You are on page 1of 20

Pure shear deformation and its

royalsocietypublishing.org/journal/rspa induced mechanical responses


in metallic glasses
Zhukun Zhou1,3 , Hao Wang2 and Mo Li1,3
Research 1 State Key Laboratory of Powder Metallurgy, Central South
Cite this article: Zhou Z, Wang H, Li M. 2019 University, Changsha 410032, People’s Republic of China
Pure shear deformation and its induced 2 Guangdong Provincial Key Laboratory of Micro/Nano
mechanical responses in metallic glasses. Proc.
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

Optomechatronics Engineering, College of Mechatronics and Control


R. Soc. A 475: 20190486.
Engineering, Shenzhen University, Shenzhen 518060, People’s
http://dx.doi.org/10.1098/rspa.2019.0486
Republic of China
3 School of Materials Science and Engineering, Georgia Institute of
Received: 31 July 2019
Accepted: 8 October 2019
Technology, Atlanta, GA 30332, USA
HW, 0000-0001-8896-5496; ML, 0000-0002-6800-8776

Subject Areas: Shear is a basic deformation mode governing


materials science, mechanical engineering, yielding, plasticity and fracture in metallic solids. For
mathematical physics amorphous metals, due to various constraints,
little work is available in addressing directly
Keywords: shear deformation and shear-induced mechanical
amorphous metals, finite deformation theory, property changes which are vital to the mechanistic
understanding of this new class of disordered
nonlinear elasticity, pure shear
materials. Here, by using a finite deformation theory,
we examine the pure shear deformation in a bulk
Authors for correspondence: metallic glass in a large range of shear strains. With
Hao Wang the continuum approach, we show systematically
e-mail: whao@szu.edu.cn for the first time the detailed shear deformation
behaviours, shear-induced normal stress and strain
Mo Li
relations, softening in the elastic constants, volume
e-mail: mo.li@mse.gatech.edu dilatation and free energy change induced by the
shear deformation. These results point to two major
consequences from the shear deformation, one is
the mechanical degradations and the other material
degradation which is responsible for the changes in
the mechanical properties of the disordered materials.

1. Introduction
Uniaxial loading, i.e. tension and compression, is the
most common deformation mode used to measure and
characterize the mechanical properties of solid materials.
From these simple tests, one can obtain directly the
stress–strain relation, Young’s modulus, and yield and
fracture stresses [1]. However, this widely used approach

2019 The Author(s) Published by the Royal Society. All rights reserved.
often falls short in providing detailed information on the deformation process and the related
2
mechanisms that are dominated by shear. In crystalline materials, yielding is caused primarily by
shear or shuffling of the atomic planes via crystal dislocations. Dislocation motion and creation,

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
therefore, determine the strength and plasticity of metallic materials [2]. In amorphous metals,
shear is equally dominant in the strength and plasticity, although there are no atomic planes,
dislocations or other extended structural defects. Under various loading modes, including tension
and compression, metallic glasses (MGs) deform elastically first and eventually yield via the
formation of shear bands. These narrow regions or bands are known to have intense local
shear strains with the magnitude well exceeding one [3–5]. It is the local shear deformation
that determines the strength, plasticity and fracture of this class of disordered materials
[3–5].
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

However, the shear deformation is often entangled with other deformation processes, and
as a result its effect on the mechanical properties and deformation mechanisms is difficult to
sort out. For example, under tension or compression, the uniaxial stress leads to not only shear
stress but also the resolved normal stresses. The interaction between these stresses influences
the shear deformation behaviour that causes the difference in the yield stresses between tension
and compression [6–8]. In the torsion and bending test, bending moments and shear strain
gradient are always present that add complications to the analysis of the shear process [9–11].
Although Hopkinson torsion bar [12,13], shear punch [14,15], high-pressure torsion [16,17] and
four-point bending [18,19] are used in shear deformation tests, these techniques are not accessible
as easily as the uniaxial tests. In addition, the samples are often subject to the superposition of the
additional and non-uniform stresses. Moreover, since the shear deformation area is small, such as
in the four-point bending and shear punch test, direct measurement in these regions is difficult
during the shear process. As a result, the properties in the sheared region are often replaced by
the sample average, such as the elastic constants, volume change and structural variation. In
theoretical modelling, besides the uniaxial loading, simple shear is often employed, mostly for the
convenience of coding and visualization. However, there are additional rotational displacements
associated with this type of shear deformation mode that need to be corrected to obtain pure shear
results [20,21]. As a result, little work is available so far that directly deals with pure shear and its
influences on mechanical properties and behaviours of MGs.
Shear is a basic deformation mode governing yielding, plasticity and fracture in amorphous
solids. Many questions fundamental to the deformation mechanisms pivot on how much
and how clearly we understand shear deformation behaviour. For instance, shear banding is
hypothesized as the result of myriad shear deformation events or shear-related processes, such
as shear-induced strain softening, shear instability, volume dilatation or proliferation of shear
transformation zones [22–28]. However, to explain these phenomenological descriptions, we must
know and understand the detailed shear process and mechanical properties in MGs subject
to shear. Quantitative measurements and characterizations are, therefore, highly desirable on
both continuum and atomic scales, from which we can connect systematically the change of the
mechanical properties to shear where no other deformation modes are present to complicate the
analysis.
This work is one of a series of papers addressing the mechanical deformation and the responses
of MGs under pure shear. We conduct our work in a finite deformation theoretical framework
primarily for the following considerations: firstly, the deformation strain in MGs is generically large.
The elastic strain usually reaches the limit at about 2–3% before yielding, which is one or two
orders of magnitude larger than that in most of the crystalline metals [29,30]. In addition, the
local deformation is also exceedingly large. For example, Donovan [4] shows that the local shear
strain in the regions in and around the shear bands before and at yielding can reach over 150%.
These strains are not necessarily all plastic in nature since the sample is not at the yield point yet.
Secondly, there should be a strong coupling between different stresses in MGs under such large
deformation strains. As mentioned above, the uniaxial stress induces normal stresses which in turn
influence the shear band behaviours and the strength. The same can be said to the hydrostatic
pressure effect on yielding and fracture in MGs. The interactions and the coupling effects are
crucial to understanding the mechanical behaviours in the multiaxial stress state loading that are
3
omnipresent, including tension and compression.
The last but not the least is the volume change induced by shear. Volume is widely recognized

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
as one of the important quantities responsible for the mechanical properties in MGs. Although
volume dilatation is known to exist, there is no volume change in the samples under shear up to
the yielding within the linear theory. The linear theory does not faithfully represent the volume
change expected from the large deformation at yielding in the shear band and surrounding
regions [4]. On the other hand, without the detailed information of the volume change induced by
shear, how the volume varies during mechanical deformations has to be described by a transition
state theory [23,24] which assumes that the volume change is an activated process rather than a
direct consequence of the mechanically driven process.
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

In passing, we note that to date, few works have been performed in pure shear in both
experiments and theories. In addition, the linear elastic theory is widely used despite its
limitations discussed above. Technical inconvenience is one of the conceivable reasons. Although
the nonlinear formulation is substantially complicated and tedious, the advantage is that we
can deal with the large deformation directly. By focusing on pure shear, we can avoid other
deformation modes involved inadvertently such as in tension and compression. This theory,
therefore, allows us to look into pure shear only and its induced mechanical responses such
as stress, strain, elastic moduli, volume and energy change, and other mechanical as well as
structural properties in this basic deformation mode. From the theory, we can obtain analytically
the changing mechanical properties in a quantitative way from small to large elastic strain limits.
This full range of shear strain allows us to gain a detailed view of the evolution of deformation
behaviours in MGs from elastic up to the yielding point. We can relate these fundamental
mechanical responses to the overall mechanical performance that have been discussed or
speculated in the vast number of literature. In addition, the coupling between the shear and other
deformation modes can be addressed conveniently.

2. Theory
(a) Finite deformation theory
For a solid such as an MG under an applied stress σ , a material point in configuration X moves
to a new configuration x with a finite, or Lagrangian strain η. At a given temperature T, the
corresponding Helmholtz free energy F(x,T) = F(η,T) at the new state x can be written as

⎡ ⎤
1 1
⎢τij (X)ηij + 2! Cijkl (X)ηij ηkl + 3! Cijklmn (X)ηij ηkl ηmn ⎥
F(X, η, T) = F(X, 0, T) + V(X) ⎣ 1 ⎦ (2.1)
+ Cijklmnpq (X)ηij ηkl ηmn ηpq + . . .
4!
 
