Hydrogen embrittlement in nuclear and
bearing applications: from quantum
mechanics to thermokinetics and alloy
design
Miles Alexander Stopher
This dissertation is submitted for the degree of Doctor of Philosophy
Jesus College
University of Cambridge
May 2017
Preface
This dissertation is submitted for the degree of Doctor of Philosophy at the University of
Cambridge. The research reported herein was conducted under the supervision of Dr. P. E. J.
Rivera-Diaz-del-Castillo in the Department of Materials Science and Metallurgy, University
of Cambridge, between April 2014 and April 2017.
This work is to the best of my knowledge original, except where acknowledgement and
references are made to previous work. Neither this, nor any substantially similar dissertation
has been or is being submitted for any degree, diploma or other qualification at any other
university or institution. This dissertation does not exceed the word limit of 60,000 words.
Part of this work has been submitted in the following publications:
• M.A. Stopher. “The effects of neutron radiation on nickel-based alloys.” Materials
Science and Technology 33; 5 (2017): 518-536.
• M.A. Stopher, P. Lang, E. Kozeschnik and P.E.J. Rivera-Diaz-del-Castillo. “Modelling
hydrogen migration and trapping in steels.” Materials & Design 106 (2016): 205-215.
• M.A. Stopher, and P.E.J. Rivera-Diaz-del-Castillo. “Hydrogen embrittlement in bearing
steels.” Materials Science and Technology 32; 11 (2016): 1184-1193.
Abstract
Hydrogen embrittlement in ferrous and non-ferrous alloys is, and has been for over a
century, a prominent issue within many sectors of industry. Despite this, the mecha-
nisms by which hydrogen embrittlement occurs and the suitable means for its prevention
are yet to be fully established. As hydrogen fuel becomes a prominent feature in modern
concepts of a sustainable global energy infrastructure and nuclear power enters its renais-
sance, with commercially viable fusion plants on the horizon, hydrogen embrittlement
is becoming an ever more pertinent issue. This has led to a considerable demand for
novel alloys resistant to hydrogen embrittlement, notably within the bearings industry,
where the commonly conflicting properties of high strength and hydrogen embrittlement
resistance are required.
This work investigates the mechanisms through which hydrogen embrittlement and ir-
radiation damage occur in steels and nickel-based alloys respectively, with novel alloys
designed for improved resistance. Through the engineering of secondary phases, opti-
mised for helium and/or hydrogen trapping capacity, the novel alloys present the bene-
fits of such trapping species with respect to embrittlement resistance. Such species have
been studied in depth with respect to their interactions with hydrogen, establishing a
novel mechanism of hydrogen embrittlement - the hydrogen enhanced dissolution and
shearability of precipitates, leading to enhanced localised plasticity.
Acknowledgments
I would like to thank Prof. M. Blamire for the provision of laboratory facilities in the
Department of Materials Science and Metallurgy at the University of Cambridge, and
express my sincere gratitude to my supervisor Pedro Rivera-Diaz-del-Castillo for his
support, and affording me the freedom to conduct such varied and interesting research.
I also thank all the persons involved in this work who have dedicated their time: Peter
Lang for his support during my first year, Sarah Driver and Michael Carpenter for
their assistance with resonance ultrasound spectroscopy, David Bombac for his guidance
in density functional theory, Mark Rainforth for his supply of industrial casts, Sue
Gymer and Kevin Roberts for their assistance and entertainment in the process lab, Bill
Clegg for his enthusiasm and encouragement during our collaboration, Tom Edwards,
Tarlan Hajilou and Afrooz Barnoush for our interesting discussions and their assistance
with the production and characterisation of nanopillars, Tom Depover for his assistance
with TDA when time was running short, Wenwen Song for her technical guidance and
assistance with synchrotron radiation experiments and atom probe tomography, Noel
Rutter for his confidence in my teaching abilities, and all my Part III students over
the years that have done so well. I would like to also acknowledge my colleagues and
friends across the University of Cambridge for their help, stimulating discussions and
dependable entertainment value. Finally, this work would have never have been possible
without the sponsorship of SKF, for which special thanks goes to Erik Vegter and
Sebastian Echeverri. This work was supported by the SKF Engineering and Research
Centre and financed by AB SKF and the Engineering and Physical Sciences Research
Council [Programme grant number EP/L014742/1].
Nomenclature
A Proportionality constant
Bs Bainite start temperature
CL Lattice concentration of hydrogen
CT Trapped hydrogen concentration
CTk Hydrogen concentration in k traps
Dapp Apparent diffusion coefficient
Dlatt Diffusion coefficient in the trap-free lattice
Ea Activation energy
Eb Binding energy
Ed Lattice saddle energy
Ef Energy of formation
EH Energy of atomic hydrogen
Es Trap saddle energy
G Gibbs free energy
GL Lattice site energy
GT Trap site energy
Ho Initial hydrogen concentration
Ht Hydrogen concentration at time t
J Nucleation rate
Kk Equilibrium constant
Ms Martensite start temperature
NL Moles of interstitial positions in the lattice
NTk Moles of possible trap positions
P Pressure
Rg Gas Constant
T Absolute temperature
Tp Temperature at which hydrogen desorption rate peaks
TQ Temperature below Ms to which the steel is quenched
VL Volume of lattice containing one mol of interstitial positions
VTk Volume of lattice containing one mol of potential trap sites
yL Lattice hydrogen concentration
yTk Trapped hydrogen concentration
V α′ Martensite content after quenching from austenite temperature
β* Atomic attachment rate taking into account long-range diffusive transport
of atoms in the nucleation rate equation
∆Ek Binding Energy
∆Gs Misfit energy
∆GV Chemical free energy change per unit volume
4
∆Gvol Volume free energy change
θT Fraction of traps occupied
φ Heating rate
5
Glossary
AM Air melted
APT Atom probe tomography
CANDU Canada deuterium uranium
DEA Dark etching areas
DER Dark etching regions
DFT Density functional theory
dpa Displacements per atom
DPS Dual-phase steel
EDX Energy-dispersive X-ray spectroscopy
EP Electrochemical permeation tests
HCR High carbon steel
HED Hydrogen enhanced decohesion
HELP Hydrogen enhanced localised plasticity
HESIV Hydrogen enhanced strain induced vacancies
HFIR High flux isotope reactor
IASCC Irradiation assisted stress corrosion cracking
IGSCC Intergranular stress corrosion cracking
LAS Low alloy steel
LCS Low carbon steel
LR Literature review
LWR Light water reactor
MAP Materials algorithm project
MCS Medium carbon steel
MET Melt extraction technique
NI-AFM Nano-indentation atomic force microscopy
PAW Projector augmented wave method
PI Pure iron
PKA Primary knock-on atom
ppm Parts per million
ppmw Part per million weight
PWR Pressurised water reactor
PWSCC Pressurised water stress corrosion cracking
SCC Stress corrosion cracking
SEM Scanning electron microscopy
STEM Scanning transmission electron microscopy
SG Steam generator
TDA Thermal desorption analysis
6
TEM Transition electron microscopy
VASP Vienna ab initio simulation package
WEA White etching area
WEC White etching crack
XRD X-ray diffraction
7
Contents
1 Introduction 13
I Literature review 15
2 Hydrogen embrittlement mechanisms 16
2.1 Hydride induced embrittlement . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Hydrogen enhanced localised plasticity . . . . . . . . . . . . . . . . . . . 17
2.3 Hydrogen enhanced decohesion . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Hydrogen enhanced strain induced vacancies . . . . . . . . . . . . . . . 21
3 Steels 23
3.1 Mechanisms of hydrogen ingress . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 Ingress from lubricant decomposition . . . . . . . . . . . . . . . . 23
3.1.2 Ingress during processing . . . . . . . . . . . . . . . . . . . . . . 25
3.1.3 Ingress due to general corrosion and internal decarburisation effects 27
3.2 Designing hydrogen embrittlement resistant microstructures . . . . . . . 28
3.2.1 Microstructural interactions with hydrogen . . . . . . . . . . . . 33
3.2.2 Modelling trapping . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.3 Mechanical testing . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.4 Bearing steels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4 Nickel-based alloys 54
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
8
4.3 Pre-irradiated microstructure and corresponding stress corrosion cracking 58
4.4 Fundamentals of radiation damage in nickel-based alloys . . . . . . . . . 59
4.5 Effects of radiation on mechanical properties . . . . . . . . . . . . . . . 64
4.5.1 Tensile properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.5.2 Fracture toughness and creep . . . . . . . . . . . . . . . . . . . . 66
4.5.3 Swelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.6 Radiation induced microstructural changes and irradiation-assisted stress
corrosion cracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.6.1 Microstructural changes . . . . . . . . . . . . . . . . . . . . . . . 71
4.6.2 Irradiation assisted stress corrosion cracking . . . . . . . . . . . . 74
4.6.3 Radiation induced segregation . . . . . . . . . . . . . . . . . . . 76
4.7 Helium and hydrogen embrittlement . . . . . . . . . . . . . . . . . . . . 76
4.8 Trapping of interstitial elements . . . . . . . . . . . . . . . . . . . . . . . 80
4.8.1 Helium trapping . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.8.2 Hydrogen trapping . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.9 Methods of remediation . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.10 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
II Microstructure-hydrogen interactions 92
5 Modelling hydrogen migration and trapping in steels 93
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2.1 Nucleation of precipitates . . . . . . . . . . . . . . . . . . . . . . 94
5.2.2 Growth and coarsening - The SFFK model . . . . . . . . . . . . 95
5.2.3 Dislocation evolution model . . . . . . . . . . . . . . . . . . . . . 96
5.2.4 Dislocation generation . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2.5 Dislocation annihilation . . . . . . . . . . . . . . . . . . . . . . . 96
5.2.6 Trapping of interstitial elements . . . . . . . . . . . . . . . . . . 97
5.3 Simulation setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
9
5.4.1 Case 1: hydrogen trapping in pure iron . . . . . . . . . . . . . . 100
5.4.2 Case 2: hydrogen redistribution in ferritic and martensitic mi-
crostructures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.4.3 Case 3: hydrogen redistribution in ferritic steels under deformation105
5.4.4 Case 4: hydrogen trapping at NbC nanoprecipitates . . . . . . . 107
5.4.5 Case 5: hydrogen trapping at coherent and incoherent TiC pre-
cipitates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6 The role of cementite in hydrogen embrittlement 116
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.2 Computational model and methods . . . . . . . . . . . . . . . . . . . . . 119
6.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.3.1 Differential scanning calorimetry . . . . . . . . . . . . . . . . . . 131
6.3.2 Resonance ultrasound spectroscopy . . . . . . . . . . . . . . . . . 133
6.3.3 In situ hydrogen charged nanopillar compression testing . . . . . 135
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7 The optimisation of vanadium carbides for hydrogen trapping in marten-
sitic steels 139
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.2 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . . . . . 141
7.3 Experimental procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.3.2 Transmission electron microscopy . . . . . . . . . . . . . . . . . . 144
7.3.3 Synchrotron X-ray diffraction . . . . . . . . . . . . . . . . . . . . 144
7.3.4 Atom probe tomography . . . . . . . . . . . . . . . . . . . . . . . 144
7.3.5 Hydrogen trapping characterisation and thermal desorption analysis145
7.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.4.1 Effect of tempering . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
10
8 The optimisation of gamma prime for hydrogen trapping in nickel-
based alloys 161
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.2 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . . . . . 162
8.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.3.1 Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.3.2 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
8.3.3 Hydrogen charging and thermal desorption analysis . . . . . . . 171
8.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
III The design of novel hydrogen embrittlement resistant alloys 174
9 Nanocarbide containing alloys 175
9.1 Modelling and experimental methodology . . . . . . . . . . . . . . . . . 180
9.1.1 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . 180
9.1.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
9.1.3 Hydrogen charging and thermal desorption analysis . . . . . . . 182
9.2 100Cr6 steel variants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.2.1 100Cr6+0.5V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.2.2 100Cr6WV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
9.3 Offshore Flexpipe X steel . . . . . . . . . . . . . . . . . . . . . . . . . . 210
9.3.1 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . 211
9.3.2 Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
9.3.3 Hydrogen charging and thermal desorption analysis . . . . . . . 216
9.3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.4 Case carburised Grade 159V steel . . . . . . . . . . . . . . . . . . . . . . 217
9.4.1 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . 218
9.4.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
9.4.3 Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
9.4.4 Hydrogen charging and thermal desorption analysis . . . . . . . 223
9.4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
9.5 Steel G5V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
11
9.5.1 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . 228
9.5.2 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
9.5.3 Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
9.5.4 Hydrogen charging and thermal desorption analysis . . . . . . . 237
9.5.5 Mechanical testing . . . . . . . . . . . . . . . . . . . . . . . . . . 240
9.5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
9.6 Nickel-based Alloy 600V . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
9.6.1 Thermodynamic and kinetic modelling . . . . . . . . . . . . . . . 242
9.6.2 Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
9.6.3 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
9.6.4 Thermal desorption analysis . . . . . . . . . . . . . . . . . . . . . 253
9.6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
IV Conclusions and future work 258
10 Conclusions and future work 259
10.1 Microstructure-hydrogen interactions . . . . . . . . . . . . . . . . . . . . 259
10.2 The design of novel hydrogen embrittlement resistant microstructures . 262
10.3 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
12
Chapter 1
Introduction
Hydrogen-induced degradation in metallic materials manifests itself in numerous ways,
such as the catastrophic failure of high-strength steels, the contribution to stress cor-
rosion cracking of various materials, and the failure of ZircaloyTM fuel cladding in nu-
clear reactors by hydride formation [1] among others. Despite being first identified in
1875 by Johnson [2], with significant research in succeeding decades, hydrogen-induced
degradation is still not fully understood. Although a number of mechanisms postulated
and consequently extensively investigated in the literature present valid evidence on
the fundamental mechanisms, the true causes of hydrogen-induced degradation remain
inconclusive. Hydrogen-enhanced fracture is a considerably complex problem, one of
micro-plasticity, with the contradiction of ductile processes leading to brittle fracture.
An example of such hydrogen embrittlement is shown in Figure 1.1 for nickel-based Al-
loy 625 and iron during tensile testing of hydrogen charged and uncharged specimens,
discussed further in the proceeding chapter.
Figure 1.1: Stress-stress data from tensile testing of hydrogen-charged and uncharged speci-
mens of a) nickel-based Alloy 625 b) Iron [3].
13
This complexity is further extended by the ingress-dependence and microstructural-
dependence of the resultant mechanisms. For example, the hydrogen-trapping charac-
teristics of features such as grain boundaries and precipitate interfaces will depend upon
the coherency of the respective boundaries. Coupling these structural features’ trapping
behaviour with the effects hydrogen has on the features’ strength, predicting hydrogen-
enhanced crack initiation and propagation becomes problematic. In addition to these
microstructural dependencies, the nature of crack initiation and subsequent growth is
governed by the external stresses applied, the temperature and the alloy strength, re-
sulting in a range of possible mechanical reactions from grain boundary decohesion to
precipitate shear. This results in a challenging division for stress to strain criteria in
hydrogen-induced failures, and indicates that various phenomena could be postulated
for the same failure. It is this complexity, and the lack of conclusive hydrogen-induced
failure mechanisms, which has resulted in hydrogen embrittlement remaining a largely
misunderstood phenomenon.
The first part of this thesis provides a review of the current literature regarding hydro-
gen embrittlement of relevance to designing hydrogen embrittlement resistant bearing
steels and nickel-based alloys for nuclear applications, including the fundamentals of
rolling contact fatigue, irradiation embrittlement, microstructural design and the exper-
imental techniques available for the assessment of hydrogen embrittlement. An in-depth
description of the primary mechanisms of hydrogen ingress and resulting embrittlement
as postulated in the literature, identifying corresponding experimental evidence, dis-
crepancies and relevant knowledge gaps is provided. Part II presents investigations on
the effects of hydrogen on individual phases found within bearing steels and nuclear
nickel alloys, establishing the embrittlement mechanism by which such alloys fail in hy-
drogen containing environments. As a product of this work, six new alloy compositions
have been developed to achieve the high strength, hydrogen embrittlement resistant
properties desired. The complete design process through which these alloys were con-
ceptualised, modelled and characterised, both in terms of microstructure and hydrogen
trapping capacity, is presented in Part III.
14
Part I
Literature review
15
Chapter 2
Hydrogen embrittlement
mechanisms
It has been extensively documented that hydrogen, interacting with steel, can result in
the catastrophic degradation of mechanical properties, often represented by the reduc-
tion in fracture resistance, a phenomenon known as ‘hydrogen embrittlement’. The term
itself implies a hydrogen-induced transition from ductile to brittle behaviour. However,
as will be shown, this is not necessarily true. It is possible to distinguish materials
with respect to their microstructural stability in the presence of hydrogen. Many ma-
terials, most notably the hydride-forming group IV and V transition metals [4], may
be characterised by hydrogen-induced changes in phase stability. This requires a dif-
ferent evaluation to that of phase-stable metals only involving metal-solute interaction.
This will be covered in the following section. This thesis will focus on non-hydride-
forming microstructurally stable alloys. As such, there is only limited applicability of
the developed models to hydride-susceptible alloys.
The fundamental mechanisms of hydrogen embrittlement in metals have been reviewed
extensively in recent decades [1, 3, 5–7]. Over this time, four proposed mechanisms
have come to prominence: hydride-induced embrittlement (HIE), hydrogen enhanced
localised plasticity (HELP), hydrogen enhanced decohesion (HED) and hydrogen en-
hanced strain-induced vacancies (HESIV). In the following sections, each of these mech-
anisms is briefly described, with supporting theoretical and experimental evidence iden-
tified.
2.1 Hydride induced embrittlement
Hydride-induced failure is a well-documented phenomenon [1] [5] [8], notably in the nu-
clear industry where hydride cracking is a prominent concern in zirconium fuel cladding
16
[1]. Hydride nucleation and growth occurs in the stress-field surrounding a crack tip [9].
It has been observed in beta-phase titanium that hydrides grow through the nucleation
of new hydrides in the local vicinity of hydrides already precipitated, combining to
form larger hydrides [10]. This mechanism is the primary cause of hydride-induced em-
brittlement in hydride-forming elements, such as zirconium, niobium and titanium. The
structure of hydrides and the resultant cracking behaviour are strongly microstructurally
dependent and extensive research has been carried out to evaluate these effects [11]. The
crystallographic structure affects hydride precipitation due to the hydride platelets hav-
ing a crystallographic relationship with the metal from which they are formed. The
transformation strains are also directionally related to crystallographic structure. The
stress field surrounding a crack tip is altered by the presence of hydrides, which is de-
pendent upon the orientation of the crack in a material [4]. The hydrides nucleated will
combine to form larger hydrides until a critical hydride size is reached upon which crack-
ing occurs, dependent on the microstructure and texture [1]. Thus, a material’s texture,
in addition to its microstructural features, can be optimised to reduce hydride-induced
embrittlement susceptibility. The description provided here is purposefully limited given
the focus of this thesis on steels, and the extensive literature available on hydride-induced
embrittlement. For a more in-depth description, see references [1] [8] [10].
2.2 Hydrogen enhanced localised plasticity
Hydrogen Enhanced Localised Plasticity (HELP) theory [6] cites that hydrogen embrit-
tlement is a result of the increased mobility of dislocations caused by hydrogen. This is
in contradiction to the well-established phenomenon that increased dislocation mobility
enhances ductility in metals. In HELP, the stacking fault energy (at least in fcc metals)
is thought to reduce with hydrogen content, increasing dislocation mobility local to the
crack tip due to the hydrogen accumulation in the plastic zone. This hydrogen decreases
the repulsive forces between dislocations, and thus the equilibrium distance, increasing
the permissible dislocation density around the crack tip due to the more concentrated
dislocation pile-ups. As hydrogen is attracted to the strain fields surrounding a disloca-
tion, it forms a Cottrell atmosphere, which results in a reduced yield strength [4]. This
is the direct opposite of that observed for carbon-dislocation interactions and the effects
of their corresponding Cottrell atmospheres.
The result of this behaviour is that less stress is required to induce a given plastic strain.
Consequently, this phenomenon has been labelled the softening effect [12], an example of
which is shown in Figure 2.1. Failure by HELP is thus due to a reduction in the plastic
strain capacity. Fracture can be either intergranular and transgranular, depending upon
the microstructure and hydrogen concentration distribution [1] [6].
Ferreira [14], using in situ hydrogen charging within an environmental TEM, evaluated
17
Figure 2.1: Results from tensile testing of high purity iron at 200 Kelvin with a strain rate of
8.3×10-5s-1 with hydrogen charging (20 Am-2) in 0.1 N CH3OH-H2O-H2SO4 solution switched
on and off [13].
the interactions between dislocations in stainless steel. The hydrogen within the TEM
chamber was shown to reduce the elastic interactions between perfect and partial dis-
locations and any obstacles to the dislocations’ motion, increasing dislocation mobility.
This results in an increased density of dislocation pile-ups, such as that shown in Figure
2.2. The initial dislocation configuration was produced by deforming the sample in a
vacuum. Hydrogen gas was added to the TEM cell and the induced increase in pressure
was seen to cause a reduction in the equilibrium distance between dislocations at the
grain boundary, as shown in Figure 2.2. It should be noted that the in situ observations
during hydrogen pumping were carried out on very thin films, with no plastic constraint,
and the dislocations observed are only those pinned on both sides of the film. Thus,
there are drawbacks to using TEM-based hydrogen evaluation, most notably its limita-
tion with respect to simulating real conditions. Worthy of note, Ferreira also discovered
that upon the introduction of hydrogen into the TEM chamber, previously stationary
cracks began to propagate along grain boundaries when in the presence of impurities
such as sulphur, and within the adjacent matrix.
18
Figure 2.2: Superimposition of two TEM images of a single dislocation pile-up upon a grain
boundary under constant stress; one within a vacuum (black), the other under 95 Torr of
hydrogen gas (white) [14].
2.3 Hydrogen enhanced decohesion
Hydrogen enhanced decohesion (HED) theory hypothesises that there is a reduction in
the bonding energy between atoms in the locality of hydrogen, consequently increasing
the risk of decohesion i.e. the surface energy reduces in the presence of hydrogen,
reducing the critical stress intensity factor as a consequence. It was initially hypothesised
due to the observed increases in crack tip opening angles, a consequence of reduced
cleavage toughness, with increasing hydrogen content [7]. It should be noted however,
that hydrogen-enhanced decohesion remains unproven by experimental methods.
The concept of hydrogen-enhanced decohesion was pioneered by Oriani et al. in the
1970s. Oriani proposed that hydrogen is driven to surpass the solubility limit within the
lattice due to its dilation, a consequence of elastic-hydrostatic stresses [7]. As work on
the phenomenon continued, trapping was identified as a method of hydrogen segregation,
such as on grain boundaries - reducing the cohesive bonding strength between metal
atoms. HED states that hydrogen embrittlement occurs within the crack tip fracture
process zone, when the tensile stress opening the crack exceeds the maximum-local
atomic cohesive strength, which, as stated, is reduced in the presence of hydrogen.
Given that the peak stress is located below the crack tip surface, hydrogen-induced
damage initially occurs below the surface, the extent of the damage can be quantified
by modelling the stress distribution around the fracture zone and the hydrogen content
within, coupled with their relationship to the inter-atomic bonding force versus atom
displacement law.
Supporting HED is the established behaviour that hydrogen tends to diffuse to areas
of high tensile stress, similar to that found ahead of a crack tip, and the increased
density of hydrogen trapping sites along a crack path [15]. Atomic simulations have
19
also validated the hypothesised reduction in atomic cohesion with increasing hydrogen
content. However, HED’s primary set-back is that this effect on atomic cohesion cannot
be validated experimentally [16]. The predominant experimental proof for HED is in
the observed reduction in the crack tip opening angle with increasing hydrogen content,
as shown in Figure 2.3. The dependence of in situ crack tip opening angle on both
hydrogen pressure and temperature for a 3 wt% Si-doped iron single crystal is shown
in Figure 2.4, indicating that as decohesion-based crack growth becomes increasingly
important, the angle between active slip planes is reduced. This implies that since the
crack planes of the samples were parallel to (100) and no dimples were observed on
the fracture surface, the decohesion mechanism becomes more dominant, replacing the
crack tip slip mechanism.
Figure 2.3: Crack tip opening angles of a Fe-Si(3 wt%) single crystal after straining in a) a
vacuum and b) hydrogen. [17]
A number of models have been developed in recent decades, based around HED us-
ing both dislocation mechanics and fundamental continuum fracture mechanics, that
have proven effective in emulating experimental results for fracture toughness and crack
growth rates [7] [18]. At the forefront of research aimed at legitimising HED is the desire
to demonstrate the dissolved atomic hydrogen’s effect on lowering the interatomic force-
displacement relationship in metals, and the resultant effects on surface energies and
elastic properties that induce the mechanical property changes observed. The design of
an appropriate experiment to resolve these issues has proven difficult due to the high
concentration of hydrogen within the fracture process zone and the difficulty of repro-
ducing this content within a characterisable bulk specimen due to the limited hydrogen
solubility within the bulk [4]. As modelling techniques and most notably computational
power improves, it is viable that a deeper insight into hydrogen’s effects on bonding may
20
Figure 2.4: Crack tip opening angle as a function of in situ hydrogen pressure (PH2) for a
Fe-Si(3 wt%) single crystal over a range of temperatures. The horizontal dashed line indicates
crack growth solely by crack tip slip [17].
be achieved in the near future. This bonding effect would not illegitimise other theories
on hydrogen embrittlement, and may validate their mechanisms as well as HED.
2.4 Hydrogen enhanced strain induced vacancies
Hydrogen Enhanced Strain Induced Vacancy (HESIV) formation theory hypothesises
that in the presence of hydrogen, the formation of vacancies is enhanced, leading to
ductile crack growth by slip localisation. Vacancies, formed such as during deformation
or neutron irradiation, are stabilised by trapping hydrogen, increasing the vacancy den-
sity and consequently plasticity [3]. These vacancies can form microvoids, which in turn
may combine to form larger voids, and initiate cracking [19]. This phenomenon can be
observed during tensile testing and fracture toughness testing of hydrogen-charged sam-
ples. The concept of HESIV was pioneered by Nagumo [3], testing hydrogen charged
nickel-based Alloy 625 and iron, identifying the influence of hydrogen on the stress-
strain relationship and fatigue life, as shown in Figure 2.5. It was found that hydrogen-
charged samples had the higher void density, as predicted by HESIV, after the same
fatigue cycles as the uncharged samples. Figure 2.6 shows fracture micrographs of
0.57C-1.42Si-0.65Mn-0.67Cr steel after fatigue testing for both hydrogen-charged and
uncharged specimens, showing the smooth cleavage-free fracture surface indicative of
HESIV in comparison with the prevalent cleavage in the uncharged sample.
The production of strain-induced vacancies has been modelled using dislocation dynam-
ics [21], incorporating their nucleation and clustering into microvoids, which in turn
21
Figure 2.5: Stress-stress data from tensile testing of hydrogen-charged and uncharged speci-
mens of a) nickel-based Alloy 625 b) Iron [3].
Figure 2.6: Fracture Micrographs from fatigue testing of a) hydrogen-charged and b) un-
charged 0.57C-1.42Si-0.65Mn-0.67Cr steel [20].
induce cracking. Takai et al. [19] showed that hydrogen adsorption capacity in ferrite
is increased due to the hydrogen enhancing production of strain-induced vacancies and
dislocations. However, during annealing at 200oC, these defects were almost entirely
annihilated, indicating that these were primarily vacancies. A considerable amount of
experimental data confirms hydrogen’s effects on vacancy production. Sakaki et al. [22]
directly observed the hydrogen-induced increase in vacancy generation during deforma-
tion using positron lifetime measurements. McLellan and Xu [23] found that the vacancy
density in iron is greatly increased in comparison to that of thermal equilibrium val-
ues when in a high-pressure and high-temperature hydrogen atmosphere, indicating the
reduction in formation energy of vacancies in iron due to iron-hydrogen interactions.
Although vacancies certainly play a role in hydrogen embrittlement, they do so for all
hypothesised hydrogen embrittlement mechanisms. Although HESIV does not present a
conclusive mechanism of embrittlement, the fundamental effects of hydrogen on vacancy
formation and stability as described in the theory still hold prevalence and thus must
be accounted for in the characterisation of hydrogen-enhanced fracture.
22
Chapter 3
Steels
3.1 Mechanisms of hydrogen ingress
The ingress of hydrogen into steels can occur in a number of ways, varying with the re-
spective environment. To develop hydrogen resistant materials in a systematic way, it is
first necessary to establish the mechanisms by which hydrogen forms, enters and diffuses
through the material. As it will be shown, almost all instances of hydrogen embrittle-
ment begin with the decomposition of hydrogen-containing molecules into hydrogen
atoms at the material surface, the exception being those already in a hydrogen-rich
environment where diffusion of atomic hydrogen directly into the bulk dominates. The
following subsections, based upon open literature, present the most common sources of
hydrogen ingress, their current means of prevention and any source-specific effects.
3.1.1 Ingress from lubricant decomposition
A mechanism of hydrogen ingress prevalent to bearings, the decomposition of lubricant
oil, occurs under the repeated stresses and resulting heat of bearing operation. As
discussed later, it is well known that white etching structures, often termed white etching
areas (WEA), form directly underneath the rolling contact surface, becoming initiation
sites for subsurface cracks leading to bearing failure. A study conducted by Kino and
Otani [24] investigated the cause of such regions with respect to bearing failure due
to flaking, focusing primarily on the effects of hydrogen content. Thrust ball bearings,
having raceways of SCM420 H steel, a surface carburisation hardened mild carbon steel,
and balls of SUJ2 steel, a high-carbon chromium alloy steel, with surface hardnesses
of around 720 and 770 HV respectively, were tested in two different traction fluid oils,
both containing different additives. The compositions of the two steels are shown in
Table 3.1. It was found that the decomposition of both oils resulted in hydrogen ingress
into the steels causing hydrogen-enhanced failure. However, oil no.1, the composition
23
of which is undisclosed, produced greater amounts of hydrogen upon decomposition in
an identical testing environment to that of oil no.2, again the composition of which
is undisclosed, indicating the significance of oil composition on the resulting hydrogen
ingress. The results are shown in Figure 3.1, indicating the hydrogen content of the
steels before and after testing in both oils.
Table 3.1: Common commercial bearing steels (in wt %; balance, Fe) [25].
Grade C Mn Si Cr Ni Mo Cu S P
100Cr6 0.98 0.28 0.28 1.38 0.18 0.06 0.21 0.02 0.12
AISI 1070 0.71 0.76 0.20 0.09 0.08 0.02 0.07 0.012 0.006
ShKh4 0.98 0.18 0.17 0.38
ShKh15 1.05 0.28 0.28 1.5 0.11 0.06 0.015 0.013
SUJ2 1.03 0.37 0.23 1.35 0.51 0.15 0.023 0.018
100CrMo7-3 0.97 0.66 0.27 1.79 0.11 0.26 0.15 0.007 0.009
52CB 0.85 0.35 0.85 0.90 0.60
SCM420H 0.20 0.75 0.25 1.05 0.25 0.23 0.30 <0.03 <0.03
Figure 3.1: Thermal desorption analysis data of SCM420 H steel subjected to hydrogen
ingress from the decomposition of two different oil compositions, the compositions of which
were undisclosed [24].
A more detailed example of oil decomposition, that of synthetic hydrocarbon oil (1, 2,
4-Tris (2-Octyl-1-Dodecyl) Cyclopentane), has been presented elsewhere, demonstrating
the process by which hydrogen ingress occurs in the most common bearing steel, 100Cr6,
the composition of which is shown in Table 3.1 [26]. Under feasible rolling conditions,
decomposition was shown to result in the formation of hydrogen and hydrogenous prod-
ucts: CH3
+, C2H3
+, C2H4 and C3H7
+ which then diffuse into the bulk of steel [27].
Diffusion into the bulk is permissible due to the electropotential with respect to the
24
nascent surface caused by the breakdown of oxides, a result of friction, vibration and/or
electrolysis. Evidently, there are multiple competing species for adsorption onto the
steel surface, and as such, hydrogen absorption is retarded, as proposed in the Iyer-
Pickering-Zamanzadeh-Al-Faqeer model [28].
As shown in Figure 3.2, active sites are generated upon the nascent surface of the
bearing steel after the removal of the oxide layer due to rubbing [27]. Tribochemical
decomposition of the oil then occurs upon these active sites with hydrogen and gaseous
hydrocarbons desorbed as reaction products. The desorbed amount of hydrogen and
hydrocarbons increased linearly with rolling velocity and parabolically with load. In
addition, there appears a critical load beyond which decomposition occurs. The ratio of
desorbed gaseous hydrogen to hydrogen remained unaffected by the variations in rolling
conditions tested. It should be noted, although not covered at present in the literature,
that lubricants containing water are also liable to induce hydrogen embrittlement due
to water splitting by electrolysis.
Figure 3.2: Chemical reactions on a 100Cr6 bearing steel’s nascent surface due to oil decom-
position [27].
In summary, it is evident that both oil composition and rolling conditions affect the
decomposition of oil and the resulting hydrogen absorption. Limited data are available
on the decomposition of oil with respect to hydrogen production and absorption and
as such, it is difficult to simulate the respective reactions of hydrogen in relevant oper-
ating environments for use in hydrogen-charging experiments and as shown, given the
numerous variables, it is likely that this should be done on a case-by-case basis. Thus,
the development of hydrogen-resistant lubricants can be of as much importance as the
development of hydrogen embrittlement resistant bearing steels in improving rolling
contact fatigue life.
3.1.2 Ingress during processing
The ingress of hydrogen into molten steel can occur through reactions with water and
hydrogen bearing compounds, often included in the feedstock [29]. The control of hydro-
25
gen content in steels is an important task for steelmakers, just a few parts per million
of dissolved hydrogen can induce hydrogen flakes (hairline cracks), blistering, loss of
ductility and increased porosity, especially in larger castings.
Steelmakers pay significant attention to the elimination of hydrogen by degassing treat-
ments, both within the steel bath and post casting. For molten steel, a vacuum treat-
ment encourages the formation of gas bubbles containing hydrogen, nitrogen and gaseous
oxides. These bubbles rise to the top of the molten steel and are separated from the
steel bath. In addition, argon bubbling will also enhance hydrogen egress by reducing
the hydrogen partial pressure in the melt [30]. For cast steel, a simple hydrogen bake-
out treatment is often carried out by heating the steel, typically at low temperatures to
prevent detrimental microstructural changes, for a period of time dependent upon the
cleanliness of the melt.
In addition to degassing treatments, steelmakers will also take into account the following
influences on hydrogen egress:
– Melt method (a steel’s hydrogen concentration at the end of an electric arc furnace
melt will on average be up to 4 ppm greater in comparison to that of top blowing
converters [29]).
– The use of lime or aluminium-lime (scavenging moisture from the environment and
into the slag [30]).
– The use of oxygen blowing (reducing hydrogen content via the combustion of carbon
[31]).
– Quality of scrap used (rusted material will increase hydrogen content [5]).
– The use of un-preheated ladles, new linings and lime during tapping (all of which
increase hydrogen content [30]).
The detrimental effects of hydrogen can begin to be observed even upon solidification,
commonly through hydrogen flake formation. Due to the higher solubility of hydrogen
in steel at higher temperature, molecular hydrogen will form upon solidification and
cooling. This gas forms pressure raisers within the matrix, inducing defects that in turn
can cause failure. Hydrogen flakes begin to appear once the steel reaches 200oC, forming
upon defects such as inclusions and areas of martensite and segregation, with manganese
additions significantly increasing the steel’s susceptibility. Sulphur also affects the steel’s
susceptibility to hydrogen cracking, as reductions in sulphur content, consequentially
reducing inclusion content, can increase hydrogen content on the now limited number
of inclusions, often reaching a critical limit beyond which cracking occurs [32] [30].
The casting process is not the only source of hydrogen ingress during processing however,
with hydrogen absorption observed after surface treatments like pickling and electro-
plating [33]. Of course, hydrogen ingress during steel processing is a far broader subject
26
than is appropriate to cover in this thesis. However, by identifying the contributing
factors to hydrogen ingress, steels can be developed alongside their respective processes,
and optimised accordingly to prevent hydrogen embrittlement.
3.1.3 Ingress due to general corrosion and internal decarburisation
effects
The formation of rust is known to induce hydrogen ingress into steels, the reaction for
which is shown below [34].
4Fe2+ + O2 + 6H2O→ 4FeOOH + 8H+ (3.1)
Fe3+ + 3H2O→ Fe(OH)3 + 3H+ (3.2)
The formation of rust is of course temperature and environment dependent. However,
as the operating temperatures increases, the formation of rust is not the only chemi-
cal reaction that needs to be considered. At high temperature, due to high hydrogen
mobility through interstitial sites, hydrogen atoms may begin to form molecular hy-
