the same glide plane by Frank–Read against the grain boundary resistance’’ section. Addi-
sources [38]. These dislocation loops then accumulate tionally, there is a lack of evidence that directly
against grain boundaries. The shear stress at the head connects the pile-up length to grain size [14, 43].
of these pile-ups increases with their length until the A Bayesian analysis of aggregated Hall–Petch data,
stress exceeds a threshold value, at which point dis- presented by Li et al. [15] and discussed in ‘‘The size
location sources are activated in the adjacent grains, effect’’ section, indicates that the pile-up mechanism
initiating plastic flow. The deformation is described does not account for the wide scatter in the data.
2664 J Mater Sci (2020) 55:2661–2681

r¼r0
þm½s0 þ ð3Gb=4plÞfð5=6Þðln½4l=b1Þ1=16gþsc ;
ð3Þ

where l is the loop diameter (taken to be equal to the

grain diameter), s0 is the multislip shear stress for
deformation within grain volumes, b is the Burgers
vector, G the shear modulus, and m is the Taylor
orientation factor. In constructing this equation, a
term sc (the shear stress required to penetrate through
the grain boundary) was added to the equation of
expansion of a circular dislocation loop [15, 44]. This
theory has been supported by several experimental
studies [45, 46], as shown in Fig. 6.
Figure 4 Plot of hardness against reciprocal square root of grain More commonly as the grain size decreases, a
size for nickel electroplated material with coarse and nano-sized lower strength is observed than the Hall–Petch rela-
grains. Note that in this paper Hughes et al. [39] and Torrents et al. tion predicts. In Fig. 7, data reanalysed by Hansen
[40]. From [13]. and Ralph can be seen to be in agreement with Eq. (2)
for coarse grains. At n = 1 (where n is the number of
Expansion of a single dislocation loop dislocation loops), the pile-up model predicts a
against the grain boundary resistance transition to a higher stress than Hall–Petch [46, 47].
The discontinuity in the prediction stems from the
As the number of dislocations in a pile-up decreases, transition from the Hall–Petch equation (which
their discreteness becomes apparent [15, 31] (see also assumes n is large) to Eq. (3), when n is small [46]. Lu
Fig. 5). The limiting case of a single dislocation loop et al. [48] tested nano-twinned copper, taking the
expanding against a grain boundary is described by twin thickness as the effective grain size. Their data
Eq. (3):

Figure 5 Graph showing that the linear Hall–Petch relationship is Figure 6 Data for electrodeposited nickel, exhibiting stress
valid for a number of dislocations, n, which is larger than 20. The values greater than those estimated from the Hall–Petch relation.
discrete nature of the dislocations in the pile-up is apparent for The two curved lines show the values of stress r calculated using
smaller n, leading to a breakdown in the Hall–Petch model for Eq. (2). The transition between them is marked at n = 1, where n
small pile-up lengths and hence small grain sizes. From [41]. is the number of dislocations in a pile-up. From [46].
J Mater Sci (2020) 55:2661–2681 2665

can be seen in Fig. 7 to initially follow the Hall–Petch (TEM), e.g. Fig. 9 [52]. In this micrograph, the dislo-
relation but with a lower gradient due to the prop- cation density was measured at distances of * 10 lm
erties of coherent twin boundaries. A reversal of the and * 1 lm from the grain boundary in a coarse-
Hall–Petch relation can be seen in their data at grained polycrystal. The dislocation density was
smaller grain sizes, which they ascribed to grain found to be roughly three times greater near the
boundary weakening. Armstrong suggested, how- boundary than in the grain interior for engineering
ever, that this may be an artefact of the preparation of strains of up to 1% in 304 stainless steel [52], aligned
the nano-twinned material [44]. This matter is dis- with the schematic dislocation distribution shown in
cussed further in ‘‘The inverse Hall–Petch phe- Fig. 10.
nomenon’’ section. In conclusion, the data presented Yang and Vehoff [53] studied the influence of grain
in Fig. 7 are not in agreement with the single dislo- size on hardness using nano-indentation and a high-
cation loop model. resolution atomic force microscope. For an indent
depth of 28 nm in ultrafine-grained nickel, they
Work-hardening models found a d-0.5 relation between the applied force and
the individual grain size (Fig. 11). Since the strain is
In these models, dislocations are produced from proportional to the dislocation density [54], the
ledge sources (Fig. 8) [51]. The stress required to pffiffiffi
increase in flow stress is proportional to q: The
move a dislocation through a forest array of extrinsic authors argued that the increase in hardness was due
dislocations is of a form similar to that of the Taylor to an increase in the dislocation density rather than a
equation which describes work hardening [52], decrease in the pile-up length. This observation
pffiffiffi
r ¼ r0 þ aGb q; ð4Þ supports the work-hardening model. Evidence of the
activation of dislocation sources in adjacent grains
where q is the average dislocation density, a is a
was gathered from the analysis of grain size-depen-
property of the material, and r0 is defined in Eq. (2).
dent ‘pop-ins’ (discontinuities in the force–displace-
Li et al [15]. proposed that the density of ledges
ment curve) [55].
scales with grain boundary area per unit volume of
In Fig. 12, the first pop-in corresponds to the initial
material. Their idea implies that fine-grained mate-
yield point of nickel. The later pop-ins were taken by
rials have a greater dislocation density and longer
Yang and Yehoff as evidence for the activation of
dislocation lines when they yield. Ledges have been
sources in grains adjacent to the indented grains. The
imaged using transmission electron microscopy
figure also shows the following: (a) the first pop-ins
(at the initial yield point) occurred at forces and
displacements that were independent of the grain
size; (b) for later pop-ins, the force increased and the
displacement decreased with decreasing grain size.
Since the pile-up length L is related to the number of
dislocations n by
L ¼ D=2 ¼ 2nA=s; ð5Þ
where D is the grain size, s is the external stress, and
A is a constant. If the stress for activation of sources
in adjacent grains is constant (as assumed by Hall–
Petch theory), the pop-in load must be higher for
smaller grains. This is supported by the experimental
data shown in Fig. 12.
The authors conclude that in ultrafine-grained
nickel, hardness scales with dislocation density
Figure 7 Log/log stress-grain size graph for copper. The black
(rather than pile-up length) where the pile-up length
line shows the theoretical values calculated using Eq. (2) for larger
is within the grain size (supporting the Orowan
grain sizes and Eq. (3) for smaller grain sizes. Note that in this
paper Lu et al. [48], Li and Liu [47], Armstrong and Smith [49], model). However, they also found clear evidence for
and Hansen and Ralph [50]. From [46]. dislocation source activation in adjacent grains, and
2666 J Mater Sci (2020) 55:2661–2681

Figure 8 Schematic diagram

of the grain boundary ledge
model. From [15].