1 ∂F  1 ∂2F 
to the fourth order in the strain η, where τij (X) = V(X) 
∂ηij X,η , C ijkl (X) = V(X) ∂ηij ∂ηkl X,η ,
 
1 ∂ F
3  1 ∂ F
4 
Cijklmn (X) = V(X) ∂ηij ∂ηkl ∂ηmn  
and Cijklmnpq (X) = V(X) ∂ηij ∂ηkl ∂ηmn ∂ηpq  
are the corresponding
X,η X,η
second Piola–Kirchhoff stress, the isothermal second-, third- and fourth-order elastic constants
at state X, respectively. V(X) is the volume of the system at state X. Here, Einstein summation
convention is automatically assumed for the subscripts.
To obtain the stress–strain relations and elastic constants at any deformed state x when an MG
is subject to a deformation strain, we apply another Lagrangian strain ς to the state x such that
x → x , a new deformed state. From this operation, we can obtain the relations for the stress and
elastic constants at state x in terms of the elastic constants at state X and the applied deformation
strain η. Using the relation between the Cauchy stress σ and the second Piola–Kirchhoff stress τ
(see appendices A–C), we have the stress–strain relations
4
⎛ ⎞
1
1 ∂F(x) V(X) ⎜ σkl (X) + C(X)klmn ηmn + C(X) klmnpq ηmn ηpq ⎟

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


σij (x) = = aik ajl ⎝ 1 2 ⎠,

..........................................................
(2.2)
V(x) ∂ςij V(x) + C(X)klmnpqrs ηmn ηpq ηrs + . . .
6
and the strain-dependent second-order elastic constants
⎛ ⎞
1 ∂ 2 F(x) V(X) Cmnpq (X) + Cmnpquv (X)ηuv
Cijkl (x) = = aim ajn akp alq ⎝ 1 ⎠, (2.3)
V(x) ∂ςij ∂ςkl V(x) + Cmnpquvrs (X)ηuv ηrs + . . .
2
where aij = aji = ∂xi /∂Xj is the element in the deformation gradient tensor a from X → x and
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

V(x) the volume in the deformed state. Here, we choose the state without external loading as
the reference state, or the undeformed state. As shown below, from equations (2.2) and (2.3),
we are able to obtain the mechanical behaviour and properties in terms of the stress–strain
relations and the elastic constants at any deformed state x under the applied strain η. As shown in
these relations, the finite deformation theory enables us to deal with strain-dependent properties,
coupling between different stresses, and most importantly, the mechanical properties at large
deformation strains such as strength at yielding or fracture and volume changes.

(b) Pure shear deformation in metallic glasses under large shear


(i) Pure shear deformation and its induced mechanical properties
The expressions for equations (2.1)–(2.3) are under the general deformation strain. The detailed
derivation can be found in equation (A 1) in appendix A. To implement pure ⎛ shear deformation

−η 0 0
⎜ ⎟
to an MG, we apply the following strain tensor from a biaxial loading, E = ⎝ 0 η 0⎠, where
0 0 0
η is the applied normal strain along x and y directions, or η1 = ⎛ −η and η2 = η,⎞ while η3 = 0. The
σ1 0 0
⎜ ⎟
corresponding stress tensor under this deformation is thus σ = ⎝ 0 σ2 0 ⎠. As shown in the
0 0 σ3
above matrix forms, the biaxial loading is composed of a normal compressive and a normal tensile
strain in x- or 1- and y- or 2-directions. The applied normal deformation strains induce the normal
stresses in the materials along the three axial directions in the Cartesian coordinates, σ 1 , σ 2 and
σ 3 . From the free energy (see equation (A 2) in appendix A), we can obtain the three normal
stresses, along with the elastic constants (see equations (A 3) and (A 4) in appendix A).

(ii) Pure shear in the rotated coordinate frame along the shear plane
To have a better understanding of pure shear deformation and possibly make direct comparison
with experiments such as the four-point bending or Hopkinson torsion bar, we shall rotate
the coordinate system to align it with the shear plane by applying
⎛ √ a rotation of ⎞π /4 for the

1/√2 −1/√ 2 0
⎜ ⎟
xy plane around the z-axis with the transformation matrix T = ⎝1/ 2 1/ 2 0⎠. The pure
0 0 1
shear deformation
⎛ ⎞ strain and stress tensor in the rotated shear coordinates become, E = T−1 ET =
0 η 0
⎜ ⎟
⎝η 0 0⎠, where all other strain components are zero except the shear strain E12 = η, and
0 0 0
⎛ ⎞ ⎛ ⎞

σ11 
σ12 0 σ1 + σ2 σ1 − σ2 0
⎜  ⎟ 1 ⎜ ⎟
σ  = T−1 σ T = ⎝σ12 
σ22 0 ⎠ = ⎝σ1 − σ2 σ1 + σ2 0 ⎠, where η, σ 1 , σ 2 and σ 3 are the
 2
0 0 σ33 0 0 2σ3
strain and stress components in the unrotated or original coordinates shown in the last section.
5
The stress tensor σ  in the rotated state can be explicitly expressed as shown in equations
(B 1)–(B 3) in appendix B, along with the elastic constants (equation (B 4)).

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
(iii) Pure shear deformation with relaxation of normal stresses
Shear deformation caused by the biaxial loading can induce normal stresses as shown in
the previous section. We can see this most clearly from the components of the stress tensor
in equations (B 1) and (B 2) along the three directions in the rotated shear coordinate frame.
Therefore, in the pure shear frame, besides the shear stress, there are three normal stress
components induced by the shear strain.
There are two ways to deal with the induced normal stresses. One is to keep them inside the
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

sample, that is allow the sample to subject to the induced normal stresses while being sheared.
In this case, the corresponding normal strains are zero, i.e. E11 → 0, E22 → 0 and E33 → 0. Our
formulations expressed in §2b or equations (B 1) and (B 2) in appendix B are for this case. The
internal normal stresses are kept as they are induced by shear. The sample is effectively under
multiaxial stress state, i.e. one shear and three normal stresses.
Another way is to relax them out, i.e. let σ1 → 0, σ2 → 0 and σ3 → 0 during shear deformation.
By doing so, the internal normal strains will be generated, that is E11 , E22 and E33 are not zero. In
the real shear test, the internal deformation or relaxation usually takes place automatically since
the samples have free boundary conditions. As seen in equations (B 1) and (B 2), as the induced
stresses are normal stress parallel and perpendicular to the shear plane, so are the induced normal
strains after releasing the corresponding normal stresses. The normal strains can be obtained
using the procedure described in the next section.
In this case, the expression for the normal stresses, σ1 , σ2 and σ3 , and the shear stress τ6 in
the rotated shear coordinates and with the non-zero-induced normal strains considered can be
obtained via equations (C 3)–(C 6) in appendix C. Using the same procedure, we can obtain the
elastic constants from equations (C 11) or equation (C 12) where the normal stresses are set to zero.

(c) Implementation of the finite deformation theory


To obtain the pure shear stress–strain relations, elastic constants, deformation energy and volume
change as the function of the applied shear strain in both the relaxed and unrelaxed cases, one
needs to carry out the following procedures:

(i) For pure shear without relaxation, one can obtain the shear-induced normal stresses via
equation (A 3) and the elastic constants via equation (A 4) in the unrotated coordinates,
and the shear stress–strain relation τ6 and the shear-induced normal stresses via
equations (B 3), (B 1) and (B 2) in the rotated coordinates where the shear plane is on
the new xy plane. The elastic constants in the rotated coordinates can be obtained via
equations (B 4). One can obtain the elastic constants in the laboratory frame by simply
transforming the elastic constants back using the method discussed in equation (B 4).
(ii) For pure shear with all shear-induced normal stresses relaxed, one needs first to obtain
the induced normal strains η1 , η2 and η3 at each given applied shear strain η. This is done
 = 0, σ  = 0 and σ  = 0
by allowing the normal stresses to be fully relaxed, i.e. letting σ11 22 33
on the left-hand side of equations (C 3)–(C 5) at each given applied shear strain η6 = η. By
solving these three auxiliary equations, we obtain η1 , η2 and η3 .
(iii) Then one can substitute the normal strains η1 , η2 and η3 back in equation (C 1) to get
the deformation gradient tensor a, and volume ratios expressed in equation (C 2) in the
rotated coordinates and unrotated coordinates respectively. With these quantities, one can
proceed to obtain the shear stress τ6 using equation (C 6) and the elastic constants in the
rotated and unrotated coordinates via equations (C 11) and (C 12).
(iv) At each given applied shear strain η6 = η, we still need another set of inputs for the
second-, third- and fourth-order elastic constants measured at the initial undeformed
state X. These data are scant for MGs, whereas widely available for crystalline metals
6
[31,32]. In this work, we use the data for Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 (Vit105) bulk MG
measured by using ultrasonic techniques [33]. However, the above given elastic constant

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
are adiabatic from the ultrasound measurement, while the finite deformation theory
uses isothermal ones. For isotropic material, the difference between the adiabatic and
isothermal elastic constants for the second-order elastic constants is CS11 − CT11 = CS12 −
CT12 = BS − BT = TVβ 2 TVβ
CV BT = CP BT BS , and C44 − C44 = 0 [34]. Here, T is temperature, V
S T

volume, β thermal expansion coefficient, Cp and Cv the heat capacities at the constant
pressure and volume, respectively, and BS and BT the adiabatic and isothermal bulk
modulus. The superscript and subscript, T and S, stand for isothermal and adiabatic.
For Vit105 in 300 K, the thermal expansion coefficient β is 10−5 K−1 [35], and the
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

specific heat capacity Cp at constant pressure Cp is 1177 Jg-atom−1 K−1 [36]. The adiabatic
bulk modulus Bs is 113.7 GPa [33]. Therefore, from these data, one can see that the
difference between the adiabatic and isothermal elastic constant in MG Vit105 is only
about 0.33 GPa. Thus, we use the adiabatic elastic constants in this work.