drogen, depending on the reaction site, causing blistering [29]. More specifically, when
at temperatures above 200oC, internal decarburisation can occur from the reaction of
molecular hydrogen with atomic carbon or carbides such as cementite, forming methane,
as shown:
C + 2H2 → CH4 (3.3)
Fe3C + 2H2 → 3Fe + CH4 (3.4)
This methane, in a similar way to molecular hydrogen, forms gas voids within the matrix
and promotes the formation of cavities, blisters and cracks due to the resulting internal
stress. Given the larger molecular size of methane, these effects are more severe than
that of molecular hydrogen. However, molecular hydrogen may also react with free
oxygen and oxides to form water, resulting in further increases in internal stress [35].
In all cases, once hydrogen has entered the material, it diffuses through a combination of
self-diffusion, driven by concentration gradients and described by Fick’s laws of diffusion,
and physicochemical interactions with microstructural features such as hydrogen traps.
Such traps have higher binding energies to hydrogen than those of regular interstitial
sites within the matrix and can trap hydrogen either reversibly or irreversibly, depending
on the magnitude of the binding energy. It is these hydrogen traps, and their ability
27
to prevent further hydrogen migration, leading to potentially detrimental interactions,
which show most promise in the search for hydrogen-resistant steels. The following two
sections outline these trapping mechanisms as presented within the literature, identifying
suitable traps for further investigation along with the modelling techniques available to
evaluate this suitability.
3.2 Designing hydrogen embrittlement resistant microstruc-
tures
It is well established that immobilised hydrogen does not contribute to hydrogen embrit-
tlement in its general case due to its inability to diffuse to regions of high stress, such as
crack tips, and enhance failure [36]. Thus, the state of the respective hydrogen within
the steel is critical with regards to its contribution to hydrogen embrittlement. The
proportions of hydrogen in the mobile and immobile states depend on the microstruc-
ture of the respective steel, defining the solubility, transport and trapping of hydrogen
within the steel under consideration. The following section explains these terms both
qualitatively and quantitively with respect to hydrogen embrittlement susceptibility and
proposes a primary route through which such susceptibility can be reduced.
Bearing steels contain complex microstructures, the modelling of which has improved
significantly in recent years with advances in thermodynamic and kinetic modelling, as
described later. These microstructures contain defects and features such as grains and
numerous phases, along with their corresponding boundaries. These features play a
crucial role in the uptake and transport of hydrogen in steels and their engineering is
crucial with respect to reducing hydrogen embrittlement susceptibility. Features that
typically attract hydrogen are classified as traps, a localised region in which an atom or
molecule, in this case hydrogen, remains for a longer period of time than it would in a
typical interstitial lattice site [37].
Such traps can be further categorised as either reversible or irreversible, depending
on their activation energy. Reversible traps are those which hydrogen frequents for a
limited time dependent upon temperature, more specifically those with an activation
energy of less than 50 kJ·mol-1. Irreversible traps on the other hand, with activation
energies greater than 50 kJ·mol-1 show a negligible probability of escaping the trap at
room temperature [38]. The consequence of such trapping sites, both reversible and to
a greater extent those irreversible, is a reduction in the diffusion rate and an increase
in the hydrogen solubility for the steel.
The schematic shown in Figure 3.3 presents the relevant energy levels with respect to
trapping, with the activation energy Ea, the lattice saddle energy Ed, the trap saddle
energy Es, and the binding energy Eb. The activation energy Ea defines the energy
28
Figure 3.3: Free energy levels of an interstitial lattice site and trapping site [39].
necessary to escape a trap. The lattice saddle energy Ed defines the mean energy
necessary for diffusion through a trap-free lattice. A wide range of hydrogen traps and
their corresponding binding energies are present within the literature, as shown in Table
3.2.
29
Table 3.2: Common hydrogen traps with their corresponding binding energies.
Class Trap Type
Binding
Energy
Eb
(kJ/mol)
Comment
Technique
(Material)
Ref.
Precip. Mo2C 22-28 peak aged EP [40]
Mo2C 21-29 fine precip. EP [41]
VC 17 coherent EP [42]
V4C3 30 coherent TDA (LCS) [43]
Fe3C 21-29 - EP [41]
Fe3C 84 incoherent LR (MCS)
[44]
[25]
Epsilon carbide 13 - TDA [45]
AlN 65 - LR [46]
AlN >83.94 - LR [46]
TiC 94 incoherent LR [46]
TiC 77-95
semi-
coherent
LR [32]
TiC 87 incoherent TDA (MCS) [47]
Phases
Retained
austenite
45 TDA (MCS) [48]
Retained
austenite
55 (DPS)
[44]
[49]
Defects Dislocations 26-29
5-15
atoms/nm
TDA+EP
[50]
[51]
[52]
[53]
[54]
Grain boundary 17 (PI) [55]
Grain boundary 53-59 high angle LR [32]
Voids 21 LR [32]
Where: TDA - Thermal Desorption Analysis, EP - Electrochemical Permeation testing, LR
- Literature Review, (MCS) - medium carbon steel, (DPS) - dual-phase steel, (LCS) - low
carbon steel, and (PI) - pure iron.
Deducing the appropriate trapping mechanisms through which to achieve hydrogen em-
brittlement resistance for a particular steel depends upon the form of hydrogen ingress
and the operational conditions. For such a selection process, one can categorise the de-
sired microstructural effects into three fundamental objectives: hydrogen trapping with
in-life degassing, through-life hydrogen trapping, and hydrogen ingress prevention.
Trapping with in-life degassing is achieved by a microstructure which contains traps of
a suitable depth at which it is viable to degas part-way through life, without detriment
to the microstructure. For example, for a particular application, if the hydrogen ingress
rate was high enough to saturate all permissible traps within a microstructure during
30
its lifetime, be it during processing and/or operation, a degassing treatment would
be required to empty the hydrogen traps, reducing the free hydrogen content during
proceeding operation through renewed hydrogen trapping capacity. To permit such a
degassing treatment, the traps must be shallow enough to be liberated during such a
treatment, the temperature of which will often be limited due to the microstructural
stability at temperature and the permissible temperatures achievable for the respective
component’s environment, if degassing is even viable in the first place. Pertinent to
bearing steels, applying degassing treatment to hydrogenated steels containing retained
austenite must be done with caution as total degassing is not guaranteed. There are
cases where hydrogen is at high concentrations within the retained austenite, and due
to the low diffusivity and high solubility of hydrogen in the austenite, that hydrogen is
retained within the phase after degassing. Consequently, retained austenite behaves as
a reservoir for hydrogen, gradually diffusing its contained hydrogen into the bulk ferrite
when under stress, thus increasing the mobile hydrogen concentration [56]. In addition,
one study by Robinson [57] found that after degassing a hydrogen charged carbon steel
wire, the embrittlement susceptibility was found to increase. It was postulated that this
was caused by the low temperature of the degassing treatment, limiting the liberation
of hydrogen to only reversible traps. After liberation, this hydrogen was then trapped
on larger defects of irreversible trapping character. As previously described, sites such
as these can act as significant stress concentrators, and if hydrogen is beyond a critical
concentration, upon reloading, rapid catastrophic failure can occur. Thus, the increased
hydrogen concentration on such defects was postulated as being beyond the critical
concentration for crack initiation under the applied stress and as a result, failure resulted
at a lower stress than that for the samples which did not receive a degassing treatment.
Through-life hydrogen trapping is achieved by a microstructure containing traps ad-
equate in number, capacity and depth to irreversibly trap all hydrogen entering the
steel during its life without saturation. This is permissible for operational environments
where hydrogen ingress is not substantial. For example, bearings, such as those used in
off-shore wind turbines, may accumulate up to 3 ppm hydrogen over their lifetime [25].
Trapping this amount of hydrogen is permissible through engineered microstructural
traps [58]. Hence, through-life hydrogen embrittlement resistance may be achieved
without in-life degassing.
Of course, preventing ingress altogether provides a solution to all operational conditions.
As such, the development of hydrogen-ingress resistant coatings and microstructures is
of great appeal. For an example relevant to various applications including bearing steels,
often cracking will initiate subsurface, with the proceeding failure enhanced by hydrogen.
To prevent such failures, it is permissible to produce a surface layer of hydrogen resistant
material which in turn reduces or eliminates hydrogen ingress. One potential ingress-
inhibitor is through the use of austenite, within which the diffusivity is up to a factor
31
of 104 less than that of ferrite, as shown in Table 3.3, a surface layer of which could
greatly limit hydrogen ingress in comparison to that of current bearing steels. Similarly,
a coating or oxide layer could be produced to the same effect. Of course, as is the case in
many applications including bearing steels, the surface morphology can play a key role
in the steel’s properties, such as rolling contact fatigue life, and as such, a protective
surface layer or coating may not be permissible.
Table 3.3: Diffusion coefficients of hydrogen in various stainless steels at 50oC [59] after [60].
Steel Matrix Phase Diffusivity (m2s−1)
405 Ferritic 2.8 x 10−12
304 Austenitic 1.1 x 10−15
316 Austenitic 2.0 x 10−15
316L Austenitic 2.4 x 10−16
Other than microstructurally-dependent hydrogen embrittlement resistance mechanisms,
reducing hydrogen content and/or production in the operational environment is of crit-
ical importance and if achievable, may negate entirely the requirement for hydrogen
embrittlement resistant microstructures.
Although absent in the majority of work within the open literature, it is critical to
distinguish external hydrogen from internal hydrogen, as both will have significant but
very different effects on a steel’s susceptibility to hydrogen embrittlement. For this
thesis, internal hydrogen is defined as the hydrogen, molecular and atomic, that is
present within the steel immediately prior to operation, which, as will be shown, is
assumed to be entirely trapped by various microstructural features. External hydrogen
is defined as that which enters the steel during operation and is thus expected to diffuse
through the material and, as shown in the following section, is again intended to be
trapped by the microstructural features engineered to optimise such trapping capabilities
of the steels concerned.
From what has been reviewed so far, it is evident that the microstructural traps engi-
neered into the steels are redundant if saturated with hydrogen upon entering operation.
Thus, case-by-case, steels must be monitored in terms of hydrogen ingress throughout
their product lifecycle, and if necessary, measures taken to outgas these steels prior
to operation. Given that the energy required to free the hydrogen will influence the
temperatures required for a degassing treatment, deeper traps will require higher tem-
peratures to the extent that cost or temperature exceeds that which is economically
viable or to a point at which microstructural stability becomes of concern. Thus, when
selecting microstructural hydrogen traps for implementation within a steel, one must
consider the hydrogen ingress and microstructural stability at all processing stages, and
ensure that the processed microstructure is suitable for the degrees of hydrogen ingress,
the temperatures and the stresses expected through-life. What follows is a comprehen-
32
sive analysis of hydrogen interactions in steels, identifying the most common forms of
hydrogen trapping and those that show most promise in future microstructural design.
3.2.1 Microstructural interactions with hydrogen
Mechanisms of hydrogen trapping
Due to hydrogen’s high diffusivity and low solubility, a hydrogen atom’s random walk,
via interstitial sites, can be of significant distance, albeit temperature dependent. How-
ever, hydrogen can become trapped, such as on dislocations, grain boundaries and pre-
cipitates, reducing the free hydrogen available to induce embrittlement. Given that it
is the mobility of hydrogen and its ability to interact with defects that causes embrit-
tlement, trapping the hydrogen prior to embrittlement is paramount. Traps can also
be categorised by their saturation point, the finite amount of hydrogen a trap can re-
tain until saturation. Some traps are un-saturable however, such as gas voids, whose
contents can continually grow. Notably, it has been shown that steels with engineered
trapping sites show little or no degradation in mechanical properties when traps are
filled with hydrogen [61] [62]. It should be noted that within materials with a high
trapping capacity, hydrogen content can notably exceed the solubility limit, increasing
the total permissible hydrogen content within the material. Hence, when calculating the
solubility, the trapping capacity as well as solubility in the lattice needs to be included.
Hydrogen may migrate into traps in three ways: by the forces of strain fields, elec-
tropotentials and by diffusion into the traps. With respect to electropotential-induced
trapping, an electrical field surrounding electronegative features, such as specific impu-
rity atoms, attracts the positively charged hydrogen towards it. Similarly, with stress
fields, hydrogen can be attracted by the potentials created within a stress field, such as
those induced by defects and incoherent precipitates. Through a random walk, hydro-
gen can fall into traps purely by chance, provided the kinetic energy of the hydrogen
permits it, these traps can be anything from dislocations to the irreversible trapping of
incoherent titanium carbides [61].
Most microstructures contain a mixture of traps. However, a particular type of trap
will result in particular corrosion and mechanical properties and thus the selected traps
must also ensure the specified requirements for the intended application. It is difficult
to produce a steel that meets all application requirements. However, by evaluating
traps which could be successfully implemented in a variety of alloys in their a respective
applications, the impact of the work will be of greater significance. In the case of bearing
steels and structural steels for use in nuclear fusion and fission plants, larger incoherent
traps are best avoided, given their negative influence on properties such as fatigue,
notably rolling contact fatigue with respect to bearings [63]. The effects of different
traps on material properties is assessed during the alloy design work of this thesis.
33
Trapping on solutes and precipitates
Hydrogen trapping has been observed on a range of precipitates. Their hydrogen trap-
ping efficiency is dependent upon chemical composition, crystal structure, coherency
with the matrix and size.
As shown in Table 3.2, the activation energy of these traps is dependent on coherency.
Small precipitates (<10 nm) are usually spherical with coherent interfaces in order to
minimise surface energy. These will generally have a lower activation energy and are
mostly reversible traps. Smaller precipitates are favoured in bearing steels due to their
permissibly large number density, strengthening effects, the evidence that vanadium
carbide has optimal hydrogen trapping efficiency when at 10 nm diameter [62], and
the detrimental effects of large incoherent carbides on the rolling contact fatigue life
of bearings [5]. For intermediate sizes (10 nm<x<1 µm), precipitates are typically
plates or needles to minimise surface energy in situations where one plane or direction
is atomically similar between parent and product phases. Larger precipitates (>1 µm)
are typically incoherent and spherical in order to minimise volumetric free energy. Al-
though 10 nm length V4C3 carbides were found to be optimised for hydrogen trapping
capacity, no work has provided a conclusive relationship of coherency-to-hydrogen trap-
ping capacity. Establishing such a relationship for the carbides focused upon in this
thesis, those incorporated into thermodynamic and kinetic models and that show suit-
ability for bearing steels; carbides containing one or more of the following solutes: Nb,
Mo, W, Ti, V, Mn, Cr would be of significant benefit in the optimisation of hydrogen
embrittlement resistant steels. As such, the effect of precipitate size and composition,
the resulting coherency and the corresponding hydrogen trapping capacity will be of
paramount importance in the proceeding work of this thesis.
Trapping behaviour varies with each type of precipitate. It remains unclear for the vast
majority whether hydrogen is trapped within the precipitates, on the matrix-precipitate
interface or within the precipitate-induced stress field, all of which will be dependent
upon multiple factors. The characterisation of this trapping behaviour can be carried
out using 3D atom probe tomography coupled with a deuterium charging cell. Hydrogen
pollution already present within the test chamber prevents accurate readings of hydro-
gen levels, hence the use of deuterium to distinguish from the background hydrogen
pollution. An example of this technique is shown in Figure 3.4 for the evaluation of
titanium carbide, indicating the trapping of deuterium on the carbide-matrix interface.
The interaction of hydrogen with carbide, nitride sulphide and oxide interfaces is com-
plex, be it the lack of conclusive evidence or means of characterising trapping locations
and atomic configurations on the interfaces, or the complications involving nonmetallic
elements and their corresponding bonding effects. Given such complexity, interfaces will
contain a range of trapping sites of varied binding energy, although general orientation
34
Figure 3.4: 3D atom probe tomography maps of steel containing nano-sized TiC precipitates
(a) after deuterium charging (b) prior to deuterium charging. D - Deuterium [61].
and coherency should be reasonably consistent across the interface.
Szost [25] showed that nanoprecipitate hardened 52100 martensitic bearing steels can
have its hydrogen trapping behaviour fully characterised using a combination of both
thermal desorption analysis and melt extraction technique. She showed that 52100 can
absorb large amounts of hydrogen, most of which is weakly trapped on dislocations at
room temperature. However, with the precipitation of fine coherent vanadium carbides,
hydrogen trapping capacity was increased with greater binding energy. Thermodynamic
and kinetic modelling was combined to develop a novel hydrogen embrittlement resis-
tant bearing steel, similar to the products intended from this thesis. Szost showed that
current hydrogen embrittlement resistant steels were inappropriate for application as
bearing steels due to inadequate hardness. As such, a V4C3 nanocarbide precipitated
martensitic steel, more precisely a 52100 steel + 0.5 wt% V, was designed and tested.
The around 10nm diameter V4C3 carbides were found to act as reversible hydrogen traps
and resulted in improved hardness and strength with increased hydrogen embrittlement
resistance. The production route remained virtually unchanged to its base 52100 steel,
except for a more complex heat treatment schedule, requiring a peak temperature of
1200oC, however modern furnaces can reach up to 1300oC [5]. The schedule was as
follows: austenitisation (860oC for 15 minutes), a subsequent temperature spike to dis-
solve coarse V4C3 (1200
oC for 1 minute), a tempering step at 600oC where around 10
nm V4C3 particles form, a subsequent temperature spike to dissolve coarse cementite,
and finally a quench and tempering at 215oC, where fine cementite strengthening parti-
cles form [25]. Given the significant improvement in hydrogen embrittlement resistance,
hardness and strength, it is clear that the development of such nanoprecipitate hardened
bearing steels is feasible and V4C3 is a proven means of achieving improved hydrogen
embrittlement resistance in such steels.
35
Tungsten carbide, with its extraordinary hardness and temperature resistance, has been
used to produce high strength steels for use in many industries including the bearings
industry, providing high hardness [5], abrasion resistance and toughness. The divertor
armour used in the design of future fusion reactors and the International Thermonu-
clear Experimental Reactor (ITER), the next-generation fusion reactor currently under
construction, consists primarily of carbon fibre composites, carbon-coated tungsten and
consequently, tungsten carbide [64]. The binding energies of tungsten carbides to hydro-
gen are currently not available in the literature, and the hydrogen trapping mechanism
remains to be characterised. Vanadium carbide and niobium carbide are assumed to
trap hydrogen through coherency strain resulting from the matrix-carbide interface [25].
This shall also be assumed for tungsten carbide. With this assumption, tungsten carbide
warrants further investigation with regard to hydrogen trapping in metals, as the sig-
nificant increase in strength permissible with fine homogeneously distributed tungsten
carbide, primarily due to coherency strains, which will be quantified using TEM, could
result in a considerable affinity for hydrogen. Molybdenum additions are well known to
increase the coherency in steels of molybdenum-doped carbides, as shown in Figure 3.5
for a quarternary Fe-C-Mo-V martensitic steel, and as such will have an effect on hydro-
gen trapping. It has been reported that the mixed carbide (V,Mo,Fe,Mn)4C3 is better
in its ability to trap hydrogen in comparison to pure V4C3, due to the contributions
to coherency of its multicomponent nature. This indicates that there is an advantage
in using multicomponent carbides to develop hydrogen embrittlement resistant steels,
manipulating chemical composition to adjust coherency within a ferrite matrix. The
density of phase boundaries and the volume fraction of the respective phases will influ-
ence the maximum hydrogen trapping capacity. However, it is thought that hydrogen
trapping is dominated by the boundaries rather than chemical binding within the re-
spective phase. This will be evaluated in the proceeding work on vanadium carbide
containing bearing steel. Worthy of note with respect to such carbide hydrogen-traps is
that more stable carbides such as those of vanadium, tungsten, titanium and niobium
reduce the risk of internal decarburisation.
36
Figure 3.5: Hydrogen trapping capacity per M4C3 particle versus fraction of Mo in M of
M4C3 [62].
Trapping on defects and grain boundaries
Hydrogen can be trapped on dislocations, grain boundaries, voids, inclusions and many
other microstructural features. Dislocations typically provide low energy (around 27
kJ mol-1) reversible trapping sites, reducing their mobility upon trapping hydrogen,
similar to the effects of carbon trapping [15]. Hydrogen trapped in dislocations is liable
to be liberated under low temperatures and/or stresses given the low binding energy, the
magnitude of which is dependent upon the type of dislocation and the host material, as
shown in Table 3.2. Lino [65] describes how dislocations carry hydrogen atmospheres and
may meet microstructural defects and exchange hydrogen depending on the hydrogen
binding energies and interactions with these defects, this is similar to carbon-dislocation
interactions such as observed in the formation of martensite.
Hydrogen can accumulate in micro-voids, form new gas voids and induce void growth.
Micro-voids are low energy traps, albeit greater than typical dislocations, with a typical
activation energy of around 35 kJ mol-1 in pure iron. As hydrogen fills a void, the
internal pressure increases, potentially inducing cracking [3].
Hydrogen can also be weakly trapped upon grain boundaries, with a typical activation
energy in pure iron of around 17 kJ·mol-1, as shown in Table 3.2. This activation energy
is dependent upon grain boundary morphology, with low angle boundaries having sig-
nificantly increased energy similar to that of some dislocations, with activation energy
reducing with increasing misorientation angle [66]. In addition to trapping hydrogen,
grain and phase boundaries have been shown to provide paths of accelerated diffusion.
However, there is very little data, both with respect to modelling and experimentation,
37
or understanding on this subject. For substitutionally-dissolved impurities, the accel-
erated diffusion is attributed to the reduced vacancy formation energy in the excess
volume of the boundary. However, for interstitial hydrogen, which does not solely rely
upon vacancy interactions, it is thought that a reduction in the activation energies at
the boundary is required to accelerate diffusion.
Reviewing the literature, it is clear that the effects of trapping on hydrogen embrittle-
ment are complex. However, some fundamental conclusions can be drawn with respect
to the desired trapping mechanisms for application in steels. The selected traps must:
– Trap a high hydrogen content and limit the critical stress concentration so crack
initiation susceptibility is reduced.
– Be high in number and homogeneously distributed to provide a high saturation level
for hydrogen and limit mobile hydrogen content throughout the microstructure.
– Be deep but reversible.
Fulfilling these objectives should result in an improved hydrogen embrittlement resis-
tance for steels, although it is viable to extrapolate these into other materials. The means
by which to design such beneficial microstructures is through modelling the behaviour
of traps both in terms of mechanical properties and hydrogen trapping efficiency. As
such, the following section describes the methods for modelling hydrogen trapping and
proposes the development of a new model, based on one by Fischer et al. [67], which
could provide the means to assess novel microstructures accurately, in terms of their
complex hydrogen trapping behaviour, and at low computational cost using established
thermokinetic frameworks.
3.2.2 Modelling trapping
The effects of trapping on the diffusion of hydrogen is well established. Oriani [36],
based upon the work of McNabb and Foster [68], presented a diffusion model, shown in
the following equation, which incorporated the effects of trapping for the case of rapid
local equilibrium between trapped and mobile hydrogen atoms, based upon a constant
activation energy for diffusion into the immediate proximity of a trapping site.
Dapp = Dlatt × (CL ÷ (CL + CT (1− θT ))) (3.5)
where Dapp is the apparent diffusion coefficient, Dlatt is the diffusion coefficient in the
trap-free lattice, CL is the lattice concentration of hydrogen, CT is the trapped hydrogen
concentration and θT is the fraction of traps occupied. Oriani stated that it is the
trapping of hydrogen on interfaces, for the case of steels with minimal cold-work, that
38
accounts for the majority of hydrogen trapping sites rather than dislocations as was the
opinion of McNabb and Foster.
However, contrary to Oriani’s local equilibrium hypothesis, Koiwa [69] claimed that
since diffusion itself determines the rate of approach to equilibrium, the diffusion rate
cannot be quantified through such a local equilibrium assumption. Instead, the varia-
tion in activation energy local to traps should be incorporated. Many models extending
upon those mentioned here have been developed in recent years, incorporating the ef-
fects of lattice stress, temperature, inhomogeneous solubility and distinguishing between
reversible and irreversible trapping. Most significant with respect to this thesis, is the
model of Fischer et al. [67], which is discussed in detail in section 3.4 alongside its novel
incorporation into the thermodynamic and kinetic modelling software package MatCalc.
Luppo et al. [70] presented that irreversible trapping of hydrogen is favourable to re-
versible trapping with respect to resisting hydrogen embrittlement, given that hydrogen
reversibly trapped, given enough energy, can migrate to a neighbouring crack tip, en-
hancing fracture. As shown in Figure 3.6, the amount of hydrogen released during
outgassing from hydrogen-charged samples of low carbon steel is greater for quenched
martensite, with increasing tempering temperature decreasing hydrogen release, with
the same trend being found for embrittlement susceptibility as for hydrogen release [38].
Figure 3.6: Hydrogen released during outgassing from hydrogen-charged samples of low carbon
steel 516-G60 in the following conditions: normalised (N), fresh martensite (FM), quenched
and tempered for 6h at 453 K (QTL), and quenched and tempered for 1h at 773 K (QTH) [70].
Pressouyre and Bernstein [43] produced a model of hydrogen-induced crack initiation
from defects, within which it is claimed a critical trapped hydrogen concentration upon
these defects exists, above which hydrogen-induced cracking will initiate. It is suggested
39
that the critical concentration corresponds to the degree of decohesion caused by HED
and hydrogen pressure reducing the required stress to overcome the cohesive strength
of the defect. Pressouyre and Bernstein investigated Fe-Ti-C steels and classified de-
fects and traps, such as titanium carbide, as good or bad according to their effects on
increasing the critical hydrogen concentration for crack initiation and/or the time re-
quired to reach this concentration. Titanium was found to be a reversible trap when
merely a solute in the lattice. Titanium carbide, here an irreversible trap, was found
to be singificantly beneficial in terms of hydrogen embrittlement resistance. Worthy of
note, Pressouyre and Bernstein found that the presence of homogeneously distributed
titanium carbide improved resistance to intergranular cracking and lowered hydrogen
diffusivity.
The difficulty in implementing this model for complex steels is how to effectively quantify
the hydrogen concentration and the stress on the numerous defects at crack initiation.
As has been shown, the trapping concentrations can be quantified for complex struc-
tures using models such as that of Fisher et al. [67]. It is feasible to model stress
distributions in complex microstructures using a finite element package with the means
of microstructural input from thermokinetics and/or microscopy. In combination with
hydrogen concentration distributions, this could provide the means to evaluate local
stresses and hydrogen contents within a microstructure. However, the primary draw-
back with this method is the lack of experimental data with respect to the critical
concentration to stress relationship to initiate a crack and the difficulty in quantifying
this experimentally for complex microstructures. In addition, an accurate hydrogen
concentration with a precise trapping location for an individual trapping species, such
as whether interfacial trapping or bulk trapping for a precipitate, something which will
be significant in its effects on crack initiation, is currently unavailable. It is implied,
given the varying susceptibility to hydrogen-induced crack initiation observed for differ-
ent microstructural features such as precipitates, that trapping location is of significant
importance. Notably, with respect to microstructural features that are known to be
potential crack initiation sites, such as large incoherent precipitates under rolling con-
tact fatigue, whether the hydrogen is trapped within the precipitate or on the interface
will likely influence the mechanical properties of the precipitate and its crack initiation
susceptibility. At present, the most feasible means of establishing atomic scale trapping
locations experimentally and by modelling respectively is through atom probe tomogra-
phy [61] and density functional theory, the limit of the latter being typically nanoscale in
its unit cell size and thus only through cluster expansion methods can larger complete
features be assessed, such as incoherent interfaces or complex crystal structures [71].
Given the evident need for the characterisation of the micro-mechanical effects and lo-
cation of hydrogen concentrations in steels, the following section provides an overview
of techniques presented in the literature, as well as proposing a novel technique, that
40
offer the means to achieve the desired characterisation.
3.2.3 Mechanical testing
Numerous experimental [17,22,24–26,72,73] and theoretical [1,15,32,64,67,74,75] works
have attempted to evaluate hydrogen embrittlement in a range of materials. By means
of a brief review of the literature, a significant degree of inconsistency can be observed
with respect to the effects of hydrogen on mechanical properties, even within single
publications [43]. Early inconsistencies are primarily due to instrumentation limitations
and the resultant scatter. Despite these inconsistencies prevalent within the literature,
a number of methods to evaluate the effects of hydrogen have been shown to produce
consistently distinguishable and more importantly, quantifiable results in a range of
materials. The following sections describe these methods and their potential application
in this work.
Comparing the mechanical properties of charged and uncharged samples is the funda-
mental method for quantifying the effects of hydrogen on such properties. However,
the wide range of set parameters for such tests have great influence on the resultant
behaviour and can illegitimise results with respect to the intended relevant applications.
As will be shown, the nature of the hydrogen charging will influence, as will the loading
conditions. It is these types of variations that have induced the discrepancies within the
literature for identical materials [76]. The effects of such parameters were evaluated by
Bergmann [77], measuring ultimate tensile strength whilst in either air, post-hydrogen
charge, or hydrogen i.e. in situ hydrogen charging (varying temperature and pressure).
Bergmann found that in situ hydrogen charging provides a more pronounced reduction
in the ultimate tensile strength than that of pre-charged samples at room temperature,
but for testing using high pressure hydrogen charging at room temperature, little reduc-
tion is observed for in situ charging compared to ex situ. These variations indicate that
any novel steels developed in this work must be tested in a prototypical environment of
the relevant applications to ensure meaningful results.
A noteworthy parameter to take into account is the temperature at which the hydrogen
is charged into the sample as hydrogen trapping efficiency is temperature dependent.
An example of the hydrogen charging temperature dependence was presented by Matsui
et al. [76], showing the effects of in situ hydrogen charging on the flow stress of high pu-
rity iron at a range of temperatures under tensile testing. Figure 3.7 shows that at low
temperature, a softening effect was observed for high purity iron, whilst at high temper-
ature, a hardening effect. Matsui et al. concluded that the corresponding causes were
interactions between hydrogen and screw dislocations and hydrogen and edge disloca-
tions respectively. It should be noted however, that the tensile testing methods used in
this study prevent any meaningful conclusions being deduced with respect to hydrogen’s
41
effects on plastic deformation as the samples cannot be affirmed as microstructurally
identical, nor can the dislocation behaviour be evaluated with enough precision.
Figure 3.7: The effect on flow stress of high purity iron for in situ hydrogen charging at
different temperatures. Specimens were tested at a strain rate of 8.3×10-5s-1 [76].
As has been shown in section 2.1.2, for use in substantiating the HELP mechanism,
environmental TEM analysis provides the means to precisely characterise the samples
microstructurally, and the dislocation behaviour, such as dislocation nucleation, multi-
plication and motion. In situ straining of metals in an environmental TEM has been
utilised successfully to study the effects of hydrogen [78,79]. By confining the hydrogen
gas to the sample region through various apparatus assemblies, one is able to perform
TEM analysis with in situ hydrogen charging. In addition to these charging systems,
various in situ straining methods have been developed and successfully used on a range
of materials [80], as shown in the in depth example of section 2.1.2. The primary draw-
back of this technique is that of the gaseous pollution of the chamber and the difficulty
in simulating real conditions.
42
The two primary experimental techniques used to evaluate hydrogen embrittlement
have resulted in two suggested mechanisms, HED from conventional mechanical testing
and HELP from environmental TEM testing. However, a promising technique, com-
bining both TEM’s micro-characterisation and conventional testing’s replication of real
conditions, is that of nano indentation atomic force microscopy (NI-AFM). Barnoush
et al. [81–83] has shown the unique benefit of NI-AFM, its ability to perform nano in-
dentations (100 nm - 10 µm diameter) submerged in liquid, or more specifically for
hydrogen charging, an electrolyte, allowing in situ hydrogen charging and characterisa-
tion for more realistic conditions. Barnoush et al. used NI-AFM on a range of materials,
including steels, identifying its ability to identify the onset of plasticity in very small
volumes of hydrogen charged material. It was identified that hydrogen reduces the
required stress to initiate plasticity, through the reduction of the shear modulus, dislo-
cation line energy and stacking fault energy, reducing the critical stress for homogenous
dislocation nucleation. Barnoush et al. stated that, for a perfect crystal exposed to hy-
drogen, the change in shear modulus was related to the reduction in crystal cohesion,
and the reduction in dislocation line and stacking fault energy are due to the reduction
in defect formation energy. This implies that neither HELP nor HED is the cause of the
embrittlement but rather the reduction in the cohesion and defect formation energy.
Although in situ electrochemical charging NI-AFM provides the desired combination
of nanoscale characterisation and realistic hydrogen-charging environment, it does not
afford the opportunity to truly isolate microstructural features with respect to their
mechanical properties in the presence of hydrogen.
Hydrogen trapping characterisation using hydrogen charged nanopillar com-
pression tests
A novel technique to be developed in this thesis is that of a hydrogen-charged nanopillar
compression test. As previously stated, it is necessary to establish where hydrogen is
trapped with respect to microstructural features such as precipitates, and the conse-
quence to the features’ mechanical properties. The proposed methodology is as follows.
A typical microstructurally complex steel will contain carbides of various sizes, which in
turn may have different hydrogen trapping characteristics both in terms of capacity and
location: interfacial or bulk trapping. A sample of such a complex material shall have
nanopillars ion-beam-milled to produce pillars of solely the microstructural feature of
interest, such as a precipitate. The size of the feature thus dictates the required pillar
size. Of course, the morphology of some microstructural features, such as martensitic
plates, will likely be impossible to produce a pillar from and as such are inappropriate for
investigation using this method. By means of a typical electrolytic hydrogen charging
cell, an ion-beam-milled sample is then charged through submersion in the electrolyte
43
with the surface of the nano-pillars shielded from hydrogen ingress through sealing the
surface in argon using a PVC cover, sealed around the nano-pillars’ location, providing a
pocket of protective argon to prevent oxidation during charging. Consequently, charging
will provide hydrogen ingress through all surfaces other than that upon which the nano-
pillars lie, preventing ingress directly into the pillars from their surface. This better
simulates the realistic hydrogen ingress from the bulk matrix steel to/into precipitates.
Once charged, the sample is then nano-indented using a flat indenter tip typical of
compression testing. Due to the method of hydrogen charging described, the pillars may
or may not show a change in properties in comparison to uncharged ones, dependent on
whether hydrogen is trapped within the precipitate or on the interface.
If hydrogen is only trapped on the interface, little change will be observed in the me-
chanical properties of the nano-pillar, more specifically the stress to induce dislocation
nucleation and slip, will be unchanged. Of course, effects of orientation, pillar size, resid-
ual stresses and compositional variation will cause changes in the resulting mechanical
response of pillars. If however, the hydrogen is trapped within the precipitate, a change
in mechanical properties should be observed in comparison to an uncharged sample,
most likely in the form of a reduced stress for dislocation nucleation and slip, imply-
ing a combination of both HELP and HED. However, until tested, the experiment’s
hypothesised conclusions cannot be assured. This experimental method affords one an
alternative to the difficult and costly method of atom-probe tomography for characteris-
ing hydrogen trapping locations in feasible microstructural features for nanopillar milling
and provides the means of nano-scale hydrogen-dependent mechanical property evalua-
tion. The results of such an experiment could provide the parameters for the intended
