2 The Model

2.1 Shear Transformation Zones

We want to simulate time-dependent shear deformation of amorphous materials
under load on a mesoscopic scale. Creep is usually associated with metals
where it depends on the motion of dislocations. Since there are no disloca-
tions in amorphous materials, other mechanisms are responsible for the plastic
deformation.
As proposed by Argon [21], and Falk and Langer [22], a sum of small isolated
events is responsible for the observable plastic deformation of amorphous solids.
These events are rearrangements of a bundle of several atoms or molecules, the
so called STZs. In the present simple model the STZs are equally oriented, and
an event causes a plastic strain increment ∆γ in always the same direction. The
plastic deformation created by one shear event in the 2-dimensional simulated
material is described by
!
∆γ 0 1
∆pl = . (2.1)
2 1 0

Every rearrangement generates a new atomic configuration that may contain

new STZs. This model assumes that the number of STZ in the simulated
material is not changed by an event.
For the activation of a STZ the overcoming of a certain energy barrier ∆H 0 is
necessary, otherwise spontaneous creep would be observed. Different STZs may
have different energy barriers. In the simulated material consisting of L2 cells
on a grid each of the cells contains a single STZ with a stochastically assigned
local energy barrier ∆H 0 .
Since creep is load dependent, the energy barrier is obviously lowered by the
application of shear stress. As proposed by Bulatov and Argon [23], ∆H 0 is
decreased by the stress acting on the STZ volume. Assuming that the reduction
in barrier energy equals half the work done by the shear stress σ in creating
the shear strain ∆γ within the STZ volume V we observe
σ∆γV
∆H = ∆H 0 − = ∆H 0 − σVa (2.2)
2

13
2 The Model

without stress
∆H0
with stress

Energy

σVa

Initial configuration Final configuration

Figure 2.1: Reduction of the energy barrier of a loaded STZ.

with the activated STZ volume Va . The relation is depicted in fig. 2.1
The local stress σ is a combination of the external stress σext , which we
assume homogeneous is space, and the internal stress σint (eq. (2.3)). The
internal stresses arise from heterogeneous deformation associated with STZ
events and will be discussed in the following.

σ = σext + σint (2.3)

2.2 The Elastic Stress Field

Local changes during a STZ event affect other STZs in the system. In our
discrete model STZs can be imagined as 2-dimensional, quadratic, and densely
packed cells. Plastic flow shows up as a shape change of one of the cells in
between the surrounding cells. The shape change can be represented by an
arrangement of edge dislocations as shown in fig. 2.2.
Due to our assumption of densely packed cells, the shape change is accommo-
dated by elastic deformation of the surrounding material plus the plastically
deforming of itself. Elastic, long-range compression and tensile stresses are
applied on the other STZs. As a consequence the local stress of the flipped cell
drops and is distributed on the other cells.
To obtain a function describing the elastic stress distribution associated with

14
 2.2 The Elastic Stress Field

Figure 2.2: Arrangement of edge dislocations representing material deficit and

excess during shear transformation in two dimensions.

a single STZ event, the shear stress field of the dislocation array representing
the deformation is computed. In the following only one of the four dislocations
located at 3, 6, 9 and 12 o’clock positions will be examined in detail as the stress
field of the others can be obtained by rotation and shifting. The dislocation at
3 o’clock is chosen for further examination.
The shear stress field σ(x, y) of a single edge dislocation can be evaluated
with eq. (2.4), where x and y are the distance to the dislocation and D is a
constant containing the material parameters shear modulus G, Burgers vector
b and Poison’s ratio ν[24, 25].
x x2 − y 2


σ(x, y) = D (2.4)
(x2 + y 2 )2

Gb
with D = (2.5)
2π (1 − v)
Due to bi-periodic boundary conditions of the simulation, an infinite number of
equal dislocations in y-direction and in x-direction can be found. To account
for the L-periodicity in y-direction the analytical solution for dislocation arrays
is used. The total shear stress field of the array is obtained by summation of
the individual dislocations stress fields (eqs. (2.6) to (2.7))[24].
 
∞ x x2 − (y − kL)2
Gb X
σ(x, y) = (2.6)
2π (1 − v) k=−∞ x2 + (y − kL)2 2
 

Gb x b) cos(yb) − 1)
b (cosh(x
= (2.7)
2L (1 − v) (cosh(xb) − cos(yb))2

2πx 2πy
with x
b= , yb =
L L

15
2 The Model

The L-periodicity in x-direction is not considered here. It is numerically

implemented for a finite number of image dislocation arrays shifted in x-direction.
This and the application of the above formulas on elsewhere located dislocations
is discussed in section 3.1.
The stress distribution function obtained by the summation of the stress fields
of all four dislocations is used as the Greens function G. This function is needed
for solving the integral arising of the elastic stress redistribution discussed in
section 2.4.

2.3 Evolution of the System

The system, containing a number of L2 STZs arranged on a 2-dimensional
grid, is evolved using a stochastic algorithm that chooses exactly one STZ per
iteration to perform an event. The indices i and j used in below equations
determine the position of the STZ on the grid. The introduction of spatially
resolved variables provides the ability of studying the localisation of strain.
To introduce a structural disorder in the system, we assume the stress free
energy barrier ∆Hij0 following a uniform distribution with an average barrier
height H. Since we want a selection of the flipping STZ with attention to
its barrier energy, the stress biased energy barrier ∆H (eq. (2.2)) is used to
calculate the rates of all potential events[5].