3. Results
(a) Shear stress and strain relations
By inputting the experimental data of the elastic constants into the stress–strain relations
(equations (B 1)–(BA3) and (C 6)) in the theory, we can obtain the stress–strain curves under a
given applied shear strain. Note that all stress–strain relations are performed in the rotated shear
coordinates. If the shear-induced normal stresses are not relaxed (equations (B 1) and (B 2)), we
could obtain the pure shear stress–strain relations (equation (B 3)) with the presence of the internal
normal stresses or pressure which is the mean value of these normal stresses. If the normal stresses
are relaxed, we can obtain the shear stress–strain relations (equation (C 6)) as seen in experiments.
Nonlinearity in large deformation is often casted as anharmonic effects, which is usually done
by including in the free energy up to the third-order (3rd) terms in the deformation strain. On
the practical side, since the higher order elastic constants are not easy to obtain, usually only up
to the third-order elastic constants are considered. Here, we have an opportunity to include the
higher order nonlinear terms up to the fourth order (4th). Therefore, we can examine how these
two anharmonic terms balance each other in affecting the mechanical properties.
The third-order terms are known to act as a destabilizing factor to the ‘harmonic’ description of
the system, which is described in the linear elastic theory. The fourth-order terms are expected to
bring some stability. Hence, under external loading, the amorphous material does not develop a
runaway instability which may occur if only the third-order terms are considered. By comparing
the two different theories, we can gain a better understanding of what order of the shear strains or
nonlinearity we should include in the theory and how relevant they are in predicting the overall
mechanical behaviours.
As a result, we shall have four outcomes for each mechanical loading shown in the figures
below. The first two are from the fourth-order theory with and without relaxation, which we
denote as ‘4th r’ and ‘4th ur’, and the other two are from the third-order theory with and without
relaxation, which are denoted as ‘3rd r’ and ‘3rd ur’. Figure 1a is the predicted shear stress–strain
relations for Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 bulk MG with the above conditions considered, i.e. relaxed
versus unrelaxed and with up to the fourth-order strain terms included versus only the third-
order ones.
The stress–strain relations show a linear elastic regime, a yielding point at 3% shear strain and
maximum stress or the fracture stress. The yielding point is measured by using the 0.2% offset
strain method, since the stress–strain relations show a nonlinear behaviour immediately outside
of the elastic regime with less than 0.2% shear strain. The obtained yield point agrees with those
observed experimentally [29]. There is no fracture at the yielding point in the theoretical result.
(a) 8 (b) 15 7
7
10

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
6
5
5 P 4th
τ12(GPa)

s (GPa)
P 3rd
4 0
3
4th ur –5 s11 4th
2 4th r s33 4th
3rd ur –10 s11 3rd
1
3rd r s33 3rd
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

0 –15
0 0.04 0.08 0.12 0.16 0.20 0 0.04 0.08 0.12 0.16 0.20
h12 h12

(c) 8 (d) 0.030


h11 4th
7
0.025 h33 4th
6
h11 3rd
0.020
normal strain

5 h33 3rd
s12(GPa)

4 –s11 4th 0.015


3 –s33 4th
–s11 3rd 0.010
2 –s33 3rd
P 4th 0.005
1
P 3rd
0 0
0 5 10 15 0 0.05 0.10 0.15
sii(GPa) h12

Figure 1. (a) The shear stress versus shear strain with and without relaxation of the induced normal stresses, (b) the induced
normal stresses and pressure without relaxation of the induced normal stresses, (c) the induced normal stresses and pressure
versus the shear stress and (d) the induced normal strains after relaxation of the induced normal stresses in the bulk MG
Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 . Here, we treat the different cases from the combination of the nonlinear terms up to the third and fourth
order labelled as ‘3rd’ and ‘4th’ and whether the induced normal stresses are relaxed or not labelled as ‘r’ and ‘ur’. (Online version
in colour.)

Instead, the sample continues the deformation and becomes softened after the stress reaches the
maximum stress. This is because the sample in the theoretical model does not have any free
boundaries. In the infinitely large sample, shear deformation continues without separating the
sample or fracture at either yield point or the maximum stress. In addition, since we do not
consider microcracks and other imperfections, the sample does not yield or fracture in these
local regions. This feature in the theoretical treatment allows us to focus on the MG to examine
systematically the evolution of deformation and its mechanical responses under the full range of
the shear strains from elastic to the maximum stress typically seen at the onset of shear banding.
The stress–strain curves with the different conditions are almost the same up to about 6% shear
strain. The closeness of the relations to each other indicates that there is little effect of the higher
order elastic constants and relaxation. Beyond the 6% strain, both factors start to show influences.
However, the order of nonlinear terms has a larger effect than the relaxation: the unrelaxed sample
with up to the third-order terms shows a larger shear stress than that with the fourth-order terms
considered. With the relaxation of the normal stresses, shear stresses drop under all conditions but
with the smallest magnitude in the fourth-order theory. For example, the maximum shear stress
for the unrelaxed sample with up to third-order terms is more than 33% larger than that without
relaxation. When the fourth-order terms are included, the shear stress drops and almost overlaps
8
with that in the relaxed sample which includes only the third-order terms. In fact, the difference
between these two is so small that the theory with the fourth-order terms and not relaxed appears

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
to be the same as that in the third-order theory with relaxation. Therefore, the higher order terms
to the fourth order in the finite deformation theory act effectively as a ‘relaxation’; they are a
stabilizing factor in reducing the internal shear stress. This is an important finding as most of the
nonlinear theories only consider anharmonic effects up to the third-order terms.
These observations can also be seen clearly in the maximum stresses and strains. Figure 1a
shows that upon relaxation, the maximum stress in the fourth-order theory does not change
much when compared with that in the unrelaxed one. Without and with relaxation, with the
inclusion of the fourth-order terms, the maximum stress values are 5.22 and 4.79 GPa with the
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

corresponding shear strains at 0.12 and 0.11. By contrast, the stress from the theory with the
third-order terms without considering relaxation is 7.01 GPa and with relaxation 5.24 GPa, and
the corresponding shear strains are 0.16 and 0.11, respectively. The maximum shear stresses and
strains reduction caused by the relaxation and the inclusion of the higher order terms are about 9
and 9.4% separately.

(b) Shear-induced normal stresses and strains, and coupling between shear and normal
stresses or pressure
Figure 1b shows the shear-induced normal stresses which are generated if the sample is not
allowed to relax. For Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 bulk MG, there are three normal stresses, one
perpendicular (σ33  ) and two parallel (σ  and σ  ) to the shear plane. They are all compressive
11 22
in nature as indicated by their negative sign. Besides, they are nearly equal to each other in
magnitude, i.e. σ11 = σ  ≈ σ  . In addition, the order of the nonlinear terms included in the theory
22 33
seems to have little effect on the magnitude of the normal stresses.
The unrelaxed internal normal stresses of the equal magnitude effectively act as a compressive
pressure applied to the MG. The pressure is obtained by taking the negative of the mean value
of the three normal stresses, P = −1/3(σ11  + σ  + σ  ). In figure 1b, we plot P versus the shear
22 33
strain. Since the three normal compressive stresses are nearly the same in magnitude, the internal
pressure is practically compressive and hydrostatic in nature. In addition, the pressure increases
rapidly versus the applied shear strain, reaching about 0.75 GPa and 4.5 GPa, respectively, at the
corresponding 5% and 12% shear strains. The generation of the pressure of this magnitude by
pure shear strain is significant by any means. The coupling between the induced normal stresses
and pressure and the shear stress is shown in figure 1c. Therefore, one should expect its impact
on mechanical deformation and related properties. In other words, one must include the shear-
induced pressure or normal stress effect, or conversely, pressure or normal stress-induced shear,
into consideration for deformation in MGs [6,37].
If these induced internal normal stresses are relaxed out, normal strains are generated. The
induced normal strains are obtained by solving equations (C 3)–(C 5) by setting the normal
stresses to zero (see appendix C). Figure 1d is the induced normal strains versus applied shear
strain after the internal normal stresses are relaxed. Same as the normal stresses, the three strain
components are nearly equal in their magnitude, i.e. η11  = η ≈ η and tension in nature, which
22 33
is natural since we relaxed out the compressive normal stresses to allow the sample to expand.
As shown below, this expansion will lead to sample volume dilation.
The order of the nonlinear terms included in the theory also shows a dramatic effect on the
magnitude of the normal strains: the fourth-order theory (the blue curves in figure 1d) shows
the same magnitude for η11  = η ≈ η , but the theory with the third-order terms gives different
22 33

strains in which η11 is the largest and η33 the smallest (the orange curves in figure 1d). The drastic

difference between the normal strains η11  and η indicates that the third-order or ‘anharmonic’
33
theory leads to artificial deformation anisotropy in MGs. And such effect is absent in the fourth-order
theory where both the normal stresses and the induced normal strains have the same magnitude.
(c) Shear-induced volume dilatation 9
One of the basic questions regarding yielding and plasticity in MGs is the free volume production