density-functional-theory calculations with respect to trapped hydrogen location, and
the mechanical property effects could be used to establish the critical hydrogen-to-stress
relationship for crack initiation, which in turn can provide input into microstructural
FEM models. The first step in affirming this method is to evaluate an established mi-
crostructural feature for which the hydrogen trapping location is affirmed, such as TiC
for deuterium trapping [61].
Given the importance of hydrogen embrittlement with respect to bearing steels, and the
need for established resistant microstructures suitable for applications in the bearing
industry, the following section provides a review of the most prominent steels in the
industry, the typical modes of failure, and the methods by which hydrogen embrittlement
susceptibility has attempted to be reduced in these steels.
3.2.4 Bearing steels
Bearing steels are typically designed for high strength, hardness and wear resistance,
and with such a diverse range of applications, including the automotive, aerospace,
44
nuclear and marine industry, many are prone to accelerated wear and damage from
the harsh environments in which they operate. The primary failure mechanism of such
bearings is rolling contact fatigue, a mechanism significantly accelerated by hydrogen
[26]. The following sections provide a brief description of bearing steel design, processing,
mechanical properties, and rolling contact fatigue.
Design and Processing
Although there are many different bearing steels with a significant range of composi-
tions, a number have come to prominence since their introduction; these are shown in
Table 3.1. Except for the less common SCM420H, these steels all share a notably high
carbon content between 0.7 to 1.05 wt% and solute contents less than 3 wt%. Typi-
cally, carbon content is reduced in high temperature bearing steels and increased when
through-hardenability is required. With increasing carbon content comes increased wear
resistance, higher phase fractions of carbides and retained austenite, and a reduction in
the martensite start temperature [5]. Through quenching in oil or salt from a mostly
austenitic microstructure, these steels can be made martensitic. Chromium, manganese
and molybdenum increase hardenability [84]. In addition to contributions to harden-
ability, Mo stabilises carbides, can increase their coherency with the matrix, provides
resistance to decarburisation, and improves toughness due to retarding P segregation in
prior austenite boundaries, as manganese does for sulphur segregation [5]. Manganese
extends the time required for bainite and pearlite transformation, as does aluminium,
improves grain refinement, improving toughness, and along with Co (although not in
bearing steels) and Cr, is used to reduce the dissolution rate of spheroidised cemen-
tite. The ferrite-austenite transformation temperature is also reduced by the addition
of manganese. Silicon has been shown to delay cementite formation, most notably at
low temperatures. The mechanism by which this occurs is not fully understood [5, 85].
Of all the popular bearing steels, 52100 type steels (100Cr6) dominate bearing steel
production, and were first employed over a century ago. Such steels are typically sup-
plied hot-rolled with a pearlitic structure and will require annealing to spheroidise the
cementite [86] for further machining, after which heat treatments are applied to create
the desired hard martensitic microstructures required for improving rolling contact fa-
tigue life. Chromium is the primary alloying element in 52100 steels. It has been shown
to retard bainite formation and improve corrosion resistance through the formation of
a chromium-rich protective oxide layer, the presence of which defines a stainless steel.
For such chromium containing steels, cobalt (though not utilised in bearing steels) and
aluminium can be added to accelerate the ferrite-to-austenite transformation. In addi-
tion, these elements influence carbide formation, with aluminium retarding cementite
precipitation, given its insolubility in this carbide, and cobalt encouraging cementite
45
formation, given it is a carbide former. Of special note for bearings operating under
neutron-irradiation, such as those used in control-rod drive mechanisms of some nuclear
reactors, cobalt is undesirable given its activation and corresponding fission products.
Nickel however is desirable as it decreases the ductile-to-brittle transition temperature,
which is increased under neutron irradiation, whilst improving toughness.
A typical heat treatment schedule for martensitic 52100 steel consists of an initial
austenitisation treatment either in the austenite-cementite field (residual cementite can
increase wear resistance) or in the austenite field of the corresponding phase diagram
shown in Figure 3.8. Once the steel is austenitised, it is quenched to form hard marten-
site and tempered at low temperatures to provide toughness and to precipitate fine iron
carbides such as cementite for increased strength [63]. The resulting microstructure,
shown in Figure 3.9, is thus a martensitic matrix incorporating some retained austenite,
large residual carbides and nano-sized transition carbides, the relevance of each with
respect to rolling contact fatigue is discussed in the following section.
Figure 3.8: Phase diagram of a 52100 steel. The dotted lines indicate the typical carbon range
of 52100 [87].
Current steels regarded as resistant to hydrogen embrittlement are typically of low
carbon content, which is typically to improve weldability, formability, ductility and
corrosion resistance [5]. Titanium, niobium and vanadium additions are used to pro-
vide hydrogen trapping sites, both in the form of carbides and solutes, and to increase
strength and wear resistance. In addition to these effects, these elements have been
46
Figure 3.9: Optical microscopy image of the microstructure of 100Cr6 steel after austenitisa-
tion at 850 C and quenching in oil; [87]
added to retard austenite grain growth at high temperatures typical of austenitisation
treatments, and improve carbide stability, which is of appeal in high temperature ap-
plications.
With current commercial as-rolled 52100 steels, the total hydrogen content achievable is
said to be around 1 ppm, although electron beam melted 52100 type steel has been shown
to contain as little as 0.4 ppmw [5] [88]. The mean hydrogen content for a 52100 15 mm
diameter bar can be as high as 8 ppm. The distribution of hydrogen is fairly uniform
but ranges from 7-15 ppm to 0.5-2.5 ppm at the surface and core respectively. These
contents are reduced during further processing [5] [72]. The hydrogen concentrations
achievable in a variety of bearing steel production processes are shown in Table 3.4.
Table 3.4: The effect of melting processes on gaseous impurity concentrations in M50 bearing
steels ( [5] after [89]).
Process
Hydrogen
(ppmw)
Nitrogen
(ppmw)
Oxygen
(ppmw)
Air melting (AM) 4.8 150 67
AM + vacuum degassing 3.2 120 46
AM + 1 vacuum arc remelt (VAC) 1.8 80 37
AM + 2 VACs <1 60 9
AM + 3 VACs <1 50 5
Vacuum induction melting (VIM) <1 85 21
VIM + VAC <1 60 6
47
Bainitic Bearing Steels
Although currently not implemented commercially, a nanostructured superbainitic steel
has been considered as a possible alternative to the market-dominating martensitic bear-
ing steels [90]. Such superbainitic steels are of great appeal with respect to hydrogen
embrittlement resistance. A softer lower-bainitic structure of 52100 steel has been com-
mented as potentially affording a longer rolling contact fatigue life when in the presence
of hydrogen due to the increased ductility and toughness ( [5] after [91]). Using a bainitic
steel is of significant benefit with respect to retained austenite, as it has been shown
in 52100 that almost the entirety of the austenite is consumed during transformation
and as such the retained austenite fraction is minute (<1 vol%). Given that bainitic
transformation heat treatments are isothermal, the risk of cracking upon quenching,
prevalent in martensitic 52100 [92], is avoided. An optimised composition, such as one
depleted in manganese but precipitate hardened, and its corresponding heat treatment
can be produced to reduce the transformation time, in turn reducing the processing cost
that is typically high for bainitic steels due to the long holding times required.
It is feasible that a nanostructured bainitic steel, consisting of extremely fine platelets
of bainite decorated with nano-sized carbides (around 10 nm in length) within a carbon-
rich austenite matrix could provide a suitable bearing steel, feasible in terms of both
manufacture and operation. As presented by Solano et al. [90], nanostructured bainitic
steels, absent of nano-carbides, have been shown to have hardnesses of up to 670 Hv,
strengths exceeding 2 GPa, and toughnesses of up to 40 MPa·m-2. For carbide-free
nanostructured bainite, retained austenite-bainite interfaces are found to act as hydro-
gen traps [25] and the retained austenite found to limit hydrogen absorption [93].
Superbainite is currently commercially available as armour but is known to have excep-
tional abrasion, rolling-sliding wear resistance, uniaxial fatigue toughness and rolling
contact fatigue toughness [90]. The austenite content, once exceeding the percolation
threshold, allows the structure to resist the penetration of hydrogen. The failure mecha-
nisms of such steel bearings were found to differ from conventional damage mechanisms
of 52100 bearings. Solano et al. concluded that the degradation mechanism is that of
ductile void formation at interfaces, followed by the growth and coalescence into larger
voids, leading to fracture along the direction of the softer phase. Due to the high density
of interfaces, a greater number of dispersed voids form and result in cracking, observed
in the area of sub-surface peak stress expected in rolling contact fatigue [90].
Rolling Contact Fatigue
Rolling contact fatigue is the predominant failure mechanism in bearing steels, typically
identified by three mechanisms of spalling: spalling due to cracks initiated below the sur-
face, due to insufficient lubrication or surface roughness/defects. Despite the significant
48
improvements in steel cleanliness of recent decades, defects, such as inclusions, continue
to play a dominant role in rolling contact fatigue failure and as such, subsurface failure
is a common failure mechanism [5].
During rolling contact fatigue, plastic deformation manifests itself through the formation
of dark-etching regions and white-etching areas, as shown in Figures 3.10(b) and (c)
respectively, due to the repeating contact of the two respective bearing components.
Prior to the formation of white etching regions, dark etching regions, around 0.3 mm
below the surface, have been observed to form on rolling contact fatigue test specimens.
Later in the bearing’s life, white etching areas, also known as butterflies, as shown in
Figure 3.10, typically initiate at defects such as voids or inclusions, but can also initiate
from carbides [93].
The term white etching comes from the white appearance of the microstructure after
being etched, typically with Nital or Picral, and observed under optical microscopy.
These regions contain nanocrystalline bcc ferrite that is supersaturated with carbon [94].
However, recent Atom Probe Tomography (APT) investigations have shown that carbon
is segregated at the grain boundaries rather than in the ferrite grains [95]. The level of
carbon saturation of ferritic nano-grains in these regions has not been confirmed. Grains
were found to be equiaxed 10 nm ferrite grains within the region, whereas the original
material has a plate-like martensite structure with carbides. The hardness of the white
etching regions has been found to be 30 to 50% higher than the surrounding matrix,
which has been suggested to be related to the ultrafine, nanocrystalline structure and
carbon supersaturation [94].
Dark etching regions are deformed microstructural features that appear dark under
optical microscopy after being etched. They are often seen at a depth from several
hundred of micrometres under the surface which corresponds to the depth of maximum
shear stresses [96]. Their typical width is 2 mm and increases with running time and
contact pressure. Since DERs manifest themselves with Nital or Picral, which are known
to reveal carbide/ferrite interfaces, it implies that stronger etching can be attributed to
the presence of larger carbides. In addition, larger carbides result in a reduction in C
concentration in solid solution which explains the observed hardness decrease in such
regions. As such, dark etching regions are believed to be regions of dislocation-assisted
tempering [96].
Rolling contact fatigue life can depend on carbide size and distribution, inclusion con-
tent, such as aluminium oxides and silicates, porosity, and other defects [5]. Such
microstructural features can act as crack initiation sites, which once initiated under
rolling contact fatigue can propagate and develop into butterfly wings, decreasing fa-
tigue life [63]. Given such dependence on inclusions, steel cleanliness is critical to fatigue
life. However, despite the efforts of bearing steel manufacturers, oxygen and sulphur
contents are still of issue and methods must thus be taken to account for inclusions in
49
Figure 3.10: Optical micrographs of a) early dark area formation b) fully developed dark
etching regions with magnified 30 degree bands and c) white etching region bands forming at
80 and 30 degree angles to the tangent of the raceway [97].
Figure 3.11: Micrograph of a butterfly wing formed parallel to the rolling direction (left to
right of micrograph) in 52100 steel during rolling contact fatigue testing [98].
designing novel bearing steels.
As with general cracking mechanisms, the role of hydrogen in initiating and progress-
ing rolling contact fatigue failure is not fully understood but its significance is well
50
established. Kino and Otani showed the reduction in rolling contact fatigue life with
hydrogen content [24], as did Murakami et al. [99]. In the early stages of rolling contact
fatigue, damage accumulation is governed by dislocation glide which in turn is controlled
by obstacles to such gliding, resulting in an accumulation of defects within the stress
fields, holding back both the yielding and proceeding work hardening mechanism [25].
This stage, known as the shakedown stage, can be regarded as a constant applied-stress
fatigue test, the oscillating stress amplitude is fixed but the plastic strain per cycle
decreases, due to the material’s cyclic hardening, until, on a macro scale, the strain
ceases to increase [5]. Elastic shakedown is said to occur if all the components of the
plastic deformation tensor become constant [31]. The shakedown limit depends upon
the steel’s residual stress and microstructure. The presence of retained austenite will
increase the shakedown period due to the transformation of austenite into martensite
under stress [100], the consequences and resulting failure of such microstructural vari-
ations and residual stress contributions are beyond the scope of this thesis. For more
information please refer to [5].
A steady-state is reached after the shakedown stage, with the rate of plastic-damage
significantly lower than that during shakedown. The control within this steady-state
stage is critical to rolling contact fatigue life, as during this stage, the damage rate is
limited by dislocation glide and is increased with dislocation mobility, and this is where
hydrogen is of great significance. Given that dislocation climb is negligible at typical
bearing operating temperatures (<100oC) due to insufficient energy to promote self-
diffusion and dislocation pinning due to carbon, dislocation mobility is limited. However,
in the presence of hydrogen, dislocation mobility is greatly enhanced, accelerating the
fatigue damage process by enhancing the self diffusion process [101].
To evaluate the rolling contact fatigue life of novel hydrogen resistant bearing steels, ide-
ally one would carry out full-scale bearing endurance testing, reproducing the operating
environment as realistically as possible, including the decomposition of lubricant and
any other hydrogen ingress mechanisms active in the intended application. However,
these tests are un-economical and as such, accelerated testing using bearing elements
is the most common form of bearing life analysis. These simpler tests have been found
to reproduce bearing lives accurately compared to full-scale testing, providing a cost-
effective, widely applied, and thus reproducible, method of evaluating the rolling contact
fatigue life of novel bearing steels. It is permissible, using such a testing method, that
in situ hydrogen charging or a precharged specimen could be utilised and have its hy-
drogen embrittlement resistance evaluated.
It should be noted, however, that results from element testing can be misleading due
to differing stress states in comparison to the intended operational environment. Mi-
crostructural effects have been shown to differ between different test methods [96] and as
such, it is essential that consistency is kept when comparing and conducting experiments
51
and in evaluating the resulting data on rolling contact fatigue life.
3.3 Summary
A review of the published literature on hydrogen embrittlement of steels has been pre-
sented, from which a number of conclusions can be taken.
The bulk of the literature concerns the effects of hydrogen on the mechanical properties
of bearing steels. Hydrogen has been shown to increase dislocation mobility, drastically
reduce the RCF life and enhance the formation of WEAs and WECs [102] [103]. The
most supported mechanisms of embrittlement are hydrogen enhanced decohesion (HED)
[104] and hydrogen enhanced localised plasticity (HELP) theory [6]. Despite the lack
of a conclusive theory, it is established that hydrogen enhances plasticity local to crack
tips, increases slip deformation and reduces the critical stress required to propagate
cracks [102].
Almost all instances of hydrogen embrittlement begin with the decomposition of molecules
into atomic hydrogen at the material surface. During the decomposition of lubricant
oil, the mechanism thought to be most prevalent to bearings, the composition of the oil
plays a crucial role on the degree of hydrogen ingress and consequent embrittlement.
Active sites are generated upon the nascent surface of the bearing steel after the re-
moval of the oxide layer due to rubbing [27]. Tribochemical decomposition of the oil
then occurs upon these active sites with hydrogen and gaseous hydrocarbons desorbed
as reaction products. The desorbed amount of hydrogen and hydrocarbons increased
linearly with rolling velocity and parabolically with load [27]. In addition, there appears
a critical load beyond which decomposition occurs [27]. Oil containing water may also
increase ingress rates due to water splitting.
The control of hydrogen content in steels is an important task for steelmakers: just a few
parts per million of dissolved hydrogen can induce hydrogen flakes (hairline cracks), blis-
tering, loss of ductility and increased porosity [105], especially in larger castings [106].
Steelmakers will often carry out a degassing treatment after solidification. Degassing at
a low temperature (<200oC) can liberate weakly trapped hydrogen and concentrate it in
deeply trapping microstructural features which can, in some cases, increase the embrit-
tlement susceptibility due to higher stress concentrations [57]. With current commercial
as-rolled 52100 steels, the total hydrogen content permissible is said to be around 1 ppm,
although electron beam melted 52100 type steel has been shown to contain as little as
0.4 ppmw [5, 88]. The mean total hydrogen content for a 52100 15 mm diameter bar
can be as high as 8 ppm, the distribution is fairly uniform but ranges from 7-15 ppm
to 0.5-2.5 ppm at the surface and core respectively, these contents are reduced during
further processing [5, 72].
52
Hydrogen trapping with in-life degassing, through-life hydrogen trapping, and hydrogen
ingress prevention are possible means to reduce hydrogen embrittlement. When selecting
microstructural hydrogen traps for implementation within a steel, one must consider
the hydrogen ingress and microstructural stability at all processing stages, and ensure
that the processed microstructure is suitable for the degrees of hydrogen ingress, the
temperatures and the stresses expected through-life [107].
RCF tests carried out on a novel vanadium containing variant of 100Cr6 bearing steel,
100Cr6+0.5V, showed that the presence of nanosized vanadium carbides resulted in
no white etching areas after similar conditions to that which produced areas in its
vanadium-free variant, indicating the reduction in hydrogen-embrittlement susceptabil-
ity with such microstructures [108]. However, more tests are required to conclusively
affirm the benefits of vanadium carbides to reducing hydrogen-embrittlement suscepti-
bility, and potentially increasing RCF life in the presence of hydrogen [108].
Although this review focuses on that of bearing steels, the application of such hydrogen
embrittlement resistant design technology is far reaching. As discussed in the following
section, a notable application is to nickel-based alloys where hydrogen embrittlement is
also of concern. The application of nanostructured microstructures containing suitable
interfaces for hydrogen trapping could be implemented in a similar way for such alloys,
with a number of hydrogen traps having already been identified, such as TiC [109,
110]. Such nanostructures are also relevant with respect to radiation embrittlement
resistance, as nanoparticles have been shown to provide defect recombination sites as
well as trapping hydrogen and helium [111]. Although significant research is being
conducted to develop technologies for resisting hydrogen embrittlement, it is evident
that the entire manufacturing process and intended operating environment must be
evaluated to develop novel microstructures that suit both the manufacturing routes and
operational environments.
53
Chapter 4
Nickel-based alloys
4.1 Introduction
Nickel-based alloys show outstanding corrosion and high temperature resistance. Such
alloys are closely related to austenitic stainless steels but more highly alloyed, primarily
with nickel, chromium and molybdenum in order to enhance their corrosion resistance.
These are termed “superalloys” when high temperature strength and oxidation resis-
tance is achieved. Superalloys incoporate chromium, cobalt, titanium, and aluminium
among others elements, and can even be produced as a single crystal for specific appli-
cations. Such alloys can have strengths above 1000◦C that exceed that of many steels
at room temperature. As such, these materials are of great appeal for application in the
corrosive and high temperature environments present in nuclear reactors.
Nuclear power plants have largely been of two design types, developed since their fruition
in the 1950s. However, new designs are coming into operation as the first generation
of reactors are decomissioned. Originally developed for the propulsion of submarines
and naval ships, pressurised water reactors (PWRs) generate the majority of the world’s
nuclear electricity. A nuclear reactor produces electricity through the controlled release
of energy from the splitting (fissioning) of atoms of certain elements. The energy released
is used as heat to produce steam to generate electricity via turbines, as with a typical
fossil fuel plant. In order to understand the proceeding chapter, one must be aware of
the following fundamental components and aspects of a PWR reactor, shown graphically
in Figure 4.1:
– The Fuel assemblies. Usually pellets of uranium oxide (UO2) are arranged in tubes
to form fuel rods. These rods are arranged into fuel assemblies in the reactor core.
A PWR may have fuel assemblies of 200-300 rods each, arranged vertically in the
core, with a large reactor having around 150-250 fuel assemblies. These assemblies
make up the reactor core.
54
– The reactor pressure vessel (RPV). Most commonly a low carbon steel vessel con-
taining the reactor core, moderator/coolant and so on. Notably, the RPV and the
components it contains, are those most at risk of radiation damage (often quantified
in terms of displacements per atom (dpa)), discussed in more detail later.
– The Coolant. A fluid (water in the case of a PWR) circulating through the nuclear
core in order to draw heat from it. In light water reactors (LWR), a type of PWR,
there is a primary and a secondary coolant circuit (where the water becomes steam).
– The steam generator (SG). The part of the cooling system for PWRs where the high
pressure primary circuit coolant, bringing heat from the core, is used to produce
steam in the low pressure secondary circuit. This steam then proceeds to feed the
turbines. This is effectively a heat exchanger, much like a radiator in one’s house.
Reactors may have multiple loops, each with its own steam generator. Many
SGs have had to be replaced prior to expected end-of-life, due to the corrosive,
high temperature environment. The primary coolant water is channeled through
SG tubes to transfer heat. These tubes are designed so not to vibrate, fret or
accumulate corrosive deposits that may impede flow or induce corrosion. Tubes
which fail/leak are plugged and the remaining tubes accommodate flow accordingly.
Figure 4.1: Schematic showing the basic design of a pressurised water reactor, indicating the
key components.
Nickel-based alloys, because of their superior high temperature strength, toughness,
creep and corrosion properties, have been proposed for various in-core applications in
55
Gen IV1 nuclear reactor systems [111, 113] fusion plants [114] and neutron spallation
sources [111], requiring structural integrity under doses of up to hundreds of dpa [111].
However, the effects of radiation on these alloys, in comparison to steels, are little un-
derstood, with many alloys having little or no associated data. As nickel-based alloy
components previously in operation are decommissioned [115] and surveillance speci-
mens are evaluated [111], several publications have emerged in recent years presenting
significant evidence on the mechanisms and effects of irradiation embrittlement.
Rowcliffe et al. [111] summarised the applications of nickel-based alloys in Gen IV reac-
tor designs and the corresponding concerns with respect to radiation embrittlement. For
the core components of the supercritical water reactor (SCWR) and sodium fast reactor
(SFR), normal operating temperatures range from 300 to 620◦C [114], whereas for the
gas-cooled fast reactor (GFR) and the lead-cooled fast reactor (LFR), the operating
temperature range is approximately 400 to 850◦C [116,117]. For the very high tempera-
ture reactor (VHTR), normal operating temperatures range from 600 to 900◦C [118]. In
the VHTR, a number of structural components are considered to be made from nickel-
based alloys, notably, in terms of extremes of dose, the core barrel and the inside shroud
where doses will likely be up to several dpa at end of life (50 to 60 years) [111].
It is seen that neutron radiation can induce significant reductions in the ductility of
nickel-based alloys over a range of temperatures, with considerable degrees of swelling
observed [119,120], typically attributed, speculatively, to irradiation-induced grain bound-
ary helium bubble formation. As such, a conclusive review is necessary to better es-
tablish the effects of neutron radiation on such alloys in the pursuit to better predict
embrittlement and identify the means by which it may be limited. By reviewing the
literature as a whole, a number of issues relevant to the embrittlement of nickel-based
alloys have been identified and methods of remediation proposed.
This chapter reviews the irradiation data considered most relevant to nickel-based alloys
presently in operation and those proposed for Gen IV reactors [111, 114, 116–118] and
other relevant applications. This work pays special attention to Alloy 600 due to its prac-
ticality for cross-alloy comparison, the large amount of data available, and its common
application in currently operating reactor systems. Irradiation assisted stress corrosion
cracking (IASCC) of nickel-based alloys in light water reactor (LWR) environments is
also reviewed.
1The International Generation-IV Initiative aims to foster the research and development necessary to estab-
lish a new generation of nuclear energy systems. Activities are managed by the Generation-IV International
Forum (GIF). The Gen IV systems, which comprise both the reactors and their associated fuel cycle facili-
ties, are intended to deliver significant advances compared with current advanced light water reactors (Gen
III) with respect to economics, safety, environmental performance, and proliferation resistance for commercial
deployment by 2030 at the earliest [112].
56
4.2 Background
Nickel-based alloys form a large class of materials possessing good mechanical strength,
toughness, corrosion resistance and creep properties at high temperatures [121]. It is
well established that the effects of neutron radiation on nickel-based alloys in thermal
reactors are defying predictions that were made based upon fast reactor irradiation
data [122,123]. As nickel-based alloy components face significant doses in ageing thermal
reactors [124] and their application becomes significant in Gen IV reactor designs [125],
the literature pertaining to the effects of radiation on such alloys must be reviewed so to
enable better structural integrity assessments for the relevant components and further
the development of radiation resistant alloys.
Alloy 600, a nickel-based alloy, the composition of which is shown in 4.1, has often
been utilised in steam generator tubing in various designs of nuclear reactor [126]. In
addition to this, Alloy 600 has been utilised in Reactor Pressure Vessel (RPV) heads
and in other regions of considerable neutron flux, such as core barrels [127]. Alloy
600 is the base composition for many nickel-based alloys, notably Alloy 690 (additions
of chromium for corrosion resistance) [127] and X-750 (additions of Ti, Al and/or Nb
added for precipitate hardening) [128]. Further additions of Mo make Alloy 718 [129].
An additional thermal treatment is given to Alloy 600, often thus notated as 600TT,
to promote intergranular carbide precipitation and reduce grain boundary chromium
depletion, increasing stress corrosion cracking (SCC) resistance [127]. For the rest of
this review, Alloy 600 in its thermally treated condition will be referred to as Alloy
600TT.
Many nickel-based alloys were investigated under neutron irradiation programmes in the
late 60s and 70s with significant reductions in ductility observed during post-irradiation
tensile testing [120, 130–134]. This effect is often attributed to helium embrittlement
due to the transmutation of nickel, as is discussed later alongside alternate mechanisms.
However, almost the entire literature does so without conclusive evidence, notably that
of microscopy.
The initial decision to investigate nickel-based alloys for use as in-core components was
taken due to their excellent high temperature creep resistance and strength [111]. Stain-
less steels had become unfavourable in such components due to void swelling observed in
type 300 stainless steels, first discovered by Cawthorne and Fulton [135]. Nickel-based
alloys were not initially considered under risk from such swelling, evidenced by charged
particle irradiations [136, 137]. However, due to fast reactor irradiations showing sig-
nificant reductions in ductility during post-neutron irradiation tensile testing, attention
turned to high chromium (9-12 wt%) ferritic-martensitic steels [138] and the devel-
opment of swelling resistance stainless steels [139]. Recently, nickel-based alloys have
become prominent once more in reactor design as the high temperature stability and
57
Table 4.1: Nominal compositions of commercial nickel-based alloys in wt% (*min , ˆmax)
Alloy Ni Fe Cr Mo Co W Ti Al Mnˆ Cˆ
Nb
+
Ta
Has. X 47 18 22 9 1.5 0.6 - - 1.0 0.1 -
X-750 70* 7 16 - 1.0ˆ - 2.26 1.0ˆ 1.0 0.08 1.0
600 72* 8 16 - - - - - 1.0 0.15 -
690 58* 9 29 - - - - - 0.5 0.05 -
PE16 43 Bal 17 3 - - 1.2 1.2 0.2 0.05 -
718 53 Bal 19 3 1.0ˆ - 0.9 0.2ˆ 0.35 0.08 5.2
625 58* 5ˆ 21 9 1.0ˆ - 0.4ˆ 0.4ˆ 0.5 0.1 3.6
903 38 Bal - - 15 - 1-4 1.15ˆ - - 3
creep resistance demanded by Gen IV and fusion reactor designs limit the applicability
of their steel counterparts [140].
Although the alloys evaluated in the literature vary considerably, the fundamental effects
of neutron irradiation on nickel-based alloys can be described with reference to only a
few. As such, for simplicity, this chapter focuses on those alloys that present the most
conclusive evidence of embrittlement, the mechanisms by which it occurs, and those of
most concern with respect to presently operating and future reactor designs. Unless
stated otherwise, neutron doses have been converted using an approximate relationship
of 1.0 x 1026 cm-2 (E > 0.1 MeV) = 5 dpa [111,141,142].
4.3 Pre-irradiated microstructure and corresponding stress
corrosion cracking
Alloy 600 has come under significant research with regard to SCC [127]. Many alloys
are susceptible to SCC in one or more operating environments, and when cracked, can
show little or no sign of plastic deformation [126]. A susceptible material will only fail
by SCC when under tension, be it applied or residual, and in conditions aggressive to
the material. Intergranular and transgranular stress corrosion cracking (IGSCC and
TGSCC respectively) may both occur in the same alloy depending on microstructure
and environment [127]. To minimise IGSCC in pressurised water reactor (PWR) envi-
ronments, a high density of grain boundary carbides is desired, reducing crack growth
rates and increasing time to initiation [126]. Minimising chromium depletion at the
grain boundaries (sensitisation) is crucial in preventing harm from extended periods
under oxidising conditions. An additional thermal treatment (TT), such that the “heal-
ing” of the chromium depletion can occur, is an effective method of achieving both these
goals [127].
The dependency of SCC resistance on intergranular carbides is still under research [126].
58
However, several studies have shown a solution annealing treatment at around 1000 ◦C
followed by a thermal treatment of 700◦C for 10-15 hours improves the SCC resistance of
Alloy 600 by ensuring the alloy is in its self-healing stage and increasing creep resistance
through pinning of mobile dislocations and the inhibition of grain boundary sliding by
intergranular carbides [126]. The consequence of these different structures will be shown
to be significant in the alloy’s susceptibility to not only SCC but irradiation hardening
and helium and hydrogen embrittlement.
4.4 Fundamentals of radiation damage in nickel-based al-
loys
Neutrons produce damage by two principal mechanisms. Firstly, direct displacement of
atoms. Here, a fast neutron-atom collision creates an energetic recoiling atom, known as
a primary knock on atom (PKA), which proceeds to displace many neighbouring atoms,
generating Frenkel pairs (vacancies and interstitials in equal numbers). However, spatial
distribution of the damage regions is highly heterogeneous and consists of single defects
followed by a localised concentration of point defects known as a collision cascade.
Secondly, creation of impurity atoms - here, neutrons are absorbed in nuclear reactions
resulting in creation of new atoms by transmutation, such as helium (alpha radiation)
and hydrogen (proton radiation) generated from (n,α) and (n,p) reactions respectively.
In a fission reactor core, radiation damage is primarily due to direct displacements caused
by fast neutrons and the resulting collision cascade. However, in nickel-based alloys,
displacement damage can also occur due to thermal neutrons. For thermal neutrons,
the majority of damage is typically due to transmutation and displacement damage from
both the resulting recoil of the nucleus, and to a lesser extent, the particles emitted [123].
The process by which fast neutrons induce damage occurs through a number of stages.
There exists a threshold energy of fast neutrons beyond which a nickel atom, upon
collision with the neutron, can be displaced. The minimum neutron energy needed to
displace a Ni atom in a single crystal, via a direct collision, is ∼580 eV [122]. If this
threshold energy is met, the Ni atom will have a recoil energy of ∼40 eV [122], which
is sufficient to result in the atom becoming a primary knock-on atom (PKA). Not all
collisions will be direct collisions. A glancing collision will require a higher neutron
energy to result in a PKA due to the reduced proportion of kinetic energy transferred
by such a collision. Provided the required neutron energy is met, the resulting PKA will
transfer its recoil energy to neighbouring nuclei. The PKA is charged and thus partially
slowed by electrical repulsion, resulting in heat. Nonetheless, much of the PKA’s energy
is transferred to neighbouring nuclei through proceeding collisions, which recoil and
are displaced from their lattice positions. Such recoiling nuclei can induce secondary
59
collisions of their own and displace further nuclei, resulting in a defect cascade.
Such cascades occur in a very short time and can involve hundreds of displaced nuclei,
dependent on the recoil energy of the PKA. Electrical forces amongst defects and the
temperature dependence of interstitial diffusion can cause the mobile interstitials to re-
combine at a nearby lattice vacancy. The degree of recombination is related to tempera-
ture and point defect sink densities. As such, for alloys of high point defect sink densities
and/or under high temperatures, the resulting regime is referred to as recombination-
dominated. Alternatively, in some alloys, the interstitial and vacancy point defects can
diffuse to seperate sinks. Such a regime is defined as sink dominated [143]. Most engi-
neering alloys implemented at typical reactor operating temperatures will be in the sink
dominated regime.
In a thermal reactor, the neutron energies can span ten orders of magnitude, 10 MeV
to 0.0001 eV, depending on the location in the reactor [122]. For thermal neutrons, the
transmutation of Ni-58 to Ni-59 occurs with subsequent (n,γ), (n,p) or (n,α) reactions,
as is presented later in this section. Thus, nickel-based alloys experience direct damage
and recoil displacement from thermal neutron capture, typically not such a concern in
nickel-free materials. The Ni-59 reactions stated previously are very exothermic and thus
produce both charged particles and heavy atomic recoils, leading to damage [144]. For
the (n, α) reaction with Ni-59, total damage is ∼176.2 keV per neutron capture [123] so
the subsequent total number of displacements per neutron capture is 1762 for Ni-59 [144].
The (n,p) and (n, γ) reactions produce 222 and 4.9 displacements respectively [145]. As
such, given that the Ni-59 cross section for absorption of thermal neutrons is high, as
shown in Figure 4.2, damage caused by particle emission and recoil is considerable upon
reaching a steady state level of Ni-59 production [144].
Figure 4.2: Neutron cross sections for the (n,α) and (n,p) transmutation reactions of nickel-59
60
Void swelling is an increase in volume and a decrease in density caused by the radiation
induced clustering of vacancies into three-dimensional cavities, stabilised by helium and
hydrogen [146, 147]. Void swelling leads to changes in component dimensions and me-
chanical properties, such as reductions in strength, fatigue resistance and thermal con-
ductivity. Nickel-based alloys are highly susceptible to this form of void swelling [148].
Following initial void formation, a process of vacancy-assisted growth occurs. Disloca-
tions act as biased sinks for the preferential absorption of interstitials as a consequence
of the differing strain fields compared with vacancies. The net excess of vacancies cluster
in association with helium, resulting in enhances void growth [122].
Void swelling is of great concern in commercial reactors, where doses to the structural
material can be as high as 40 dpa per annum in core components [111] and under
temperature ranges of typically 300◦C to 600◦C. Measured void swelling in such reactors
has been as high as 10%. Helium resulting from radiation in nickel-based alloys is of
particular issue due to the low solubility in the material, and hence its preference to form
bubbles. Although there are numerous reactions that produce hydrogen and helium in
nickel-based alloys under neutron irradiation, nickel-59 dominates production in most
applications, with boron-10 playing a significant role, under thermal neutron radiation,
in producing grain boundary helium embrittlement, as is discussed later. Following the
transmutation of nickel-58 to nickel-59, nickel-59 may transmute to form iron-56 and
helium, the neutron cross section of which is shown in Figure 4.2:
58Ni + n→59 Ni + γ (4.1)
59Ni + n→56 Ni +4 He (4.2)
The transmutation of boron-10, which comprises 19.9% of natural boron [149], results
in lithium-7 and helium.
10B + n→7 Li +4 He (4.3)
In comparison to that of the nickel-59 reaction, boron-10 has a very large capture cross
section for thermal neutrons, as shown in Figure 4.3. This is inversely proportional to
neutron velocity. Therefore, a small amount of boron impurity will burn out rapidly and
produce helium at modest cold neutron fluences (< 0.025 eV). An estimate of helium
production from the nickel-59 reaction can be calculated using the following:
∂NNi
∂t
= −NNiσNiφ (4.4)
where N is the number of atoms, σ is the neutron cross section for the corresponding
61
reaction, φ is the neutron flux and t is the time under irradiation. As shown in Figures
4.2 and 4.3, neutron cross section varies significantly with neutron velocity. For exam-
ple, the neutron cross section for Ni-59 indicates that at energies higher than ∼0.01
MeV, hydrogen production becomes more probable than helium production. Each nu-