∆Hij0 − σij Va
!
νij = ν0 exp − (2.8)
kT
(2.9)

The coefficient ν0 is the attempt frequency, that is proposed to be the Debye

frequency, kT is the Boltzmann thermal energy. The normalization of these
P
rates with the total rate νtot = ij νij results in a list of probabilities pij .
νij
pij = (2.10)
νtot
The KMC method used for the probability based selection of events is explained
in section 1.3. This method also provides the ability to calculate a time scale
once an event is fulfilled. This is necessary for simulating time-dependent
deformation.
Furthermore it must be paid attention to the substitution of the initial STZ
with a new one caused by atomic rearrangements. This is done by changing the
local energy barrier ∆Hij0 of the activated cell.

16
 2.4 Simulation Parameters

Due to a missing correlation to previously reassigned energy barriers, no

hardening resp. softening can be modelled in the evolution of the system. This
appears in the resulting creep curves: no tertiary creep can be observed. The
vanishing creep rate, i. e. hardening, during primary creep is caused “by a
progressive exhaustion of the weakest sites in the configuration”[9] of energy
barriers.

2.4 Simulation Parameters

In order to simplify the simulation’s numerics dimensionless variables are
introduced. The variables stress σ, strain , time t and Green’s function G have
normalization factors as shown in eqs. (2.11) to (2.14). Variables marked with
hat (e.g. t)
b are dimensioned, whereas unmarked variables are dimensionless.

σ=σ
b Cσ (2.11)
 = b C (2.12)
t = tbCt (2.13)
b CG
G=G (2.14)

The stress free energy barrier of each simulation cell Hij0 is assigned using
uniform distributed random numbers fij with expectancy hf i = 1 and average
barrier height H.

∆Hij0 = Hfij (2.15)

As shown in eqs. (2.16) to (2.17) the inverse temperature P , a input parameter

of the simulation, and the constant Cσ can be excerpted from eq. (2.8).
!
H Va

νij = ν0 exp − fij − σij (2.16)
kT H
= ν0 exp (−P (fij − Cσ σij )) (2.17)

Va
Cσ = is given as a fraction of two material properties. Inserting Cσ in P ,
H
the inverse temperature can also be written as
Cσ V a
P = . (2.18)
kT
Another parameter, the stress redistribution parameter C, results from the
redistribution of the internal stresses. As proposed by Bulatov and Argon [23]

17
2 The Model

the Greens’s function G is used for solving the surface integral arising from the
conserving stress redistribution of internal stresses σij in eq. (2.19). The indices
k and l determine the position of the activated cell.
ZZ
σij = σext + C CG ∆γ Gi−k,j−l dA (2.19)
1
ZZ
= σext + ∆γ Gi−k,j−l dA (2.20)
C
1 X
= σext + ∆γ Gij,kl (2.21)
C ij,kl

C combines the strain normalization C and the Green’s function normalisation

CG that arises of the function for the stress field of a single edge dislocation.
Since our array contains four dislocations the constant D must be taken four
times and be normalised by the dislocation’s distance to the center of the array.
The shear modulus G is made dimensionless by inserting Cσ .
1
= CG C (2.22)
C
Gb 1 C
=4 (2.23)
2L (1 − ν) d2 Cσ
L (1 − ν) d Cσ
C= (2.24)
4 Gb C
The externally applied stress σext given in units of Cσ and the number of
iterations of the algorithm NMCS are also input parameters. Because the MC
simulation performs a shear event in every MCS the mean strain per cell hi
can be calculated by eq. (2.25).
NMCS ∆γ
hi = . (2.25)
L2 2
In summary the input parameters of the presented model are:
• Simulation size L

• Inverse temperature P

• Stress distribution parameter C

• Externally applied load σext

• Number of MCS NMCS

Suitable values for these parameters will be discussed in the following section 2.5.

18
 2.5 Material Properties

Table 2.1: Estimated material properties for the determination of the simulations
parameters.

Material property Value

Activation volume Va 1 nm3
Boltzmann constant k 1.4 × 10−23 J K−1
Debye frequency ν0 1 × 1012 s−1
Mean energy barrier H 1 eV
Shear modulus G 10 GPa
Strain increment ∆γ 0.5
Poisson’s ratio ν 0.3

2.5 Material Properties

As explained in section 2.4 several material properties are required for the
simulation model. Using the appraised values in table 2.1 one can determine
reasonable values for the simulation parameters.
The stress norm Cσ is approximately 160 MPa. The stress redistribution
parameter C lies in the order of 1 calculated for commonly used system sizes
with L = 128 and with the assumption, that the Burgers vector b equates the
distance between two dislocations with similar oriented half plane. The strain
norm C is 1.
The simulated temperatures T residing in the range of 200 K to 870 K corre-
spond, according to equation eq. (2.26), to inverse temperatures P between 57
to 13.
Cσ Va H 1.14 × 104 K
P = = = (2.26)
kT kT T

19