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
and the related deformation mechanisms under externally applied stress. In crystalline metals
under applied stresses, yielding is connected to the atomic movement through the creation
of new dislocations or slips of dislocations already present. Both mechanisms determine the
strength of crystalline metals [2]. In amorphous metals, the atomic disorder is down to each atom,
which makes it impossible to define any extended structural defect such as dislocations. One
possible mechanism for plastic deformation is hypothesized to be related to the excess volume:
the existence of excess volumes around each atom enables plastic deformation by ‘swapping’ the
excess spaces with the atoms subject to external stress. This excess space, also called free volume,
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

is defined as an extra space around each atom [22,23]. However, for lack of direct evidence, the
free volume production is thought to be provided by an activated process via a Boltzmann factor
[23]. The role of the applied stress is to help the atoms to jump over the energy barrier.
It is obvious that mechanical deformation creates excess volumes in the sample, which may
not necessarily follow the activated process, especially at low temperature. The straightforward
example is the sample under tensile loading. The tensile strain alone can produce a large amount
of volume dilation at 3% elastic strain even at very low temperature. In pure shear, the volume
is usually considered as conserved, or not changing from the linear elastic theory. As shown
in equation (C 2), the bulk MG Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 under pure shear exhibits a parabolic
relation between the sample volume change and the applied shear strain, as well as the induced
normal strains, i.e. V(x)/V(X) ≈ (1 + 2η11 )(1 + η33 ) − 2η12 2 − (1/2)η2 . One can see that it is the
3
normal strains that contribute primarily to the volume dilatation, or material degradation. Note that
the same parabolic functional form of the volume change was observed in atomistic modelling
[38]. The rapid rise of volume dilatation under pure shear is a unique property of MGs. By
contrast, in crystalline materials, dislocation motion does not lead to much volume change. In the
following, we also compare the magnitude of the volume dilation under shear with that under
thermal expansion which is in fact thermally activated. We will see that the mechanical deformation
is far more effective in produce volume change than thermal agitation.
Figure 2a shows the relative volume change versus the applied shear strain. The volume
changes by about 0.20% when the shear strain reaches about 6% in both cases with the third- and
the fourth-order strain terms included. Beyond 6% strain, the volume changes become drastically
different. At 12% shear strain where the maximum shear stress is reached, the volume change
in the third-order theory reaches 1.5%, nearly twice as large as that in the fourth-order theory.
When the shear strain reaches 15%, the volume change nearly diverges in the third-order theory,
whereas only about 1.5% from the fourth-order theory. As already seen previously in the stress–
strain relations, the fourth-order term acts effectively as a ‘relaxation’ in lowering the shear stress
(figure 1a). Similarly, when compared with that of the third-order theory, the volume dilatation is
smaller when the fourth-order terms are considered.
The magnitude of the volume dilatation induced by the shear deformation is orders of
magnitude larger than that by thermal expansion. For bulk MG Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 when
heated from the room temperature, the experiment gives the volume expansion of about 0.007%
near the glass transition temperature [35]. As a comparison, at 0.5% shear strain, the pure
shear-induced volume expansion is about 0.003%, and at 3% shear strain, the volume dilatation
can reach 0.05%. The large volume dilatation is generated purely by mechanical deformation,
rather than thermal expansion. Note that the volume change shown here is elastic by nature.
Nevertheless, such a large volume dilation must facilitate the atomic movement in the plastic
deformation that follows.

(d) Deformation free energy due to volume dilatation induced by shear deformation
Figure 2b shows the increase in the free energy in pure shear. It can be considered as consisting
of two parts, one from the shear and another the volume dilatation associated with the induced
(a) 0.025 (b) 1.2
0.3 10
4th r 4th

(Uur–Ur)/Ur
0.020 1.0 0.2 3rd

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


3rd r

..........................................................
(U–U0)/V0(GPa)
0.1
0.8
(V–V0)/V0

0.015 0
0.6 0 0.05 0.10 0.15

0.010
0.4 4th ur
4th r
0.005 0.2 3rd ur
3rd r
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

0 0
0 0.05 0.10 0.15 0 0.05 0.10 0.15
h12 h12

(V–V0)/V0
0 0.005 0.010 0.015 0.020 0.025
(c) 0.25
4th
0.20 3rd
(Uur–Uu)/V0(GPa)

0.15

0.10

0.05
4th
3rd
0
0 0.03 0.06 0.09 0.12 0.15
h12

Figure 2. (a) The relative volume change versus shear strain, (b) energy change versus shear strain and (c) the mechanical
energies associated with the volume dilatation versus shear strain. The inside in (b) is the ratio of the mechanical energy to
the energy from the shear deformation. Here, we treat different cases from the combination of when the nonlinear terms are
considered up to the third and fourth order labelled as ‘3rd’ and ‘4th’ and whether or not the induced normal stresses are relaxed
or not labelled as ‘r’ and ‘ur’. (Online version in colour.)

normal strains. In the unrelaxed case, there is no volume change and, thus, the energy is related to
the shear strain only. However, there is a shear-induced internal pressure in the sample as shown
in figure 1b. Upon relaxing out the internal pressure, the volume dilatation occurs. Therefore,
we can associate the amount of mechanical energy with the volume dilatation by the relation,
UV = P V, where P is the internal pressure and V = V − V 0 the volume change when a specific
pressure P is induced at the applied shear strain. This ‘mechanical energy’ associated with the
volume dilation by releasing the internal pressure can be obtained from the differences between
the two energies with and without relaxation of the induced normal stresses, i.e. UV = Uur − Ur .
Figure 2c shows the mechanical energy density, UV /V 0 , associated with the volume dilatation in
the initial volume V 0 .
Once again, we give the results from the third order and fourth-order theories separately. To
check the dependence of the mechanical energy on either the shear strain or the volume dilatation,
we plot the energy against the shear strain (the x-axis on the bottom of the figure) and relative
volume change (the x-axis on the top of the figure) separately in figure 2c. One can see a slow
increase in the energy below 0.6% shear strain, or 0.1% relative volume increase. A rapid increase
follows beyond 10% shear strain, or 1.5% volume dilation from the fourth-order theory. The third-
order theory predicts a much faster rise of the mechanical energy at smaller shear strains and
volume dilatation. This is understandable as we showed that the third-order theory overestimates
11
the anharmonic effects.
To check the magnitude of the mechanical energy for the volume dilatation, we plot the ratio

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
between the mechanical energy and the shear energy, that is (Uur − Ur )/Ur . Note that the energy
associated with the pure shear deformation is the free energy after relaxation, Ur , as shown
in figure 2b. From the inset in figure 2b, one can see that the energy associated with the volume
dilatation is only a small fraction of that associated with shear deformation at a small shear strain. With the
increasing shear, however, the fraction reaches about 2, 10% and a significant fraction of 25% at
6, 10 and 14% shear strains, respectively, from the fourth-order theory. The result from the third-
order theory shows the same trend but higher values. The same magnitude of increase in the
deformation of MGs at the yielding point was observed by Dyre & Olsen [39] and Jiang & Dai [40].
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

The volume dilation and energy increase with the volume change reflect the ‘material damage’
inside the MG, although the continuum theory cannot reveal the atomic details of the damages.
Particularly, the rapid increasing energy needed in creating volume dilatation reveals that the
deformation inside the MG becomes more volume related at larger shear deformation.