clear reactor and location within the reactor has its own individual neutron spectrum.
As such, when calculating the production of transmutation products, one should use
spectrum-specific values for the neutron cross sections and flux.
Integrating the previous equation gives:
NNi = NNiinitiale
−σφt (4.5)
From this, a final helium content can be calculated:
NHe = NNiinitial
(
1− e−σφt
)
(4.6)
Similarly, an estimate can be made for the helium production from boron-10:
∂NB
∂t
= −NBσBφ (4.7)
Which when integrated gives:
NB = NBinitiale
−σφt (4.8)
The overall estimate for the magnitude of helium production, incorporating all sources
as a function of neutron fluence, is given by:
NHe = NBinitial
(
1− e−σBφt
)
+NNiinitial
(
1− e−σNiφt
)
(4.9)
In addition to the production of helium, Ni-59 may also transmute to form cobalt-59
and hydrogen, or to nickel-60, accompanied by gamma emission:
59Ni + n→59 Co + H (4.10)
59Ni + n→60 Co + γ (4.11)
Identical to that for helium production, we can predict hydrogen concentrations as a
function of fluence. The capture cross sections shown in Figures 4.2 and 4.3 indicate
the neutron spectrum dependence of the discussed transmutation reactions. As previ-
ously stated, the neutron cross section for Ni-59 indicates that at energies higher than
∼0.01 MeV, hydrogen production becomes more probable than helium production. As
62
such, helium and hydrogen production rates must be quantified on a case-by case basis,
accounting for a reactor’s individual neutron spectrum. A highly-thermal spectrum will
develop a high content of Ni-59 rapidly, and as such, helium and hydrogen production
will be non-linear. Once Ni-59 has built up to an almost steady state, the production
will become linear. The proceeding production of helium from Ni-59 is again dependent
on the neutron spectrum. Greenwood and Garner [144] published a model for calcu-
lating both helium and hydrogen content incoporating the initial Ni-58 transmutation
and all possible proceeding reactions for specified spectral-averaged cross sections, the
validation of which was attempted with irradiated X-750 spacers from a Canada Deu-
terium Uranium (CANDU) reactor, the experimental work of which is discussed later
in this review.
Figure 4.3: Neutron cross section for the (n,α) reaction of boron-10
The equations described can be used to estimate the total hydrogen and helium produced
as a function of nickel and boron content for a specified neutron spectrum. It should
be noted that matrix damage and helium will also be produced from other element
transmutations such as iron and chromium isotopes. However, their contribution to
embrittlement is minimal in comparison to that of nickel.
In fast reactor environments in which high levels of radiation hardening occur, the major
sources of helium are the threshold (n,α) reactions with the major alloying elements.
Nickel has the largest cross-section and thus the helium production rate scales with nickel
content. Additional helium, depending on the thermal neutron flux, will also arise due
to low energy (n,α) reactions with boron-10 and nickel-58 and 59. For highly-thermal
spectra, it is these low energy reactions that dominate helium production.
Although boron-10 levels are typically low in nickel-based alloys (<40 wppm), for most
63
LWR environments boron-10 will transmute to helium at a faster rate. This, coupled
with the fact that boron segregates to grain boundaries, as is discussed later, means that
boron is of concern regarding early grain boundary helium embrittlement. Although
boron transmutation will also have an effect on swelling, insufficient data prevents a
relationship being quantified.
During the latter part of the 1970s, it was found that small boron additions significantly
improved the hot workability and stress-rupture life of nickel-based superalloys [150].
Boron was found to increase grain boundary cohesiveness. thereby reducing the likeli-
hood of brittle intergranular fracture [151]. Thermodynamic modelling work by Yin et
al. [152,153] has shown that with regard to grain boundary carbide precipitation, boron
has two significant effects: increased carbide volume fraction and decreased carbon and
chromium solubilities. Boron is shown to affect chromium diffusivity, carbide precipi-
tation and change the grain boundary carbide coverage, affecting IGSCC susceptibility.
The addition of boron to Alloy 690 was modelled and shown to degrade the IGSCC resis-
tance by lowering the minimum grain boundary chromium concentration and delaying
the self-healing process [153]. This degradation in IGSCC resistance is seen to be less
significant when a high density of intragranular carbides are present, presumably due
to trapping of boron on these carbides. The effects of a lower grain boundary carbide
coarsening rate, due to boron reducing the carbon solubility, are limited with regard to
IGSCC susceptibility.
It is clear that even small traces of boron (∼10 wppm) may initially dominate helium
production for the typical neutron spectrum of LWRs, with higher levels of boron sig-
nificantly exacerbating the degradation in irradiated material properties. This, coupled
with the modelling work of Faulkner and Bajaj et al., presenting boron’s deleterious
effects on IGSCC susceptibility [152–154], indicates it is vital to ensure boron content,
especially that of boron-10, in nickel-based alloys is as low as reasonably practical. The
use of depleted boron is one option if boron additions are necessary. Depleted boron
(boron containing mostly boron-11) is already produced for a number of radiation-
related applications, such as borosilicate glasses in radiation-hardened electronics [155].
4.5 Effects of radiation on mechanical properties
4.5.1 Tensile properties
The effects of neutron radiation on the tensile properties of nickel-based alloys are cov-
ered elsewhere, for both thermal and fast reactor irradiations, in considerable detail and
are well established in the literature [111,131,133,156,157] and as such, it is unnecessary
to provide more than a summary for this review.
64
For both fast and thermal reactor irradiations, nickel-based alloys produce similar ef-
fects to those seen in austenitic stainless steels [111]. Low temperature irradiations
cause increases in yield strength and ultimate tensile strength, alongside reductions in
ductility [131,133,156], as shown in Figure 4.4 for solution annealed Alloy 600. However,
post-irradiation annealing has been shown to almost entirely reverse the changes [156].
Embrittlement is also seen at higher irradiation temperatures but is limited due to an-
nealing [157]. Creep rates are not affected, but the rupture life is significantly reduced
due to a loss in ductility, as will be described later [158]. Low strain rates and increased
nickel content also appear to increase these detrimental effects [156].
Figure 4.4: Tensile properties of irradiated solution annealed Alloy 600 (4 hrs at 1080◦C) with
predicted helium contents of 600-1700 appm, data taken from [157].
The work of Kangilaski [156] shows that irradiation temperatures below 300◦C result in
65
considerable increases in strength and decreases in ductility for Alloy 600. Both these
effects increase with neutron dose. As the irradiation temperature is increased, matrix
damage is annealed out and post-irradiation properties depend more on the ageing
characteristics of the alloy [156].
The effects of neutron irradiation on the tensile properties of numerous nickel-based
alloys, including Alloy 600, were examined in the 1960s by Claudson and Pessl [131],
including specimens irradiated at temperatures ranging from 50◦C to 740◦C in water.
Unirradiated control samples were thermally aged to replicate the times at temperature
seen in the irradiated samples. Tensile tests were then carried out at 50, 300 and 650◦C.
Room temperature yield strength was three times higher for samples irradiated at 50◦C
compared to 740◦C, though ductility levels were similar. The losses in strength for
samples irradiated at 740◦C were shown to be independent of irradiation and were due
to thermal ageing, as similar losses were observed in the unirradiated thermal control
specimens. No improvement in ductility due to overaging was observed.
The following year, Claudson [133] provided some additional tensile test data. Alloy
600 was tested in the solution treated, mill annealed and 20% cold-worked condition.
Samples were irradiated at 280◦C to a maximum of 5.7x1020 n cm-2 (> 1 MeV) and
subsequently tested at 650◦C. A notable increase in yield strength was seen in both the
solution treated and mill annealed samples, but a decrease in the cold worked samples.
Cold work was found to produce greater levels of irradiation induced strengthening at
650◦C than mill-annealed or solution annealed material. The post irradiation ductility
of all specimens was drastically reduced when tested at 650◦C. There was no microstruc-
tural characterisation post-testing. However, the tensile data was said to be consistent
with that of grain boundary failure at high temperatures due to helium embrittlement,
as irradiation has little effect at ambient temperature but more significant effects at
elevated temperatures, with decreased ductility and strength.
Kangilaski [156] presented thermal reactor irradiated welded Alloy 600 data. Speci-
mens irradiated at 50◦C to 2.3x1018 n cm-2 (>1MeV) with 6.7 x 1019 n cm-2 (<1MeV)
indicated no change in strength with irradiation at temperatures of 500-900◦C. No mea-
surable irradiation-induced changes in uniform elongation were observed, but significant
changes in the necking elongation were identified. There was very little change in weld
ductility up to 600◦C. However, at 700◦C, only 28% of the original unirradiated ductility
remained.
4.5.2 Fracture toughness and creep
Modelling by Baskes [159] has indicated that fracture stress is reduced by an order
of magnitude by the presence of small helium-vacancy complexes. Baskes suggested
helium atoms behave as crack nucleation sites, enhancing the fracture process. Helium
66
was also identified as having a strong cohesion to grain boundaries, and can thus totally
restructure these boundaries, reducing cohesive strength.
Fracture toughness data is scarce for irradiated nickel-based alloys. However, the detri-
mental effects observed in the tensile properties of irradiated nickel-based alloys, such
as Alloy 600, raise concerns for fatigue and crack growth performance. If sufficiently
high, a loss in ductility can lead to channel fracture with local separation along planar
slip bands. Channel fracture is associated with localised separation along dislocation
channels formed by an extensive heterogenous slip mechanism. Lead dislocations cut
through a localised region, clearing out the irradiation-induced defects, producing a
narrow defect-free path considerably weaker than its surrounding matrix [160]. All dis-
locations are then channeled along these narrow paths until localised shear cracks form
and propagate through these channels.
Mills [160] reviewed the effects of irradiation on the toughness properties of Alloy 600.
Two mill-annealed plates of 25.4 and 63.5 mm thickness were tested and their mi-
crostructural evolution documented. The initial microstructures varied between plates,
with the thicker plate having larger and more numerous carbides throughout the grain.
Irradiations were carried out in the Experimental Breeder Reactor - II (EBR-II) at 400-
427◦C to doses of 7 to 24 dpa on material from the thinner plate. The results indicate a
three-fold increase in yield strength and a factor of two reduction in ductility for tensile
properties measured at 7 dpa, irradiated at 427◦C.
Figure 4.5 shows the J -based crack extension resistance JR versus crack extension, where
J is the J-integral (strain energy release rate for a crack), indicating the reduction in
initiation toughness and tearing modulus at 7 dpa, further reducing at 24 dpa. Mills
stated that when observing the fracture surfaces, they were highly faceted, indicating
channel fracture. As this failure mechanism was dominant at the doses tested, and little
change in toughness was seen for an increase of 7 to 12 dpa, it may be assumed that
once channel fracture becomes dominant, ductility and fracture toughness are saturated.
This concurs with the findings of Hunter and Fish [161,162] and the saturation identified
for uniform elongation at high fluences in the work of Mills [160].
Fatigue test data compiled by James [163] at ∼430◦C for Alloy 600 irradiated from 2.5
to 6 x 1022 n cm-2, shows that fatigue crack growth behavior was only slightly degraded
from its unirradiated condition.
Angeliu et al. studied the creep behavior of Nimonic PE16 and Hastelloy X [158],
evaluating both in-pile creep and post-irradiation tensile creep experiments. The results
were mixed, however, none of the results showed that radiation significantly affected the
creep behaviour for typical PWR environmental conditions, concurring with the findings
of others on various nickel-based alloys [111].
67
Figure 4.5: The effect of neutron irradiation on the JR curve behaviour for Alloy 600, data
taken from [160].
4.5.3 Swelling
As stated previously, general trends are observed in nickel-based alloys with respect to
the effects of neutron irradiation on mechanical properties. However, notable differences
in the degree of swelling are observed in the range of alloys for which swelling data is
available.
For simplicity, swelling behavior in nickel-based alloys can be divided into two stages;
an initial stage of little or no swelling, referred to as the incubation period and a second
stage with a significantly higher rate of swelling [111]. The critical dose beyond which
second stage swelling occurs, and the rate of this swelling, is dependent upon alloy
composition, microstructure and neutron spectrum [119]. As it will be shown, it is the
number of initiation sites for cavity growth that dominate swelling susceptibility.
The work of Wiffen [157] and Kai and Lee [164] shows that mixed spectra fast reactor
and proton irradiation respectively of Alloy 600, in the absence of boron, for doses
up to 8.5 dpa, results in <1% swelling for temperatures in the range of 55 to 500◦C.
Below 500◦C, swelling appears to be relatively temperature independent. Above, 500◦C,
swelling increases notably with dose, becoming more sensitive to temperature. This
temperature dependence indicates that helium bubble formation may play a significant
role in swelling.
68
Porollo et al. [165] showed that for some austenitic stainless steels, void swelling, in the
typical range of temperatures and dpa rates of a LWR, occurs down to ∼305◦C. At
330◦C, the swelling reaches ∼1% at 20 dpa. Comparing this data with other published
results from Russian LWR reactors at<10 dpa confirms that the lowest temperature that
stainless steels can begin swelling appears to be ∼300◦C [165]. The upper temperature
limit of void formation is related to the lower supersaturations of vacancies at high
temperatures. The lower temperature limit of void formation is thought to arise from
low vacancy mobility and/or the inability of vacancies (and helium atoms) to aggregate
and nucleate voids (and helium bubbles). Although no data is available to indicate
an upper temperature limit for nickel-based alloys, given the mechanistic similarities
with regards to irradiation embrittlement of stainless steels and nickel-based alloys, the
assumption that one exists for nickel-based alloys would be justified.
Helium plays a considerably more significant role in nickel-based alloy embrittlement
than it does in stainless steel [119,165,166]. It is expected that the increase in swelling
identified in Alloy 600 for temperatures >500◦C, discussed previously, is due to increased
vacancy and helium mobility. It would be logical to thus assume an increase in the rate
of helium bubble growth. However, the location of the helium can play a significant
role in resulting embrittlement and swelling. Nickel-produced helium will be produced
intragranularly. If intragranular carbides are present, which can behave as helium traps,
as is discussed later, they may limit the helium accumulation rate on grain boundaries.
Boron however, which preferentially segregates to grain boundaries, is able to nucleate
and grow bubbles much faster along these grain boundaries. Alloy 600TT, with its
grain boundary carbides, will be more susceptible to grain boundary boron segregation,
as relatively high boron concentrations would be expected in these carbides due to
boron’s affinity to inhabit carbides [167]. Although a model could be produced to
calculate bubble growth, as will be discussed, it requires the complex combination of a
range of modelling techniques. However, a simple model could be produced by solving a
two-dimensional master equation to describe vacancy-helium cluster evolution. For an
example of this modelling in stainless steels, see [168].
A number of High Flux Isotope Reactor (HFIR) irradiations, a mixed spectrum reactor,
conducted to investigate radiation effects in Alloy 600 under high helium production
rates are presented by Wiffen [157], providing insight on the potential effects of helium at
lower irradiation doses and/or temperatures. Alloy 600 showed good swelling resistance
under a limited range of irradiation conditions. Samples were irradiated in both the
solution-annealed and 20% cold-worked condition, at 300 to 700◦C under doses of 4
to 9 dpa. It should be noted however, that helium production rates in the HFIR are
considerably higher than would be expected in a typical LWR, this is due to the unique
neutron spectrum of the HFIR, most notably its very high thermal flux [169]. The alloy
tested did not have a boron content stated in its composition. As such, all helium is
69
assumed to be a product of nickel transmutation.
The swelling values were found to be considerably high for the dpa levels. This indicates
that higher helium production rates may result in increased swelling rates. Cold work
was found to be ineffective at suppressing swelling during irradiation with high helium
production rates [157]. Swelling was only weakly dependent on irradiation tempera-
ture for 55 to 600◦C, but increasing considerably for irradiations at 650◦C and 700◦C.
Increasing helium content increased the susceptibility to intergranular fracture at irradi-
ation temperatures of 600-700◦C, decreasing ductility. This is likely due to the increased
diffusion of the helium, increasing bubble growth rate. Although the data indicates that
as irradiation temperature reduces, helium induced swelling reduces, boron-induced he-
lium will likely increase helium effects at lower temperatures due to boron’s segregation
to grain boundaries prior to irradiation. All of this evidence supports the thesis that, in
the presence of high helium production rates, swelling of Alloy 600 is driven by helium
content, and for high helium contents, comes a largely intergranular fracture mode.
Assuming that swelling at a fixed irradiation temperature is controlled by helium con-
tent, the helium behaves according to the ideal gas law, and there is a constant concen-
tration of equilibrium helium bubbles containing all helium, then swelling is found to be
proportional to helium concentration. This reinforces the temperature dependence for
swelling due to nickel-induced helium. Using the data of Wiffen [157] and these assump-
tions, the irradiation temperature dependence of swelling can be estimated, as shown in
Figure 4.6, indicating that it is the ability of helium to form bubbles, commonly on grain
boundaries, that governs its influence. At 700◦C for a 1541 appm helium content, the
swelling reaches over 3%. Extrapolation of the data to higher helium contents suggests
a significant, almost exponential, increase in swelling with helium bubble volume.
70
Figure 4.6: Swelling of solution annealed Alloy 600 irradiated to 4-9 dpa, normalised to the
average helium content of 1541 appm, data taken from [157].
4.6 Radiation induced microstructural changes and irradiation-
assisted stress corrosion cracking
4.6.1 Microstructural changes
A 1993 study by Kai and Lee [164] evaluated the effects of radiation on the microstruc-
ture of two Alloy 600 heats. Noting that irradiation-assisted stress corrosion cracking
(IASCC) was becoming a significant concern in the nuclear community, a number of
specimens were irradiated with 1 MeV protons at 450◦C to doses of 0.01, 0.1 and 1.0
dpa, at around 6 µm depth, and their microstructural evolution analysed. It has been
reported elsewhere, that even under low stress and no sensitisation, intergranular IASCC
can occur in some austenitic stainless steel structures above a threshold dose of 5x1024
n cm-2 (>1MeV) [170].
Kai and Lee irradiated two Alloy 600 heats; mill-annealed and solution-annealed + sen-
sitised, both followed by a water quench. The mill-annealed specimens contained many
M7C3 precipitates located along the grain boundaries and in the matrix. There was
no chromium depletion near the grain boundaries. The sensitised material was deco-
rated with small semi-continuous M7C3 grain boundary carbides. Significant chromium
depletion was observed near the grain boundaries.
In the mill annealed specimens, irradiated at 450◦C to various doses, there is no signif-
icant alteration in carbide morphology except for a slight degree of coarsening after 5
dpa. As expected, the dislocation density increased with dose, up to a saturation level
71
of 1.2 x 1014 m-2. Voids were found to form at 0.05 dpa, although small. At 0.5 dpa,
voids grew to ∼90 nm, but with low density and primarily on grain boundaries or twins.
For 5 dpa, void number density greatly increased, with an average void size of <25 nm,
but with some larger voids of ∼100 nm. It should be noted for these specimens, that
as proton irradiation was used, the transmutation effects of neutron irradiation will not
occur. Hence, helium production and its effects are not considered.
The sensitised samples, irradiated to 5 dpa, showed prominent rod-shaped M23C6 fcc car-
bides. Their size and number density increased with dose, indicating they are radiation-
induced. Dislocation density increased with dose to saturation at 6 x 1013 m-2 in the
5 dpa sample. In the specimens irradiated to 5 dpa, there were still only a few voids
randomly distributed in the matrix and attached to grain boundary carbides. The grain
boundary chromium concentration was around 11 wt% for unirradiated specimens, and
this decreased to ∼10 wt% in irradiated specimens of 5 dpa.
The results further validate that heat treatments are influential on the effects of ir-
radiation on Alloy 600. The differences between mill-annealed and sensitised samples
are due to the initial dislocation densities, and the formation of small M23C6 carbides
during irradiation, which act as point defect sinks, increasing the recombination rate,
and thus limiting the accumulation of defects. This concurs with the observed reduction
in the saturated dislocation density for the sensitised samples. The mill annealed sam-
ples showed only irradiation enhanced coarsening. However, sensitised samples showed
additional radiation-induced fine, uniformly distributed M23C6 carbides in the matrix,
considered primarily irradiation induced [164]. The lesser void density seen in the sen-
sitised samples at 5 dpa indicates a lower swelling susceptibility. This is likely due to
the high density of M23C6 carbides formed during irradiation, providing recombination
sinks.
In the mill annealed samples, 1 MeV proton irradiation showed no effect on chromium
depletion at grain boundaries. In the 5 dpa sensitised samples however, it enhanced
it. This was due to the mill annealed samples having already reached their self-healed
stage [127]. The sensitised samples however, had supersaturated carbon content and
could thus develop further carbides on the grain boundaries, reducing the chromium
content. This indicates that irradiation can induce chromium depletion, provided the
material has not previously reached its self-healed stage. Ensuring the material is well
into its chromium self-healing stage, through thermal treatment such as that of Alloy
600TT, will thus reduce IASCC susceptibility [127]. The void swelling observed in Kai
and Lee’s study [164] indicated that Alloy 600 is highly susceptible to void formation
during irradiation. As proton irradiation was used in the study, helium production and
any corresponding voids and resulting enhancement of swelling have not been taken into
account.
Rowcliffe and Horak [171] reported grain boundary failures in double aged Inconel 706
72
after irradiation to 20-34 dpa in the range of 500-625◦C. The failures were attributed
to separation between the matrix and the continuous grain boundary eta phase, and
the formation of voids associated with this interface. For tensile tests carried out above
the irradiation temperature, high levels of sulphur and phosphorus were detected on the
fracture surfaces and helium release was detected during fracture [142]. Additionally,
post-irradiation ring crush testing of solution annealed Alloy 706 to 25 dpa showed
brittle intergranular failure for irradiation and testing temperatures above 560◦C. These
were attributed to the radiation enhanced formation of continuous intergranular gamma
prime phase [172]. Contrary to this, solution treated 706 fuel pins have also shown
intergranular eta phase and no gamma prime when irradiated to the same dose [173].
Vaidyanathan et al. [174] reported brittle grain boundary fracture in solution treated
PE16 following irradiation to 35 dpa at 554◦C, with mechanical testing above this tem-
perature. It was suggested that irradiation induced intergranular films of gamma prime
are responsible for the observed embrittlement. Clearly, the initial microstructure plays
a role on the resulting embrittlement mechanisms with any grain boundary eta or gamma
prime films appearing to cause embrittlement, further enhanced by helium accumula-
tion on the interface of such films. Grain boundary embrittlement in precipitate hard-
ened alloys such as PE16 have been attributed to one or a mix of several mechanisms;
radiation-induced hardening and the limited permissible relief of triple point stresses
due to grain boundary sliding [175], brittle failure of radiation-induced continuous grain
boundary lamellae of gamma prime (or eta phase in the case of Alloy 706) and grain
boundary helium bubbles [176, 177]. However, Boothby [178] emphasises that any ra-
diation induced formation of continuous intergranular eta or gamma prime lamellae is
strongly dependent on the initial alloy condition; with such films only observed in solu-
tion treated materials and not in fully aged material. Further to this, grain boundary
embrittlement of PE16 has been observed both with and without irradiation-induced
intergranular gamma prime, with the dominant embrittlement mechanism being most
likely related to intergranular helium bubbles.
Radiation induced losses in ductility in PE16, the composition of which is shown in
Table 4.1, in mixed spectrum test reactors have been shown to be due to the rapid
burn-up of boron-10 in boron-rich grain boundary precipitates, resulting in significant
helium contents and thus grain boundary helium embrittlement [179,180]. Such thermal
spectra, high boron levels and resulting helium bubbles have been shown to reduce the
fracture toughness [181] with helium contents ranging from 17-24 appm. Further to
this, for Alloy 718, there is no evidence linking radiation induced ductility loss to any
grain boundary embrittlement phenomena. Instead, embrittlement has been attributed
to the weakening of interfaces between eta phase and the matrix due to sulphur, phos-
phorus and helium. Such interfaces were observed to separate during tensile testing at
the intersection of localised slip bands and the eta platelets [142]. Regardless of any
73
contradictions, the studies reviewed clearly indicate that the initial microstructure prior
to irradiation is critical in governing the dominant embrittlement mechanisms, although
in almost all cases helium is thought to play a role.
4.6.2 Irradiation assisted stress corrosion cracking
Bajaj et al. [154] investigated the microstructural and microchemical changes in irradi-
ated high strength Alloy X-750 and 625 heats of varying boron content with respect to
IASCC. X-750 is essentially Alloy 600 with additions of Ti, Al and/or Nb for precipitate
hardening through the presence of gamma prime precipitates, Ni3(Ti, Al) [182]. X-750,
in addition to gamma prime within the matrix, has a high number density of intergran-
ular Cr23C6 carbides [182]. Ni3Ti, η-phase and (Ti,Nb)C precipitates are also observed
along grain boundaries, though low in number [115]. Both of these precipitates have a
relatively low number density in comparison with Cr23C6 carbides.
SCC tests were performed on pre-cracked compact tension (CT) specimens exposed
to water in both in-flux and out-of-flux locations at 360◦C. Figure 4.7 shows that the
low boron X-750 heats (< 20 ppm) were not susceptible to IASCC, with no specimens
exhibiting any high temperature SCC. The high boron heats (>40ppm) and process vari-
ation heats with 20 and 30ppm boron showed extensive IASCC under comparable and
less severe irradiation and KI conditions. This affirms that boron is critically influential
on IASCC susceptibility. Bajaj et al. observed that once the boron content threshold
for IASCC is exceeded, little additional effect is seen for increasing boron content with
regard to SCC susceptibility. Notably however, Alloy 625 showed no SCC in any of the
heats, indicating its superior SCC resistance in comparison to X-750. Fractographic ex-
amination of the X-750 specimens showed a mixed transgranular/intergranular fracture
mode with secondary cracking. A-10, a high boron heat, showed the most intergranular
fracture, which was suggested to be due to helium bubbles weakening grain boundary
bonding. The grain facets in the irradiated samples indicated increased planar slip.
This, coupled with the evidence of channel fracture in irradiated specimens, indicates a
significant increase in planar slip due to irradiation, an effect of grain boundary helium
embrittlement.
STEM-EDS analysis revealed only a slight increase in chromium segregation and de-
pletion (2-2.5 wt%) due to irradiation in the grain and phase boundaries. Where sig-
nificant IASCC was identified, thermal flux analysis revealed that approximately 50%
of the boron-10 originally present was transmuted to helium at an equivalent fluence
of 2 x 1020 n cm-2 (E>1 MeV). Figure 4.8 shows the secondary ion mass spectroscopy
(SIMS) examination of the high boron heats prior to irradiation showing segregation of
boron to grain boundaries, as evidenced in other studies [183]. The low boron heats
did not show significant boron concentrations at the grain boundaries. TEM studies
74
Figure 4.7: The IASCC in X-750 of various boron contents, data taken from [154].
were carried out on the irradiated and 816◦C annealed heats to determine the helium
distribution developed during irradiation. The high boron heat A1 showed the highest
density of helium bubbles, with an average bubble size of ∼2nm decorating the grain
boundaries, with very few bubbles observed within the grain.
Figure 4.8: SIMS images showing boron segregation to grain boundaries in two separate heats
of X-750 containing a) 28 wppm boron b) 2 wppm boron [160].
The significance of boron content with respect to unirradiated SCC susceptibility was
also investigated. There appeared to be no effect of boron content on SCC initiation
susceptibility for levels <40ppm (the maximum boron content tested). However, only
three out of 18 specimens initiated, indicating the significance of neutron irradiation in
regard to SCC initiation susceptibility [154].
It has been speculated that helium accumulates away from grain boundaries due to re-
coil, with a calculated distance of 2 µm [154]. However, given that helium is produced
75
from boron on grain boundaries within a 2µm radius, helium will still preferentially
diffuse to grain boundaries and thus will likely remain on these boundares [180]. Re-
gardless, helium, along with vacancies, will diffuse to sinks such as grain boundaries,
thus increasing the probability of helium/vacancy complex formation on grain bound-
aries. This all results in an increased probability of grain boundary helium bubble
formation [156].
4.6.3 Radiation induced segregation
Radiation-induced segregation is believed to be a factor in IASCC of nickel-based alloy
components in LWRs [184]. However, limited data is available for neutron irradiated
material with respect to segregation. Due to the cost of generating such data, much
effort has been put into modelling, typically using the Perks model [185], which assumes
segregation is caused by differences in the participation of atoms in the vacancy flux.
Work by Yang et al. used an altered Perks’ model to include ordering energies and
compared their results with experimental 3.2 MeV proton irradiations of nickel-based
alloys. Due to lack of data, the significance of segregation in comparison to other factors,
will not be covered in this review. However, the recent work of Yang et al. has further
improved the modelling techniques for irradiation-induced segregation, which may prove
useful for application in future evaluations of nickel-based alloys [186].
4.7 Helium and hydrogen embrittlement
The effects of hydrogen on nickel-based alloys is an extensive subject in itself, and as
such, inappropriate to cover in-depth in this review. However, despite the well es-
tablished detrimental effects of hydrogen on the mechanical properties of nickel-based
alloys and steels [187, 188], the mechanisms behind hydrogen embrittlement are little
understood [189].
In the cores of CANDU reactors (a design of PWR using heavy water), nickel-based
alloys are often used as spacers in fuel channel assemblies [122]. Alloy 600 was used to
sheath paired coiled detectors in these reactors in early designs [123]. Post-irradiation
characterisation found that the Alloy 600 wires had fractured after several years in the
reactor, at 9.4 and 11.15 effective full power years (EFPY) [123]. All fractures, which oc-
curred during operation, were found to be intergranular. Post-fracture characterisation
indicated that sections operating at the highest temperatures were the most severely
embrittled, with low temperature sections remaining ductile [123]. This temperature-
dependence of embrittlement was stated as being the result of a transition from the
recombination dominated to the sink-dominated regime [122].
76
Modern CANDU reactor fuel channel assemblies use X-750 springs as spacers [122].
These spacers operated between 200 to 350◦C, depending on the circumferential location.
The end of life dose was stated as ∼55 to 60 dpa, including the effects of both fast and
thermal flux. Given that the Ni-59 cross section for absorption of thermal neutrons is
high, damage caused by particle emission and recoil is considerable upon reaching a
steady state level of Ni-59 production. It was found that this Ni-59 isotope and the
proceeding reactions made up the majority of the dpa in the X-750 spacers [122].
Experimental validation of the modelled helium and hydrogen production rates, us-
ing the models of Greenwood and Garner [144], was attempted for the spacers using
hot vacuum extraction mass spectroscopy, for quantification of helium and hydrogen.
Quantification was unsuccessful. However, one specimen (11 EFPY) had >2500 appm
detected, with no indication of exhaustion during testing. As such, further validation of
these models is required. It is permissible that a combination of the original methods
used, at reduced doses, and/or thermal desorption spectroscopy [190,191] could be used
to establish both hydrogen and helium contents of irradiated samples.
An additional set of spacer specimens were characterised in a later study to better
link grain boundary embrittlement with the presence of helium bubbles [115]. The two
samples were irradiated to ∼55 dpa and 18,000 appm helium, at 120-280◦C and 300-
330◦C respectively. TEM lamellae were extracted directly from an intergranular fracture
surface produced during post-irradiation examination. The results of the TEM analysis
are shown in Figures 4.9, 4.9, 4.11 and 4.12. Figure 4.9 indicates that with increasing
temperature, the mean bubble size increases and number density reduces. However,
most notable in this work is the direct evidence that helium bubbles concentrate along
grain boundaries and result in intergranular fracture, as shown in Figures 4.10 and 4.11,
providing the evidence which previous studies lacked with respect to helium induced
grain boundary embrittlement [111,148].
Electron energy loss spectroscopy (EELS) affirmed the higher concentration of helium
in the bubbles in comparison with the matrix, and established that cracking propagated
around the intergranular carbides and along the boundary due to reduced boundary
strength. The significance of secondary phases present within the spacers with respect
to helium bubble formation is of much importance, and is discussed in detail later in
this review.
Increased grain boundary hydrogen concentrations are known to promote intergran-
ular fracture [192]. Given the evidence that carbides, such as those present on the
grain boundaries in X-750, have been seen to strongly trap hydrogen, it was stated
that hydrogen may have also contributed to the grain boundary embrittlement ob-
served. Intergranular carbides are preferable in reducing SCC susceptibility. However,
it would appear detrimental to have such grain boundary networks of carbides with
respect to irradiation embrittlement as these enhance hydrogen bubble formation on
77
Figure 4.9: TEM Fresnel Contrast Imaging in the underfocus condition for material with
∼55 dpa and 18000 appm He and reactor operating temperature at a) 120-280◦C and b)
300-330◦C [115].
Figure 4.10: TEM Micrograph in the underfocused condition, showing bubble conglomeration
on grain boundaries and matrix-precipitate interface after neutron irradiation to ∼55 dpa and
18,000 appm He at 300-330◦C [115].
the grain boundaries. As will be discussed later, it is proposed that by engineering the
microstructure to result in a high number density of nanosized intragranular carbides,
ideally in addition to grain boundary carbides for SCC resistance, hydrogen and helium
may be trapped upon the precipitate-matrix interfaces within the grains, preventing any
significant bubble formation throughout the microstructure and reducing or eliminating
the risk of grain boundary embrittlement altogether.
Zhang et al. investigated cavity formation in Inconel X-750 using heavy ion (1 MeV
Kr2+) irradiation and pre-injected (cold) helium, with in situ characterisation using
an intermediate voltage electron microscope [193]. By comparing helium injected and
helium free samples [128], it was found that pre-injected helium was essential in cavity
78
Figure 4.11: SEM image of a fracture surface of X-750 irradiated to ∼55 dpa at 300-330◦C.
The intergranular fracture surface shows features indicative of pull out and remaining grain
boundary precipitates. The sample contained 18,000 appm helium [115].
Figure 4.12: TEM image of an η-phase precipitate in X-750 with local bubbles larger in
comparison with the matrix [115].
nucleation. Cavities began to be identifiable after 2.7 dpa at around 200◦C in sam-
ples containing 200, 1000, and 5000 appm helium respectively, but were not seen under
lower temperatures. Alongside cavity formation, a reduced number of vacancy type
defects and stacking fault tetrahedral were observed. With increasing pre-injected he-
lium content, a higher density of smaller cavities was seen, assumed to be the result
of local helium trapping, limiting larger void formation. With increasing irradiation
temperatures, the mean cavity size increased, number density of cavities reduced and
79
cavities became more heterogeneously distributed. Notably, no grain boundary cavity
sinks were identified, even at elevated temperatures, with MC carbides identified as the
leading cause behind the lack of bubble formation due to their trapping efficiency with
respect to helium and point defects.
Another study [193], rather than using cold injection, injected helium at 400◦C to simu-
late the neutron induced damage in the CANDU reactor’s X-750 spacers. Cavities were
seen to form during hot injection, the size and density of which increased with helium
content. The cavities were found to be consistent with neutron-induced cavities as ob-
served in the CANDU spacers [122], indicating that the helium distribution is dependent
upon the injection temperature, with pre-injection at high temperatures preferable for
simulating the migration of helium produced by transmutation in a thermal reactor like
the CANDU.
4.8 Trapping of interstitial elements
It is generally accepted that the non-trapped fraction of hydrogen and helium dominate
embrittlement. As such, a higher trapping capacity results in a lower susceptibility to
embrittlement [189, 194]. However, it is speculated that only trapping away from grain
boundaries is beneficial with respect to preventing grain boundary embrittlement, as will
be discussed. Materials with a high defect density have an increased trapping capacity,
resulting in an increase in hydrogen and helium content exceeding the solubility limit of
the perfect lattice [189]. This is because when the chemical potential of hydrogen, for
example, reaches equilibrium, a material can absorb hydrogen to the solubility limit of
the host lattice, after which additional hydrogen will occupy any available traps. The
free hydrogen in the lattice and the trapped hydrogen equilibrate with each other and
the apparent solubility or total hydrogen concentration can be significantly higher than
the lattice solubility [195].
The presence of traps will also reduce the apparent diffusivity, as atoms require energy
greater than the lattice migration energy to leave the trap and, consequently, the appar-
ent diffusivity will be lower than in the undisturbed lattice diffusivity. The decrease in
total energy of a system by traps with a trap-interstitial binding enthalpy is described
by Fischer et al. [67]:
G = Go +RgT