Figure 9 Bright-field TEM image of 304 stainless steel at a strain

of 0.8%. Grain boundary ledges are visible. From [52].

hence, a higher external load is needed to nucleate

dislocation sources in adjacent grains for smaller Figure 10 Schematic diagram of the distribution of dislocations
grain sizes. produced by ledge sources in fine-grained polycrystalline
Meyers et al. [18] argued against this idea, pointing materials. From [52].
out that when the spacing of ledges is in the
nanometre range, there would be a grain size below grain boundary ledges (which are affected by grain
which this deformation mechanism is no longer size) to the density of the dislocations produced.
operational. We would postulate that for grains a few
nanometres in diameter, the grain boundaries may no The core and mantle model
longer be sharp and therefore point defects could be
seen rather than the ledges that can be seen in Fig. 9. According to several other models, including those
Cordero et al. [14] also claimed in their study that due to Meyers and Ashworth [56] and Raj and Ashby
there was no direct evidence that links the density of [57], a grain can be treated as a composite, which is
J Mater Sci (2020) 55:2661–2681 2667

Figure 12 Graphs showing the forces and displacements at which

pop-ins occur for nano-indentation on nano-nickel. From [53].

Figure 11 Nano-indentation of ultrafine-grained nickel. a A

force–displacement curve with increasing indentation depth. The
pop-in indicates the activation of dislocation sources in adjacent
grains. b Variation in applied force with grain size for an
indentation depth of 28 nm. An inverse square root relationship is
observed. From [53].

an improvement on ‘rule of mixtures’ models [58].

The composite model consists of the grain interior
(the ‘core’), which is under homogeneous stress, and
a work-hardened layer (the ‘mantle’) in which
impurity segregation may occur adjacent to grain
boundaries (Fig. 13).
The predicted stress–grain size relationship for the
model shown schematically in Fig. 13 is given by the
following equation [18]:
 
ry ¼ rfG þ 8kMA rfgb  rfG d0:5 Figure 13 Simplified schematic of the Meyers–Ashworth model.
 
 16k2MA rfgb  rfG d1 ; ð6Þ From [56, 59].

where rfG is the flow stress of the dislocation-free micron-sized grains, but it predicts a reduction in
interior, rfgb is the flow stress of the grain boundary slope for smaller grain sizes. Figure 14 appears to
region, and kMA is a fitted parameter. Equation (6) is show an agreement between experimental data and
in agreement with the Hall–Petch dependence for
2668 J Mater Sci (2020) 55:2661–2681

the Meyers–Ashworth model. However, Li et al. [15]

recently concluded after analysis of a larger body of
data that there is little correlation between experi-
mental data and elastic anisotropy models such as
those discussed above.
In the Ashby plastic anisotropy model [57], defor-
mation takes place in two stages [14, 60]. Firstly, grain
boundary shearing occurs along glide planes. This
leads to voids and overlaps between grains (Fig. 15).
Then, to restore compatibility at grain boundaries,
Figure 15 Schematic diagram for elastic and plastic
arrays of ‘geometrically necessary’ dislocations
accommodation models, such as the Meyers–Ashworth and
(GNDs) are generated from the grain boundaries. The
Ashby models. From [15].
deformation is described by
rffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi be relations. They attributed the lack of a clear parabolic
r ¼ r0 þ aGb qGN þ qSS  r0 þ aGb ; ð7Þ dependence to their invalid assumption of small
4D
strain in the historic studies they examined. Cordero
where qSS refers to ‘statistically stored’ dislocations
et al. ascribed the small number of cases where k did
(as would build up in a uniformly strained single
not increase either to sample processing effects or to
crystal) and qGN refers to GNDs. The approximation
the effects of twinning (as opposed to glide) as a
shown in Eq. (7) is valid in the limit of small strains
deformation mechanism. Despite this, Cordero et al.
where qGN  qSS. The model predicts the following
[14] argued that Ashby’s model is the most consistent
relationship between the Hall–Petch coefficient k and
pffiffi overall with their examination of the literature on the
the plastic strain e namely k / e in the limit of small
strain dependence of the Hall–Petch coefficient and
plastic strains [14].
with experimental observations of dislocation sub-
Figure 16, taken from Cordero et al. [14], is a
structure. However, against this Li et al. [15] argued
summary of six decades of investigations of the Hall–
that the Ashby model is not consistent with the
Petch effect in which k was measured as a function of
experimental data they obtained.
plastic strain. Cordero et al. note that a parabolic
strain dependence was not found experimentally. The size effect
Rather in the majority of cases, although k increases
with strain, it did so according to a number of other Recently Li et al. [15] conducted a Bayesian meta-
analysis on the body of available Hall–Petch data.
They concluded that there was no experimental evi-
dence for the 60-year-old Hall–Petch effect (Fig. 17).
The authors made use of Matthews critical thickness
theory [61] for thin metallic multilayers (Fig. 18a), to
derive a relation by which the grain diameter is
inversely proportional to dislocation curvature and
hence to stress (Fig. 18b). This dictates the minimum
strength for dislocation plasticity. For nano-grained
materials, other mechanisms (discussed in Table 1)
can result in data points that lie in the ‘no data’
region of the log(strength) - log(size) region of
Fig. 17. These data points fall below the minimum
strength predicted by the size effect equation (which
is based on a dislocation curvature model). This is
Figure 14 Plot of yield stress, ry, against the inverse root of the due to the onset of grain boundary sliding in some
grain size (D-0.5) for iron. The dotted line shows the ry values nano-grained materials and the transition away from
estimated using Eq. (5). For details of the papers from which the dislocation-based plasticity. Li et al. concluded that
data in this plot were obtained, consult Ref. [59].
J Mater Sci (2020) 55:2661–2681 2669

defined as the tendency on the part of researchers to

publish articles based on the perceived strength of
the findings of a particular study [63]. So for the
phenomenon under review in the present article,
sampling is skewed towards results that are in
agreement with the long-standing Hall–Petch rela-
tionship. Additionally, Li et al. assumed that since
there was no underlying physics governing the dis-
tribution of k and r0, their values were uniformly
distributed in log r - log d space (Fig. 17), in
agreement with Benford’s Law [64, 65]. Benford’s law
is applicable to large data sets where the data points
come from many different distributions that span
several orders of magnitude [66, 67]. The law states
that the probability that the first digit of a number is p
is given by
 