(e) Change of elastic constants by shear deformation


Another important mechanical property of MGs is the change of the elastic constants in
deformation. The elastic constants measure the mechanical response between stress and strain
directly in both linear and nonlinear elastic regimes. The change reflects both the mechanical
degradation and structural damage discussed above. However, there are only scant measurements
of strain-dependent elastic constants under simple loading conditions [41,42]; no quantitative
measure has been available under pure shear nor over a large range of strains.
Figure 3 shows the shear strain dependence of the second-order elastic constants of
Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 calculated from the theory including up to the third- and fourth-order
strain terms. The elastic constants with relaxed and unrelaxed normal stresses are computed
from equations (A 4) and (C 12), respectively. Figure 3a is the second-order elastic constants from
the fourth-order theory without relaxation of the induced normal stresses and figure 3b is after
relaxation. The results from the third-order theory are included in figure 3c,d. Note that all elastic
constants are in the unrotated, or laboratory coordinates. One benefit for this is that we can
compare the elastic constants with those in the samples measured in the laboratory coordinates.
In the following, we describe the changes in the elastic constants and their physical meanings:
(1) All elastic constants change under applied shear strain, some of which dramatically. Both
decrease and increase trends show up, depending on the status of relaxation and whether or not
we include the third (figure 3c,d) or the fourth-order terms in the theory (figure 3a,b). C11 , C12 and
C13 in which the shear direction is along x- or 1-direction all increase with shear without relaxing
the normal stresses. But upon relaxation, C11 shows a decreasing trend, C12 and C13 still increase
but with a smaller amplitude. The rest of the elastic constants all show a decreasing trend and
becomes more so upon relaxation. Softening or decreasing trend after relaxation happens in all of the
elastic constants, which is a direct result of the volume dilation after relaxation.
(2) The two non-zero, independent second-order elastic constants in the originally isotropic sample
without deformation evolve into nine under pure shear deformation, indicating the breaking of the isotropic
symmetry by shear. The degeneracy can be seen more clearly in the relaxed case (figure 3b). The
increase in the number of independent elastic constants indicates that MG can develop mechanical
anisotropy induced by shear, or the MG is no longer isotropic once deformed by shear.
However, as seen in figure 1c, the elastic anisotropy can be artificially enhanced in the third-
order theory. The fourth-order terms, on the other hand, lower the mechanical anisotropy. This
effect can be seen again by comparing the variations of C11 , C22 and C33 in figure 3b,d. The very
large decrease in C22 with shear strain in the third-order theory is much significant (figure 3c,d
when compared with those in the fourth-order theory in figure 3a,b. Figure 3c,d shows the elastic
constants from the third-order theory without and with the internal normal stresses relaxed. The
elastic constants do not show a significant difference before and after the relaxation.
180
(a) (b) 12
C≤11

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
150 C≤12

C≤22
elastic constants (GPa)

120 C≤13
C≤23
90 C≤33
C≤44
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

60 26
C≤55
23
0.07 0.08 0.09 C≤66
30

0 0.05 0.10 0 0.05 0.10


h12 h12

180
(c) (d)
C≤11
150
C≤12

C≤22
elastic constants (GPa)

120
C≤13
C≤23
90
C≤33
33 C≤44
60 31
29 C≤55
27
0 0.05 0.10 C≤66
30

0 0.05 0.10 0.15 0 0.05 0.10


h12 h12

Figure 3. The strain-dependent second elastic constants versus shear strain in the case with the fourth-order terms considered
in (a) the unrelaxed stress state and (b) the relaxed stress state. The elastic constants calculated from the third-order theory in
(c) the unrelaxed and (d) the relaxed stress state. The insets show the elastic constants C44 , C55 , and C66 which are nearly equal
in magnitude. (Online version in colour.)

(3) The largest decreases are associated with the elastic constants along the three normal strain
directions, i.e. in C11 , C22 and C33 , especially after relaxation. Apparently, the elastic constants
involving the strain or stress in y-direction all decrease, including the components C22 , C23 , C44 ,
C12 and C66 . This is understandable since the strain in y-direction is tensile. The large decrease
in elastic constants up to 50% in C22 suggests that pure shear can couple to the uniaxial loading
significantly, or vice versa.
Figure 3c,d shows the shear strain-dependent elastic constants predicted from the theory with
13
only up to the third-order strain terms included without and with normal stress relaxation. One
can see from comparing figure 3b,d that softening happens in both relaxed and unrelaxed samples

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
when the fourth-order and thirrd-order terms are considered. In fact, in the third-order theory,
the relaxation plays a relatively stronger role when compared with the case with the fourth-order
terms considered. This is consistent with the results observed in the predicted shear stresses,
which shows the fourth-order terms in the finite deformation function effectively lowering the
stress at a given shear strain. Moreover, the shear elastic constants in relaxed and non-relaxed
samples in the third-order theory are nearly the same. The large difference, however, happens
due primarily to the omission of the fourth-order terms, rather than relaxation. For examples,
C22 in figure 3c decreases by 80% and in figure 3d is nearly zero both at 15% shear strain in the
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

unrelaxed and relaxed cases. Even worse is the predicted increase in C33 in figure 3c from the
third-order theory without relaxation.
The softening of the elastic response to shear is influenced by the volume dilatation. We
call this the material damage. From the finite deformation theory, we can predict quantitatively
that the shear modulus can decrease by about 50% at 8% shear strain while the corresponding
volume dilatation is just slightly less than 0.4%. The shear modulus can be reduced to zero when
the shear strain reaches about 11%, or at about 0.7% volume dilatation. These magnitudes of shear
strains can be easily reached in MGs, considering the large local deformation and the presence
of sample imperfections such as microcracks, voids and different phases. The volume dilatation
is introduced by the release of the induced normal stress or pressure in pure shear deformation.
Therefore, the softening effect from the volume dilatation, on the other hand, reflects the strong
pressure dependence of the mechanical properties in MGs. This coupling between the pressure
and shear is explored in a separate work [37].

4. Conclusion
We have applied a finite deformation theoretical framework to investigate the mechanical
behaviours of a Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 MG as a specific example under pure shear deformation.
The theoretical work gives several predictions and rationales to the mechanical properties in MGs
that still have no qualitative interpretations.
The theory gives the first estimate for theoretical shear strength, the magnitude of which
is only about 20–39% larger than that from experimental results. The theory also predicts the
softening of the elastic constants induced by shear deformation and the symmetry breaking from
the mechanical isotropy to anisotropy. In addition, the volume dilatation is obtained that follows
a parabolic function of the shear strains, along with the detailed description of the shear-induced
normal stress (and pressure) and normal strains (or volume change). The induced stresses
and strains can interact with the applied shear strain, which is expect to lead to complicated
coupling phenomena such as the shear band inclination angle change and pressure sensitivity
in mechanical properties. The large volume expansion induced by shear deformation is captured
and the softening of the elastic constants is found to be related to the volume dilation.
Although unable to give more microscopic details, the continuum theory and its results from a
bulk MG point to two types of damages in the material by shear deformation. One is the material
or structural damage represented by the volume dilatation and free energy increase, and the other
is the mechanical damage in terms of the softening of the elastic constants. It is the accumulation
and the interplay of the two that eventually leads to the failure of the material.
Finally, we should mention that the theory is at room temperature for now. To apply
it to different temperatures, such as close to the glass transition temperature, we need the
corresponding elastic constants measured. In addition, by including spatial variation of the
deformation strains in the theory, we can extend the theory to address the size-dependent
mechanical properties in MGs such as shear band formation and cavitation.
Data accessibility. This is an analytical theory. The input data can be found in [33].
Authors’ contributions. M.L. conceived the idea, carried out derivation and writing. Z.Z. participated in derivation,
14
analysis of results, and writing and H.W. participated in discussion and writing.
Competing interests. We declare we have no competing interests.

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
Funding. M.L. would like to acknowledge the financial support from the National Thousands Talents
Program of China. Z.Z. would like to thank the support provided by China Scholarship Council (grant
no. CSC 201606370100). H.W. would like to thank the partial supported by National Natural Science
Foundation of China (grant no. 51771123) and the Shenzhen Peacock Innovation Project (grant no.
KQJSCX20170327151307811).

Appendix A. Finite deformation theory: stress, strain and elastic constants


Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

In general, we can write the free energy of a solid under deformation in equation (2.1) as the
function of the deformation strain components, η = {η1 , η2 , η3 , η4 , η5 , η6 }, such that

F = F0 + V(X)(ψ2 + ψ3 + ψ4 . . .), (A 1)