[yL ln yL + (1− yL)ln(1− yL)]NL
+
m∑
k=1
{
[yTk ln yTk + (1− yTk)ln(1− yTk]NTk − ∆Ek
RgT
yTkNTk
}

(4.12)
80
This model is used to describe the interaction of different trapping sites with interstitial
hydrogen, though the model could potentially be applied to helium trapping. The
approach assumes a system of NL moles of interstitial positions in the crystal lattice
and m different types of trapping sites, each sort of trap involving NTk (k=1,...,m)
moles of possible trap positions. yL is the fraction of hydrogen in the lattice and yTk
the fraction of hydrogen in a trap of type k.
In this model, these relations are used to describe the interaction of different trap-
ping sites with interstitial hydrogen. Knowledge of the effective binding enthalpy ∆Ek
between a k trap and a hydrogen atom enables one to calculate the site fraction of
trapped hydrogen. This method is designed to fit in a general thermokinetic simula-
tion framework, based on the existence of energetic traps following the hypothesis of
local hydrogen equilibrium, and has been implemented in the commercial thermokinetic
modelling software package MatCalc to model hydrogen trapping in novel hydrogen re-
sistant steels [189]. A more in depth explanation of this model and its applications is
presented elsewhere [87]. Although this method does not incorporate irradiation en-
hanced kinetics or damage, it can be implemented to model the trapping efficiency of
novel microstructures.
Both helium and to a lesser extent hydrogen behave as large interstitials within the
nickel lattice. As such, the primary driving force for trapping by crystal defects appears
to be the size effect. For grain boundaries, both elements move to the sites associated
with local expansion. However, despite these similarities, their different atomic sizes
result in significantly different trapping behaviours. During early investigations into
helium and hydrogen embrittlement, it was shown that, for hydrogen in nickel, grain
boundary trapping was almost negligible and often less than at dislocations and va-
cancies [196]. For helium however, grain boundary traps were strong, often stronger
than that of vacancies and dislocations [197]. However, as will be shown, more recent
work has suggested that hydrogen may behave in a more similar manner to helium with
respect to embrittlement and trapping than first observed. There is little data on the
trapping characteristics of secondary phases in nickel-based alloys with respect to he-
lium. However, a greater amount of data exists for the binding energies of helium to
microstructural features in ferritic materials [198] and some exists for beryllium [199],
such as various morphologies of grain boundary. If such data existed for nickel-based
alloys, it could prove of great benefit in developing novel radiation-resistant alloys and
assessing the trapping characteristics of established and novel alloys.
81
4.8.1 Helium trapping
Trapping on defects
Thomas et al. [200] measured helium-3 concentrations by thermal desorption analysis
for high-purity annealed, cold-worked, and single-crystal nickel samples. Helium-3 was
introduced by natural tritium decay to eliminate the implantation produced defects of
irradiation. Helium release was detected in the ranges of ∼100 to 300 K and above 800
K. All low temperature releases were only ∼1% of the total helium generated. As such,
helium was measured to be strongly trapped within the microstructure with a binding
energy of ∼2 eV, even for an effectively defect free material. The low temperature data
was shown to fit with a modelled bulk mobility of ∼0.35 eV and a lesser energy for pipe
diffusion.
Later work by Ehrenberg et al. [201] investigated the thermal desorption of helium from
nickel after implantation with fluences of >1017 He-ions cm-2. Again, it was identified
that desorption can be divided into two temperature dependent regions. At <300◦C,
helium desorbs only if it is implanted above a critical dose for blistering. Desorption
at these temperatures can be described by a first order single-step mechanism with a
Gaussian distribution of activation energies around 1eV. This desorption was speculated
to be from surface related traps which are created at the critical fluence, however low
temperature peaks were also seen which were speculated to be due to trapping on dislo-
cations. At >300◦C, helium desorption was attributed to the opening of helium bubbles
at the surface. It was proposed that diffusion appeared not to be a release controlling
process. Noteworthy for future work is that the trapping and release behaviour did not
appear to depend on implantation energy, and as such, the depth to which helium is
implanted, further affirming the conclusion that diffusion does not determine the release
process. From this work, it can be concluded that as bubbles grow, they become more
stable and helium desorption initiates at a higher temperature.
Trapping on secondary phases
Precipitates play an important role in the formation of voids, their subsequent swelling
and resulting alloy embrittlement. The composition, morphology, size, distribution,
matrix-precipitate interface and stability of these precipitates will dictate whether they
retard or enhance irradiation-induced swelling [115,202,203].
One possible way by which precipitates may retard swelling is through acting as sinks
for point defects, preventing the coalescence of vacancies into voids. Similarly, a high
number density of precipitates may be used to trap helium on precipitate-matrix inter-
faces [202], thus preventing helium bubble formation and stabilisation of voids. These
precipitates can also pin dislocations, causing a steady state dislocation structure,
82
thereby retarding swelling [111]. However, if precipitates are large but low in num-
ber, it is believed that these precipitates will enhance swelling by acting as nucleation
points for vacancies and helium to form helium-stabilised voids/bubbles, a product of
the precipitate-point defect collector effect [204]. Large volume fractions of precipi-
tates, depleting the matrix of solutes such as carbon that can inhibit swelling, have also
proven to be detrimental [205]. This effect is postulated to be behind the observed loss
in swelling resistance for Alloy 800 and Hastelloy X under irradiation at temperatures
of 510◦C and 593◦C respectively by Garner and Gelles [194]. Both these alloys precipi-
tate M6C and M23C6 carbides throughout the microstructure at these temperatures, as
evidenced in thermal ageing investigations [206,207]. It may be that long hold times at
these temperatures, where carbides may grow and coarsen, had resulted in large isolated
carbides forming, favouring precipitate-cavity formation, resulting in the more efficient
sinks necessary to induce rapid cavity growth and swelling. However, no microstructural
characterisation was ever conducted, and as such, there is no evidence to validate this
hypothesis.
A key piece of evidence with respect to hydrogen and helium trapping on secondary
phases, is that of the CANDU X-750 spacers investigated in [115], providing the critical
microscopy evidence of trapping on select phases. As shown in Figure 4.12, intragranular
η-phase was found to host a higher density of bubbles on its interface than that seen in
the matrix. The observed increase in bubble density on the interface affirms that helium
and hydrogen are trapped on such interfaces, and as such, should not induce grain
boundary embrittlement. In addition, the more abundant Cr23C6 carbide precipitates
located at grain boundaries were observed to contain a higher density of smaller bubbles
within the precipitate than in the matrix.
Nanostructured alloys, containing many homogeneously distributed intragranular parti-
cles such as those seen to trap helium in X-750, could provide the high number density
of sinks/traps necessary to limit cavity growth, resultant swelling and embrittlement.
However, it is evident that great care must be taken in developing such alloys so to
ensure that precipitate phases remain stable at reactor operating temperatures, taking
note of the enhancement of precipitation kinetics during irradiation [208,209], and thus
retain their radiation embrittlement resistance.
The US National Cladding/Duct Programme [194] evaluated the swelling behavior of
a wide range of ternary model alloys and commercial compositions, within which the
effects of precipitates can be observed. The gamma prime Ni3(Ti,Al) and gamma dou-
ble prime Ni3Nb precipitates of Alloy 718 and 706 appeared to significantly retard
swelling [194]. However, Alloy PE16, with an intergranular network of carbides, is more
susceptible to grain boundary embrittlement and swelling, most likely due to grain
boundary helium bubble formation [111].
As previously described, boron tends to segregate to grain boundaries, aiding grain
83
boundary helium embrittlement due to boron-10 transmutation [183]. As such, elimi-
nating boron or trapping boron within the grains may prevent such embrittlement. A
number of studies have identified the benefits of reducing boron levels on the irradiation
embrittlement of nickel-based alloys. Reductions in boron content from a typical com-
mercial cast level to < 0.2 wppm for Hastelloy X, were shown to increase post-irradiation
ductility significantly [210]. The most notable piece of work however, is that of Barnaby
et al. [202], where low-boron heats of PE16, coupled with varying thermal treatment,
were shown to have a beneficial effect on ductility after an exposure of 20-29 dpa at
850◦C in the Dounreay Fast Reactor. Reducing boron levels from 82 to 18 wppm re-
sulted in significant improvements in post-irradiation total elongation, over 10% for one
heat treatment variant. For this heat treatment variant, they observed an almost inter-
granular carbide free structure, with significant numbers of intragranular TiC and 30 nm
gamma prime precipitates. This suggests that the intragranular carbides are trapping
boron and/or helium, preventing grain boundary embrittlement. Further evidence of
helium trapping on TiC precipitates was presented by Harman [203] on Alloy 800 using
incremental Ti additions. An optimised Ti content of 0.1 wt% resulted in the optimum
post-irradiation creep-rupture ductility, the reasons behind which, although speculative
due to no characterisation being conducted, were a fine distribution of intragranular
Ti-rich precipitates, trapping helium.
Trapping on phosphides
Post-irradiation fracture surfaces of Alloys 718 and 706 were found to have high concen-
trations of sulphur and phosphorus [142]. In both alloys, fracture was assumed to be a
result of weakening of the matrix-eta phase platelet interfaces due to radiation induced
sulphur and phosphorus segregation, and potentially helium bubble formation [142],
affirming the significance of both elements on the radiation embrittlement of nickel-
based alloys. The effects of phosphorus and sulphur on the embrittlement of steels and
nickel-based alloys are well understood, and as a consequence, their levels are closely
monitored by manufacturers [211, 212]. As such, the effects of these elements will not
be covered in detail. In general, sulphur has been seen to segregate to carbide-matrix
interfaces as well as grain boundaries [129, 213], whereas phosphorus segregates largely
to grain boundaries [129], both inducing embrittlement. Phosphorus has also been seen
to enhance hydrogen embrittlement, likely due to grain boundary decohesion [214]. Seg-
regation of phosphorus to grain boundaries has been observed in Alloy 600, along with
boron segregation [211], where phosphorus was found to embrittle the alloy through
reduced boundary strength. In spite of the deleterious effects of phosphorus segrega-
tion, several phosphorus-modified austenitic stainless steels have been shown to provide
efficient helium trapping at particle-matrix interfaces due to the formation of fine phos-
phide particles, delaying the transition to steady-state swelling [166,215]. A more recent
84
paper has also shown that smaller grains, with their higher grain boundary surface area,
capture segregating self-interstitials and vacancies, encouraging recombination [216].
Trapping on oxides
Through achieving a high number density of nanoprecipitates, for example, 1x1024 m-3,
resulting in an inter-precipitate spacing of ∼10 to 15 nm, one could enhance point
defect, helium and hydrogen capture [87]. The work of Gan et al. and Hoelzer et al.
[217, 218] presents two ferritic oxide-dispersion strengthened (ODS) alloys, MA957 and
14YWT developed upon this methodology, the nominals compositions of which are Fe-
14Cr-1Ti-0.3Mo-0.25Y2O3 (wt %) and Fe-14Cr-0.4Ti + 0.25Y2O3 respectively. These
alloys present a promising method of inducing the desired nanoparticles for irradiation
embrittlement resistance. Through ball milling powders of Ti, Y, and O, or Zr, La and
O, with high purity Ni powder, the powders can be annealed at 800◦C and hot pressed,
after which, a high number density of nano-size particles were identified in the Ni lattice.
A number of studies have been carried out on both these alloys with respect to radia-
tion embrittlement, including their interactions with helium through helium implanta-
tion [219]. MA957 was neutron irradiated to 9 dpa at 500◦C with ∼380 appm helium
implanted into TEM specimens. Post-irradiation characterisation showed evidence that
the irradiation induced point defects and helium atoms were trapped by the 2 to 8 nm
diameter, 3 x 1023 m-3 oxide particles with no helium bubbles identified anywhere within
the microstructure [216]. These nanoparticles were found to be unaffected by radiation.
Post-fracture creep specimens showed the nanoclusters retained their stability, with no
change in structure even after 88,555 hours at 800◦C and 100 MPa [111,220], indicating
these nanoparticles are of significant appeal in developing high temperature creep and
radiation embrittlement resistant microstructures. There is potential to replicate this
production method for producing nickel-based irradiation embrittlement resistant alloys
with a similar nanoparticle distribution through pressing powders.
4.8.2 Hydrogen trapping
Trapping on MeH complexes and general defects
Kronmuller and Vargas [221], using magnetic relaxation experiments, evaluated the
binding energy of diatomic metal (Me) hydrogen complexes in nickel. They identified
that due to a modification of the dipole tensor components, the elastic interactions
between H and Me impurities in Ni favours trapping, regardless of whether the substi-
tutional impurity atom, Me, expands or contracts the lattice. A number of impurity
elements were evaluated, notably vanadium, which has been shown to have a strong
85
binding energy for hydrogen in steels [189]. indicating its trapping abilities in both steel
and nickel-based alloys.
Dislocations in nickel act as trapping sites for hydrogen, with a hydrogen trap activation
energy lower than that for the bulk diffusion of hydrogen [222]. The binding energy for
the trapping of hydrogen at grain boundaries is estimated to be 20.5 kJ mol-1 [222].
A recent study [223] has shown that, for Alloy 690, hydrogen causes Ni+ ion radiation
induced defects to reorganise, similar to the effects of temperature. Through SIMS
and TEM investigation, Jambon et al. [223] suggested that the main mechanism behind
hydrogen-assisted defect motion is through the lowering of lattice friction, decreasing the
activation volume for the unfaulting or motion of point defect clusters, up to dislocation
loops. These effects were found to be significant up to ∼325◦C (within the range of
typical LWR temperatures). Hydrogen was found to also affect the morphology and
size distribution of cavities, Frank faulted loops and line dislocations. As such, these
results suggest that hydrogen-defect interactions play a role in the intergranular stress
corrosion cracking of nickel-based alloys.
Angelo et al. [224] computationally investigated hydrogen interactions with edge, screw,
and a Lomer dislocation, in the locked configuration, in nickel. For edge and screw
dislocations, the maximum trapping energy was found to be ∼0.1 eV, where the lattice
is in tension, 3-4 angstroms from the core. For the Lomer-Cottrell lock, the maximum
binding energy was 0.33 eV and was at the core of the a/6(110) dislocation. A number of
low-index grain boundaries were also evaluated, Sigma 3(112), Sigma 9(221) and Sigma
11(113) tilt boundaries. All showed a maximum binding energy of ∼0.25 eV at the tilt
boundary. The results, when compared to experimental studies on hydrogen-initiated
failure on the iron based superalloy IN903, indicating an activation energy for failure
of 0.3-0.4 eV, show that the embrittlement process is directly linked to the trapping of
hydrogen at grain boundaries and Lomer-Cottrell locks [224].
Trapping on precipitates
As previously stated, the thermal treatment of Alloy 600 is conducted to produce grain
boundary carbides in order to reduce SCC susceptibility [127]. Materials with such
intergranular carbides have been shown to be susceptible to hydrogen embrittlement
[115, 192]. These intergranular carbides, which are typically incoherent, could provide
strong traps for hydrogen on their interface, and will concentrate hydrogen on grain
boundaries as a result. Such grain boundary hydrogen concentrations have been shown
to promote intergranular fracture [192].
Turnbull et al. [225], using permeation tests, evaluated the effects of thermal aging on
hydrogen diffusion in alloy X-750 at ∼810◦C, using single and double ageing treatments.
Diffusivity was found to reduce with ageing due to the formation of gamma prime
86
precipitates, trapping hydrogen upon the precipitate-matrix interface with a binding
energy of ∼32 kJ mol-1. Symons [226] however, found a smaller binding energy of ∼15
kJ mol-1 for the same alloy and treatment. Validating this, work by Moody et al. [227]
on Alloy 903 identified a binding energy of ∼19 kJ mol-1 for the gamma prime-matrix
interface. Turnbull et al. proposed that these variations were due to differences in misfit
strain of the gamma prime precipitates between the alloys studied. Regardless, these
precipitates appear to provide irreversible trapping in all cases for the (Ti,Nb(C,N))
precipitates. However, Turnbull et al. recognised that these precipitates may in fact
be irreversible traps, with higher binding energies than observed, but were too few in
number to have a significant effect on the transients, one of the detriments of using
permeation testing as a means to calculate binding energies [228].
The work of Pound [228] evaluated the trapping of hydrogen in X-750 using a potentio-
static pulse technique. With different ageing treatments, he found a significant degree
of trapping present and concluded that the carbide particles (Ti,Nb(C,N)) and grain
boundary chromium carbides, are in fact irreversible traps but coarse gamma prime
is not. Turnbull et al. stated that the intergranular carbides had little influence on
trapping due to their fraction being < 10% of the total carbide fraction. Nonetheless,
given that hydrogen migrates to grain boundaries in nickel [229], and as such, is likely
to behave the same in nickel-based alloys, the location of these carbides means they do
play a prominent role in the trapping of hydrogen, being critical to the hydrogen embrit-
tlement of X-750 [230]. Using an electron microautoradiographic technique, the binding
energy of these intergranular chromium carbides has been shown to be dependent upon
their coherency, with incoherent and coherent carbides being strong and weak traps re-
spectively [229]. This indicates that through novel heat treatments, microstructures can
be engineered for hydrogen embrittlement optimisation. For Inconel 718 and Incoloy
925, Nb,Ti(C,N) and TiC particles behave as irreversible traps respectively. For Alloy
625, TiC particles are the only irreversible traps [231].
As shown, the binding energies for hydrogen in nickel-based alloys are available within
the literature and the opportunity for reducing hydrogen embrittlement susceptibility
through microstructural engineering is clear. Recent developments have made con-
siderable progress in identifying hydrogen interactions in steels and in the design of
hydrogen embrittlement resistance variants of commercial steels [189]. The technology
behind these novel hydrogen embrittlement resistant steels has potential transferability
to nickel-based alloys [87]. Novel steel compositions, alongside their associated ther-
momechanical processing routes, have been developed to resist hydrogen embrittlement
through the use of a high number density of nanoscale precipitates homogeneously dis-
tributed through the microstructure, trapping hydrogen and preventing it from inducing
deleterious effects [87]. These designs are industrially feasible and use established pro-
cessing routes, and as such provide an appealing alternative to other proposed hydrogen
87
embrittlement resistant alloys such as those using oxide dispersion strengthening. It is
feasible that commercial nickel-based alloy microstructures could be engineered, through
novel heat treatments or compositional alterations, to produce similar hydrogen trap-
ping species. As has been shown, this design methodology has successfully been applied
to the trapping of helium and has been shown to limit radiation damage accumulation.
Given the similarities in hydrogen and helium trapping mechanisms and the hydrogen
trapping seen for such nanoprecipitates in steels, this methodology could be applied to
nickel-based alloys to reduce the susceptibility to all key aspects of radiation embrittle-
ment.
4.9 Methods of remediation
The means by which the effects of neutron radiation on nickel-based alloys can be
reduced have been investigated by numerous researchers [111,158,197,232]. Suggestions
include that of reducing boron-10 content in the alloy, reducing the grain size and
trapping helium using secondary phases, which also provide defect recombination sinks.
The most promising method of improving radiation embrittlement resistance appears
to be through the development of a nanostructured microstructure, free of boron-10,
sulphur and phosphorus, containing a high number density of intragranular, and inter-
granular if needed for SCC resistance, nanosized radiation-stable secondary phases with
particle-matrix interfaces optimised for helium and hydrogen trapping. These phases
will increase the swelling incubation dose and reduce embrittlement by preventing high
concentrations of helium/defects to form at isolated sinks by sharing them amongst
the multiple sinks provided by these particles. Grain boundary engineering to produce
boundaries of low coincidence site lattice relationships has also been shown to reduce
an alloy’s susceptibility to intergranular sliding and cavitation. Notably, recent work
has shown the benefits of grain boundary engineering on the resistance to SCC in PWR
primary and secondary water [127]. Finally, the presence of a complex dislocation net-
work may provide defect sinks, reducing the formation rate of voids and bubbles and
limiting swelling.
There are several methods through which the desired high number density nanoparticles
could be achieved; oxide-dispersion via the pressing of nanopowders, Ti additions to
precipitate titanium carbides, other Group IV and V element additions to precipitate
similar, yet unverified helium and hydrogen trapping particles, and through novel heat
treatments for commercial alloys, with little or no compositional alterations. Through
novel heat treatments, nanosized chromium carbides and/or gamma prime precipitates
have been identified as promising trapping and recombination sites. Although this
methodology could be applied to a range of environments, several issues will need to
be evaluated on a case-by-case basis, depending on the operating environment; the
88
hydrogen and helium generation rate, radiation induced segregation and changes in
phase stability, and the effects of thermal ageing.
4.10 Future work
The effects of secondary phase composition and the particle-matrix coherency on defect
recombination and trapping of helium and hydrogen must be evaluated further to better
optimise any future irradiation embrittlement resistant alloys. This could be done via
hydrogen and helium charging, proceeding thermal desorption analysis and irradiation
of heats with different secondary phase characteristics. The long-term stability of such
nanosized secondary phases needs to be established to ensure irradiation embrittlement
resistance is retained throughout the design life in the respective environment. This
could be done using thermal ageing and irradiation experiments. One known method
of stabilising carbides is to add group IV and group V elements such as titanium and
vanadium. Vanadium is of notable interest given vanadium carbide’s proven efficiency
in trapping hydrogen in steels, with high stability resulting in dissolution temperatures
of around 1200◦C for modest additions. The effects of any novel microstructure on SCC
susceptibility will also have to be established. Low temperature neutron irradiations of
high and low boron heats of a nickel-based alloy, ideally Alloy 600 for consistency, would
be of great benefit with respect to providing conclusive evidence on the effects of boron
on grain boundary helium embrittlement and the effects on PWSCC susceptibility. To
develop materials for use in fusion reactors, the effect of impurities generated by trans-
mutation will need to be evaluated in more detail as the rates of nickel transmutation
will be considerably higher than typical of fission environments.
It would appear that a considerable redesign is required for commercial alloys, poten-
tially with respect to both composition and thermomechanical processing, to optimise
irradiation embrittlement resistance in their intended applications. Radiation induced
segregation and solute concentration gradients, inducing non-equilibrium phases, will
need to be accounted for during modelling, as will radiation enhanced kinetics and defect
cascade modelling. This is a complex problem but achievable with current established
modelling techniques.
Hydrogen and helium trapping energies derived from atomic scale simulations and/or
thermal desorption analysis, coupled with with kinetic modelling, incorporating diffusion
data for helium and hydrogen, enables one to model the microstructural and trapping
evolution simultaneously. This allows a designer to optimise an alloy with respect to
both trapping capacity and the depth of such traps. The results of such thermokinetics,
when combined with defect cascade modelling, should provide an accurate indication of
an alloy’s irradiation embrittlement resistance. Neutron irradiation experiments of the
novel alloy should then be conducted to validate this modelling.
89
4.11 Summary
The effects of radiation on nickel-based alloys have been reviewed with emphasis on Al-
loy 600, indicating transmutation induced helium and hydrogen, as the primary cause
of embrittlement. Although the detrimental effects of these elements on the mechanical
properties of nickel-based alloys and steels are well established, the mechanisms behind
the resulting embrittlement are little understood. Recent developments in the design
of hydrogen embrittlement resistance variants of commercial steels present a promising
means to achieve irradiation embrittlement resistance in nickel-based alloys. Novel alloy
compositions, alongside their associated thermomechanical processing routes, could be
developed to resist hydrogen embrittlement through the use of a high number density of
nanoscale precipitates homogeneously distributed through the microstructure, trapping
hydrogen and preventing it from inducing deleterious effects [87]. The designs can be
industrially feasible and use established processing routes, and as such, provide an ap-
pealing alternative to other proposed methodologies, such as the use of oxide dispersion
strengthening. It is feasible that commercial nickel-based alloy microstructures could be
engineered, through novel heat treatments or compositional alterations, to produce sim-
ilar trapping precipitates seen in hydrogen embrittlement resistant steels. The benefits
of such a design methodology have already been partially validated in nickel-based al-
loys through TEM evidence of helium trapping on secondary phases, preventing helium
induced grain boundary embrittlement, and the limiting of radiation damage accumu-
lation through these phases behaving as recombination sinks. Alongside the successful
application of the methodology to hydrogen embrittlement steels, and the similarities
in hydrogen and helium trapping mechanisms, this indicates that the proposed method-
ology is the most promising means to reduce the susceptibility to all key aspects of
radiation embrittlement in nickel-based alloys.
Titanium carbides (TiC), chromium carbides (Cr23C6), γ′, γ′′ and η-phase precipitates,
among others, have all been shown to trap helium on the precipitate-matrix interface.
Each one of these precipitates is already implemented in various commercial nickel-based
alloys, albeit not at the nanoscale desired for irradiation embrittlement resistance. In
addition to helium trapping, these carbides have also been shown to trap hydrogen, al-
though limited data is available on gamma double prime. In addition, although niobium
carbides (NbC) have not been evaluated for helium trapping, they have been shown to
trap hydrogen. As such, there are several carbides already implemented in commercial
alloys that may be engineered to provide the desired high number density of nano-
sized precipitates. Thus, it is feasible to modify existing alloys through small variations
in composition and/or thermomechanical processing to generate the desired radiation
embrittlement resistance microstructure.
Although there is considerable variation in the literature with respect to the hydro-
90
gen and helium binding energies of microstructural features such as these carbides, the
potential for optimising precipitates with respect to trapping energies affords alloy de-
signers a variety of trapping behaviours to choose from, depending on the carbide size,
location and the surrounding matrix, the effects of which will need to be evaluated in
more detail to better optimise future alloys.
Boron, added for hot workability [233] has proven to be significant with respect to
radiation-induced property changes. Helium, resulting from the transmutation of boron-
10 by mainly thermal neutrons, increases intergranular stress corrosion cracking (IGSCC)
susceptibility due to the formation of intergranular bubbles, promoting planar slip. It
is likely that higher boron content will lead to significantly higher radiation-induced
swelling, although no data is available to validate this. This review has identified that
both embrittlement and stress corrosion cracking susceptibility can be reduced through
lowering boron-10 contents through the use of depleted boron or reducing to levels that
are as low as reasonably practical.
It is evident that the mechanisms by which radiation embrittlement occurs in nickel-
based alloys are not solely that of helium embrittlement, and are microstructure and
environment specific. The significant losses in ductility observed for solution annealed
Alloy 706 and PE16 are attributed to radiation-induced grain boundary films of gamma
prime/eta phase and eta phase respectively, indicating the importance of microstructure
prior to irradiation with respect to what embrittlement mechanisms will likely dominate.
Difficulties in applying the proposed methodology have been reviewed, the most sig-
nificant being the application of thermokinetic models to operational conditions where
defect cascade kinetics must be incorporated. This chapter has indicated that although
significant future work is required to ensure the optimised design of future radiation em-
brittlement alloys, the methods by which such work can be achieved, both experimental
and computational, are well established, and as such, present no significant reason why
such a methodology could not be implemented.
91
Part II
Microstructure-hydrogen
interactions
92
Chapter 5
Modelling hydrogen migration
and trapping in steels
5.1 Introduction
The main aim of the present computational approach is to study the evolution equations
for hydrogen diffusion and for trapping particle distribution kinetics in ferrous alloys. A
model for hydrogen reduced mobility due to trapping is introduced. The proposed trap-
ping approach is combined with physical modelling of nucleation, growth and coarsening
of second phase particles within the matrix. In these multicomponent microstructure
simulations, the hydrogen-trap binding enthalpies and mechanical loading conditions
are of relevance too and are taken into account. Thermokinetic approaches to diffusion
and precipitation evolution have already been proposed by several authors [234, 235].
The proceeding work links thermodynamics and kinetics to nanoscale mechanisms using
analytical models for hydrogen transport in the presence of energetic traps, and thus
are able to identify the relationship between microstructure and the ability for hydrogen
to become trapped. Thermokinetic modelling is used to predict the hydrogen binding
properties of microstructures with variations in grain size, matrix system (austenite,
ferrite or martensite) or different compositions for a given hydrogen loading condition.
Very few studies have investigated the role of microstructure on hydrogen diffusion
5.2 Theory
In this chapter, a novel modelling approach has been developed to evaluate hydrogen
trapping phenomena. The approach focuses on hydrogen interactions with multiple
trap types in a thermokinetic simulation framework. Firstly, it is necessary to establish
the fundamental models for nucleation, growth and coarsening. Then, the microstruc-
93
tural evolution equations and the proposed numerical methods for computing hydrogen
interactions are explained.
Precipitation from a supersaturated solid solution is a diffusion controlled reaction in-
volving three processes: nucleation, growth and coarsening. Phase transformation from
supersaturated solid solution via nucleation and growth is driven by reductions in Gibbs
free energy, while coarsening is driven by reductions in the overall interfacial energy of
the precipitates with the matrix. Given that precipitation influences material proper-
ties such as strength, toughness and creep resistance, precipitation kinetics has been an
area of considerable research. When combined with thermodynamic data for secondary
phases, kinetic models provide a useful tool for evaluating material properties evolution,
such as the fraction and size of precipitates with respect to time and temperature. In this
section, expressions for the stages of precipitate nucleation, growth and coarsening are
introduced. Subsequently, these models are integrated to account for multiprecipitation
kinetics to predict microstructure evolution.
5.2.1 Nucleation of precipitates
Thermokinetic simulations were carried out employing MatCalc software [236, 237]. In
this approach, precipitation nucleation kinetics were calculated employing classical nu-
cleation theory (CNT) [238] for multicomponent systems [236]. Accordingly, the tran-
sient nucleation rate J is given by:
J = NoZβ
∗ · exp
( −G∗
kB · T
)
· exp
(−τ
t
)
(5.1)
which describes the rate at which nuclei are created per unit volume and time. N0
represents the total number of potential nucleation sites and Z, the Zeldovich factor,
quantities the ability for a nucleus to be destabilised by thermal excitation in comparison
to its inactive state; this is given by [238]:
Z =
( −1
2pikBT
· ∂
2∆G
∂n2
) 1
2
for n = n∗ (5.2)
where ∆G represents the total free energy change for nucleus formation, n is the number
of atoms in the nucleus, kB is the Boltzmann constant and T is the temperature. The
derivative of ∆G is taken at the critical nucleus size n∗. The atomic attachment rate
β∗ takes into account the long-range diffusive transport of atoms, which is needed for
nucleus formation if the chemical composition of the matrix is different from the chemical
composition of the precipitate. A suitable multicomponent formulation for β∗ has been
94
derived by Svoboda et al. [236], and is given by:
β∗ =
4piρ∗2
a4Ω
[
n∑
i=1
(cmi − c0i)2
c0iD0i
]−1
(5.3)
where ρ∗ is the critical nucleation radius, a is the lattice constant, Ω is the molar
volume of the precipitate, cki and c0i are the concentrations in the m precipitate and
in the matrix in equilibrium, D0i is the diffusivity of element i in the matrix. The
incubation time τ is given by [238]:
τ =
1
2β∗Z2
(5.4)
The critical energy ∆G∗ for the formation of a spherical nucleus, with interfacial energy
γ and volume free energy change ∆Gvol is:
∆G∗ =
16pi
3
· γ
3
(∆Gvol)
2 (5.5)
MatCalc incorporates CALPHAD-assessed thermodynamic model parameters for the
evaluation of chemical potentials, driving forces and interfacial energies between the ma-
trix and the precipitate. The latter is based on the generalised broken-bond model [239]
and takes into account capillarity effects [240]. In this work, thermodynamic database
mc fe v2.021.tdb was used [241]. The diffusion data required for both nucleation and
growth kinetics are based on the SFFK model [26,236], as is described in the following
section. In this work, diffusion database mc fe v2.006.ddb was used [241].
5.2.2 Growth and coarsening - The SFFK model
Once a precipitate nucleates its growth stage follows. This situation can be described
by the equations of Svoboda et al. [236] for precipitate radius and composition using a
mean-field approach known as the SFFK model [236, 237]. This model states that in a
system containing m chemical elements and k randomly distributed precipitates with
spherical diffusion fields, the total free energy can be described by [235]:
G =
∑m
1
N0iµ0i +
∑ n
1
4piρ3m
3
(
λm +
∑ n
i=1
cmiµmi
)
+
∑ n
1
4piρ2mγm, (5.6)
where µ0i is the chemical potential of the component i in the matrix and µmi is the
chemical potential of the component i in the precipitate m. The chemical potentials are
expressed as a functions of the concentration cmi, λm takes into account the contribu-
tions from the elastic energy due to the molar volume change due to precipitation. ρm
is the radius of the discrete precipitate class. The numerical time integration is carried
95
out following the Kampmann-Wagner approach [242]. This evaluation routine provides
the evolution of size distributions, mean radii and phase fractions of precipitates for
arbitrary heat treatment schedules. Further details of the models used and the numer-
ical treatment of the evolution equations were presented by Svoboda et al. [236] and
Kozeschnik et al. [237].
5.2.3 Dislocation evolution model
In this work, an enhanced one-parameter model is used to describe work hardening
and softening. The term “one-parameter” emphasizes that the stress-strain response
is controlled by the evolution of the total dislocation density, ρ. This evolves as a
consequence of the dislocation generation and annihilation terms during deformation
and is expressed as:
dρ
dt
=
dρ+
dt
− dρ
−
dt
. (5.7)
The first term in the right hand side of the equation accounts for the dislocation accu-
mulation. The second term accounts for the dynamic recovery and is dependent on the
applied strain rate and spontaneous annihilation.
5.2.4 Dislocation generation
Based on the Kocks-Mecking formalism [243] and assuming that mobile dislocations
move a mean free path before they become immobilised or annihilated, the increase in
the dislocation density ρ is given by [243]:
dρ+
dt
=
Mϕ˙
dL
(5.8)
where M is the Taylor orientation factor, ϕ˙ is the plastic strain rate and b is the
magnitude of the Burgers vector. The dislocation travelling distance L is limited by the
average spacing between dislocations [244]:
L =
A√
ρ
(5.9)
where A is a material constant dependent on composition.
5.2.5 Dislocation annihilation
When two dislocations with antiparallel Burgers vectors approach a critical distance,
they can become annihilated, reducing the dislocation density [245]:
dρ−
dt
= B · 2dann
b
ρMϕ˙ (5.10)
96
where B is a material constant, dependent on the number of activated slip planes, and
the critical distance dann [246]:
dann =
Gsb
4
2pi (1− ν)Qfv
(5.11)
where Gs is the shear modulus of the matrix, ν is the Poisson’s ratio and Qv
f is the
vacancy formation energy. The model describing recovery by dislocation climb is based
on the assumption that thermally activated climb is controlled by the self-diffusion
coefficient along dislocations Dd (pipe-diffusion) [247]:
dρ
dt
= −C2DdGsb
3
kBT
(
ρ2 − ρ2eq
)
(5.12)
where ρeq is the equilibrium dislocation density and C is a calibration parameter, which
accounts solute trapping effects. The dislocation core diffusivity is coupled to the bulk
diffusivity via a temperature-dependent factor αd as [248]:
Dd = αdD
∗
i (5.13)
The multi-component diffusion model describes the tracer diffusion coefficient, D∗i , by
the relation [236]:
D∗i = RgTMi (5.14)
with Mi being the diffusional mobility of component i. In MatCalc Mi is stored in the
mobility databases and D∗i is uniquely given as a function of composition and tempera-
ture. with Rg is the universal gas constant and T the absolute temperature. For more
information on the models, algorithms and approaches implemented in MatCalc, please
see [249].
5.2.6 Trapping of interstitial elements
Materials with a high defect density have an increased hydrogen trapping capacity lead-
ing to an increase in hydrogen content exceeding the solubility limit of the perfect
lattice. This is because when the chemical potential of hydrogen reaches equilibrium,
a material can absorb hydrogen to the solubility limit of the host lattice, and then ad-
ditional hydrogen can occupy the available traps. The free hydrogen in the lattice and
the trapped hydrogen equilibrate with each other and the apparent solubility or total
hydrogen concentration can be significantly higher than the lattice solubility [195].
The presence of traps will also reduce the apparent diffusivity, as atoms require energy
greater than the lattice migration energy to leave the trap and, consequently, the ap-
parent diffusivity will be lower than in the undisturbed lattice diffusivity. The decrease
97
in total energy G of a system by traps with a trap-interstitial binding enthalpy ∆Ek is
described by Fischer et al. [67] and shown in eq. (5.15). This model is used to describe
the interaction of different trapping sites with interstitial hydrogen.
G = Go +RgT