1
PbenðpÞ ¼ log 1 þ p ¼ 1; 2. . .; 9; ð9Þ
p
although its physical significance is not well under-
stood [68]. The validity of this assumption requires
further investigation in light of the behaviour of k in
Fig. 16 and the higher k values reported for fcc metals
as compared to bcc metals [69].
Bayesian updating resulted in On = 2nP0, where On
are the odds that Eq. (8) is true and n are the number
of datasets that fall above the 1/d line (Fig. 19).
Therefore, even with a low prior probability P0 that
such a well-established equation is incorrect, the
odds were found to be overwhelmingly in favour of
Eq. (8). Hence, Li et al. [15] argued that the grain size
strengthening of metals is driven by constraints on
the dislocation curvature and therefore that the pile-
up, grain boundary ledges, and core and mantle
Figure 16 The strain dependence of the Hall–Petch coefficient models make a much weaker contribution to grain
for several fcc, bcc, and hcp metals. In the majority of cases, k can size strengthening than the dislocation curvature.
be seen to increase with strain. From [14].

klnd The inverse Hall–Petch phenomenon

e ð dÞ ¼ e 0 þ ; ð8Þ
d
Since Chokshi et al. first reported a negative slope in
where k * 1 and a variable e0 best described the
a Hall–Petch plot for nanocrystalline copper and
data and where e = r/Y is the stress normalised by
palladium in 1989 (Fig. 20) [25], there have been
the elastic modulus. They also postulated that a
several reports of grain size softening and hardening
random error in grain size determination explains the
in the grain size range below 100 nm (Fig. 21). One of
apparent agreement with Eq. (2).
the first models to capture the Hall–Petch transition
Li et al. [15] claimed that their data were not subject
theoretically for grain sizes less than 100 nm was
to any sampling bias, as all available data from the
published by Konstantinidis and Aifantis [70].
published studies were considered. However, the
Although Chokshi et al. attributed the softening
published data are subject to publication bias,
effect to the onset of Coble creep [71], researchers
2670 J Mater Sci (2020) 55:2661–2681

Mechanisms governing grain size

weakening

Several theoretical models that have been proposed

over the years to explain grain size softening are
summarised in Table 1. The articles referred to in that
table should be consulted for more information on
Figure 17 Graphical representation of the distribution of
the derivation of and evidence that supports each
expected data if Eq. (8) holds. The absence of experimental data theory (Figs. 22, 23, 24).
below the minimum strength strongly suggests that grain Zhang and Aifantis [79] built on the individual
boundary effects are an extension of the ‘smaller is stronger’ mechanisms shown in Table 1. They examined the
size effect seen in micromechanical testing. If the values of k and mechanical grain boundary energy in order to
r0 are uniform in log r - log d space, there is a relative explain both the conventional and inverse Hall–Petch
probability density of 2 above the line and 0 below. Figure from relation for numerous experimental investigations.
the graphical abstract of [62]. They also provided a theoretical expression to predict
the grain size at which the transition occurs from
strengthening to weakening with decreasing grain
have since suggested several other theories to explain size. They used a gradient plasticity framework to
it including flaws in the synthesis of the nanocrys- capture the softening behaviour by treating the grain
tallites [26], the presence of disordered grain boundaries as a separate phase with a finite thickness
boundaries [31], or a transition to and from disloca- rather than as a surface as is normally done for larger
tion-based deformation to grain boundary sliding or grain sizes. The grain structure assumed is shown in
rotation [72–74]. Fig. 25 where it can be seen that they identified a
Koch et al. argued [26] that incomplete densifica- ‘grain interior’ (GI) phase (which is purely elastic), a
tion during synthesis of nanocrystallites via inert gas ‘grain boundary’ (GB) phase of thickness Lgb (which
condensation (as employed by Chokshi et al. [25]) is assumed to be soft and prone to deformation
and ball milling methods [75] can lead to residual through rotation and sliding), and a plastic ‘GI–GB’
porosity in nano-grained materials and consequently phase of thickness Lg adjacent to the grain boundary
to poor bonding between particles, resulting in a which accounts for the transition from the ductile GB
decrease in the strength of these materials [26]. to the rigid GI phase due to the limited diffusion into
Armstrong argued that a reduction in Hall–Petch the grain interior of dislocations and disclination
slope could be caused by the presence of disordered dipoles generated at the grain boundary.
grain boundaries in nanocrystalline materials, which Zhang and Aifantis’ [79] gradient plasticity model
would allow plastic flow to be transmitted more includes an interface energy term cgb which allows
easily between grains [31].
the interface itself to follow its own yield behaviour.

Figure 18 Schematic diagrams of the size effect. a An illustration substrate. b A Frank–Read source operating inside a grain. A
of Matthews critical thickness concept [61] for a spiral and Frank– smaller grain diameter requires greater dislocation curvature. From
Read dislocation source [38] in a strained epitaxial layer on a [15].
J Mater Sci (2020) 55:2661–2681 2671

Table 1 Summary of proposed mechanisms responsible for the grain size weakening effect

Mechanism Mechanism origin Governing equation References

name
 
Grain Independent atomic shear events at the grain s  s0 ¼ kT dvd
þ DF kT [18, 34]
V ln c_ V þ V lnd
boundary boundary. Thermally activated shear. Does not
where DF is the Helmholtz free energy, d is the
sliding account for compatibility of deformation
grain boundary width, vd is the Debye frequency
and V = b3 is the activation volume (b being the
Burgers vector)
   
Grain boundary shear dominates over dislocation c_ ¼ 3d 3d _
d cgb þ 1  d cD
_ [76, 77]
plasticity as volume fraction of grain boundary where 3d
d is the volume fraction of the grain
increases. Predicts a ‘strongest size.’ Assumes boundary region
dislocations are emitted from triple grain
boundary junctions to satisfy compatibility.
Grain boundary sliding described in terms of a s0 ¼ ðgi þ gD Þc_ þ sp [18, 59]
viscous and a plastic accommodation term sp . where s is the shear stress, gi is the intrinsic
Grain boundary sliding accounts for a third of the component of grain boundary viscosity, and gD is
behaviour (see Fig. 22) the diffusional component
1
Accommodation between adjacent grains through c_ ¼ 64dXD
kT
B
s
d3 0
[18, 57]
diffusional creep where X is the atomic volume and DB is the
boundary diffusion coefficient
Grain Competition between lattice dislocation slip and s ¼ s0 þ kd0:5 þ k1 þ Ad þ Bd3 [19, 25, 43, 76]
boundary Coble creep mechanisms (see Fig. 23) kðd Þ0:5 ¼ dA þ Bðd Þ3
diffusion
where k1, A, B are constants and d* is the critical
grain size at which the classical Hall–Petch
mechanism switches to Coble creep
0:4ad
Grain Grain rotation and translation through motion of s ¼ 2pðGb
1mÞd ln b
[72–74]
rotation dislocation quadrupoles and dislocation dipoles where G is the shear modulus, m is the Poisson
(see Fig. 24a, b) ration, a is the dislocation core parameter (which
varies between 1 and 4 for metals), and b is the
Burgers vector
Amorphous Transition to glasslike deformation behaviour. Rate [78]
limit and pressure sensitivity of nano-grained materials
are characteristic of amorphous solids. For the
smallest grain size, unstable localised plasticity
occurs (shear banding)