where
1 1
ψ2 = C11 (η12 + η22 + η32 ) + C12 (η1 η2 + η2 η3 + η3 η1 ) + C44 (η42 + η52 + η62 ),
2 2
1 1
ψ3 = C111 (η13 + η23 + η33 ) + C112 (η2 η12 + η3 η12 + η1 η22 + η1 η32 + η2 η32 + η3 η23 )
6 2
1 1
+ C123 η1 η2 η3 + C144 (η1 η42 + η2 η52 + η3 η62 ) + C155 (η2 η42 + η3 η42 + η1 η52
2 2
+ η3 η52 + η1 η62 + η2 η62 ) + C456 η4 η5 η6
1 1
and ψ4 = C1111 (η14 + η24 + η34 ) + C1112 [η13 (η2 + η3 ) + η23 (η3 + η1 ) + η33 (η1 + η2 )]
24 6
1 1
+ C1122 (η1 η2 + η2 η3 + η32 η12 ) + C1123 (η12 η2 η3 + η22 η3 η1 + η32 η1 η2 )
2 2 2 2
4 2
1 1
+ C1144 (η12 η42 + η22 η52 + η32 η62 ) + C1155 [η12 (η52 + η62 ) + η22 (η42 + η62 ) + η32 (η42 + η52 )]
4 4
1
+ C1255 [η1 η2 (η42 + η52 ) + η2 η3 (η52 + η62 ) + η1 η3 (η42 + η62 )]
2
1
+ C1266 (η1 η2 η62 + η2 η3 η42 + η1 η3 η52 ) + C1456 [η4 η5 η6 (η1 + η2 + η3 )]
2
1 1
+ C4444 (η44 + η54 + η64 ) + C4455 (η42 η52 + η52 η62 + η62 η42 ),
24 4
in terms of the second-, third and fourth-order elastic constants in the state X. The Cartesian
coordinates are labelled as x, y and z, or 1, 2 and 3. Here, the Voigt notation is used for the
subscripts, i.e. 11 → 1, 22 → 2, 33 → 3, 23 and 32⎛
→ 4, 13 and⎞
31 → 5, and 21 and 12 → 6.
−η 0 0
⎜ ⎟
For pure shear under the applied strain, E = ⎝ 0 η 0⎠, we have the deformation gradient
0 0 0
⎛√ ⎞
1 − 2η 0 0
⎜ √ ⎟
tensor a = ⎝ 0 1 + 2η 0⎠. The general expression for the internal energy in equation
0 0 1
A 1 can be expressed as
1 1
ρ0 U = C11 (η12 + η22 ) + C12 η1 η2 + C111 (η13 + η23 )
2 6
1 1 1
+ C112 (η12 η2 + η22 η1 ) + C1111 (η14 + η24 ) + C1112 [η13 η2 + η23 ]
2 24 6
1 2 2
+ C1122 η1 η2 . (A 2)
4
Here, we use Cauchy stress σ which can be obtained directly by using the relation between the
15
Cauchy stress and the second Piola–Kirchhoff stress, σ = J−1 aτ aT , or more specifically σij (x) =
(V(X)/V(x))aki alj τkl (x), where τij (x) = 1/V(X)∂F(x)/∂ηij |X,η is the Piola–Kirchoff stress, J = det(a) =

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
V(x)/V(X). The results are
  ⎫
(1 − 2η) 1 ⎪
σ1 (x) = −(C11 − C12 )η + (C111 − C112 )η2 ⎪ ⎪


(1 + 2η) 2 ⎪



 ⎪

1 ⎪

3
+ (−C1111 + 4C1112 − 3C1122 )η + . . . , ⎪

6 ⎪



  ⎪

(1 + 2η) 1 2 (A 3)
σ2 (x) = (C11 − C12 )η + (C111 − C112 )η ⎪
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

(1 − 2η) 2 ⎪



 ⎪

1 ⎪

3
+ (C1111 − 4C1112 + 3C1122 )η + . . . ⎪



6 ⎪



1 ⎪

and σ3 (x) =  2
((C112 − C123 )η + . . .), ⎪

(1 + 2η)(1 − 2η)
and the second-order elastic constants under pure shear
   
 (1 − 2η)3/2 1 1
C11 = C11 − (C111 − C112 )η + C1111 − C1112 + C1122 η2 ,
(1 + 2η)1/2 2 2
   
(1 + 2η)3/2 1 1
C22 = C 11 + (C 111 − C 112 )η + C1111 − C1112 + C1122 η2 ,
(1 − 2η)1/2 2 2
1
C33 = 1/2
[C11 + (C1122 − C1123 )η2 ],
(1 − 2η) (1 + 2η)1/2
C12 = (1 − 2η)1/2 (1 + 2η)1/2 [C12 + (C1112 − C1122 )η2 ],
 
(1 − 2η)1/2 1
C13 = C 12 + (−C 112 + C 123 )η + (C 1112 − C 1123 )η 2
,
(1 + 2η)1/2 2
 
(1 + 2η)1/2 1
C23 = C 12 + (C 112 − C 123 )η + (C 1112 − C 1123 )η 2
,
(1 − 2η)1/2 2
 
(1 + 2η)1/2 1
C44 = C 44 + (−C 144 + C 155 )η + (C 1144 + C 1155 − 2C 1255 )η 2
,
(1 − 2η)1/2 2
 
(1 − 2η)1/2 1
C55 = C3131 = C 44 + (C 144 − C 155 )η + (C 1144 + C 1155 − 2C 1255 )η 2
(1 + 2η)1/2 2

and C66 = (1 + 2η)1/2 (1 − 2η)1/2 [C44 + (C1155 − C1266 )η2 ]. (A 4)

Note that for isotropic MG, there are only two independent second-order elastic constants.
Under pure shear, there are nine as shown in equation A 4. Also, all quantities are measured in
the laboratory coordinates labelled as x (1), y (2) and z (3).

Appendix B. Stresses and strains in rotated or shear coordinates


It is convenient to examine the stress–strain relation ⎛ in the ⎞rotated, or shear frame. In
0 η 0
⎜ ⎟
the new coordinate system, the strain tensor is E = ⎝η 0 0⎠ and the stress tensor σ  =
0 0 0
⎛ ⎞

σ11 
σ12 0
⎜   ⎟
⎝σ12 σ22 0 ⎠, where η is the applied shear strain. The strain and stress tensors in the rotated
0 0 
σ33
frame show that the biaxial deformation in the original laboratory coordinates is equivalent to
pure shear in the rotated coordinates, or the shear frame. The strain tensor E is thus simply
16
expressed with only the applied shear strain γ = E12 = η.
The stress component in the rotated state can be explicitly expressed as

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................

  σ1 + σ2 (1 − 2η)
σ1 (x) = σ2 (x) = = ((C111 − C112 )η2 + . . .), (B 1)
2 (1 + 2η)
1
σ3 (x) = σ3 (x) =  ((C112 − C123 )η2 + . . .) (B 2)
(1 + 2η)(1 − 2η)
−σ1 + σ2
and τ6 = σ12

=
2
 
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

1
= (C11 − C12 )η + (2(C11 − C12 ) + (C111 − C112 ) + (C1111 − 4C1112 + 3C1122 ) η3 + . . .
6
(B 3)

The shear stress τ6 and shear strain constitute the basic stress–strain relation of mechanical
response in MGs under pure shear. And, σ1 = σ2 = σ3 are three shear-induced normal stresses.
The elastic constants tensor under the rotated coordinate in the pure ⎛ shear configuration
⎞ can
√ √
1/√2 −1/√ 2 0
⎜ ⎟
be obtained straightforwardly by using the transformation matrix a = ⎝1/ 2 1/ 2 0⎠. We
0 0 1
have the new set of nine elastic constants in the rotated shear frame expressed as

Cijkl = ami anj apk aql Cijkl ,

or explicitly

⎛  ⎞
C C C
11 12 13 0 0 0
⎜ ⎟
⎜C C C 0 0 0 ⎟
⎜ 12 22 23 ⎟
⎜ 0 0 0 ⎟
⎜  ⎟
⎜C13 C C ⎟
⎜ 23 33 ⎟,

⎜ C 0 0 ⎟

⎜ 0 0 0 44 ⎟

⎝ 0 0 0 0 C 0 ⎟

55
0 0 0
0 0 C
66
⎛1  ⎞
C + 1 C + 1 C + C 1 C + 1 C + 1 C − C 1 (C + C ) 0 0 1 (C − C )
⎜ 4 11 4 22 2 12 66 4 11 4 22 2 12 66 2 13 23 4 22 11 ⎟
⎜1  ⎟
⎜ C + 1 C + 1 C − C 1 C + 1 C + 1 C + C 1 (C + C ) 0 0 1 (C − C ) ⎟
⎜ ⎟
⎜ 4 11 4 22 2 12 66 4 11 4 22 2 12 66 2 13 23 4 22 11 ⎟
⎜ 1 (C + C ) 1 (C + C ) ⎟

⎜ C 0 0 − 1 (C − C ) ⎟

2 13 23 2 13 23 33 2 13 23

=⎜
1 (C + C ) 1 (C − C )

⎟ (B 4)
⎜ 0 0 0 0 ⎟
⎜ 2 44 55 2 44 55 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 1 (C − C ) 1 (C + C ) 0 ⎟
⎜ 2 44 55 2 44 55 ⎟
⎝ ⎠
1 (C − C ) 1 (C − C ) − 1 (C − C ) 0 0 1 C + 1 C − 1 C
4 22 11 4 22 11 2 13 23 4 11 4 22 2 12

in terms of the elastic constants in the laboratory frame in equation (A 4). Note that the number of
independent elastic constant components is still nine and the symmetry is kept in the share plane.