[yL ln yL + (1− yL)ln(1− yL)]NL
+
m∑
k=1
{
[yTk ln yTk + (1− yTk)ln(1− yTk]NTk − ∆Ek
RgT
yTkNTk
}

(5.15)
The approach assumes a system of NL moles of interstitial positions in the crystal lattice
and m different types of trapping sites, each sort of trap involving NTk (k=1,...,m)
moles of possible trap positions. yL is the fraction of hydrogen in the lattice and yTk
the fraction of hydrogen in a trap of type k. In this model, these relations are used to
describe the interaction of different trapping sites with interstitial hydrogen. Knowledge
of the effective binding enthalpy ∆Ek between a k trap and a hydrogen atom enables
one to calculate the site fraction of trapped hydrogen. This approach is reviewed, next.
yL and yTk can be related by [67]:
yl (1− yTk)
yTk (1− yl) = exp
(−∆Ek
RgT
)
= Kk (5.16)
If ∆Ek is known, eq.(5.16) can be resolved with respect to the site fraction of trapped
hydrogen yTk:
yTk =
yL
Kk + yL (1−Kk) (5.17)
Accordingly, the effective diffusion coefficient in the matrix is described by equations
(5.18) through (5.20) [235]:
D˜eq =
(
D∗i +
∑ n
i=1
DT i
dcT i
dcL
∣∣∣∣∣ x,t
)
dcL
dc
∣∣∣∣∣ x,t, (5.18)
dcL
dc
=
[
1 +
∑m
1
VL
VTk
Kk
[Kk + VLcL (1−Kk)]2
]−1
, (5.19)
dcTk
dcL
=
VL
VTk
Kk
[Kk + VLcL (1−Kk)]2
(5.20)
with DT i being the effective diffusion coefficient of trap element i, n the total number
of trapping elements, VL the volume of the lattice containing one mol of interstitial
positions in the lattice (where hydrogen can diffuse freely), VTk the volume of the lat-
tice containing one mol of possible trap positions (where hydrogen becomes immobile).
Since the unit volume of material contains 1/VL moles of interstitial lattice positions
and 1/VTk moles of possible trap positions, the volume containing one mol of pos-
98
sible hydrogen positions is given as V=1/(1/VL+1/VTk). cL is the concentration of
hydrogen in the lattice (cL=yL/VL), cTk the concentration of hydrogen in the k traps
(cTk=yTk/VTk), and Kk the equilibrium constant.
In this model, the contributions to the diffusion coefficient are separated into free and
trapping components, which contain all the positions of the lattice in the vicinity of a
trapping element. The non-trapped fraction (free hydrogen) is generally considered to
dominate embrittlement. The relationship in eq. (5.15) through (5.17) does explicitly
consider the effect of different trapping sites and the related specific binding energies,
but it does not distinguish between reversible and irreversible traps, as all energetic
traps associated to a specific binding energy are characterised by a critical tempera-
ture, where hydrogen ejection from the trap is possible. To simulate hydrogen diffusion
throughout processing, a mean field approach to describe the microstructure evolution
is used. A unit volume with evolving trapping properties can be occupied by the diffus-
ing hydrogen atoms. The calculation is divided into time and temperature increments.
Dynamic equilibrium between hydrogen in the lattice and in trapping sites develops.
The concentration ratio at these sites depends upon the relative binding energy for a
hydrogen atom. So the trapped phase fraction of initially homogeneously distributed
diffusive hydrogen atoms over the microstructure (t=0) at every time/temperature step
is considered and used in the subsequent time/temperature steps to update the hydrogen
trap concentration state. Trapping is conducted in a single simulation cell to observe
the synchronised evolution of microstructure and corresponding trapping within a single
precipitation domain and as such, Fick’s law across multiple cells is not considered. The
apparent diffusion coefficient at the very beginning of the diffusion process is taken as
that when all hydrogen is mobile. If all traps are filled, yTk is close to 1, the apparent
diffusion coefficient is equal to the tracer diffusion coefficient Di
∗.
This method is designed to fit in a general thermokinetic simulation framework; based on
the existence of energetic traps following the hypothesis of local equilibrium as described
above.
5.3 Simulation setup
The examples shown next are extracted from literature. The goal is to reproduce and
simulate experimental results to test the model accuracy. Some input parameters need
to be entered, and are used for all cases considered here. Table 5.1 summarises the input
parameters; the majority were taken from earlier work and their sources are indicated.
99
Table 5.1: Summary input parameters
Parameter Value / [unit] Cf. equation Reference
A 90 [250] this work
B 2 [251] this work
C 10−3 [252] this work
∆EH dislocations 20.6 / kJ/mol [234] and [235] [253]
∆EH grain boundaries 58.6 / kJ/mol [234] and [235] [253]
∆EH TiC (size dep.) 22.0 + 5 10
7rmean(T iC)/kJ/mol [234] and [235] this work
∆EH TiC(coherent) 22.0 / kJ/mol [234] and [235] this work
∆EH TiC(incoherent) 98.0 / kJ/mol [234] and [235] [46]
∆EH NbC 59.3 / kJ/mol [234] and [235] [254]
5.4 Results
This section examines the accuracy of the combined models for precipitation and mi-
crostructural evolution. Quantitative and qualitative aspects are compared to data from
the literature. The material, processing and loading conditions are considered in each
case. The most important input values used for the thermokinetic simulations are given
in Table 5.1, more detailed information regarding simulation methods is given for each
example. The concentration of trapped and free hydrogen over a wide range of different
microstructures in pure iron and for various steel grades is outlined next.
5.4.1 Case 1: hydrogen trapping in pure iron
The pure Fe-H system is the subject of this first case study as it is critical to examine
the potency of grain boundaries and dislocations as hydrogen traps in the absence
of secondary phases. The relative amount of trapped hydrogen and its evolution in
the presence of various lattice defects and for a variety of temperatures is calculated.
In MatCalc, the polycrystalline grain microstructure is modelled as an ensemble of
space-filling tetrakaidecahedra with minimum separating area and 120 degree angles
between faces. In MatCalc, a slightly simplified version of the tetrakaidecahedron is
used with planar faces, which makes it easier to evaluate its geometrical features. From
these objects, the total area, line length and number density of grain boundaries, grain
boundary edges and grain corners can be evaluated. From these, the density of potential
trapping sites is calculated by evaluating the number of atoms located on the junctions.
For this case study, the grain size is varied by altering the area of these grains and
evaluating the corresponding trapping capacity.
As the trapping mechanisms and trapping at different sites is not clearly established, we
use the characteristic trapping energies of hydrogen to dislocation cores ∆EH to dislocations
20.g kJ/mol and to grain boundaries ∆EH to grain boundaries 58.6 kJ/mol as suggested
by Gaude-Fugarolas [253]. These parameters describe experimental results accurately.
100
Hydrogen is the lightest and smallest of solute elements and is extraordinarily mobile,
especially in iron and other body-centred cubic structures. Assuming a perfect iron
crystal with no defects, the solubility limit of hydrogen in iron is approximately 5x10−5
at 700◦C [255] and 2x10−5 at 300◦C [256]. Below 300◦C, the degree of scatter in the
measured hydrogen solubility in iron becomes significantly large, with the reported data
at around 25◦C spanning a range of more than 4 orders of magnitude [257]. Several
investigations have ascribed the scatter in solubility and diffusion to the presence of
trapping sites in bcc iron [255,257].
The change in hydrogen trapping efficiency in iron introduced by grain refinement or
cold work depends on the concentration of lattice defects and their interaction energies
with hydrogen. In this case study, the cold work is accounted for by simply altering the
dislocation density, thus increasing the density of trapping sites. Some non-equilibrium
defects can be removed upon annealing as a function of temperature and time. In
the first part of this case study, low temperature (25◦C) is considered for a system in
quasistatic equilibrium with stable defects. The statistical thermokinetic calculations
involve three energetically distinct sites: trapping sites at dislocations, trapping sites at
grain boundaries and non-trapping sites in the bulk. The calculations are performed with
a total of 1 ppm hydrogen uniformly distributed across the bulk. All calculations shown
in this case study are performed for equilibrated conditions at the given temperatures.
Figure 5.1 depicts the predictions of the relative amount of H trapped at various grain
sizes at 25◦C. The grain size ranges from 1 µm to 10 mm. From Figure 5.1, it is
clear that with decreasing grain size the amount of trapping sites increases, and the
percentage of trapped hydrogen reaches its maximum at smaller grain sizes (less than
10 µm). Grain sizes of 10 to 100 µm are of great interest as these are very common in
industrial steels. The trapping efficiency within this range is given in Figure 5.1. The
simulation results predict a drop in the hydrogen trapping efficiency at grain boundaries
by 70% between 1 µm (100% of hydrogen is trapped) and 100 µm (30% of hydrogen is
trapped) at 25◦C. At a grain size of 1 mm and above, the trapping efficiency of grain
boundaries is negligible.
Figure 5.1: Calculated amount of trapped hydrogen for grain sizes ranging from (a) 1 to 104
µm and (b) 10 to 100 µm.
101
Figure 5.2 depicts the calculated amount of trapped hydrogen as a function of dislo-
cation density at 25◦C. The results highlight that hydrogen desorption during plastic
deformation is strongly influenced by interactions between hydrogen atoms and dislo-
cations. As the simulations were performed at an isothermal temperature of 25◦C and
a constant grain size of 100 µm, the concentration of trapped hydrogen around disloca-
tions depends on the binding energy ∆EHtodislocations to dislocations and the magnitude
of the dislocation density. Accordingly, the concentration of trapped hydrogen is greater
for heavily deformed materials. This results from the build-up of Cottrell atmospheres
around dislocation cores.
Figure 5.2: Calculated amount of trapped hydrogen for dislocation densities ranging from
1010 m−2 to 1017 m−2 at a temperature of 25◦C and for a grain size of 100 µm.
There is a decrease in hydrogen diffusivity with increasing dislocation density. This
is translated into an increase in the activation energy for H diffusion from 5.4 to 11
kJ/mol in well-annealed iron [257] and for high dislocation densities (3x1014 m−2) at
25◦C [257], respectively. The present results confirm such observations, which indicate
that when the dislocation density is 1015 m−2, hydrogen diffusion is almost suppressed,
whereas for a dislocation density of 1012 m−2 approximately 70% of the hydrogen is
diffusible. Such a relationship of dislocation density with trap density is consistent
with the work of Sofronis and McMeeking [258], who proposed a relationship between
the percent equivalent plastic strain and trap density in iron, affirming the presence of
an increase in trap density with plastic strain from the strain-free trap density, and a
maximum limit of trapping capacity as strain reaches around 100%. This affirms the
direct relationship between dislocation density and trapping capacity, which in turn, are
functions of plastic strain. However, as shown later, grain boundaries are also involved
in the trapping and dominate at low strains.
The temperature dependence of the trapping behaviour in pure iron is now investigated.
As shown in the previous examples, the temperature and the binding energy between
a hydrogen atom and a specific lattice defect play a key role. Figure 5.3 illustrates the
trapping behaviour for microstructures with grain sizes (φ) of 10, 100 and 1000 µm and a
102
constant dislocation density of 1010 m−2. The simulations are performed with a stepped
equilibrium calculation from 100 to 600◦C, prescribing null trapping for temperatures
in excess of 400◦C.
Figure 5.3: Amount of trapped hydrogen at temperatures ranging from 100 to 600◦C for grain
sizes, φ, of 10, 100 and 1000 µm.
At temperatures between 200◦C and 400◦C, the influence of finer grains on the trapping
behaviour of hydrogen is markedly reduced. Above 400◦C the trapping behaviour seems
to be completely insensitive to the presence of grain boundaries. For the trap densities
and interaction energies used in the simulations, it can be concluded that trapping
sites with these assumed energies do not accommodate hydrogen at temperatures above
500◦C. This temperature effect is described by da Silva and McLellan [255] validating
the insensitivity of hydrogen solubility to the presence of grain boundaries in the range
300◦C up to the α →γ transformation.
5.4.2 Case 2: hydrogen redistribution in ferritic and martensitic mi-
crostructures
To understand the flux of hydrogen atoms within a certain matrix, we must first know
the influence of microstructure including the distribution of lattice defects, as well as
their interactions with hydrogen. Gaude-Fugarolas [253] described the redistribution
of hydrogen as a function of thermal agitation and atom mobility and related it to
a random walk, taking into account the diffusivity and saturation of hydrogen during
the cooling process of two different low alloy steels. The two materials are different in
their composition and resulting microstructural evolution: one having austenite trans-
forming into ferrite and the other into martensite. The initial hydrogen content for
both materials is assumed to be around 5 ppm. The model input data (e.g. austenite
transformation temperature, mean grain size and dislocation density) are taken from
ref. [253]. Figure 5.4 shows the modelled redistribution of hydrogen into the various
trapping sites for both steels during cooling compared to the results of ref. [253]. The
simulations show good agreement with the results, in that the redistribution of hydrogen
103
is strongly influenced by the microstructural evolution during cooling. The two steels
are good examples because their difference in microstructure is large and so the charac-
teristics of hydrogen redistribution are clearly noticeable. As shown in Figure 5.4, the
hydrogen atoms partition from the free interstitial sites towards the trapping sites in
dislocations and grain boundaries. The difference in hydrogen redistribution between
the two steels is attributed to their differences in dislocation density and grain size. The
ferritic matrix is assumed to have a dislocation density of 1010 m−2 and a mean grain
size of 100 µm [253]. The martensitic matrix is assumed to have a dislocation density
of 1015 m−2 and a mean grain size of 10 µm [253]. At the end of the cooling process,
all the hydrogen is trapped in the martensitic microstructure, whereas in the ferritic
microstructure some diffusible free hydrogen (around 33%) is still present.
Figure 5.4: Hydrogen redistribution into different trap sites for idealised ferritic and marten-
sitic steel. Reference data is taken from [253].
The modelled redistribution shown in Figure 5.4 disregards the effects of precipitates and
only accounts for the effects of the two different idealised microstructures. Nevertheless,
104
the models provide a description of hydrogen trapping effects and redistribution along
dislocations and grain boundaries. A further reduction in the amount of free hydrogen
can be assumed for the interaction of precipitates to the free hydrogen atoms in the
ferritic system. The presence of precipitates in the martensitic matrix is of moderate
relevance with respect to the hydrogen trapping as, due to the cooling process, all the
hydrogen is trapped in the final microstructure already.
Although in this example, the trapping of hydrogen is not rate limited by diffusion,
there may be situations where the time at temperature, and thus diffusion distance, is
insufficient to ensure the filling of traps, and as such, the model would be unsuitable for
such an application. One example would be that of very rapid quenching to cryogenic
temperatures where hydrogen is unable to diffuse to newly created traps.
5.4.3 Case 3: hydrogen redistribution in ferritic steels under deforma-
tion
The effect of plastic deformation on hydrogen desorption in ferritic steels is investigated
and analysed using thermokinetic modelling. The evolution of the dislocation density is
calculated according to equation (5.7). The experimental data are taken from ref. [259].
The chemical composition in wt.% of the steel is Fe-0.004C-0.078Mn-0.015Cr-0.046Al-
0.015Ti. The mean grain size is 120.3 ± 19.6 µm [259]. The dislocation density of the
deformed steel was estimated by converting tensile stress into shear stress τ :
τ = τo + αGb
√
ρ (5.21)
where α is a numerical constant with a value of 0.3 to 0.6 (0.6 in this case), G is the
shear modulus (81.7 GPa) and b is the magnitude of the Burgers vector, equal to 0.287
nm for ferrite. The calculated dislocation density changes for 10 and 20% deformed
sample are 3.5313 and 6.31x1013 m−2 respectively, with τ o = 48.5 MPa (calculated as
half the tensile stress at yielding) [259]. It has been shown that for annealed iron, the
measured dislocation density is practically naught when compared with iron subjected
to the plastic strains of the magnitudes considered here [259], as such, it has been
assumed that the undeformed sample contains ρ = 1010 m−2.
Although in this case, strain is assumed to be homogeneous, in reality, a strain gradient
will result in a corresponding gradient of hydrogen trap density and thus a gradient of
free hydrogen content yL. To further develop the applicability of the model described,
one possible area of future work would be to incorporate it into a finite element model,
allowing one to evaluate the effects of concentration gradients such as that of trapping
density, stress, and the resulting hydrogen concentration gradients.
The material was deformed at 25◦C. Further details about sample preparation and
105
experimental conditions can be found in ref. [259]. For the simulations, the heat treated
matrix is assumed to have a mean grain size of 120 µm. The interaction energies of
hydrogen with dislocations and grain boundaries are taken, as with cases 1 and 2, to be
20.6 kJ/mol and 58.6 kJ/mol, respectively.
The relationship between the redistribution of hydrogen and the applied plastic strain is
illustrated in Figure 5.5. The solid line represents the total amount of trapped hydrogen
while the triangles are the experimental data taken from ref. [259]. The horizontal line
at 0.26 ppmw represents the amount of trapped hydrogen at grain boundaries. The
dotted line represents the amount of free hydrogen in the system.
Figure 5.5: Simulated (lines) and measured (triangles [259] evolution of hydrogen content (a)
and computed dislocation density evolution (b) for ferritic steel as a function of plastic strain.
The evolution of the dislocation density, evaluated by equation (5.7) and illustrated in
Figure 5.5(b), shows significant increase in the dislocation density with plastic strain.
The total trapped hydrogen content can be related to the dislocation density evolution,
106
with the assumption of a constant grain size of 120 µm. The resulting simulations agree
well with the experimental data and assumptions given in ref. [259].
5.4.4 Case 4: hydrogen trapping at NbC nanoprecipitates
The effect of NbC particles on hydrogen trapping in steel is considered here. Hydrogen
trapping by NbC particles has been identified and measured quantitatively by several
authors [260–263]. Wei and Tsuzaki [262] investigated the hydrogen trapping character-
istics of nanosized NbC precipitates in a 0.05C-0.41Nb-2.0Ni tempered martensitic steel
using thermal-desorption analysis and atomic-scale microstructural observation. The
major processing conditions of the material are quenching from 1350◦C and tempering
at 700◦C. Ohnuma et al. [263] provides direct structural evidence for the trapping of
hydrogen by NbC with the use of small-angle neutron scattering (SANS). Case study 4
is based on the experiments of Wei [262], which are reproduced by computational mod-
elling. The mean dislocation density is assumed to be 1015 m−2 for the sample having
a mean grain size of 100 µm.
The hydrogen trapping character of NbC nano-particles during the tempering stage is
investigated. The precipitation evolution of the NbC particles is calculated first for the
isothermally heat-treated state at 700◦C for 100 h. The specimen has been cathodically
charged with hydrogen for 48 h, which is assumed to be long enough for the specimen
to become saturated [262].
The volume-equivalent precipitate radius rmean is determined by assuming a spherical
shape for the precipitates and is shown in Figure 5.6(a). Amounts to around 20 nm
with a number density of around 1020 m−3 after 100 hours isothermal aging at 700◦C
are obtained and shown in Figure 5.6(b).
Figure 5.6(c) shows that the major part of the trapped hydrogen in the tempered marten-
site is related to the NbC growth, and the hydrogen content is quickly released back into
the matrix upon tempering at 700◦C. This behaviour can be explained by the simulta-
neously fast decrease in the number density of an initially large amount of small NbC
clusters (radius < 1 nm). This result, as shown in Figure 5.6, suggests that the early
formed NbC clusters are only able to trap hydrogen weakly, while the coarser (around
20 nm) NbC precipitates are significantly stronger traps. The traps associated with
the microstructure of tempered martensite, such as grain boundaries and dislocations,
contribute little to the total trapped hydrogen in this steel, and their interaction with
hydrogen at elevated temperatures is weaker compared to that of the NbC precipitates.
Precipitates can be associated with various trap sites with different interaction energies
with hydrogen. These include precipitate/matrix interfaces, the coherency strain fields
or even crystal defects within the precipitates. The calculations are performed with one
single NbC trapping energy value, which is taken as 59.3 kJ/mol. This value is close
107
Figure 5.6: Calculated evolution of the (a) mean radius, (b) number density of NbC, and (c)
the calculated (solid line) and measured (triangles) content of hydrogen remaining in energetic
traps as a function of exposure time for tempering at 700◦C in 0.05C-0.41Nb-2.0Ni steel.
to 56 kJ/mol reported by Wei and Tsuzaki [262] after charging the specimen for 96 h.
The calculations reproduce the trapping behaviour of NbC for the temperature range
of 300◦C to 800◦C, as depicted in Figure 5.7, for which the tempering time is 3 h. The
initial maximum amount of solute hydrogen is assumed to be 1.0 ppm after charging
for 1 h. The trapping capacity around 500◦C can be explained by the precipitation
reactions of the NbC particles [262]. From Figure 5.7, it is obvious that the maximum
hydrogen trapping capacity results at a tempering temperature of 600◦C, where the
NbC precipitation is complete, reaching a highest number density and optimal radii.
108
No coarsening of the NbC precipitates is observed in the simulations at 600◦C for 3 h.
Figure 5.7: Calculated (solid line) and measured (triangle) content of hydrogen remaining
in energetic traps at NbC precipitates as a function of different isothermal tempering stages.
Hydrogen pre-charging was performed for 1 h. Reference data is taken from [262].
It can be concluded that the trapped hydrogen content at NbC particles decreases with
increasing tempering temperature and/or storage time due to the resultant coarsening
of the particles.
5.4.5 Case 5: hydrogen trapping at coherent and incoherent TiC pre-
cipitates
The effect of TiC particles on hydrogen trapping in iron is considered here. Hydrogen
trapping by TiC particles has been identified and measured quantitatively by several
authors [61, 254, 264–268]. In order to clarify the binding conditions and hydrogen
trapping efficiency for different precipitate sizes, thermokinetic precipitation reactions
are simulated for a 0.42C-0.30Ti (wt.%) steel. Experimental data are taken from ref.
[254]. The input parameters for the simulations are the mean grain size, which is taken
as 100 µm, the dislocation density, which is taken as 1012 m−2, and the interaction of
hydrogen with dislocations, grain boundaries, TiC (large incoherent particles) and TiC
(small coherent particles), which are respectively taken as 20.6 kJ/mol, 58.6 kJ/mol,
98.0 kJ/mol [46] and 22.0 kJ/mol. The hydrogen trapping efficiency of TiC is assumed
to be dependent on the coherency of the precipitate with the surrounding matrix and
on the precipitate size. Pressouyre and Bernstein [269] have argued that small TiC
precipitates have a smaller interaction energy with hydrogen compared to larger ones,
but Takahashi and Kawakami [61] have indicated that a high number density of small
coherent TiC particles can be most effective in trapping hydrogen. To take the size
effect into account the mean radius of the precipitates is related to the binding energies,
formulated as:
109
∆EH−T iC = 22.0 + 7 · 107 · rmean(T iC) [kJ/mol] (5.22)
The thermal treatment and experimental details are only briefly discussed here. Full de-
tails are available from ref. [254]. The austenitisation of the specimen was performed at
the relatively high temperature of 950◦C in order to obtain the incoherent TiC particles.
According to ref. [254], about 83% of the total 0.43 vol.% TiC will remain undissolved
in austenite at 950◦C. This is in good agreement with this model, as depicted in Figure
5.8.
Figure 5.8: Simulated TiC phase fraction (vol.%) from Thermocalc (triangles) and MatCalc
(line). Reference data is taken from [254].
The purpose of these simulations is to identify the origin of hydrogen trapping for TiC
particles. Figure 5.9 shows the hydrogen content trapped by TiC precipitates in an
iron lattice as a function of the tempering temperature. The solid line is the calculated
hydrogen content in ppm trapped at TiC, and the triangles represent the experimental
measurements given in ref. [254].
As mentioned previously, incoherent TiC particles, have strong attractive interaction en-
ergy with hydrogen, trapping hydrogen during heat treatments at 600, 800 and 950◦C.
The differences in the amount of trapped hydrogen, dependent on the aging temperature,
can be correlated to the precipitate distribution at the corresponding tempering condi-
tions. Figure 5.10 shows the scaled number density for TiC at 600, 800 and 950◦C. It is
clear that with decreasing tempering temperature, the mean precipitate size decreases,
giving the solute hydrogen atoms more possible sites to be trapped. The correlated
hydrogen trapping capacity with the precipitate distributions, shown in Figures 5.10, is
depicted in Figure 5.9.
The simulated results are in good agreement with the experimental observations, indicat-
ing that the large amount of nano-sized TiC precipitates, together with grain boundaries
and dislocations, are able to trap hydrogen at room temperature, even when the inter-
110
Figure 5.9: Hydrogen content trapped by TiC particles, showing the experimental values
(triangles) and calculated trapped content (line) as a function of temperature and related
particle distribution of TiC. Reference data is taken from [254].
action energy of these coherent particles is assumed to be approximately three times
smaller compared to the incoherent TiC precipitates (Table 5.1). The trapping energy
of large incoherent TiC particles is identified as being able to trap hydrogen even at
elevated temperatures (600 to 950◦C).
111
Figure 5.10: Scaled number density of TiC at (a) 600, (b) 800 and (c) 950◦C for a 0.42C-0.30Ti
(wt.%) steel.
5.5 Summary
Thermokinetic computer simulations have been carried out to capture, for the first
time, the essential elements of hydrogen trapping in complex iron and steel microstruc-
tures. The model is formulated on classical laws of precipitate nucleation and growth,
combined with descriptions of energetic trap effects on hydrogen mobility. Therefore,
the presented simulation approach involves predictive precipitation reactions and mi-
crostructural evolution combined with the ability to predict the redistribution of solute
hydrogen throughout real processing conditions in pure iron and complex steels. The
112
simulations involve the identification of potential energetic trap sites which precede the
diffusion behaviour of interstitial hydrogen, and their relationship to the characteris-
tic features of different microstructures, quantifying the concentration of diffusible and
trapped hydrogen. Computing the effects of hydrogen trapping in iron and steel is
essential for the determination of trapping sites and depths to accurately predict and
quantify the redistribution of hydrogen.
The thermokinetic mean-field approach is important as it provides the link between
statistical distributions of microstructural defects, such as precipitate size distributions,
dislocation densities and the presence of grain boundaries with the hydrogen redistri-
bution during heat treatment.
Two- or three-dimensional lattice defects, such as dislocations and grain boundaries,
have a two-fold effect on hydrogen diffusion: they enhance diffusion along the disturbed
lattice regions (pipe diffusion), but also display a hydrogen trapping effect. As the
grain size decreases, or the dislocation density increases, the mobility of hydrogen will
increase with increasing the pipe diffusion area. However, finely grained or highly de-
formed microstructures have higher densities of potential trap sites for hydrogen atoms,
leading to a reduction in the hydrogen mobility. The compromise between contradic-
tory effects is outlined in the first two case studies presented in this work, indicating
that very fine grains (< 10 µm) or high dislocation densities (> 1015 m−2) result in
the highest hydrogen trapping ability for pure iron systems. The presence of precipi-
tates, carbides in this case, may change this trend, as some carbides, such as TiC and
NbC, show strong interaction energies with soluble hydrogen in steel. The influence of
such trapping sites on precipitate/matrix interfaces is introduced in case studies 4 and
5. The proposed model is able to predict the global partitioning of trapped and free
hydrogen as a function of interaction energies, temperature and number densities of the
three major types of trap sites: dislocation cores, grain boundaries and precipitates.
Moreover, the model provides a useful interpretation of the global trapping ability for a
given material. The approach enables one to correlate the ratio of trapped/free hydro-
gen in a system with distinctive trap sites, which vary in binding energy and number
density. The model’s ability to simulate microstructural evolution, including the evolu-
tion of precipitate number densities and sizes with the simultaneous calculation of e.g.
dislocations pile-up during plastic deformation, is unique in the literature, and provides
a tool for predicting the mechanistic behaviour of a given steel in hydrogen-containing
environments, prevalent in both academic and industrial applications.
In the presented modelling examples (Case Studies 1 through 5) we depict five differ-
ent scenarios, where the trapping capacity of different microstructures is calculated.
The simulated results are compared and validated by experiment in all five cases. Ac-
cordingly, we conclude that the trapping energies used are sufficient to describe the
interactions of hydrogen with carbides, grain boundaries and dislocations in a qualita-
113
tive and quantitative way. The analyses clearly show that in the presence of trapping
sites, the partial functions of the interstitial solute can exhibit large effects, especially
at low temperatures, which are due to both kinetic (lower diffusivity at lower tempera-
tures) and thermodynamic (temperature dependency of Gibbs free energy) effects. The
distribution functions, assuming local equilibrium, describe the partitioning of solute
hydrogen atoms between trapping sites and normal sites.
The case studies include annealed microstructures with bcc ferritic matrices with large
grain sizes and low dislocation densities, as well as quenched and tempered microstruc-
tures with body-centred tetragonal (bct) martensitic matrices with relatively small
grains and high dislocation densities. Austenitic face-centred cubic systems, pure iron
systems and chemically more complex steels, with the influence of carbides, complete
the extensive range of application possibilities through the model’s ability to predict the
redistribution of hydrogen under the presence of energetic traps.
It should be noted however, that the mechanisms proposed, demonstrated and vali-
dated here may not explain all experimental observations reported in the literature.
Uncertainties regarding precise trapping energies and diffusion coefficients are common.
Modelling transients in the evolution of microstructures and the corresponding hydro-
gen trapping transients, although only briefly covered in this chapter for grain boundary
and dislocation evolution, are key in understanding time dependent evolutions in the
presence of multiple competing trapping species. It is intended, in future work, to evalu-
ate, experimentally and through modelling, the evolution of hydrogen trapping for such
transients. For example, during the nucleation and growth of NbC and cementite in
steel to better understand the behaviour of hydrogen during tempering.
Given that predictions are not given about the main driver of the embrittlement process
or mechanisms identified leading to crack formation, and knowing that the free hydrogen
content is not always the decisive quantity that determines the failure of a specific steel,
the presented approach does not provide predictions about failure. However, the mech-
anisms and observations based on the trapping of hydrogen, as identified and quantified
in this work, are consistent throughout the microstructures evaluated, providing coher-
ent explanations for the experimental and calculated findings in the literature. In this
respect, the model needs to be optimised to correlate the amount of hydrogen trapping
and crack nucleation. The present approach considers the interaction between multiple
trapping sites and atomic hydrogen, and is able to predict the redistribution behaviour
for typical microstructures. The presented analysis could be used as part of an efficient
prediction procedure, such as for the design of novel hydrogen embrittlement resistant
steels and the development of new processing routes for steels. Although this chapter
focuses solely on steels, the presented approaches could be applied to other materials
where hydrogen embrittlement is of concern, such as nickel-based [270–272] and alu-
minium alloys [273–276]. For such applications, the necessary diffusion thermodynamic
114
and diffusion databases are available through the MatCalc licence. The binding energies
and microstructural information necessary can be found within literature surrounding
hydrogen permeation and desorption experiments on the respective alloys.
115
Chapter 6
The role of cementite in hydrogen
embrittlement
6.1 Introduction
As discussed in the literature review, the most accepted mechanisms of hydrogen embrit-
tlement are that of hydrogen enhanced localized plasticity (HELP), hydrogen induced
decohesion (HID) and hydrogen enhanced strain induced vacancy formation (HESIV).
In bearing steels, HELP has been hypothesised as the principal hydrogen embrittle-
ment mechanism due to the controlled formation of cracks in hydrogenated steel [108].
However, mechanical degradation of cementite under rolling contact fatigue in bearing
steels happens much sooner than cracks initiate, suggesting cementite degradation plays
a formative role in hydrogen enhanced fatigue failure.
It is well understood that heavy deformation during wire drawing can cause partial
dissolution of cementite. Such cementite dissolution is a product of destabilisation due
to an increased free energy arising from thinning of cementite lamellae and the creation
of slip steps during wire drawing [277]. A similar decomposition of cementite spheroids
was noticed during rolling contact fatigue tests of 100Cr6 bearing steel [108, 278], with
several studies stating that, within the damage region at which cracks initiate and
grow, resulting in fatigue failure, cementite is observed to dissolve, and such dissolution
is seen to be accelerated in the presence of hydrogen [108]. As such, in this work, it is
hypothesised that hydrogen reduces the stability of cementite under plastic strain, both
in terms of its free energy and Peierls stress, enhancing its dissolution during fatigue
and increasing failure rates as a result. This could provide further evidence for the
embrittlement mechanism of hydrogen enhanced localised plasticity.
The following work is part of a collaboration with Dr David Bombac of Cambridge
University (assisting in the density functional theory calculations with the Author),
116
Prof. Afrooz Barnoush and Tarlan Hajilou of NTNU (leading the nanopillar compression
testing with the Author) and Prof. Bill (W.J.) Clegg, Rob Thompson (continuum
modelling) and Tom Edwards of Cambridge University (characterisation and FIB milling
of nanopillars with the Author). In this chapter, theoretical calculations based on
density functional theory are employed to determine the elastic properties of cementite
in the presence and absence of hydrogen. The results are compared to data found in the
literature and experimentally confirmed using resonant ultrasound spectroscopy (RUS)
and differential scanning calorimetry (DSC). Figure 6.1 shows the described programme
of work as a flow chart.
117
Figure 6.1: Flow diagram for the programme of work on the role of cementite in the hydrogen
embrittlement of steels, indicating the findings in red.
118
6.2 Computational model and methods
The crystal structure of cementite shown in Fig. 6.2 is orthorhombic with a space group
Pnma and contains four formula units of Fe3C with 12 Fe atoms and 4 C atoms per
unit cell.
Figure 6.2: The crystal structure of orthohombic cementite (Fe12C4).
Table 6.1 provides the lattice coordinates for the primitive cementite unit cell shown
in Figure 6.2. The unit cell is scaled by the lattice constants (a0, b0, c0). The listed
basis vectors determine the lattice positions of each of the atoms, as they depend on
the parameters xi, yi, zi, numbers on [0,1]. The unit cell contains eight iron atoms
in ”general” positions, four in ”special” positions and four carbon atoms in the iron
interstices. The general and special position iron atoms are 14-coordinate. The general
position iron atoms have 11 Fe-Fe and 3 Fe-C bonds. The special position iron atoms
have 12 Fe-Fe and 2 Fe-C bonds. The carbon atoms are 8-coordinate and are contained
within six iron atoms in a triangular prism structure with two other iron atoms slightly
beyond this.
Atomistic calculations were performed using the projector augmented wave (PAW) and
plane wave basis implemented in the Vienna ab initio simulations package (VASP) [279,
280]. The exchange-correlation functional was described within the generalised gradient
approximation (GGA) as parametrised by Perdew, Burke and Ernzerhof (PBE) [281].
Spin-polarised calculations were performed to model the ferromagnetic influence. The
k-point meshes for the Brillouin zone sampling were constructed using the Monkhorst-
Pack scheme [282]. The Brillouin zone integrations were performed using a Methfessel-
Paxton [283] method with a smearing width of 0.1 eV for structural relaxations. After
119
Table 6.1: The lattice coordinates for the primitive unit cell of cementite shown in Figure 6.2,
where x1 = 0.89, z1 = 0.45, x2 = 0.036, z2 = 0.852, x3 = 0.186, y3 = 0.063 and z3 = 0.328.
Atom Basis vectors (a0x, b0y, c0z)
C (+x1,+
1
4 ,+z1), (−x1,+34 ,−z1), (12 − x1,+34 , 12 + z1), (12 + x1,+14 , 12 − z1)
Feg (+x3,+y3,+z3), (−x3,−y3,−z3), (12 + x3, 12 − y3, 12 − z3),
(12 − x3, 12 + y3, 12 + z3), (−x3, 12 + y3,−z3), (+x3, 12 − y3,+z3),
(12 − x3,−y3, 12 + z3), (12 + x3,+y3, 12 − z3)
Fes (+x2,+
1
4 ,+z2), (−x2,+34 ,−z2), (12 − x2,+34 , 12 + z2), (12 + x2,+14 , 12 − z2)
Feg and Fes are iron atoms in the general and special positions, respectively.
the relaxation was completed, a static calculation was conducted to obtain accurate total
energies and stresses using the linear tetrahedron method with Blo¨chl corrections [284].
Following convergence tests for the 16 atom unit cell of cementite (cf. Fig. 6.2), a plane-
wave cut-off energy of 800 eV and 10× 8× 11 k-point mesh were found to be sufficient
to converge the total energy to better than 10−7 meV/atom. The calculated converged
plane-wave cut-off energy and k-point mesh were used in all proceeding calculations.
To calculate the formation enthalpies of Fe3C and other structures at nought Kelvin,
structure optimisations and total energy calculations were performed for carbon in the
graphite and diamond crystal structures, the ferromagnetic phase of α-Fe, and the
hydrogen molecule H2. The formation enthalpies of cementite and all crystal structures
with hydrogen were calculated using following equations:
∆Hf = Etot(Fe12CxH)− 12Etot(Febcc)− xEtot(Cgraphite)− 1
2
Etot(H2) (6.1)
and
∆Hr = Etot(Fe12CxH)− 12Etot(Febcc)− xEtot(Cdiamond)− 1
2
Etot(H2) (6.2)
where ∆Hr, with the total energy of diamond, is used only as a reference. For hydrogen,
the reference was assumed to be a H2 molecule. The total energy was calculated from a
fully relaxed supercell of 2 hydrogen atoms in a box 10 A˚in length in all directions. The
calculated bond length of the H2 molecule was 0.748 A˚, with a vibrational frequency of
4335 cm−1, as in agreement with previous studies [285,286].
6.3 Results and discussion
Table 6.2 shows the lattice constants of cementite at the ground state. The structural
minimisation was performed using a conjugate-gradient algorithm to relax the ions into
their instantaneous ground state. The limit for relaxation of forces was set to 1 ×
10−6 eVA˚−1. Results in Table 6.2 were compared to other theoretical [287–292] and
120
experimental [293–298] values found in the literature and are in good agreement to
previous studies.
Table 6.2: Comparison of calculated Fe3C lattice parameters in A˚.
Current Theory Experiment
5.038, 6.717, 4.484
5.038, 6.727, 4.484 aRef. [287] 5.082, 6.733, 4.512 cRef. [293]
4.997, 6.702, 4.444 aRef. [288] 5.088, 6.748, 4.522 dRef. [294]
5.035, 6.716, 4.480 aRef. [289] 5.088, 6.742, 4.525 dRef. [295]
5.037, 6.720, 4.482 aRef. [290] 5.040, 6.730, 4.480 dRef. [296]
5.058, 6.703, 4.506 aRef. [291] 5.090, 6.765, 4.258 cRef. [297]
5.068, 6.714, 4.513 bRef. [292] 5.091, 6.743, 4.526 dRef. [298]
a PAW/PBE
b FP-LAPW/PW91
c Neutron Powder Diffraction
d X-ray Powder Diffraction (XRD)
The relaxed structure was then populated with hydrogen atoms at different positions
and allowed to relax into its new ground state. Gamma point calculations of ground
state structures were then performed to obtain total energies and formation enthalpies
using equations (6.1) and (6.2), calculated using both graphite and diamond forms of
carbon. The results for cementite and the structures with hydrogen are given in Table
6.3. Structure A represents cementite with a H atom in the centre of the supercell.
Structure B represents cementite with a H atom in a carbon vacancy at (+x1,+
1
4 ,+z1).
Structure C represents cementite with two H atoms in adjacent carbon vacancies, one
at (+x1,+
1
4 ,+z1) and (
1
2 − x1,+34 , 12 + z1). Cementite is orthohombic. Structure A is
triclinic. Structures B and C are both monoclinic. This is due to the smaller H atom
pulling other Fe atoms closer, breaking the symmetry in the fully relaxed condition.
Table 6.3: Enthalpy of formation (∆H) for different structures calculated with graphite
(∆HC) and diamond (∆HD) forms of carbon.
Structure ∆HC / eV atom
−1 ∆HD / eV atom−1
Fe3C -0.004 -0.157
A -0.005 -0.149
B 0.028 -0.088
C 0.04 -0.037
Calculating the elastic constants is critical in order to understand the structural stabil-
ity and mechanical properties of cementite, and to confirm the hypothesis that H has
a negative effect on the shearability of the cementite structure. The elastic constants
are calculated using the stress-strain relationship [299] as implemented in VASP. The
elastic tensor is determined through six finite distortions of the lattice and deriving elas-
121
tic constants from the stress-strain relationship. The final elastic constants include the
contributions for distortions with rigid ions and the contributions from the ionic relax-
ations. The following are the necessary and sufficient Born criteria for an orthorhombic
system [300]:
cii > 0, c11c22 > c
2
12,
c11c22c33 + 2c12c13c23 − c11c223 − c22c213 − c33c212 > 0,
c44 > 0, c55 > 0c66 > 0.
(6.3)
The elastic constants for all structures are given in Table 6.4, and full elastic tensors
in Tables 6.4 through 6.7. The elastic constants c11, c22 and c33 are related to the
deformation behaviour and atomic bonding characteristics. For all structures except C,
it can be seen that c11 > c22 > c33, indicating that atomic bonds along the (010) plane
between the nearest neighbours are stronger than the atomic bonds along the (100)
plane and (001) plane.
Table 6.4: Calculated elastic constants (cij) in GPa and elastic compliance constants (sij) in
GPa−1 for all structures considered
Fe3C A B C
ij cij sij cij sij cij sij cij sij
11 387.2 0.0035 373.7 0.0035 341.3 0.0038 331.4 0.0041
22 342.7 0.0041 343.6 0.0046 306.1 0.005 222.8 0.0072
33 308.5 0.0047 286.6 0.0053 288.6 0.0053 279.5 0.0055
12 153 -0.001 155.3 -0.0011 134.2 -0.0011 127.2 -0.0017
13 155.4 -0.0013 128.4 -0.001 127.6 -0.0011 130.5 -0.0011
23 158 -0.0016 173.8 -0.0023 162.6 -0.00235 140.1 -0.0029
44 133.1 0.0075 128.9 0.0079 64.6 0.0155 95.3 0.0105
55 16.8 0.06 32.3 0.032 27.7 0.036 61.4 0.016
66 132 0.0076 130.1 0.0077 103.5 0.0098 44.3 0.023
The full elastic tensor Cij in GPa is for Fe3C given in equation (6.4) and for structures
with hydrogen in equation (6.5) for structure A, in equation (6.6) for structure B and
in equation (6.7) for structure C:
CFe3Cij =