cgb is positive for microscopic grains as grain k cgb

r ¼ r0 þ pffiffiffi þ ; ð10Þ
boundaries inhibit plastic flow (the yield stress of the d 2ad
GB phase is greater than the yield stress of the GI
where the grain boundary thickness is some fraction
phase), whereas for nanometre-sized grains, cgb is
of the grain size (i.e. Lg = ad, where d is the grain size
negative because the grain boundaries behave plas- and a is a constant with a value lying between 0 and
tically and are softer than the grain interior (the yield 1) and r0 is defined in Eq. (2). Equation (10) was used
stress of the GB phase is less than the yield stress of by Zhang and Aifantis to analyse the data published
the GI phase). by a number of authors (see Fig. 26). As the fig-
The stress of the unit cell shown in Fig. 25 upon ure shows, it provides a good fit for seven different
yielding of the grain boundary and adjacent grain nanocrystalline metals and alloys.
boundary layers is given by However, it was pointed out by Zhang and
Aifantis that the processing methods were not the
2672 J Mater Sci (2020) 55:2661–2681

that they used Tabor’s hardness-yield stress relation

to compare experimental data which could poten-
tially introduce errors for nano-grained materials
(this matter is discussed further in the ‘‘Discussion of
the evidence’’ section).
The critical grain size at which material behaviour
transitions from the ‘normal’ Hall–Petch to the
inverse Hall–Petch relation can be calculated using
Eq. 11 which gives the grain size dc at which peak
material strength occurs. This critical size can be
directly computed if the GB energy cgb , the fraction a
of the GB thickness that yields and the Hall–Petch
coefficient k are known for the material.
 2
dc ¼ cgb =ðakÞ : ð11Þ

There is currently no general consensus on the

mechanism that gives rise to the inverse Hall–Petch
Figure 19 Comparison of the pile-up and grain boundary ledge
model with aggregated Hall–Petch data. The heavy black line behaviour exhibited by some nano-grained materials
shows the predictions of the pile-up model calculated using as a number of different theories including Coble
Eq. (2). The dark blue dashed lines show the range of theoretical creep, grain rotation, and gradient plasticity have
values calculated on the basis of the grain boundary ledge model some experimental support. The experimental evi-
described by Eq. (4). The red chain dotted lines show the range of dence used to support the theories discussed in this
predictions of the slip-distance model, which is not evaluated in section is considered in ‘‘Discussion of the evidence’’
this review paper. The intensity of the shading corresponds to the section.
probability of finding data in each region according to the models
discussed where white represents a probability close to zero. From Discussion of the evidence
[15].
Due to the difficulty in obtaining bulk defect-free
nanocrystalline specimens for testing, there is a
dearth of reliable strength and hardness measure-
ments for materials with nano-sized grains [14].
Stress-induced coarsening, due to either grain
boundary migration or grain rotation, can occur
during indentation testing of nanocrystalline materi-
als (Fig. 27), which can introduce errors into the
experimental estimation of hardness.
Brooks et al. [2] argued that Eq. (1) (Tabor’s classic
hardness-yield stress formula) overestimates the
yield stress in electrodeposited nanocrystalline
materials. They reported that the ratio of hardness to
yield stress lay between 4 and 8.6 for the materials
they studied rather than 3 (see Fig. 28a). However,
Figure 20 The first experimental results indicating a negative Zhang et al. [3] estimated, more conservatively, that
Hall–Petch slope for nanocrystalline copper and palladium at the ratio of hardness to yield stress was between 2.3
room temperature. From [25]. and 3.7 for a number of nanocrystalline copper and
copper-zinc alloys (Fig. 28b).
same for the various experimental data sets they
The Tabor relation is widely used in papers that
compared in Fig. 26 (see ‘‘Synthesis of ‘super-hard’
argue for and against the inverse Hall–Petch effect in
nanocrystalline materials’’ section of our review for a
nanocrystalline materials (e.g. [14, 79]). For example,
discussion about synthesis methods). We also note
Cordero et al. used Eq. (1) to convert nano-
J Mater Sci (2020) 55:2661–2681 2673

Figure 21 Plot of data from a

number of publications
showing the range of grain
size strengthening and
weakening behaviour for
nanocrystalline copper. For
details of the papers from
which the data in this plot
were obtained, consult Ref.
[18].

Figure 22 Computed stress against plastic strain for nano-grained Figure 23 Model for grain size dependence of shear stress (s). At
copper with a grain size of 20 nm. The solid line shows perfect large grain sizes, the Hall–Petch relationship holds. At d*, a
grain boundary bonding. The dashed line is computed assuming transition occurs at maximum strength. For d \ d*, Coble creep
that the boundaries have ggb = 0, so that free slip can occur. The diffusion dominates. In reality, there is competition between the
strength for free slip is two-thirds that of the equivalent material two mechanisms due to the grain size distribution in
with perfect grain boundaries, therefore showing that grain nanocrystalline materials. From [19].
boundary sliding accounts for one-third of the resulting stress
whereas plastic deformation within grains accounts for the Molecular dynamics (MD) simulations [27, 84, 85]
remaining two-thirds. From [59]. have predicted a peak in hardness for copper with
grain sizes in the range 10 nm \ d \ 15 nm. The
indentation hardness measurements into yield simulations also support the existence of the inverse
strengths (Fig. 29). They then plotted this data Hall–Petch slope and deformation via grain bound-
alongside yield strengths measured using compres- ary slip. Although MD simulations allow researchers
sion or tension tests. Given the results of Brooks et al. to directly model atoms and investigate grain
[2] and Zhang and Aifantis [79], this methodology boundary structure for grains less than 10 nm in size
should be used with caution since as just discussed [86], the simulated strain rates are so high as to be
the classic Tabor relation appears not to hold for all inaccessible experimentally [87] (see the discussion of
nanocrystalline materials. Fig. 3 in ‘‘Introduction’’ section). Also due to
2674 J Mater Sci (2020) 55:2661–2681

Figure 24 Rotational deformation via the motion of disclinations. disclinations (labelled II) are rotated relative to those outside the
a Drawing showing the distortion of {110} planes seen in a high- disclination dipole (labelled III). From [72]. b A schematic of a
resolution TEM image of mechanically milled, nano-grained iron nanocrystalline solid under mechanical tension. Motion of
powder. It shows direct observation of partial disclination dipoles. disclination dipoles (triangles) results in rotational deformation.
The set of terminating planes making up an individual partial The non-parallel lines in the inset represent the disclinations. From
disclination is labelled I. The planes located between two partial [73].

