Kinetic Monte Carlo Simulation of

Creep Processes in Amorphous

Solids
Contents

Contents 1

List of Abbreviations 3

List of Figures 4

List of Tables 6

1 Introduction 7
1.1 Amorphous Solids . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Time Laws Of Creep . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 The Monte Carlo Method . . . . . . . . . . . . . . . . . . . . . 10
1.4 Scope Of The Thesis . . . . . . . . . . . . . . . . . . . . . . . . 12

2 The Model 13
2.1 Shear Transformation Zones . . . . . . . . . . . . . . . . . . . . 13
2.2 The Elastic Stress Field . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Evolution of the System . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Simulation Parameters . . . . . . . . . . . . . . . . . . . . . . . 17
2.5 Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . 19

3 Implementation 20
3.1 Calculation of the Elastic Stress Field . . . . . . . . . . . . . . 20
3.2 The Kinetic Monte Carlo Method . . . . . . . . . . . . . . . . . 21
3.3 Numerics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.1 Ensemble Average of Simulations . . . . . . . . . . . . . 23
3.3.2 Transient Creep Analysis . . . . . . . . . . . . . . . . . 24
3.3.3 Determination of the Steady State Creep Rate . . . . . 24

4 Results and Discussion 27

4.1 Correlated Shear Banding . . . . . . . . . . . . . . . . . . . . . 27
4.2 Collapse Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3 Transient Creep . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

1
Contents

4.4 Steady State Creep . . . . . . . . . . . . . . . . . . . . . . . . . 37

5 Outlook 41

Bibliography 42

2
List of Abbreviations
BMG bulk metallic glass

KMC kinetic Monte Carlo

MC Monte Carlo

MCS Monte Carlo steps

STZ shear transformation zone

3
List of Figures

1.1 Cooling paths of a liquid to crystalline and amorphous solids[11]. 8

1.2 Stages of creep at an exemplary creep curve. . . . . . . . . . . 9
1.3 Picking events in a KMC calculation[20]. . . . . . . . . . . . . . 12

2.1 Reduction of the energy barrier of a loaded STZ. . . . . . . . . 14

2.2 Arrangement of edge dislocations representing material deficit
and excess during shear transformation in two dimensions. . . . 15

3.1 Stress distribution of a shear transformation zone (STZ) satisfy-

ing bi-periodic boundary conditions computed for L = 16 and
NImg = 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Ensemble average of 5 independent creep curves obtained with
L = 128, P = 38, C = 1.0, σext = 0.3 and NMCS = 131 072. . . . 25
3.3 Definition of the length of transient creep by calculation of two
intersections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.1 Evolution of the local plastic strain of a system obtained with

stress distribution parameter C = 0.4. . . . . . . . . . . . . . . 28
4.2 Evolution of the local plastic strain of a system obtained with
stress distribution parameter C = 1.0. . . . . . . . . . . . . . . 29
4.3 Evolution of the local plastic strain of a system obtained with
stress distribution parameter C = 4.0. . . . . . . . . . . . . . . 30
4.4 Creep curves obtained with different external stresses plotted
with double logarithmic scaling. . . . . . . . . . . . . . . . . . . 32
4.5 Offset as a function of normalized load for different inverse tem-
peratures P . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.6 Linear dependency between slope m and inverse temperature P . 33
4.7 Collapsed curves of fig. 4.4 with reference time scale defined by
the curve with σext = 0.5. . . . . . . . . . . . . . . . . . . . . . 35
4.8 Creep curves obtained with different stress redistribution param-
eters C plotted with double logarithmic scaling for P = 38 . . . 36
4.9 Transient creep exponent as a function of inverse stress distribu-
tion parameter C1 for different inverse temperatures P . . . . . . 36

4
 List of Figures

4.10 Length of transient creep range as a function of inverse stress

distribution parameter C1 for different inverse temperatures P . 37
4.11 Arrhenius plot of steady state creep rates observed with different
temperatures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.12 Activation Energy as a function of external stress. . . . . . . . 40

5
List of Tables

2.1 Estimated material properties for the determination of the simu-

lations parameters. . . . . . . . . . . . . . . . . . . . . . . . . . 19

4.1 Simulation parameters of the obtained stress maps. . . . . . . . 27

4.2 Simulation parameters for the determination of the collapse law. 31
4.3 Simulation parameters for analysis of the transient creep. . . . 34
4.4 Simulation parameters for analysis of the steady state creep. . . 37