387.2 153 155.4 0 0 0
153 342.7 158 0 0 0
155.4 158 308.5 0 0 0
0 0 0 133.1 0 0
0 0 0 0 16.8 0
0 0 0 0 0 132

(6.4)
122
CAij =

373.7 155.3 128.4 −1.7 3.4 6.3
155.3 343.6 173.8 3.3 −7.3 −0.8
128.4 173.8 286.6 2.6 −9.9 −1.7
−1.7 3.3 2.6 128.9 −8 0.2
3.4 −7.3 −9.9 −8 32.3 −1
6.3 −0.8 1.7 0.2 −1 130.1

(6.5)
CBij =

341.3 134.2 127.6 0 0 4.4
134.2 306.1 162.6 0 0 3.4
127.6 162.6 288.6 0 0 18.9
0 0 0 64.6 0.8 0
0 0 0 0.8 27.7 0
4.4 3.4 18.9 0 0 103.5

(6.6)
CCij =

331.4 127.2 130.5 0 0 −6.8
127.2 222.8 140.1 0 0 1.3
130.5 140.1 279.5 0 0 −0.3
0 0 0 95.3 3.8 0
0 0 0 3.8 61.4 0
−6.8 1.3 −0.3 0 0 44.3

(6.7)
Polycrystalline moduli can be calculated using the Voigt-Reuss-Hill approximation [301]
from Voigt [302] and Reuss [303] averages. The Voigt average represents the upper bound
for polycrystalline material and can be determined directly from the elastic tensor Cij .
Bulk (BV) and shear (GV) moduli in Voigt notation are calculated using equations (6.8)
and (6.9), respectively.
BV =
1
9
((C11 + C22 + C33) + 2(C12 + C13 + C23)) (6.8)
GV =
1
15
((C11 + C22 + C33)− (C12 + C13 + C23) + 3(C44 + C55 + C66)) (6.9)
The Reuss average represents the lower bound for polycrystalline material and is cal-
culated from the compliance tensor sij = C
−1
ij . Bulk (BR) and shear (GR) moduli in
Reuss notation are calculated using equations (6.10) and (6.11), respectively.
1
BR
= (s11 + s22 + s33) + 2(s12 + s23 + s31) (6.10)
15
GV
= 4(s11 + s22 + s33)− 4(s12 + s23 + s31) + 3(s44 + s55 + s66) (6.11)
From the upper and lower bounds, one can obtain the Voigt-Reuss-Hill approximation
123
of bulk, shear and Young’s moduli and the isotropic Poisson ratio as:
BVRH =
BV +BR
2
(6.12)
GVRH =
GV +GR
2
(6.13)
E =
9BVRHGVRH
3BVRH +GVRH
(6.14)
ν =
(3BVRH − 2GVRH)
(6BVRH + 2GVRH)
(6.15)
Using the equations above, the calculated bulk modulus BV RH , shear modulus GV RH ,
Young’s modulus E and Poisson ratio ν for Pnma Fe3Cat nought Kelvin, and structures
containing hydrogen, are summarised in Table 6.5. The results show two different effects
of H on cementite. If hydrogen occupies a carbon vacancy, or even two carbon vacancies,
both bulk and shear moduli decrease by approximately 10 %. On the other hand, if
hydrogen is pushed to the middle of the supercell, the shear modulus increases.
Table 6.5: The isotropic bulk moduli (BV RH in GPa) and shear moduli (GV RH in GPa)
obtained from the single crystal elastic constants from the stress-strain relationship using
Voigt, Reuss and Hill’s approximations. The Young’s modulus (E in GPa) and the Poission’s
ratio (ν) are estimated from Hill’s approximation.
Structure BV BR BVRH GV GR GVRH E ν
Fe3C 219 217.4 218.2 94.5 52.1 73.3 197.7 0.35
A 213 209.3 211.3 94.7 70.35 82.5 219 0.33
B 198.3 197.1 197.1 73.3 58.1 65.7 177.4 0.35
C 181 176.8 178.9 69.3 62.9 66.1 176.6 0.34
The ratio of bulk to shear moduli in polycrystalline material (BV RH/GV RH) can be
utilised to determine whether a material is ductile or brittle using the criteria intro-
duced by Pugh [304]. The bulk modulus BV RH is considered to represent resistance
to the fracture, whilst shear modulus GV RH represents resistance to plastic deforma-
tion. The critical value of BV RH/GV RH below which a material is considered brittle is
approximately 1.75. Our calculations show BV RH/GV RH > 1.75, hence all structures
should behave in a ductile manner.
The Young’s modulus provides a measure of material stiffness. The larger the Young’s
modulus, the stiffer the material. The Poisson’s ratio provides information about the
characteristics of bonding forces. It is very small for brittle materials, and around 0.33
124
for ductile metallic materials. Values for the Poisson’s ratio are in our case between 0.33
and 0.35, indicating ductile behaviour.
The crystal anisotropy of a system will be reflected by its atomic arrangements along
different directions. Chung and Buessem [305] introduced the degree of anisotropy in
compressibility and shear, given by:
AB =
(BV−BR)
(BV+BR)
,
AG =
(GV−GR)
(GV+GR)
(6.16)
In equation (6.16), 0 is associated with elastic isotropy and 1 (the largest possible) with
anisotropy. However, recognising the tensorial nature of the elastic stiffness, instead of
using ratios of individual stiffness coefficients to define anisotropy, Ranganathan and
Ostoja-Starzewshi [306] introduced universal anisotropy, given as:
AU =
5GV
GR
+
BV
BR
− 6 (6.17)
The universal anisotropy index (equation (6.17)) provides the means to quantify the
single crystal elastic anisotropy and can be interpreted as a generalisation of the Zener
anisotropy index. The calculated anisotropy index AB, AG and A
U are listed In Table
6.6. As shown, when hydrogen is present in cementite, it reduces anisotropy in the shear
AG and the universal anisotropy index A
U .
Table 6.6: Calculated anisotropy index AB, AG and A
U .
Structure AB AG A
U
Fe3C 0.0037 0.29 4.1
A 0.01 0.15 1.7
B 0.003 0.12 1.3
C 0.012 0.05 0.53
Crystals are, in practice, subjected to loads in different directions. The directional
dependence of the Young’s modulus and linear compressibility (β) in different directions,
are very useful. Figure 6.3 shows the spatial dependence of the Young’s moduli in 3D
for the structures studied.
The linear compressibility of a crystal describes its relative decrease in length when it
is subjected to a degree of hydrostatic pressure, and in general, it varies with direction.
It is defined for triclinic systems in terms of compliances as [307]:
β = (s11 + s12 + s13)l
2
1 + (s16 + s26 + s36)l1l2 + (s15 + s25 + s35)l3l1+
+ (s12 + s22 + s23)l
2
2 + (s14 + s24 + s34)l2l3+
+ (s13 + s22 + s33)l
2
3
(6.18)
125
(a) (b)
(c) (d)
Figure 6.3: 3D spatial dependence of Young’s moduli for a) cementite, b) structure with
hydrogen in the middle of supercell (A), c) structure with hydrogen in the carbon vacancy
(B) and d) structure with 2 hydrogens in the carbon positions (C).
126
where l1, l2 and l3 are the direction cosines. In Table 6.7, the minimum and maximum
values of linear compressibility for all structures studied are given. Projections in the
planes of elastic mirror symmetry are shown in Figure 6.4. The direction cosines are
l1 = cos θ, l2 = sin θ and l3 = 0 for the ab-plane, where the axis is rotated from [100] to
[010]. In the bc-plane, where the tensile axis is rotated from [010] to [001], the direction
cosines are l1 = 0, l2 = cos θ and l3 = sin θ. For the ac-plane, with the tensile axis
rotated from [001] to [100], the direction cosines are l1 = sin θ, l2 = 0 and l3 = cos θ.
Table 6.7: Minimum and maximum values of linear compressibility β in TPa−1.
Structure βmin βmax
Fe3C 1.2467 1.8381
A 1.1737 2.321
B 1.3458 1.9087
C 1.3496 2.7231
Figures 6.5 and 6.6 show the projections in the planes of elastic mirror symmetry of the
Poisson’s ratio and shear modulus respectively. Both figures show maximum (orange)
and minimum (blue) values of the Poisson’s ratio and shear modulus.
127
(a) (b)
(c)
Figure 6.4: Projections of linear compressibility in the planes of elastic mirror symmetry in
the a) ab-plane with the tensile direction rotated from [100] to [010], b) bc-plane with the
tensile direction rotated from [010] to [001] and c) ac-plane where the tensile direction rotates
from [100] to [001].
128
(a) (b)
(c)
Figure 6.5: Projections of the Poisson’s ratio maximum (orange) and minimum positive (blue)
in the planes of elastic mirror symmetry, showing its extreme anisotropy for the a) ab-plane
with the tensile direction rotated from [100] to [010], b) bc-plane with the tensile direction
rotated from [010] to [001] and c) ac-plane where the tensile direction rotates from [100] to
[001].
129
(a) (b)
(c)
Figure 6.6: Projections of the shear modulus in the planes of elastic mirror symmetry, showing
extreme anisotropy in the maximum (orange) and minimum (blue) for the a) ab-plane with
the tensile direction rotated from [100] to [010], b) bc-plane with the tensile direction rotated
from [010] to [001] and c) ac-plane where the tensile direction rotates from [100] to [001].
130
6.3.1 Differential scanning calorimetry
A commercial 100Cr6 (composition in wt %; Fe, 0.97 C, 0.28 Si, 0.28 Mn, 1.38 Cr, 0.18
Ni, 0.21 Cu, 0.01 P, 0.017 S) bearing steel was used for the proceeding experiments.
Differential scanning calorimetry (DSC) was used to evaluate the effects of hydrogen on
the phase transformations of 100Cr6. A TA intruments Q600 was used for simultaneous
DSC and thermogravimetric analysis (TGA) with argon carrier gas and a heating rate
of 20 K/min up to 950◦C. A calibration was run prior to each test. Figure 6.7 shows
the results for 100Cr6 in the following states: hydrogen-free spheroidised (with the
maximum phase fraction of cementite), hydrogen-free quenched and tempered (with
temper cementite), and hydrogen-charged quenched and tempered (two samples). 2
mg small cylindrical samples of 3 mm diameter were used for all tests. Quenched
and tempered samples were charged for 48 hours at 10 mA·cm−2 using an electrolytic
solution of 3 wt% NaCl and 0.3 wt% NH4SCN in 2 dm
3 distilled water. After charging,
samples were lightly polished and cleaned in ethanol, totalling 1 minute, and submerged
in liquid nitrogen until testing.
0 
0.1 
0.2 
0.3 
0.4 
0.5 
0.6 
0.7 
0.8 
0.9 
1 
0 
0.2 
0.4 
0.6 
0.8 
1 
1.2 
0 100 200 300 400 500 600 700 800 900 
Ph
as
e 
fr
ac
tio
n 
/-
 
H
ea
t fl
ow
 / 
W
 g
−1
 
Temperature / °C 
tempered 
spheroidised 
tempered, H charged 1 
tempered, H charged 2 
ferrite (eq.) 
austenite (eq.) 
cementite (eq.) 
M7C3 (eq.) 
Figure 6.7: Differential Scanning Calorimetry results for spheroidised, tempered martensitic
and charged tempered martensitic 100Cr6, using Argon carrier gas and a heating rate of 20
K/min.
Figure 6.7 also presents the equilibrium phase fractions as a function of temperature for
100Cr6 to better indicate the transformations responsible for each peak in heat flow.
The most prominent peak is that of the ferrite to austenite transformation at around
131
760◦C, an exothermic reaction resulting in increased heat flow from the sample. Factors
that will affect the enthalpy of this transformation are: the amount of retained austen-
ite (more austenite means less martensite/ferrite to transform, reducing the enthalpy),
the annealing of dislocations (more dislocations annealed means more released energy,
increasing enthalpy), dissolution of any carbides (an endothermic reaction, reducing the
enthalpy). As shown, the heat flow is consistent amongst samples at temperatures be-
low around 400◦C. Above this temperature, small M7C3 particles begin to dissolve, an
endothermic reaction, thus reducing the heat flow. As evidenced by the lesser change in
heat flow observed for the charged samples at this stage, it can be deduced that hydro-
gen inhibits the dissolution of the carbides. The following peak at around 760◦C, the
ferrite to austenite transition, is greatly reduced in the spheroidised sample due to the
absence of retained austenite, reduced dislocation density, the dissolution of M7C3 car-
bides and the lack of cementite growth (with both phases having around the maximum
equilibrium phase fraction due to the nature of spheroidisation treatments, discussed in
part II, chapter 1 of this thesis), in comparison to the quenched and tempered sample.
However, the presence of hydrogen in the quenched and tempered sample appears to
reduce the enthalpy of the reactions occuring from 600◦C, implying it makes the austen-
ite transformation easier (all alloys were observed to be fully austenitised during this
stage). However, there are two postulated mechanisms behind the observed effect: a
lower content of retained austenite due to storage in liquid nitrogen (reducing the energy
released during transformation), and most notably, hydrogen inhibiting the growth of
cementite and causing its dissolution at lower temperatures, thus reducing the enthalpy
in this region. Given the evidence shown through DFT that hydrogen destabilises the ce-
mentite phase, the inhibition of cementite growth and its lower temperature dissolution
observed via DSC would validate this.
Of most relevance to this work is that of the enthalpy and corresponding temperature for
the dissolution of cementite. Although the equilibrium phase fraction gives an indication
of where dissolution of secondary phases may initiate, in reality, due to kinetic effects,
dissolution will only be significant at higher temperatures in hydrogen-free 100Cr6.
From experience, as evidenced in the proceeding chapter, rapid cementite dissolution in
hydrogen-free 100Cr6 variants is observed above 860◦C in a matter of minutes. As such,
the trough proceeding austenitisation is that of cementite dissolution, an endothermic
reaction. The spheroidised sample shows a lower enthalpy for this reaction due to the
larger phase fraction of cementite, necessitating more energy for full dissolution. As
such, the trough is broadened to higher temperatures, and the minimum heat flow is
lowered. The presence of hydrogen has a similar effect. By destabilising the cementite,
its dissolution is enhanced, reducing the energy necessary for its dissolution. As such,
the cementite may well be dissolved once the steel reaches full austenitisation.
132
6.3.2 Resonance ultrasound spectroscopy
A commercial 100Cr6 bearing steel in its spheroidised condition was used for all pro-
ceeding experiments. Rectangular parallelepipeds with dimensions of 3 × 4 × 5 mm3
were machined by wire electrical discharge machining (EDM) and polished to remove
the recast surface layers. Hydrogen charging was performed in a gas chamber with
9 MPa H2 gas at 37
◦C for 168 h. The charged condition was found to contain 0.33
wppm, determined by thermal desorption analysis up to a maximum of 285◦C (TDA).
After charging, the samples were stored in liquid N2 before RUS measurements were
performed. The uncharged samples were also kept in liquid nitrogen for consistency.
The hydrogen content is significantly lower than desired - that observed in bearings
at end-of-life (often greater than 3 wppm). Low pressure gas charging was chosen as
electrochemical charging could accumulate oxide and samples would thus require un-
precise polishing, reducing parrallelism, preventing accurate results. As such, it was
expected that any changes measured in the elastic tensor would be relatively limited in
comparison to that predicted by the higher hydrogen contents in cementite simulated
using DFT.
The methodology of RUS has been described elsewhere by Migliori and Sarrao [308].
RUS measurements were performed at room temperature with the sample held directly
between driving and receiving PZT-5A piezoelectric transducers, coated with gold to
reduce radio interference in the output signal. The transducers are directly connected
to the signal generator and detector electronics. DRS M3odulus II electronics were used
to generate the signal and process the output spectrum. Spectra were acquired in the
frequency range of 0.1 − 1 MHz, with 50000 data points in each spectrum. Similarly,
spectra were acquired for charged samples. For the charged samples, after removal from
the liquid N2, they were allowed to equilibrate to room temperature, spectra were then
acquired several times in the first 60 minutes, followed by further acquisitions 24 hours
after charging.
The RUS spectra were analysed using the software package Wavemetrics IGOR PRO.
Each resonance peak was fitted with an asymmetric Lorentzian function to accurately
determine the peak frequency, f, and the full width at half of the maximum height, ∆f .
In RUS, f2 scales with the elastic constants of the material, with the main contribution
coming from the shear modulus in the case of a polycrystalline sample. In order to
determine average bulk and shear moduli, the frequencies of up to 58 resonance peaks
are obtained from each spectrum. For each resonance, the inverse mechanical quality
factor, Q−1, which may be taken as Q−1 = ∆f/f , provides a measure of the acoustic
dissipation.
The rectangular parallelepiped resonance (RPR) code was used to convert the exper-
imental resonant frequencies into elastic moduli. It operates by estimating the elastic
133
constants for the sample and using these to calculate the expected resonant frequencies,
which are then compared to those obtained through experiment. The elastic constant
estimates are then modified iteratively until the calculated and experimental frequencies
match.
Figure 6.8: RUS results showing the shift in the resonance peak representative of the shear
modulus for spheroidised 100Cr6 a) 24 hours after H charging, b) 1 hour after charging, and
c) without charging.
Figure 6.8 shows the change in one resonance peak that depends solely on shear modulus.
As shown, in the presence of hydrogen, the peak shifts to lower frequencies, resulting
in a reduction in shear modulus. Little change is seen in the resonance peak for the
samples tested immediately after being removed from the liquid nitrogen and 24 hours
after being removed. This indicates that hydrogen is trapped in the cementite at room
temperature and that hydrogen within the bulk, if any, during testing has had very little
effect on this peak, with only a tiny increase in shear modulus seen for the immediately
tested sample. As shown, 10 runs were conducted for the immediate case, and 3 for
the 24 hours case. The variability in results is evidently very small, with variability
a result of aligning the sample. Figure 6.9 presents the results in terms of shear and
bulk moduli. The error in the quantification of shear and bulk modulus is 0.2% and
2% respectively. The rms error for the fits is 0.08%, implying that the sample is truly
134
isotropic. The density at room temperature, based on dimensions and mass, is 7.814
g·cm−3.
Figure 6.9: The shift in a spheroidised 100Cr6 sample’s shear and bulk moduli 24 hours after,
1 hour after, and without hydrogen charging.
As shown, a small reduction in the shear and bulk moduli is observed in the presence
of hydrogen. Again, although this shift is small in comparison to that calculated using
DFT, this is expected as the lower hydrogen content from charging prevents the ce-
mentite from being saturated it is expected to be significantly smaller, as the cementite
will not be saturated. As such, work is currently underway on in situ electrochemi-
cally charged nanopillar compression tests on cementite particles in steel and within the
grains of a bulk cementite sample in order to quantify the shear and bulk moduli, and
the Peierls stress, for higher hydrogen contents.
6.3.3 In situ hydrogen charged nanopillar compression testing
A 12 mm diameter rod was machined from the same industrial cast of 100Cr6 used
for the RUS and DSC work described previously. This rod was sealed in a glass tube
backfilled with argon to prevent oxidation and decarburisation during the proceeding
heat treatment shown in Figure 6.10, to precipitate large cementite particles on the
grain boundaries for the focused ion beam (FIB) milling of nanopillars in preparation
for compression testing.
Samples were ground and polished to a colloidal silica finish at the top surface. After
polishing, back-scattered electron (BSE) imaging was used to identify cementite, and
135
10 days
Figure 6.10: Heat treatment of 100Cr6 for the precipitation of large grain boundary globular
cementite
EBSD used to quantify the phase fractions and crystallographic orientations of the par-
ticles. Nanopillars of 350 nm diameter by 1.37 µm were focused ion beam (FIB) milled
using a dual beam FIB/SEM (Helios NanoLab, FEI, USA) with Ga+ ions. As shown,
in Figure 6.11, the heat treatment produced large globular cementite particles across
the grain boundaries of around 5 microns in diameter. EDX was performed to char-
acterise the carbides, indicating chromium-rich cementite. The matrix was martensitic
with carbon enriched regions surrounding the carbides. 10 µm diameter troughs were
milled, within which the nanopillars are centred, as shown in Figure 6.12. 31 pillars
were FIB milled in total, each with varying orientation, and thus different active slip
systems, permitting the quantification of Peierls stress for cementite.
In addition to the 100Cr6 sample, a sample of bulk cementite was also FIB milled to
produce identical pillars for further testing. The bulk polycrystalline cementite was
produced at the Department of Production Systems Engineering, Toyohashi University
of Technology, using a mechanical alloying and spark plasma sintering technique, dis-
cussed in more detail elsewhere [309]. As shown in Figure 6.13, the grain size of the bulk
cementite sample was 10 microns and consisted of 8% ferrite and 92% Fe3C. However,
the measured porosity was 4%, thus the real fraction of ferrite will be lower. XRD
analysis measured only 3% ferrite. The density was found to be 7.386 gcm−3, based on
dimensions and mass. RUS found a bulk and shear modulus at room temperature of 171
±2 GPa and 73.6 ±0.1 GPa respectively. The theoretical values from DFT at 0 Kelvin
for bulk and shear moduli were 198.6 (Voigt) and 86.5 GPa, without Hill averaging,
respectively.
The described pillars require indentation, both in bulk cementite and 100Cr6, in order
136
Figure 6.11: SEM micrographs of 100Cr6 showing a) a back scattered diffraction image of
large globular carbides, for which b) is an EDX map of Fe, c) C, d) Cr, e) an EBSD phase
map with cementite (purple) and martensite (green), and f) a FIB milled nanopillar within a
carbide prior to compression testing.
1 μm
350 nm diameter x 1.37 μm length (2:3 ratio is ideal)
(a) (b)
Figure 6.12: SEM micrographs of 100Cr6 showing a) nanopillars milled within large globular
carbides and b) a higher magnification image of one nanopillar prior to compression testing.
to validate the effects of hydrogen on the mechanical properties of cementite using an
environmental cell nano-indentation atom-force microscopy system with a combined
displacement controlled flat-punch indentor. Within the environmental cell, electrolytic
charging will be conducted during the compression tests. The experimental set-up is
described in detail elsewhere [81].
137
Figure 6.13: EBSD orientation and phase maps of bulk cementite, indicating a) the polycys-
talline isotropic structure, and b) The phase distribution, indicating degradation to ferrite
(blue) around the grain boundaries of cementite (red) with pores (black).
6.4 Summary
Ab initio simulations have been conducted to evaluate the effects of hydrogen on the
stability and mechanical properties of cementite to establish the mechanisms of hydrogen
embrittlement. The results show that in the presence of hydrogen, cementite’s stability
is notably reduced and its shearability increased, resulting in the enhanced dissolution of
cementite under fatigue, as evidenced in hydrogen charged RCF tests. The shearing of a
cementite particle destabilises the particle. As hydrogen enters the carbide, in addition
to a reduction in its overall stability, such shearing becomes more favourable, further
increasing the likelihood of dissolution. Regardless of whether cementite particles are
dissolved, the enhanced plasticity in the presence of hydrogen results in a reduction in
strength, and a consequent reduction in the fatigue life of the steel.
RUS and DSC results on hydrogen charged cementite containing 100Cr6 bearing steel
validate the findings with respect to hydrogen’s effects on shearability and overall stabil-
ity respectively. However, given the limited cementite phase fraction in the steels tested,
the effects of hydrogen on the shearability of cementite cannot be accurately quantified
by RUS. Similarly, DSC cannot accurately quantify the reduced stability of cementite
in the presence of hydrogen due to hydrogen egress from the sample during testing.
However, the preliminary results described suggest the hypothesised effects could be
correct. The nanopillar compression testing is necessary to confirm this. One method
by which the changes in mechanical properties and stability could be more accurately
quantified, is through the use of bulk cementite for RUS testing. Additionally, conduct-
ing DSC with hydrogen as a carrier gas could reduce the effects of hydrogen egress from
the sample during testing. Work is ongoing to conduct RUS testing on bulk cementite,
as is work on in situ electrochemically charged nanopillar compression testing for both
bulk cementite and cementite containing 100Cr6.
138
Chapter 7
The optimisation of vanadium
carbides for hydrogen trapping in
martensitic steels
As discussed, one proposed method by which hydrogen embrittlement resistance could
be obtained is through the use of hydrogen traps. One such trapping species is that of
vanadium carbide. However, little is known about the optimum morphology, distribution
or size of such carbides for hydrogen embrittlement resistance. In this chapter, vana-
dium carbide evolution is evaluated using synchrotron X-ray diffraction, combining both
in situ and ex situ heat treated samples of a martensitic and austenitic 100Cr6+0.5V.
Martensitic samples with varying vanadium carbide sizes were hydrogen charged and
analysed using transmission electron microscopy and thermal desorption analysis to es-
tablish the optimum size and distribution for hydrogen trapping with respect to binding
energy and trapping capacity. The carbide morphology at 600 ◦C and 860 ◦C within
martensitic and austenitic matrices respectively are evaluated and the resulting hydro-
gen trapping characteristics established. Finally, the contribution of temper carbides is
assessed.
7.1 Introduction
As discussed in section 3.2.1, Szost [25] showed that nanoprecipitate hardened 100Cr6
martensitic bearing steel can have its hydrogen trapping behaviour fully characterised
using a combination of both thermal desorption analysis and melt extraction technique.
Thermodynamic and kinetic modelling was combined to develop a novel hydrogen em-
brittlement resistant bearing steel, similar to the products intended from this thesis.
Szost showed that current hydrogen embrittlement resistant steels were inappropriate
139
for application as bearing steels due to inadequate hardness. As such, a V4C3 nanocar-
bide precipitated martensitic steel, more precisely a 100Cr6 steel + 0.5 wt% V, was
designed and tested. The 10 nm diameter V4C3 carbides were found to act as re-
versible hydrogen traps and resulted in improved hardness and strength with increased
hydrogen embrittlement resistance. Given the significant improvement in hydrogen em-
brittlement resistance, hardness and strength, it is clear that the development of such
nanoprecipitate hardened bearing steels is feasible and V4C3 is a proven means of achiev-
ing improved hydrogen embrittlement resistance in such steels. As such, the vanadium
containing 52100 variant, 100Cr6+0.5V, was chosen for investigation in this work.
The optimisation of nanoprecipitates, such as carbides, as hydrogen traps is crucial
for future alloy design. However, the effects of coherency and carbide composition on
the hydrogen trapping capacity and binding energy are little understood. Evidence
shows that the activation energy of such traps is dependent on carbide-matrix interface
coherency [310]. The effects of coherency for TiC-matrix interfaces on the hydrogen
trapping behavior in steel was studied by Lee [310]. Incoherent titanium carbides were
found to have a larger trap activation energy than semi-coherent ones, with activation
energy increasing with particle size. Smaller precipitates are favoured in bearing steels
due to their permissibly large number density, strengthening effects, and the detrimental
effects of large incoherent carbides on the rolling contact fatigue life of bearings [5].
Yamasaki and Bhadeshia [62] studied the effects of particle size and coherency on the
trapping capacity of M4C3 carbides in Fe-C-Mo-V model steels, showing an optimum
trapping capacity was achieved with a carbide size of around 10 nm for vanadium rich
carbides. Mo additions were also studied, and were shown to increase the coherency of
vanadium-rich carbides. The trapping of hydrogen on carbide-matrix interfaces has been
validated using 3D atom probe tomography [61], showing evidence of deuterium trapping
on the matrix-carbide interfaces of around 10 nm titanium carbides. The interaction
of hydrogen with the various interfaces present in typical bearing steels is complex and
largely not understood, be it the lack of conclusive evidence or means of characterising
trapping locations and atomic configurations on these interfaces, or the complications
involving nonmetallic elements and their corresponding bonding effects. Given such
complexity, interfaces will contain a range of trapping sites of varied binding energy,
although general orientation and coherency should be reasonably consistent across the
interface. Although suggested [62], no work has presented a conclusive relationship of
coherency-to-hydrogen trapping capacity, though it is evident that a size-effect exists
and so must be considered in the design of hydrogen embrittlement resistant steels.
140
7.2 Thermodynamic and kinetic modelling
Thermodynamic and kinetic simulations were carried out using MatCalc, a commercial
programme for solid state and kinetic precipitation modelling (version 5.6, with ther-
modynamic database mc fe v2.021 and diffusion database mc fe v2.006). Simulations
were carried out to generate an initial calculation state, replicating the microstruc-
ture of the as-received commercial cast, discussed later, in a spheroidised state. The
replicated microstructure was used as the initial calculation state from which kinetic
simulations would continue for the primary heat treatment schedules. The key fac-
tors of influence on the proceeding microstructural evolution were the carbide fractions
formed from the liquid (primary carbides), calculated using the Scheil-Gulliver method,
the carbide fractions and distribution formed from the solid solution (secondary car-
bides), and the resulting matrix composition, both of which are calculated using kinet-
ics. Figure 7.1a shows the equilibrium phase diagram for 100Cr6 + 0.5V. All samples
were subjected to an initial 10 ◦C·s−1 heat up to a 5 minute hold at 1200 ◦C to dis-
solve all carbides present and ensure the steel is fully austenitised. The samples were
then quenched at -10 ◦Cs−1 to room temperature, as shown in Figure 7.2 and Table 7.1.
Time-temperature-transformation (TTT) information was calculated using the freeware
programme MUCG83 [311]. The resulting TTT diagram for 100Cr6+0.5V is shown in
Figure 7.1b.
For all simulations, vanadium carbides were assumed to nucleate on both dislocations
and grain boundaries. The post-quench (fresh martensite) dislocation density was taken
as 1015 m−2 with an initial grain size of 10 µm.
The quench from 1200 ◦C to room temperature is to form martensite and increase the
dislocation density, creating more nucleation sites for the proceeding precipitation of
vanadium carbides during primary treatments, and to increase the driving force for
the carbide formation, a result, among others, of the supersaturation of carbon. The
martensite start temperature was calculated to be approximately 200 ◦C which is in
agreement with literature [63].
Table 7.1: Hold times used to precipitate different sizes of vanadium carbides (cf. Fig. 7.2)
Size /nm Hold1 / s Hold2 / s
5 1800 180
10 3600 180
15 3600 540
20 6400 700
If vanadium carbides are formed in both austenite and ferrite, it is thought that the
carbides will form preferentially within the ferrite due to the reduced carbon solubility
141
(a)
(b)
100
200
300
400
500
600
700
10 100 1000 10000 100000 1000000
Te
m
p
e
ra
tu
re
 /
 