Figure 25 Unit cell model

with grain boundary (GB)
phase, grain interior (GI) phase
comprised of GI-GB layers,
and elastic GI cores. From
[79].

For all the reasons mentioned above, the experi-

mental evidence currently available for the inverse
Hall–Petch relationship is inconclusive. So in order to
prove the existence of the inverse Hall–Petch effect,
and the mechanisms behind it, many more experi-
mental studies need to be performed in which (1)
consistent material processing methods are used, (2)
direct yield stress measurements are made (rather
than assuming the Tabor hardness-yield stress rela-
tion), and (3) attention is given to grain size coars-
ening effects during testing. If these investigations
are carried out, the uncontrollable factors in the
experiments performed so far will be minimised and
Figure 26 Plots that fit Eq. (13) to experimental data for reliable data will be generated for materials with
nanocrystalline metals and alloys published by a number of typical grain sizes in the nanometre range.
different authors. For details of the papers from which this data
came from, consult [79].

computational limitations, simulations cannot handle

samples larger than a few hundreds of nanometres in
size and therefore cannot be related simply to
macroscopic experiments [84].
J Mater Sci (2020) 55:2661–2681 2675

Figure 27 Plot of hardness (calculated using Tabor’s formula

H = 3rY) against the dwell time of an indenter in the sample. The
grain growth zone is assumed to spread from beneath the indenter
and eventually occupy the whole plastic zone. tf is the time taken
for the volume fraction of the grain growth zone to increase
linearly from 0 to 100% relative to the plastic zone. The graph
shows that the calculated hardness decreases with time, which is
consistent with the experimental data. The decrease in hardness
with dwell time is monotonic and does not saturate as time is
increased. The paper referred to in this figure as Zhang et al. is
Ref. [80]. From [81].

Synthesis of ‘super-hard’ nanocrystalline

materials
Figure 28 Plots showing the results of an investigation of the
The nature of the grain boundaries within relationship between hardness and yield strength in
nanocrystalline materials is influenced by the syn- nanocrystalline materials. a Graph showing the overestimation
thesis method used (a summary may be found in of the yield strength predicted using Eq. (1) compared to
experimental data for nanocrystalline materials from Cahoon
Table 2) and can be modified to cause ‘super’-hard-
et al. [82] and Gao [83]. Figure from [2]. b Graph showing the
ness. Severe plastic deformation (SPD) has produced
ratio of hardness to yield strength for copper and copper-zinc
materials which have a hardness greater than the alloys with grain sizes in the range 34 \ d \ 200 nm, subjected to
extrapolated Hall–Petch relationship [22]. Balasubra- different pre-treatments such as equal channel angular pressing
manian and Langdon [22] attributed this to three (ECAP), high-pressure torsion (HPT) and annealing. Figure from
main effects: [3].

A larger fraction of high-angled grain Segregation of impurity and alloying

boundaries (HAGBs) elements at grain boundaries

HAGBs are more effective in impeding dislocation Precipitation of alloying elements in grain boundary
slip as there is greater crystallographic misalignment regions suppresses the emission of dislocations from
across the grain boundary. The fraction of HAGBs grain boundaries. Additionally, the precipitates cause
can be increased from 55 to 80% by increasing the drag on GNDs [22, 89].
number of high-pressure torsion (HPT) turns from
one half to ten [88].
2676 J Mater Sci (2020) 55:2661–2681

Figure 29 Aggregated Hall–Petch data for body centred cubic correspond to yield strengths measured by compression or tensile
(bcc) metals. The red dashed lines are best fits to the data using testing. Cordero et al. ascribed the scatter seen to differences in
Eq. (2). The closed points are Vickers or nano-indentation sample preparation and testing, rather than an indication that
hardness measurements converted using Eq. (1). The open points Eq. (2) may not be valid. From [14].

Non-equilibrium grain boundaries (NGBs) grain boundaries, causing an increase in grain

boundary energy [22, 89].
SPD produces more dislocations than geometrically Hu et al. [88] recently showed that careful use of
necessary to accommodate plastic deformation at annealing can result in the doubling of hardness of
J Mater Sci (2020) 55:2661–2681 2677

Table 2 Summary of common synthesis methods for nanocrystalline materials

Mechanism Description of process(es) Sample characteristics

Inert gas Metal is evaporated, condensed into a fine powder, and compacted Porosity. Poor bonding between
condensation particles [18]
Mechanical alloying Powder particles are repeatedly ground in a dry, high energy mill Porosity [26]
Electrodeposition A current is pulsed to deposit metal cations in crystalline and amorphous Low porosity. Improved ductility due
regions to growth twins [18]
Crystallisation from Heat treatments crystallise metallic glasses into nano-polycrystalline Residual amorphous regions can
amorphous solids solids remain [26]
Severe plastic Two main methods: equal channel angular pressing (ECAP) and high- High proportion of HAGBs, NGBs.
deformation (SPD) pressure torsion (HPT). The sample of subjected to large plastic strains Impurity segregation resulting in
to break down the microstructure ‘super-hardness’ [22]