Appendix C. Stresses and strains in rotated shear coordinates with the normal
stresses relaxed
From the finite deformation formulation, the relaxed state is in a new state x different from x
which is under pure shear but without releasing the internal normal stresses (see⎛appendix⎞B).
0 η 0
⎜ ⎟
The corresponding strain tensor of the sample under pure shear changes from E = ⎝η 0 0⎠in
0 0 0
⎛ ⎞
η11 η12 0
⎜ ⎟
the pure shear coordinates without relaxation to E = ⎝η21 η22 0 ⎠ after relaxation of the
0 0 η33
normal stresses, where E12 = E21 = η12 = η6 /2 = γ = η is still the applied shear strain which
17
does not change upon relaxation, and the induced normal strains are η11 , η22 and η33 along
the three normal directions in the rotated shear coordinates which are absent in the pure

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
shear coordinates without relaxation. Due to the symmetry of the sample, for pure shear
under relaxation, the induced normal strains in the rotated shear frame are E11 = E22 = η11 ,
or η11 = η22 = η33 , and the shear strain η4 = η5 = 0. Therefore, the corresponding deformation
gradient
matrix is
⎛ ⎞
1 + η11 − 12 (η11
2 + η2 )
12 η12 (1 − η11 ) 0
 ⎜ ⎟
a =⎝ η12 (1 − η11 ) 1 + η11 − 12 (η11
2 + η2 ) 0 ⎠. (C 1)
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

12
1 2
0 0 1 + η33 − 2 η33

When the stress relaxation takes place, from the induced normal strains, we have immediately
the volume change induced by the pure shear

V(x) 1 2
= det a ≈ (1 + 2η11 )(1 + η33 ) − 2η12
2
− η33 , (C 2)
V(X) 2

which is related to the normal strains η11 and η33 as well as the applied shear strain η12 up
to the second order. Note from equation (C 2) that in the sample under pure shear without
releasing the normal stresses, the normal strains are zero, i.e. η11 and η33 are zero. The relative
volume change is V(x)/V(X) ≈ 1 − 2η12 2 . In the linear elastic approximation, volume change

is zero.
In the relaxed case, the normal stresses and shear stress components are substantially lengthy
and tedious but can be expressed in the rotated shear coordinates as
   
1 ∂F  V(X)    1 ∂F 
σ1 = = a a
V(x ) ∂η11 x V(x ) 1k 1l V(X) ∂η 
kl X
kl
         
V(X)   1 ∂F    1 ∂F    1 ∂F 
= a a + a a + 2a a ,
V(x ) 11 11 V(X) ∂η11 X 12 12
V(X) ∂η22 X 11 12
V(X) ∂η12 X
(C 3)
    
1 ∂F   V(X) 1 ∂F 
σ2 = = a2k a2l
V(x ) ∂η22 x V(x ) V(X) ∂ηkl X
kl
      
V(X)   1 ∂F    1 ∂F 
= a a + a a
V(x ) 21 21 V(X) ∂η11 X 22 22
V(X) ∂η22 X
  
1 ∂F 
+ 2a21 a22 = τ11 , (C 4)
V(X) ∂η12 X
   
1 ∂F  V(X)    1 ∂F 
σ3 = = a a
V(x ) ∂η  
33 x V(x) 3k 3l V(X) ∂η 
kl X
kl
  
V(X)   1 ∂F 
= a a (C 5)
V(x ) 33 33 V(X) ∂η33 X
   
1 ∂F  V(X)    1 ∂F 
and τ6 = = a a
V(x ) ∂η12 x V(x ) 1k 2l V(X) ∂η 
kl X
kl
⎛       ⎞
  1 ∂F    1 ∂F 
V(X) ⎜ a11 a21 V(X) ∂η11 X + a12 a22 V(X) ∂η22 X ⎟
= ⎝      ⎠ , (C 6)
V(x ) + a a ∂F  ∂F 
1
11 22 V(X) ∂η12  + a12 a21 V(X)
1
∂η21 
X X
where the strain derivative of internal energy are
18

1 ∂F  1 1
= C11 η1 + C12 (η2 + η3 ) + C111 η12 + C112 (2η12 η2 + 2η1 η3 + η22 + η32 )
V0 ∂η11 X

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


2 2

..........................................................
1 1 1
+ C123 η1 η3 + C155 η62 + C1111 η13 + C1112 (3η12 (η2 + η3 ) + η23 + η33 )
2 6 6
1 1 1
+ C1122 η1 (η2 + η3 ) + C1123 (2η1 η2 η3 + η22 η3 + η2 η32 ) + C1155 η1 η62
2 2
2 2 2
1 1
+ C1255 η3 η62 + C1266 η2 η62 , (C 7)
2 2
 
1 ∂F  1 ∂F 
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

= , (C 8)
V(X) ∂η22 X V(X) ∂η 
11 X

1 ∂F  1 1
= C11 η3 + C12 (η1 + η2 ) + C111 η32 + C112 (η12 + η22 )
V0 ∂η33 X 2 2
1 1
+ C112 (η1 η3 + η2 η3 ) + C123 η1 η2 + C144 η62 + C1111 η33
2 6
1 1 1
+ C1112 (η13 + η23 ) + C1112 η32 (η1 + η2 ) + C1122 η3 (η12 + η22 )
6 2 2
1 1 1
+ C1123 (η12 η2 + η1 η22 + 2η1 η2 η3 ) + C1144 η3 η62 + C1255 (η2 η62 + η1 η62 )
2 2 2
(C 9)
 
1 ∂F  1 ∂F 
and = = C44 η6 + C144 η3 η6 + C155 (η1 + η2 )η6
V(X) ∂η12 X V(X) ∂η6 X
1 1
+ C1144 η32 η6 + C1155 (η12 + η22 )η6 + C1255 (η1 η3 + η2 η3 )η6
2 2
1
+ C1266 η1 η2 η6 + C4444 η63 , (C 10)
6
in terms of the deformation strains in the rotated shear frame and the second- and high-order
elastic constants in the undeformed state X in the laboratory coordinates.
The shear stress τ6 and the applied shear strain η constitute the desired stress–strain relation
with all internal induced strains considered. Equation (C 6) reduces to equation (B 3) if the shear-
induced normal stresses are not relaxed. As seen later, we shall compare τ6 (equation (C 6)) to the
unrelaxed shear stress τ6 (equation (B 3)) to examine the effects of the normal stresses, or pressure
induced by shear.
The elastic constants C in the relaxed case can be obtained by substituting the strain tensor
E and its deformation gradient a into equation (2.3). Since the explicit expression for the
components of the elastic constant matrix C is lengthy, we give the matrix form here,
⎛ ⎞
C
11 C
12 C
13 0 0 C
16
⎜  ⎟
⎜C C C 0 0 C ⎟
⎜ 12 11 23 16 ⎟
⎜  ⎟
⎜C13 C C 0 0 
C36 ⎟
⎜ 23 33 ⎟ (C 11)
⎜ ⎟
⎜ 0 0 0 C C 0 ⎟
⎜ 44 45 ⎟
⎜ 0 0 0 C C 0 ⎟
⎝ 45 44 ⎠
C
16 C
16 C
36 0 0 C
66

However, since elastic constants are usually measured in the unrotated coordinates in
experiment, it is desirable to have the elastic constants C in the relaxed case and in the
rotated frame expressed in the laboratory coordinates. This can be done straightforwardly by
applying the inverse rotation transformation using the matrix a which is the transpose of the
transformation matrix a to the elastic constant tensor in equation (C 11). Hence, we can get the
elastic constants C in the sample in the unrotated coordinate under shear deformation which
19
are shown below

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


⎛ ⎞

..........................................................
C
11 C
12 C
13 0 0 0
⎜  ⎟
⎜C12 C C 0 0 0 ⎟
⎜ 22 23 ⎟
⎜  0 0 0 ⎟
⎜C13 C C ⎟
⎜ 23 33 ⎟
⎜ ⎟
⎜ 0 0 0 C 0 0 ⎟
⎜ 44 ⎟
⎝ 0 0 0 0 C
55 0 ⎠
0 0 0 0 0 C
66
⎛ ⎞
C   
11 /2 + C12 /2 − 2C16 + C66 C  
11 /2 + C12 /2 − C66 C 
23 − C36 0 0 0
⎜ ⎟
⎜ C  
11 /2 + C12 /2 − C66 C   
11 /2 + C12 /2 + 2C16 + C66 C 
23 + C36 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 ⎟
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

⎜ C 
23 − C36 C 
23 + C36 C ⎟
=⎜ 33 ⎟
⎜ ⎟
⎜ C  ⎟
⎜ 0 0 0 44 + C45 0 0 ⎟
⎝ C  ⎠
0 0 0 0 44 − C45 0
0 0 0 0 0 C 
11 /2 − C12 /2

(C 12)

Therefore, if we know the experimentally measured elastic constants on the left side of equation
(C 12), we can obtain those on the right-hand side in the rotated shear coordinates, and vice versa.