6
1 Introduction

1.1 Amorphous Solids

Amorphous solid is a term of condensed matter physics. Amorphous is a

combination of the Greek words a, without and morphé, shape, form. It stands
in great contrast to the crystalline solid.
A well known characteristic of amorphous solids is the lack of long-range
order. This is the main difference to crystals with their periodically repeated
unit cell. The atoms or molecules in amorphous solids are arranged in the same
manner as they are in liquids, but the material shows solid like behaviour due
to more closely packed molecules. Furthermore it exhibits elastic response to
applied shear stresses.
Under the different types of amorphous solids the most common one in
technical use is glass in all forms, shapes and applications, which is studied
since several hundred years. A nowadays largely studied area of amorphous
solids are the recently developed amorphous metals or so called bulk metallic
glasss (BMGs).[1]

Mechanical Properties

BMGs have elastic moduli in the same order as their crystalline counterparts.
In contrast they exhibit a large yield strength approaching the theoretical limit,
high hardness and therefore a good abrasion resistance, and also low mechanical
damping. On the other side these materials have very low ductility, due to
localization behaviour of plastic flow. The altering properties result of the
different atomic mechanisms leading to plastic deformation. The deformation
does not relay on dislocations. Instead rearrangements of several atoms are
responsible for deformation.[1, 2]
A comprehensive understanding of the shear banding behaviour, necessary
for the application of amorphous solids as structural materials, is so far missing
but it is subject of many publications[3, 4, 5, 6, 7, 8, 9, 10].

7
1 Introduction

Enthalpy or free Volume

Liquid

Glas

Crystal
Tg Tm
Temperature

Figure 1.1: Cooling paths of a liquid to crystalline and amorphous solids[11].

Formation
The basic technique for the production of solids is cooling a liquid below a
certain temperature. During this cooling process properties of the material
change. Cooling curves can be presented as the change of free volume or
enthalpy as a function of the temperature. Figure 1.1 shows exemplary cooling
curves for the formation of a crystalline and an amorphous solid. One can see
that the curves are identical for cooling the liquid until the melting temperature
Tm is reached. At this temperature the curve splits into two branches.
The lower curve shows the discontinuous transition of the liquid phase to a
crystalline phase (crystallization). An abrupt drop of the free volume occurs
because the crystal with its ordered atoms has a higher density than the liquid.
Also the enthalpy drops due to required crystallisation enthalpy. This route is
taken if the cooling rate that is strongly dependent on the chemical structure
of the material is sufficiently low.
If cooling rates are large enough in comparison with crystallization kinetics,
the functions of free volume and enthalpy follow the upper route. In this case
no drop of free volume happens at Tm because no crystallization occurs. Instead
a continuous transition between liquid and solid state can be seen reaching
a temperature Tg . The so called glass transition temperature should not be
understood as a well defined temperature, also it is often declared as a single
temperature value. It is a temperature range during which the properties of
the material change.
A advanced technical process that allows bypassing the crystallization during
the cooling is called melt spinning. It is used for the production of metallic
glasses. A jet of fused a metal alloy is sprayed on a cooled, rotating copper

8
 1.2 Time Laws Of Creep

III
Strain
II

Time

Figure 1.2: Stages of creep at an exemplary creep curve.

wheel that is able to deduce the heat in a very short time. Cooling rates of
1 × 106 K s−1 can be achieved.[11, 12]

1.2 Time Laws Of Creep

Creep curves show the plastic strain  of a sample loaded with a time-independent,
uniaxial stress in dependency of the time t. An exemplary creep curve is shown
in fig. 1.2. The creep curve has a standard shape and exhibits all stages of creep
that will be described in the following.
After the initial loading the specimen has a very large creep rate, i. e. the
plastic flow occurs instantaneously. The primary creep (I ) of the curve, which
is also called transient creep, is characterized by an decreasing creep rate ˙
due to work hardening. Reaching the minimum creep rate ˙min the secondary
creep (II ) of the curve is entered. It is also called steady state creep due to the
balance of hardening and softening processes responsible for its constant creep
rate. The steady state creep rate is the technically relevant material constant
representing the creep resistance. An increasing creep rate is observed during
tertiary creep (III ). This is caused by softening mechanisms in the material.
The development of cracks or other defects leads to failure of the sample.[13,
14, 15]
According to Cottrell [15] the first region of creep can be described by a power
law (eq. (1.1)). A and n are experiment dependent parameters. The exponent n
lays in the interval [0, 1], but commonly values of 1 or 2/3 are observed leading

9
1 Introduction

to logarithmic creep behaviour (eq. (1.2)) resp. Andrade’s creep law (eq. (1.3)).
˙ = At−n (1.1)
 = α log t (1.2)
1/3
 = βt (1.3)
The constant creep rate during steady state creep leads to a linear increase of
the strain(eq. (1.4)).
 = κt (1.4)
The overall creep curve can be represented by a summation of the one of the
two creep laws given in eq. (1.2) resp. 1.3 and 1.4 plus an instantaneous strain
0 .