nano-grained nickel and nickel-molybdenum without material. Figure 28b, c could suggest that while a
altering the grain size (Fig. 28a). They found that the combination of relaxation and segregation at grain
indentation produced little coarsening for their boundaries plays a role in the reversal of the inverse
annealed samples. They thus concluded that struc- Hall–Petch behaviour seen upon annealing, the
tural relaxation and segregation of the molybdenum dominant mechanism is in fact molybdenum segre-
in the alloy causes relaxation of local stress levels at gation since they reported a much greater increase in
grain boundaries, which then become more stable to peak hardness for their nickel-molybdenum samples
straining. This could reduce the threat of grain compared to the pure nano-grained nickel they tested
coarsening to the refinement process. (Fig. 30).
They also argued that grain boundary mediated
deformation (which can cause softening) is replaced
by deformation by the generation of extended partial Summary and conclusions
dislocations at grain boundaries. The emission of
partials is suppressed due to impurity segregation, This article has reviewed the hardness of nanocrys-
similar to the suggestions by Valiev mentioned above talline metals, focusing on the theories describing
[89], enhancing the formation of extended stacking dislocation plasticity, grain size weakening, and
faults. The large stresses required for nucleation of super-hardness effects. The main conclusions
dislocations from stable grain boundaries results in a reached are outlined below.
high hardness and a (1/d) grain size dependence. Hu
et al. [88] argued that differences in grain boundary Deformation mechanisms
structure can explain the controversy over hardening
100 nm \ d \ 1 lm: Core and mantle type models
and softening behaviour reported with decreasing
best describe the deformation behaviour.
grain size in previous studies. Their results could
30 nm \ d \ 100 nm: Dislocation ledge spacing
lead to the synthesis of further ‘super’-hard materials.
becomes large compared to the grain size; therefore,
They showed that the inverse Hall–Petch effect was
there is a transition from a dislocation-based plastic-
eliminated by annealing their samples. If this result
ity to grain boundary sliding as the main mechanism
was only due to structural grain boundary relaxation,
responsible for hardness. There is a dearth of reliable
then this could imply that the inverse Hall–Petch
hardness measurements in this grain size range, and
effect is simply a result of processing defects. How-
therefore, the main accommodation mechanism can-
ever, if impurity segregation after annealing was the
not be distinguished.
dominant mechanism for the hardening behaviour
d \ 30 nm: Transition from nanocrystalline to
they saw, then the inverse Hall–Petch relation (gov-
amorphous behaviour.
erned by the mechanisms discussed in ‘‘The inverse
Hall–Petch phenomenon’’ section) would still be
valid for a pure single-phase nanocrystalline
2678 J Mater Sci (2020) 55:2661–2681

b Figure 30 a Increased hardness for annealed nickel and nickel-

molybdenum nanocrystalline samples. The open symbols show
as-deposited samples, and the closed symbols are samples that
have been annealed. The blue solid line follows the relation
pffiffiffi
Hv  1= d and the red line follows Hv * 1/d. b Plot of the
variation in microhardness as a function of annealing temperature
for nano-grained nickel and nickel-molybdenum with varying
concentrations of molybdenum. c Plot of the variation in
maximum microhardness increment produced by annealing as
function of initial grain size. Open symbols represent data for
other types of nano-grained materials. From [88].

The relationship between hardness

and grain size

Although Li et al.’s [15] arguments for a (log d)/d

relationship are compelling, they require further
statistical analysis and corroboration in order to
overturn the large body of evidence that supports the
d0:5 relationship, which has been added to recently
by Armstrong [13] and Cordero et al. [14].

The inverse Hall–Petch effect

A transition from dislocation-based plasticity to a

grain boundary sliding mechanism could explain the
reversal in Hall–Petch slope. This transition has been
seen to occur at grain sizes from around 100 nm [18]
down to 10 nm [27, 84, 85, 88]. An analytical
expression to predict the theoretical critical grain size
was devised by Zhang and Aifantis on the basis of
the grain boundary plasticity theory [79]. However,
the inverse Hall–Petch effect could also result from
processing artefacts or stress-induced grain growth
during testing. Based on the available experimental
evidence, the existence of the inverse Hall–Petch
effect cannot be confirmed.

Processing methods

Inert gas condensation and mechanical alloying can

result in grain size weakening due to incomplete
densification resulting in porosity. Severe plastic
deformation can result in HAGBs, segregation of
alloying elements, and NGBs which produce ‘super-
hardness.’ Short annealing treatments have recently
been used to increase the hardness of nickel-molyb-
denum alloys by up to 120% [88] by reducing the
local stress levels at grain boundaries.
J Mater Sci (2020) 55:2661–2681 2679