References
1. Pelleg J. 2013 Mechanical properties of materials. New York, NY: Spinger, The Netherlands.
2. Hull D, Bacon DJ. 2011 Introduction to dislocations. Oxford, UK: Butterworth-Heinemann.
3. Pampillo CA. 1975 Flow and fracture in amorphous alloys. J. Mater. Sci. 10, 1194–1227.
(doi:10.1007/BF00541403)
4. Donovan PE. 1989 A yield criterion for Pd40 Ni40 P20 metallic glass. Acta Metall. 37, 445–456.
(doi:10.1016/0001-6160(89)90228-9)
5. Schuh C, Hufnagel T, Ramamurty U. 2007 Mechanical behavior of amorphous alloys. Acta
Mater. 55, 4067–4109. (doi:10.1016/j.actamat.2007.01.052)
6. Zhang ZF, Eckert J, Schultz L. 2003 Difference in compressive and tensile fracture mechanisms
of Zr59 Cu20 Al10 Ni8 Ti3 bulk metallic glass. Acta Mater. 51, 1167–1179. (doi:10.1016/
S1359-6454(02)00521-9)
7. Chen Y, Jiang MQ, Wei YJ, Dai LH. 2011 Failure criterion for metallic glasses. Philos. Mag. 91,
4536–4554. (doi:10.1080/14786435.2011.613859)
8. Chen LY, Li BZ, Wang XD, Jiang F, Ren Y, Liaw PK, Jiang JZ. 2013 Atomic-scale
mechanisms of tension–compression asymmetry in a metallic glass. Acta Mater. 61, 1843–1850.
(doi:10.1016/j.actamat.2012.11.054)
9. Fleck NA, Muller GM, Ashby MF, Hutchinson JW. 1994 Strain gradient plasticity; theory and
experiment. Acta Metall. Mater. 42, 475–487. (doi:10.1016/0956-7151(94)90502-9)
10. Ravichandran G, Molinari A. 2005 Analysis of shear banding in metallic glasses under
bending. Acta Mater. 53, 4087–4095. (doi:10.1016/j.actamat.2005.05.011)
11. Zhang Y, Wang WH, Greer AL. 2006 Making metallic glasses plastic by control of residual
stress. Nat. Mater. 5, 857–860. (doi:10.1038/nmat1758)
12. Campbell JD, Dowling AR. 1970 The behaviour of materials subjected to dynamic incremental
shear loading. J. Mech. Phys. Solids 18, 43–63. (doi:10.1016/0022-5096(70)90013-X)
13. Xue Q, Shen LT, Bai YL. 1995 A modified split hokinson torsional bar in studying shear
localization. Meas. Sci. Technol. 6, 1557–1565. (doi:10.1088/0957-0233/6/11/002)
14. Guduru RK, Darling KA, Kishore R, Scattergood RO, Koch CC, Murty KL. 2005 Evaluation
of mechanical properties using shear–punch testing. Mater. Sci. Eng. A 395, 307–314.
(doi:10.1016/j.msea.2004.12.048)
15. Guduru RK, Darling KA, Scattergood RO, Koch CC, Murty KL, Bakkal M, Shih AJ. 2006
Shear punch tests for a bulk metallic glass. Intermetallics 14, 1411–1416. (doi:10.1016/
j.intermet.2006.01.052)
16. Bridgman PW. 1935 Effects of high shearing stress combined with high hydrostatic pressure.
Phys. Rev. 48, 825–847. (doi:10.1103/PhysRev.48.825)
17. Zhilyaev A, Langdon T. 2008 Using high-pressure torsion for metal processing: fundamentals
and applications. Prog. Mater. Sci. 53, 893–979. (doi:10.1016/j.pmatsci.2008.03.002)
18. Yoshihara H, Kubojima Y. 2002 Measurement of the shear modulus of wood by asymmetric
20
four-point bending tests. J. Wood Sci. 48, 14–19. (doi:10.1007/BF00766232)
19. Morrison ML, Buchanan RA, Liaw PK, Green BA, Wang GY, Liu CT, Horton JA. 2007 Four-

royalsocietypublishing.org/journal/rspa Proc. R. Soc. A 475: 20190486


..........................................................
point-bending-fatigue behavior of the Zr-based Vitreloy 105 bulk metallic glass. Mater. Sci.
Eng. A 467, 190–197. (doi:10.1016/j.msea.2007.05.066)
20. Segal VM. 2002 Severe plastic deformation: simple shear versus pure shear. Mater. Sci. Eng. A
338, 331–344. (doi:10.1016/S0921-5093(02)00066-7)
21. Moreira DC, Nunes LCS. 2013 Comparison of simple and pure shear for an incompressible
isotropic hyperelastic material under large deformation. Polym. Test. 32, 240–248.
(doi:10.1016/j.polymertesting.2012.11.005)
22. Argon AS. 1976 Plastic deformation in metallic glasses. Acta Metall. 27, 47–58.
(doi:10.1016/0001-6160(79)90055-5)
Downloaded from https://royalsocietypublishing.org/ on 26 December 2021

23. Spaepen F. 1977 A microscopic mechanism for steady state inhomogeneous flow in metallic
glasses. Acta Metall. 25, 407–415. (doi:10.1016/0001-6160(77)90232-2)
24. Demetriou MD, Harmon JS, Tao M, Duan G, Samwer K, Johnson WL, 2006 Cooperative
shear model for the rheology of glass-forming metallic liquids. Phys. Rev. Lett. 97, 065502.
(doi:10.1103/PhysRevLett.97.065502)
25. Cohen MH, Turnbull D. 1959 Molecular transport in liquids and glasses. J. Chem. Phys. 31,
1164–1169. (doi:10.1063/1.1730566)
26. Schall P, Weitz DA, Spaepen F. 2007 Structural rearrangements that govern flow in colloidal
glasses. Science 318, 1895–1899. (doi:10.1126/science.1149308)
27. Pan D, Inoue A, Sakurai T, Chen MW. 2008 Experimental characterization of shear
transformation zones for plastic flow of bulk metallic glasses. Proc. Natl Acad. Sci. USA 105,
14 769–14 772. (doi:10.1073/pnas.0806051105)
28. Greer AL, Cheng YQ, Ma E. 2013 Shear bands in metallic glasses. Mater. Sci. Eng. R 74, 71–132.
(doi:10.1016/j.mser.2013.04.001)
29. Johnson WL, Samwer K. 2005 A universal criterion for plastic yielding of metallic
glasses with a (T/Tg)2/3 temperature dependence. Phys. Rev. Lett. 95, 195501. (doi:10.1103/
PhysRevLett.95.195501)
30. Trexler MM, Thadhani NN. 2010 Mechanical properties of bulk metallic glasses. Prog. Mater.
Sci. 55, 759–839. (doi:10.1016/j.pmatsci.2010.04.002)
31. Zhao J, Winey JM, Gupta YM. 2007 First-principles calculations of second- and third-
order elastic constants for single crystals of arbitrary symmetry. Phys. Rev. B 75, 094105.
(doi:10.1103/PhysRevB.75.094105)
32. Wang H, Li M. 2009 Ab initiocalculations of second-, third-, and fourth-order elastic constants
for single crystals. Phys. Rev. B 79, 224102. (doi:10.1103/PhysRevB.79.224102)
33. Kobelev NP, Kolyvanov EL, Khonik VA. 2007 Higher order elastic moduli of the
bulk metallic glass Zr52.5 Ti5 Cu17.9 Ni14.6 Al10 . Phys. Solid State 49, 1209–1215. (doi:10.1134/
S1063783407070013)
34. Wallace DC. 1998 Thermodynamics of crystals. New York, NY: Dover.
35. Luckabauer M, Kühn U, Eckert J, Sprengel W. 2014 Specific volume study of a bulk metallic
glass far below its calorimetrically determined glass transition temperature. Phys. Rev. B 89,
174113. (doi:10.1103/PhysRevB.89.174113)
36. Glade SC, Busch R, Lee DS, Johnson WL, Wunderlich RK, Fecht HJ. 2000 Thermodynamics of
Cu47 Ti34 Zr11 Ni8 , Zr52.5 Cu17.9 Ni14.6 Al10 Ti5 and Zr57 Cu15.4 Ni12.6 Al10 Nb5 bulk metallic glass
forming alloys. J. Appl. Phys. 87, 7242. (doi:10.1063/1.372975)
37. Zhou ZK, Wang H, Li M. 2019 Hydrostatic pressure effect on metallic glasses: a theoretical
prediction. J. Appl. Phys. 126, 145901. (doi:10.1063/1.5118221)
38. Wang YJ, Jiang MQ, Tian ZL, Dai LH. 2016 Direct atomic-scale evidence for shear-dilatation
correlation in metallic glasses. Scripta Mater. 112, 37–41. (doi:10.1016/j.scriptamat.2015.09.005)
39. Dyre JC, Olsen N. 2004 Landscape equivalent of the shoving model. Phys. Rev. E 69, 042501.
40. Jiang M, Dai L. 2007 Intrinsic correlation between fragility and bulk modulus in metallic
glasses. Phys. Rev. B 76, 054204. (doi:10.1103/PhysRevB.76.054204)
41. Testardi LR, Krause JT, Chen HS. 1973 Large anharmonicity of amprphous and crystalline
phases of a Pd-Si alloy. Phys. Rev. B 8, 4464. (doi:10.1103/PhysRevB.8.4464)
42. Lambson EF, Lambson WA, Macdonald JE, Gibbs MRJ, Saunders GA, Turnbull D. 1986 Elastic
behavior and vibrational anharmonicity of a bulk Pd40 Ni40 P20 metallic glass. Phys. Rev. B 33,
2380. (doi:10.1103/PhysRevB.33.2380)

You might also like