1.3 The Monte Carlo Method

Monte Carlo methods are algorithms for solving computational problems using
random numbers. Since there is a large variety of problems different branches of
Monte Carlo (MC) methods have been developed[16]. Beyond materials science
MC algorithms are e. g. used for stock market analysis[17] or to simulate the
properties of galaxies[18]. A requirement for the adaptability of this method
are “fundamental processes and elementary mechanisms that, when viewed at
the appropriate time scale, behave stochastically (or at least appear to)”[19].
The MC algorithm can also be used to solve higher dimensional, definite
integrals with limited computational resources[20]. The principle of this method
is shown below in the simplified case of solving the one dimensional integral
S(f ) in eq. (1.5).
Z b
S(f ) = f (x) dx (1.5)
a
An estimation for this integral can be obtained by choosing N random values for
x. The mean value of the evaluated function at these values gives the estimated
integral Sest (f ) as shown in eq. (1.6).
N
b−aX
Sest (f ) = f (xi ) (1.6)
N i=1
The area of the estimated integral converges to the exact result with an increasing
number of evaluation points N [16]. Although the Monte Carlo method is not
competitive with standard numerical methods as e. g. Simpsons rule for one
dimensional integrals, it can be used to maintain significant results for the
integration when more complicated higher dimensional integrals are studied.

10
 1.3 The Monte Carlo Method

Kinetic Monte Carlo Method

The kinetic Monte Carlo (KMC) method is used for the time-dependent simula-
tion of plasticity because it provides the ability to create a connection between
the MC time that simply counts the Monte Carlo steps (MCS) and a real time
scale. The application of rates is responsible for this characteristic.
A system suitable for MC simulation must be discrete and stochastic. Discrete
means, that the difference of two following configurations of a system can
be precisely expressed by a single event. Different events lead to different
configurations and only one of these events can happen at the same time. The
transitions between configurations are not infinite fast, so some time passes
during the change. The rate at which a certain event leading to the new
configuration occurs is the inverse of that time. The order of the possible
events occurring during the evolution of the system is stochastic, but the events
are selected with respect to their rates. Thus, the basic ingredient of a KMC
algorithm is a list of rates connected to certain events.[20]
The KMC method establishes a connection between the rate νi of a certain
event and the total rate of the system νtot . The probability pi of a certain event
with rate νi is given by eqs. (1.7) and (1.8) with M possible events.

νi
pi = (1.7)
νtot
M
X
νtot = νi (1.8)
i

Since eq. (1.7) equals the normalization of the rates with the total rate, eq. (1.9)
applies.
X
pi = 1 (1.9)
i

A (one dimensional) list containing the cumulative sums of these probabilities

is created, i. e. the first list item is p1 , the second p1 + p2 , . . . , and the last list
P
item is i pi = 1. A uniform distributed random number R in the interval
(0, 1) (i. e. hRi = 0.5) is selected and an event is picked by searching the element
pi that satisfies eq. (1.10). The procedure is illustrated in fig. 1.3. An event
with a large rate is not necessarily chosen with this method, but it has a grater
probability to be chosen than an event with small rate.

pi−1 < R ≤ pi (1.10)

11
1 Introduction

p1 p2 p3 p4

0 0.25 0.5 0.75 1

Figure 1.3: Picking events in a KMC calculation[20].

After choosing an event a time increment ∆t can be computed by eq. (1.11).

ln(R)
∆t = − (1.11)
νtot
The configuration of the system and therefore possible events in the next MCS
change with the execution of the event. Thus, the a new list of rate must be
determined and the whole process is repeated.[19, 20]

1.4 Scope Of The Thesis

In the present work the time-dependent plasticity of amorphous solids is ex-
amined. As adressed in section 1.1 the atomic mechanisms of plasticity in
amorphous solids defer from their crystalline counterparts. A mesoscopic model
introduced by [21] describing the macroscopic deformation behaviour with an
atomic unit deformation(STZ) will be explained in chapter 2. In the following
necessary steps for evolving a system of these STZ in time using the KMC
algorithm will be explained. Chapter 3 shows how the model is implemented in
two dimensions.
Finally the output of the simulations is analysed in chapter 4. The influence
of the simulation parameters on the resulting strain distribution maps and the
obtained creep curves is examined. The thesis is closed in chapter 5 giving
remarks on the results and ideas for future work.

12