Acknowledgements [11] Hall EO (1951) The deformation and ageing of mild steel. 3:
discussion of results. Proc Phys Soc Lond B 64:747–753
The authors thank the referees for making useful [12] Petch NJ (1953) The cleavage strength of polycrystals. J Iron
suggestions for improving the paper. Steel Inst 174:25–28
[13] Armstrong RW (2014) 60 years of Hall–Petch: past to pre-
Compliance with ethical standards sent nanoscale connections. Mater Trans 55:2–12
[14] Cordero ZC, Knight BE, Schuh CA (2016) Six decades of
Conflict of interest The authors declare that they
the Hall–Petch effect: a survey of grain-size strengthening
have no conflict of interest.
studies on pure metals. Int Mater Rev 61:495–512
Open Access This article is distributed under the [15] Li Y, Bushby AJ, Dunstan DJ (2016) The Hall–Petch effect
terms of the Creative Commons Attribution 4.0 as a manifestation of the general size effect. Proc R Soc A
International License (http://creativecommons.org/ 472:20150890
licenses/by/4.0/), which permits unrestricted use, [16] Holden AN, Hollomon JH (1949) Homogeneous yielding of
carburized and nitrided single iron crystals. J Met 1:179–185
distribution, and reproduction in any medium, pro-
[17] Gleiter H (1989) Nanocrystalline materials. Prog Mater Sci
vided you give appropriate credit to the original
33:223–315
author(s) and the source, provide a link to the Crea-
[18] Meyers MA, Mishra A, Benson DJ (2006) Mechanical
tive Commons license, and indicate if changes were
properties of nanocrystalline materials. Prog Mater Sci
made.
51:427–556
[19] Pande CS, Cooper KP (2009) Nanomechanics of Hall–Petch
References relationship in nanocrystalline materials. Prog Mater Sci
54:689–706
[1] Gouldstone A, Chollacoop N, Dao M, Li J, Minor AM, Shen [20] Veprek S, Mukherjee S, Karvankova P, Mannling HD, He
Y-L (2007) Indentation across size scales and disciplines: JL, Moto K, Prochazka J, Argon AS (2003) Limits to the
recent developments in experimentation and modeling. Acta strength of super- and ultrahard nanocomposite coatings.
Mater 55:4015–4039 J Vac Sci Technol, A 21:532–544
[2] Brooks I, Lin P, Palumbo G, Hibbard GD, Erb U (2008) [21] Brookes CA (1992) Indentation hardness, plasticity and
Analysis of hardness-tensile strength relationships for elec- creep of diamond. In: Field JE (ed) The properties of natural
troformed nanocrystalline materials. Mater Sci Eng, A and synthetic diamond. Academic, London, pp 515–546
491:412–419 [22] Balasubramanian N, Langdon TG (2016) The strength-grain
[3] Zhang P, Li SX, Zhang ZF (2011) General relationship size relationship in ultrafine-grained metals. Metall Mater
between strength and hardness. Mater Sci Eng, A 529:62–73 Trans A 47:5827–5838
[4] Unwin WC, Longridge M (1918) The experimental study of [23] Koch CC (2007) Structural nanocrystalline materials: an
the mechanical properties of materials. Proc Inst Mech Eng overview. J Mater Sci 42:1403–1414. https://doi.org/10.100
95:405–439 7/s10853-006-0609-3
[5] Norbury AL, Samuel T (1924) Experiments on the Brinell– [24] Koch CC, Ovidko IA, Seal S, Veprek S (2007) Structural
tensile relationship. J Iron Steel Inst 109:479–491 nanocrystalline materials: fundamentals and applications.
[6] Greaves RH, Jones JA (1926) The ratio of the tensile Cambridge University Press, Cambridge
strength of steel to the Brinell hardness number. J Iron Steel [25] Chokshi AH, Rosen A, Karch J, Gleiter H (1989) On the
Inst 113:335–353 validity of the Hall–Petch relationship in nanocrystalline
[7] Ruffle TW (1944) Relationship of Brinell hardness and yield materials. Scr Metall Mater 23:1679–1684
stress in certain cast steels. Proc Inst Br Foundrym 37:A117– [26] Koch CC, Narayan J (2000) The inverse Hall–Petch effect:
A121 fact or artifact? Mater Res Soc Symp Proc 634:paper B.5.1
[8] Tabor D (1951) The hardness of metals. Clarendon Press, [27] Schiøtz J, Jacobsen KW (2003) A maximum in the strength
Oxford of nanocrystalline copper. Science 301:1357–1359
[9] Walley SM (2012) Historical origins of indentation hardness [28] Huang C, Peng XH, Zhao YB, Weng SY, Yang B, Fu T
testing. Mater Sci Technol 28:1028–1044 (2018) Flow strength limit of nanocrystalline tantalum pre-
[10] Walley SM (2013) Addendum and correction to ‘Historical dicted with molecular dynamics simulations. Mater Sci Eng,
origins of indentation hardness testing’. Mater Sci Technol A 738:1–9
29:1148
2680 J Mater Sci (2020) 55:2661–2681

[29] Xu WW, Davila LP (2018) Tensile nanomechanics and the [44] Armstrong RW (2013) Hall–Petch analysis of dislocation
Hall–Petch effect in nanocrystalline aluminium. Mater Sci pileups in thin material layers and in nanopolycrystals.
Eng, A 710:413–418 J Mater Res 28:1792–1798
[30] Kuhr BR, Aifantis KE (2019) Interpreting the inverse Hall– [45] Choi HJ, Lee SW, Park JS, Bae DH (2009) Positive devia-
Petch relationship and capturing segregation hardening by tion from a Hall–Petch relation in nanocrystalline aluminum.
measuring the grain boundary yield stress through MD Mater Trans 50:640–643
indentation. Mater Sci Eng, A 745:107–114 [46] Armstrong RW (2016) Hall–Petch description of nanopoly-
[31] Armstrong RW (2016) Dislocation pile-ups, material crystalline copper, nickel and aluminum strength levels and
strength levels, and thermal activation. Metall Mater Trans A strain rate sensitivities. Philos Mag 96:3097–3108
47:5801–5810 [47] Li JCM, Liu GCT (1967) Circular dislocation pileups. 1:
[32] Carsley JE, Ning J, Milligan WW, Hackney SA, Aifantis EC strength of ultrafine polycrystalline aggregates. Philos Mag
(1995) A simple mixtures-based model for the grain size 15:1059–1063
dependence of strength in nanophase metals. Nanostruct [48] Lu L, Chen X, Huang X, Lu K (2009) Revealing the max-
Mater 5:441–448 imum strength in nanotwinned copper. Science 323:607–610
[33] Carsley JE, Fisher A, Milligan WW, Aifantis EC (1998) [49] Armstrong RW, Smith TR (1996) Dislocation pile-up pre-
Mechanical behavior of a bulk nanostructured iron alloy. dictions for the strength properties of ultrafine grain size fcc
Metall Mater Trans A 29:2261–2271 metals. In: Suryanarayana C, Singh J, Froes FH (eds) Pro-
[34] Conrad H (2004) Grain-size dependence of the flow stress of cessing and properties of nanocrystalline materials. The
copper from millimeters to nanometers. Metall Mater Trans Minerals, Metals and Materials Society, Warrendale,
A 35:2681–2695 pp 345–354
[35] Gurrutxaga-Lerma B (2019) A stochastic study of the col- [50] Hansen N, Ralph B (1982) The strain and grain size
lective effect of random distributions of dislocations. J Mech dependence of the flow stress of copper. Acta Metall
Phys Solids 124:10–34 30:411–417
[36] Gurrutxaga-Lerma B, Balint DS, Dini D, Eakins DE, Sutton [51] Esquivel EV, Murr LE (2005) Grain boundary contributions
AP (2013) A dynamic discrete dislocation plasticity method to deformation and solid-state flow in severe plastic defor-
for the simulation of plastic relaxation under shock loading. mation. Mater Sci Eng, A 409:13–23
Proc R Soc A 469:20130141 [52] Murr LE (2016) Dislocation ledge sources: dispelling the
[37] Gurrutxaga-Lerma B, Balint DS, Dini D, Eakins DE, Sutton myth of Frank–Read source importance. Metall Mater Trans
AP (2014) Dynamic discrete dislocation plasticity. Adv Appl A 47:5811–5826
Mech 47:193–224 [53] Yang B, Vehoff H (2005) Grain size effects on the
[38] Frank FC, Read WT (1950) Multiplication processes for mechanical properties of nanonickel examined by nanoin-
slow moving dislocations. Phys Rev 79:722–723 dentation. Mater Sci Eng, A 400:467–470
[39] Hughes GD, Smith SD, Pande CS, Johnson HR, Armstrong [54] Cottrell AH (1953) Dislocations and plastic flow in crystals.
RW (1986) Hall–Petch strengthening for the microhardness Clarendon Press, Oxford
of 12 nm grain diameter electrodeposited nickel. Scr Metall [55] Yang B, Vehoff H (2007) Dependence of nanohardness upon
20:93–97 indentation size and grain size: a local examination of the
[40] Torrents A, Yang H, Mohamed FA (2010) Effect of interaction between dislocations and grain boundaries. Acta
annealing on hardness and the modulus of elasticity in bulk Mater 55:849–856
nanocrystalline nickel. Metall Mater Trans A 41:621–630 [56] Meyers MA, Ashworth E (1982) A model for the effect of
[41] Pande CS, Masamura RA, Armstrong RW (1993) Pile-up grain size on the yield stress of metals. Philos Mag A
based Hall–Petch relation for nanoscale materials. Nanos- 46:737–759
truct Mater 2:323–331 [57] Raj R, Ashby MF (1971) On grain boundary sliding and
[42] Pande CS, Masumura RA (1996) A model for flow stress diffusion creep. Metall Trans 2:1113–1127
dependence on grain size for nanocrystalline solids. In: [58] Kim HS, Estrin Y, Bush MB (2000) Plastic deformation
Suryanarayana C, Singh J, Froes FH (eds) Processing and behaviour of fine-grained materials. Acta Mater 48:493–504
properties of nanocrystalline materials. The Minerals, Metals [59] Fu H-H, Benson DJ, Meyers MA (2001) Analytical and
and Materials Society, Warrendale, pp 387–397 computational description of effect of grain size on yield
[43] Pande CS, Masumura RA (2005) Grain growth and defor- stress of metals. Acta Mater 49:2567–2582
mation in nanocrystalline materials. Mater Sci Eng, A
409:125–130
J Mater Sci (2020) 55:2661–2681 2681

[60] Sun SS, Adams BL, Shet C, Saigal S, King W (1998) dislocations, triple junctions and Coble creep. Philos Mag
Mesoscale investigation of the deformation field of an alu- 84:847–863
minum bicrystal. Scr Mater 39:501–508 [77] Argon AS, Yip S (2006) The strongest size. Philos Mag Lett
[61] Matthews JW, Crawford JL (1970) Accommodation of misfit 86:713–720
between single-crystal films of nickel and copper. Thin Solid [78] Trelewicz JR, Schuh CA (2007) The Hall–Petch breakdown
Films 5:187–198 in nanocrystalline metals: a crossover to glass-like defor-
[62] Dunstan DJ, Bushby AJ (2013) The scaling exponent in the mation. Acta Mater 55:5948–5958
size effect of small scale plastic deformation. Int J Plast [79] Zhang X, Aifantis KE (2011) Interpreting the softening of
40:152–162 nanomaterials through gradient plasticity. J Mater Res
[63] Dickersin K (1990) The existence of publication bias and 26:1399–1405
risk-factors for its occurrence. J Am Med Assoc [80] Zhang K, Weertman JR, Eastman JA (2005) Rapid stress-
263:1385–1389 driven grain coarsening in nanocrystalline copper at ambient
[64] Newcomb S (1881) Note on the frequency of use of the and cryogenic temperatures. Appl Phys Lett 87:061921
different digits in natural numbers. Am J Math 4:40–93 [81] Dao M, Lu L, Asaro RJ, De Hosson JTM, Ma E (2007)
[65] Benford F (1938) The law of anomalous numbers. Proc Am Toward a quantitative understanding of mechanical behavior
Philos Soc 78:551–572 of nanocrystalline metals. Acta Mater 55:4041–4065
[66] Formann AK (2010) The Newcomb-Benford law in its [82] Cahoon JR, Broughton WH, Kutzak AR (1971) Determi-
relation to some common distributions. PLoS ONE 5:e10541 nation of yield strength from hardness measurements. Metall
[67] Berger A, Hill TP (2017) What is Benford’s law? Not Am Trans 2:1979–1983
Math Soc 64:132–134 [83] Gao X-L (2006) An expanding cavity model incorporating
[68] Shao LJ, Ma BQ (2010) The significant digit law in statis- strain-hardening and indentation size effects. Int J Solids
tical physics. Phys A 389:3109–3116 Struct 43:6615–6629
[69] Wu D, Zhang JY, Huang JC, Bei H, Nieh TG (2013) Grain- [84] Derlet PM, Hasnaoui A, Van Swygenhoven H (2003) Ato-
boundary strengthening in nanocrystalline chromium and the mistic simulations as guidance to experiments. Scr Mater
Hall–Petch coefficient of bcc metals. Scr Mater 68:118–121 49:629–635
[70] Konstantinidis DA, Aifantis EC (1998) On the ‘anomalous’ [85] Wolf D, Yamakov V, Phillpot SR, Mukherjee A, Gleiter H
hardness of nanocrystalline materials. Nanostruct Mater (2005) Deformation of nanocrystalline materials by molec-
10:1111–1118 ular-dynamics simulation: relationship to experiments? Acta
[71] Coble RL (1963) Model for boundary diffusion controlled Mater 53:1–40
creep in polycrystalline materials. J Appl Phys [86] Hasnaoui A, Derlet PM, Van Swygenhoven H (2004)
34:1679–1682 Interaction between dislocations and grain boundaries under
[72] Murayama M, Howe JM, Hidaka H, Takaki S (2002) an indenter: a molecular dynamics simulation. Acta Mater
Atomic-level observation of disclination dipoles in 52:2251–2258
mechanically milled, nanocrystalline iron. Science [87] Voyiadjis GZ, Yaghoobi M (2016) Role of grain boundary
295:2433–2435 on the sources of size effects. Comput Mater Sci
[73] Ovidko IA (2002) Deformation of nanostructures. Science 117:315–329
295:2386 [88] Hu J, Shi YN, Sauvage X, Sha G, Lu K (2017) Grain
[74] Romanov AE, Kolesnikova AL, Ovidko IA, Aifantis EC boundary stability governs hardening and softening in
(2009) Disclinations in nanocrystalline materials: manifes- extremely fine nanograined metals. Science 355:1292–1296
tation of the relay mechanism of plastic deformation. Mater [89] Valiev RZ (2014) Superior strength in ultrafine-grained
Sci Eng, A 503:62–67 materials produced by SPD processing. Mater Trans
[75] Sopicka-Lizer M (2010) High energy ball milling: 55:13–18
mechanochemical processing of nanopowders. CRC Press,
Boca Raton Publisher’s Note Springer Nature remains neutral with
[76] Gutkin MY, Ovidko IA, Pande CS (2004) Yield stress of regard to jurisdictional claims in published maps and
nanocrystalline materials: role of grain-boundary institutional affiliations.