C
Time / s
Shear transformation
Diffusional transformation
Ms
Figure 7.1: 100Cr6 + 0.5V bearing steel; (a) equilibrium phase diagram and (b) TTT diagram.
142
Figure 7.2: Heat treatment schedule used to precipitate different sizes of vanadium carbides
in 100Cr6 + 0.5V. Details for hold times (Hold1 and Hold2) are given in Table 7.1.
and high dislocation density. However, after quenching to martensite, the high density
of dislocations will provide nucleation sites for vanadium carbide formation. Notably,
the increased carbon content within Cottrell atmospheres surrounding the dislocations
will enhance carbide formation. As such, it is feasible that vanadium carbides could
form in both phases but a higher number can be nucleated in martensite than austenite.
As such, the heat treatments are conducted to grow vanadium carbides in austenite, but
these carbides are initially nucleated in martensite. Through vanadium additions, one
can increase the supersaturation of vanadium within the phases and engineer an alloy
to form vanadium carbides above or below the austenitisation temperature depending
on the desired final microstructure and/or maximum temperature limitations during
processing. Higher vanadium contents result in higher dissolution temperatures of the
resulting carbides and to a lesser extent, an increased number density of carbides.
It is common industrial practice to temper martensitic bearing steels such as 100Cr6
prior to operation. An example of a typical tempering treatment is that of 215◦C for
2 hours. As such, the effects of tempering on 100Cr6+0.5V will also be studied with
respect to hydrogen trapping behaviour for this treatment.
143
7.3 Experimental procedures
7.3.1 Materials
An industrial hot-rolled cast of 100Cr6+0.5V was produced by Tata Steel. The cast
was electro-discharge machined into cylinders of 8 mm diameter by 12 mm length for
dilatometry and hydrogen desorption analysis. Samples for synchrotron X-ray analysis
were cylinders of 3 mm diameter by 5 mm length. The primary heat treatment of
dilatometry samples was conducted in a thermal simulator Thermecmastor-Z to the
temperature-time schedules shown in Figure 7.1 and Table 7.1.
7.3.2 Transmission electron microscopy
Thin films were prepared for transmission electron microscopy (TEM) electropolished
(Struers Tenupol 5) with a solution of 5 % perchloric acid, 25 % glycerol and 70 %
ethanol at 21.5 V and 8 ◦C. An FEI Tecnai F20 TEM (200 kV) was used for all
TEM-based techniques.
7.3.3 Synchrotron X-ray diffraction
Synchrotron diffraction experiments were performed at Deutsches Elektronen-Synchrotron
(DESY), Hamburg, Germany, using the PETRA P02.1 beam line. The X-rays (wave-
length 0.207150 A˚) were measured by a flat 2D solid-state area detector mounted per-
pendicular to the incident beam and recorded as a TIFF with 32-bit dynamic range.
The data were analysed using the Materials Analysis Using Diffraction (MAUD) pro-
gramme [312, 313]. Data were recorded as 2D images with 32 bit dynamic range and
were integrated over 360◦ using ImageJ plugins within MAUD. It was observed that
the peaks in spectra converted from the silicon standard, which have face-centred cubic
crystal structures, were symmetrical and well-fitted using a face-centred cubic phase
during Rietveld refinement.
7.3.4 Atom probe tomography
Atom probe tomography (APT) specimens were prepared by focus ion beam (FIB, FEI
Helios NanoLab 660). APT analyses were performed using a Local Electrode Atom
Probe (LEAP 4000X HR, Cameca Instruments) in pulsed voltage mode at a pulse
fraction of 15 % and a pulse frequency of 200 kHz. The specimen base temperature was
approximately 60 K.
144
7.3.5 Hydrogen trapping characterisation and thermal desorption anal-
ysis
Dilatometry samples were hydrogen charged at a surface current density of 10 mA cm−2
for 96 hours at room temperature to prevent any microstructural changes. Cathodic
electrolysis was implemented using an electrolytic solution of 70g NaCl and 6g NH4SCN
in 2 dm3 distilled water. The charging cell used to introduce hydrogen consisted of
a large glass beaker of the electrolytic solution, with an anodic platinum wire wound
around its inner surface. A Gamry Interface 1000E galvanostat was used to provide a
stable power source to the cathodic sample electrode. Electrodes for the samples were
made of AISI 316 stainless steel spot-welded onto the samples. The submerged regions
of the electrodes were covered with a Lacomite varnish to ensure that hydrogen did not
diffuse into the electrodes and reduce ingress rates to the sample. A magnetic stirrer
and Ar bubble feed were used to mix the electrolytic solution and prevent bubble growth
on the samples which reduces ingress rates. Before and after charging, the steel samples
were polished to remove any oxide and then cleaned with ethanol. The samples were left
at room temperature for 5 days to allow all diffusible hydrogen to egress. All diffusible
hydrogen was considered lost once the hydrogen peaks reached background level, at
which point all hydrogen within the sample is trapped.
A thermal desorption analysis system and its corresponding operating procedure were
developed within the Phase Transformations Complex Properties Research Group lab-
oratory. The system was developed principally to determine the nature of the prevailing
hydrogen trap active at a given temperature. A Carbolite HST 12/200 tubular furnace
heated up the specimens at a rate of 25◦C hour−1 during analysis, and hydrogen content
was detected by a pulsed discharge detector with helium carrier gas in an Agilent Tech-
nologies 7890A gas chromatograph connected to a Aalborg GFC17 mass flow controller
(cf. Figure 7.3). Hydrogen desorption data were collected every 3 min from which the
hydrogen content was calculated as the area under the curve divided by the heating
rate, and the desorption rate as the evolved hydrogen per minute. High purity helium
was used as the carrier gas and a standard mixture of He and 60.7 ppmw H2 was used
for the system calibration.
In order to allow the comparison of materials with respect to their hydrogen trapping
characteristics, experimental procedure must be kept consistent. The most essential
variables to ensure consistency are that of heating rate and He carrier gas flow rate.
The electrolytic hydrogen charging conditions, including sample size, surface current
density and electrolyte composition must also remain consistent across all comparable
experiments. As such, for all experiments involving the use of TDA presented in this
dissertation, preliminary calibration tests were run to ensure the chosen experimental
parameters were appropriate. Such calibration tests must be conducted on each heat
145
Figure 7.3: Schematic of the TDA system used, where HID is helium ionisation detector and
PID is photoionisation detector.
treatment for a given composition unless microstructural variation is limited such as for
small deviations in carbide morphology. The parameters utilised for each experiment
are stated in all proceeding work.
7.4 Results and discussion
Fig. 7.4 presents the kinetic modelling results of one heat treatment variant, with the
evolution of phase fractions, the mean radius of precipitates and the number density
of these precipitates during the heat treatment. The heat treatment shown produces
an average vanadium carbide size of 10 nm. Fig. 7.2, in conjunction with Table 7.1,
presents all the heat treatment variants utilised in this work and the intended diameter
of the resulting vanadium carbides for these treatments. The kinetic modelling results
for all heat treatment schedules are summarised in Table 7.2.
The results of the kinetic modelling indicate that the mean number density of vanadium
carbides for the various heat treatments is approximately 1.8× 1021 m−3 for the larger
predicted carbide sizes and 1×1023 m−3 for 5 nm vanadium carbides. Vanadium carbides
are homogeneously distributed throughout a martensitic matrix. The size distribution
histograms for vanadium carbide and cementite are shown in Figure 7.5 and Figure 7.6
respectively. It can be seen that vanadium carbide sizes do not deviate too much from
the mean size specified in Table 7.2. In general, cementite particles are approximately
double in size compared to vanadium carbide and their phase fraction increases with the
146
(a)
(b)
(c)
Figure 7.4: Kinetic simulations for the microstructural evolution of 100Cr6 + 0.5V, expecting
10 nm vanadium carbides, showing the evolution of (a) phase fractions (b) number density of
precipitates and (c) mean radius of precipitates.
147
Table 7.2: Results of kinetic modelling precipitating different sizes of vanadium carbides,
based on selected heat treatment schedules.
Desired Phase percent / % Number density / m−3 Mean diameter / nm
size /nm V4C3 Fe3C V4C3 ×1021 Fe3C ×1018 V4C3 Fe3C
5 0.6 0.7 1.9 0.2 4.8 9.8
10 0.1 0.3 1.8 200 10.8 24.4
15 0.4 2.9 1.8 200 16 24
20 0.7 3 1.8 0.003 23 30
size of vanadium carbides. This is expected, since to obtain larger vanadium carbides,
the heat treatment step at 600 ◦C must be longer, which also allows cementite to grow.
With the subsequent temperature spike, the steel is reaustenitised and the cementite
is partially dissolved. Thermokinetic simulations confirm the reduced phase fraction of
cementite with respect to this temperature spike. However, thermokinetic simulations
were optimised to obtain the required size and number density of vanadium carbides
and thus, simulation results for cementite, which are strongly correlated with vanadium
carbide, are expected to be less reliable.
Figures 7.7 through 7.10 show TEM images of 100Cr6 + 0.5V heat treated to the
treatments described in Figure 7.1 and Table 7.1. Figure 7.7a shows a bright field
image of the sample heat treated to grow 5 nm diameter vanadium carbides, showing
martensite plates and the fine homogeneously precipitated V4C3 particles throughout
the microstructure. Figure 7.7b presents a higher magnification image of such carbides.
Compositional evidence of vanadium carbides were obtained using EDX mapping of V
and Fe, as shown in Figs. 7.7c and 7.7d respectively. The EDX maps were taken from
the lower central section of the image in Figure 7.7a. Similar micrographs and EDX
elemental maps are presented for the 10 nm sample, shown in Figure 7.8. Due to the
size of the nano carbides, the authors were unable to obtain selective area diffraction
patterns for these carbides. In this work, vanadium carbide is expected to be of a non-
stoichiometric composition VC0.75 with a rock-salt crystal structure (face centred cubic)
[108, 314] with a high number of carbon vacancies. The lattice parameter a is around
4.13 A˚. The orientational relationship that exists between vanadium carbide and ferrite
can be used to determine whether precipitation occured in ferrite or austenite [315].
Vanadium carbide may be identified as precipitating in ferrite if they are related by the
Baker-Nutting orientation relationship [316]; {100}αFe||{100}V4C3, < 011 > αFe|| <
010 > V4C3. When vanadium carbide is related to ferrite by the Kurdjumow-Sachs
orientation relationship [317]; {110}αFe||{111}V4C3, < 111 > αFe|| < 110 > V4C3, it
has precipitated in austenite.
Figures 7.9 and 7.10 show the characteristics of the carbides present in samples heat
treated to 15 nm and 20 nm, respectively. Similar to the micrographs of the 5 nm and 10
148
(a) (b)
(c) (d)
Figure 7.5: Size distribution histograms of V4C3 in 100Cr6 + 0.5V for (a) 5 nm heat treatment
(b) 10 nm heat treatment (c) 15 nm heat treatment and (d) 20 nm heat treatment.
(a) (b)
(c) (d)
Figure 7.6: Size distribution histograms of cementite in 100Cr6 + 0.5V for (a) 5 nm heat
treatment (b) 10 nm heat treatment (c) 15 nm heat treatment and (d) 20 nm heat treatment.
149
(a) (b)
(c) (d)
Figure 7.7: TEM micrographs of 100Cr6 + 0.5V heat treated to produce 5 nm vanadium
carbides showing (a) distribution of carbides (b) size of vanadium carbides (c) EDX map of
V and (d) EDX map of Fe.
150
(a) (b)
(c) (d)
Figure 7.8: TEM micrographs of 100Cr6 + 0.5V heat treated to produce 10 nm vanadium
carbides showing (a) distribution of carbides (b) size of vanadium carbides (c) EDX map of
V and (d) EDX map of Fe.
151
(a) (b)
Figure 7.9: TEM micrographs of 100Cr6 + 0.5V heat treated to produce 15 nm vanadium
carbides showing (a) distribution of carbides and (b) size of vanadium carbides.
(a) (b)
Figure 7.10: TEM micrographs of 100Cr6 + 0.5V heat treated to produce 20 nm vanadium
carbides showing (a) distribution of carbides and (b) size of vanadium carbides.
152
nm samples, small vanadium carbides and large cementite particles can be seen. Using
the higher magnification micrographs (Figs. 7.7b to 7.10b), the size of the carbides can
be seen to increase, validating the kinetics of the chosen heat treatments. Details of
the mean radii obtained from TEM are presented in Table 7.3. TEM results confirm
the desired size of vanadium carbides is achieved, as predicted by thermokintic simu-
lations. However, TEM only shows a 2D projection, making estimation of the number
density of carbides difficult. From TEM, we confirm that cementite particles are larger
in size compared to vanadium carbides. The cementite sizes obtained are in line with
the predictions from thermokinetics. However, its predicted size distribution was not
confirmed, at least in inspected TEM films. Number densities estimated through com-
paring numbers of vanadium carbide and cementite particles in TEM film projections
should be much smaller for cementite compared to vanadium carbide.
Table 7.3: Results of quantitative analysis of 100Cr6 + 0.5V using TEM.
Desired Number density ×1021 / m−3 Mean diameter / nm
size /nm V4C3 Fe3C V4C3 Fe3C
5 1.3 23.6 6.1± 1.2 21± 0.4
10 3.6 18.2 9.8± 2.8 20± 0.4
15 5.6 2.53 16.4± 3.3 24± 0.4
20 4.8 0 26± 7 36± 0.4
TEM micrographs show only the projection of the film surface. Therefore, to obtain the
correct volume representation of carbides, APT is employed. In Figure 7.11, vanadium
carbides highlighted with isosurfaces are presented along with maps for C and V. Given
the concentrations of C and V from the martensitic matrix to the centre of each particle,
we identified the carbides to be V4C3 (ct. Figure 7.12). Average carbide spacing was
measured to be 45 nm, corresponding to the average number density of approximately
2×1021 m−3. As such, the number density estimation is in line with that predicted from
thermokinetic simulations, although the small volume analysed by APT was used in the
estimation without accounting for other larger particles expected in the microstructure.
The characteristics of the vanadium carbides were confirmed by means of X-ray diffrac-
tion using synchrotron radiation. The spectra obtained for all heat treatments analysed
are given in Figure 7.13. Each spectrum was fitted using a Rietveld refinement, assum-
ing that the phases present were: martensite, austenite, cementite, M7C3 and V4C3.
The polynomial background function, total incident X-ray intensity, austenite, cemen-
tite, M7C3 and V4C3 phase fractions (the residual phase being martensite), microstrain,
crystallite size and applicable lattice parameters were included in the refinement. The
most relevant parameter regarding the refinement is considered to be the weighted profile
153
Figure 7.11: APT elemental maps for C and V along with isosurfaces representing V4C3.
Images were collected with the assistance of Dr Wenwen Song at RWTH, Aachen University.
b)a)
Figure 7.12: Detailed analysis of a carbide embedded in a martensite matrix; a) detail of
analysed carbide and b) change in concentration from matrix to centre of carbide.
154
factor Rwp [318],
Rwp =
[∑
wi(xi − yi)2∑
wix2i
]1/2
(7.1)
where xi is the observed intensity, yi the corresponding calculated intensity at the i-th
data point and wi is the observed weight equal to the inverse of the standard uncertainty
in the observed intensity (σ2[xi])
−1 = x−1i . The summation is taken over all data
points considered in the refinement. Details of the quantitative analysis are presented
in Table 7.4.
Figure 7.13: Indexed X-ray diffraction spectrum for 2θ between 4◦and 12◦.
Table 7.4: Results of quantitative analysis of 100Cr6 + 0.5V using synchrotron X-ray diffrac-
tion.
Desired Phase percent / wt % Rwp
size /nm Martensite Austenite Fe3C M7C3 V4C3 / %
5 73.6 24.9 0.5 0.4 0.6 13.4
10 76.8 19.8 2 0.7 0.7 14.6
15 82.6 14.3 0.9 1.5 0.7 11.8
20 77.6 17.3 3 1.4 0.7 12.2
The success of the treatments to precipitate different sizes of vanadium carbides was
confirmed by subsequent experimental work, spanning from TEM, APT and synchrotron
X-ray diffraction. Using synchrotron X-ray diffraction, the vanadium carbides in 100Cr6
+ 0.5V steel are thought to be V4C3 in type, in agreement with the literature [62, 108,
155
319]. However, Epicier et al. [320] come to the conclusion that precipitates reported
in the literature as V4C3 are actually of a monoclinic V6C5 structure. Their study,
using diffraction analysis in TEM, seems to show the existence of precipitates with this
monoclinic structure. However, the steel used in the study had a much lower carbon
content and did not contain any other alloying element. Oba et al. [321] assigned a
non-NaCl type structure to early stages in the vanadium carbide precipitation process.
Furthermore, during the analysis of the synchrotron data, the monoclinic structure
proposed by Epicier et al. [320] was unable to be fitted successfully using Rietveld.
Takahashi et al. [319] used APT for direct observations of hydrogen trapping sites. Their
conclusion was that deuterium used in APT experiments is most likely trapped by misfit
dislocations. They state that the misfit at the (001) interface between vanadium carbide
and the ferrite matrix is approximately 2.8 % and that a misfit dislocation is released
when the size of the carbide platelet is lower than 8 nm. Recently, Chen et al. [322]
showed deuterium was also inside the vanadium precipitate, suggesting that hydrogen
can be strongly trapped in carbon vacancies. The hydrogen trapping capacity of differ-
ent sized vanadium carbides in 100Cr6 + 0.5V bearing steel was evaluated by means
of TDA. As 100Cr6 + 0.5V consists of martensite, retained austenite, cementite and
V4C3, different grown sizes of vanadium carbides should provide different total hydrogen
trapping capacities. The desorption results for all studied carbide sizes are shown in
Figure 7.14. In order to extract peak temperatures from the TDA data, the following
assumptions were made: the hydrogen detrapping process dominates hydrogen desorp-
tion kinetics, the obtained desorption profile can be decomposed to multiple peaks with
different peak temperatures (different detrapping energies), and the decomposed profile
is approximated by Gaussian peaks to simplify the fitting process and simply acquire
the decomposed hydrogen content. Similar assumptions were used in order to study
vanadium precipitate trapping by Yokota and Shiraga [323]. The results obtained are
discussed in light of their study, although direct comparison is impossible due to dif-
ferent TDA sample size and its resulting effects on peak location. The TDA data are
sufficiently described by three Gaussian peaks. Fitting was performed using the Leven-
berg Marquardt algorithm. Table 7.5 presents the desorption rate peak temperatures
and the normalised area of each peak. Figure 7.14 shows the results of Gaussian func-
tion fitting analysis and the Gaussian peaks for the sample with 10 nm sized vanadium
carbides. Peaks are fitted to minimise the weighted sum of squared residuals, which in
all cases was lower than 0.5 %.
Similar fitted peak temperatures were observed in all samples. The first peak, with a
peak desorption rate temperature between 70 ◦C and 80 ◦C is due to trapping at low
binding energy traps such as dislocations and/or grain boundaries. A peak between
130 ◦C and 140 ◦C was assigned to vanadium carbide. At this peak, we can see that 10
nm carbide size is optimal with respect to trapping the most hydrogen. The third peak
156
Table 7.5: Total amount of trapped hydrogen, peak desorption temperatures and areas of
deconvolved peaks.
Size Total H Temperature / ◦C Trapped H / ppmw (%)
/ nm /ppmw Peak 1 Peak 2 Peak 3 Peak 1 Peak 2 Peak 3
5 1.41 79 133 243 0.18 (12.9) 1.11 (78.7) 0.12 (8.4)
10 2.52 76 140 254 0.30 (12.1) 1.95 (77.3) 0.27 (10.4)
15 2.41 66 131 255 0.25 (10.4) 1.80 (74.5) 0.36 (15.1)
20 2.01 67 132 253 0.27 (13.5) 1.26 (62.8) 0.48 (23.7)
between 245 ◦C and 255 ◦C was assigned to cementite. As can be seen from Table 7.5,
this is a reasonable explanation as the cementite phase fraction is expected to be largest
in the 20 nm sample. A similar explanation is also valid for the first peak, explained
by the lower dislocation density in the 5 nm sample due to the much shorter spike to
austenitisation, consequently lowering the carbon content of the austenite prior to final
quenching. Similar temperatures were reported for hydrogen trpaping on vanadium
carbides elsewhere [323]. However, as previously stated, direct comparison is impossible
due to the different size of TDA samples and the different steel composition.
157
(a)
0 
0.001 
0.002 
0.003 
0.004 
0.005 
0.006 
0.007 
0.008 
0 25 50 75 100 125 150 175 200 225 250 275 300 
H
 d
es
or
pt
io
n 
ra
te
 / 
w
pp
m
 m
in
−1
 
Temperature / °C 
5 nm (1.41 wppm) 
10 nm (2.52 wppm) 
15 nm (2.41 wppm) 
20 nm (2.01 wppm) 
(b)
Figure 7.14: Thermal desorption analysis of 100Cr6 + 0.5V showing a) heat treated to produce
different vanadium carbide sizes and b) details of peak deconvolution and calculated fit for
10 nm experimental data.
158
7.4.1 Effect of tempering
A sample, produced to the heat treatment in Table 7.1 for forming 10 nm vanadium
carbides, was subjected to an additional tempering treatment of 215◦C for 2 hours to
investigate the effects of tempering on the microstructure and its corresponding hydro-
gen trapping capacity. The experimental procedure was identical to the tests described
previously in this chapter, except each sample, both tempered and untempered, was
charged individually at 10 mA cm−2 for 4 days and left at room temperature for 5 days
prior to TDA. Such samples were found to be saturated with hydrogen by the end of
charging. The results of TDA are shown in Figure 7.15.
262
166
149
92
87 265
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0 25 50 75 100 125 150 175 200 225 250 275 300
H
 d
e
so
rp
ti
o
n
 r
at
e
 /
 w
p
p
m
m
in
Temperature / C
Untempered (3.06 wppm)
Tempered (2.37 wppm)
Figure 7.15: TDA data for tempered (215◦C for 2 hours) versus untempered 100Cr6+0.5V,
both subjected to the 10 nm treatment of Table 7.1. Both curves are deconvoluted into
individual trapping sites and their respective Gaussian curves.
Tempering is found to reduce the total trapped hydrogen. There is little or no change
in retained austenite content during tempering [5]. The key microstructural changes of
tempering are the reduction of dislocation density and formation of temper carbides.
The bright field TEM images shown in Figure 7.16 indicates the presence of small
temper carbides, providing additional trapping sites. As shown in Table 3.2, the binding
energy of small coherent cementite, as found after tempering, is around 25 kJ/mol,
similar to that of dislocations and vanadium carbides (around 30 kJ/mol). However,
large incoherent cementite is of 85 kJ/mol. The increase in the peak desorption rate
temperature for the tempered sample is due to the reduced dislocation density (low
temperature traps with a peak at around 92◦C), and the increase in the total trapped
159
hydrogen by vanadium carbides as a result (with a peak at around 166◦C) which were
evidently not saturated. In addition, the high temperature peak at around 262◦C is
thought to be that of temper cementite, providing deeper trapping sites but limited
in capacity. Possibly, a slight increase in the size of vanadium carbides occurs during
tempering and may result in a reduced trapping capacity of the carbides as they move
away from their optimum size. However, no significant growth was observed using TEM.
Regardless, tempering is seen to be of detriment for hydrogen trapping efficiency, with
the reduction in trapping capacity from the loss of dislocations dominating over the new
source of trapping from the temper carbides.
Figure 7.16: Bright field TEM images of tempered (215◦C for 2 hours) and untempered
100Cr6+0.5V, both subjected to the 10 nm treatment of Table 7.1, indicating temper carbides
7.5 Summary
Thermokinetic simulations have been optimised and employed to design heat treatments
resulting in different vanadium carbide sizes in 100Cr6 + 0.5V bearing steel, providing
an accurate method for modelling vanadium carbide evolution in similar high carbon
steels. TEM and APT was used to validate the modelling with respect to number
density and size of precipitates and, in combination with synchrotron X-ray diffraction,
to establish the V4C3 crystal structure of the vanadium carbides. Finally, TDA was used
to establish the optimum vanadium carbide diameter of 10 nm for hydrogen trapping
capacity. APT shows that the vanadium carbides are still in their growth stage, with
a carbon rich region surrounding the carbides. Tempering is found to reduce hydrogen
trapping capacity due to the reduction in dislocation density. Any cementite formed
during the tempering process has a negligible effect on hydrogen trapping capacity.
160
Chapter 8
The optimisation of gamma
prime for hydrogen trapping in
nickel-based alloys
8.1 Introduction
The literature surrounding the hydrogen trapping efficiency of γ′ in nickel-based alloys
is mixed. The work of Turnbull et al. [225] has demonstrated that γ′ precipitates can act
as hydrogen traps with large binding energies in Alloy X-750. Symons’ [226] work on the
same alloy and treatment found distinctly lower hydrogen binding energies. Thompson
and Brooks [324] found that γ′ precipitates did not act as hydrogen traps at all until
deformation began. Turnbull et al. [225] proposed that the observed variation in hy-
drogen trapping capacity is caused by differences in misfit strain of the γ and γ′ phases
between the studied alloys.
To investigate this claim, as shown in the proceeding sections, several model alloys con-
taining γ′ precipitates, with differing degrees of misfit between the γ and γ′ phases, have
been developed using thermodynamic modelling. Corresponding heat treatments have
been designed so that each alloy will contain similar volume fractions and sizes of γ′
precipitates. These alloys are characterised using electron microscopy, X-ray diffraction
and thermal desorption analysis to determine whether gamma prime does trap hydro-
gen, and if so, whether a relationship between lattice mismatch and hydrogen trapping
efficiency exists, both with respect to capacity and binding energy.
161
8.2 Thermodynamic and kinetic modelling
To produce γ′ precipitates in a nickel alloy, a significant portion of the composition
is required to be aluminium or titanium. The exact stoichiometry of the alloy will
determine the coherency of the γ′ phase with the matrix. Three alloy compositions
were supplied by Prof. Franck Tancret of The University of Nantes using Thermocalc
modelling software, coupled with a neural network to predict the lattice misfit of several
model alloys. Three such alloys (Table 8.1) were selected for investigation of lattice misfit
effect on hydrogen trapping efficiency. These compositions were selected to provide one
sample with no misfit, a sample with maximum misfit and a sample with intermediate
misfit. All compositions and heat treatments were designed to produce a volume fraction
of around 25 % γ′.
All the model alloys must undergo an initial solutionisation at 1200 ◦C so that, during
the proceeding heat treatment steps, precipitate size, distribution and volume fraction
of precipitates can be controlled. The chosen Alloys A, B and C are presented in Table
8.1.
Alloy Ni Fe Cr Al Ti Nb Predicted Misfit [A˚]
A Bal. 17.0 6.0 3.9 0.1 0.3 0.000
B Bal. 14.4 6.0 3.6 0.4 0.6 0.010
C Bal. 14.6 6.0 3.3 0.5 1.7 0.023
Table 8.1: The compositions in wt% of model nickel alloys designed to produce γ′ precipitates
with a range of different lattice misfits. The lattice misfits were calculated using Thermocalc
modelling software.
It is essential that there is minimal variation in volume fraction and size of the precip-
itates between each alloy to ensure that any significant variation in hydrogen trapping
capacity and binding energy can be attributed to the lattice coherency effect. MatCalc
was used with the database “mc ni v2.033” for thermodynamic calculations and the
diffusion database “mc ni v2.007” for thermokinetic calculations. γ′ was set to nucle-
ate homogeneously within the austenitic matrix, with full equilibrium assumed which
in practice is rarely achieved. Nonetheless, these conditions proved the most effective
parameters for fitting kinetics data to experimental data. Such fitting was established
through iterative heat treatments of experimental casts and their proceeding character-
isation to develop the optimised simulation parameters. In doing so, the optimised heat
treatment was designed; a solutionisation at 1200 ◦C, followed by a 10 hour hold at 750
◦C, as shown in Figure 8.1 for Alloy B, with further plots showing the evolution of phase
fraction, number density and mean radius of precipitates with time. Table 8.2 shows
the predicted final phase fraction, number density and mean radius of γ′ precipitates
for Alloys A, B and C. As can be seen, there is little deviation in the value of any of
these properties between each alloy, affirming the treatment’s suitability.
162
(a) (b)
(c) (d)
Figure 8.1: Thermokinetic plots produced by MatCalc for Alloy B subject to a heat treatment
of 1200 ◦C for 2 minutes, followed by 750 ◦C for 10 hours. (a) the temperature, (b) the phase
fraction, (c) the number density and (d) the mean radius of γ′ precipitates as a function of
treatment time.
Alloy Phase Fraction Number Density
[m−3]
Mean Radius [nm]
A 0.291 1.67×1022 15.4
B 0.242 1.65×1022 14.9
C 0.233 1.80×1022 14.0
Table 8.2: Predicted γ′ precipitate information for alloys A, B and C subject to a dissolution
hold at 1200 ◦C followed by a 10 hours hold at 750 ◦C. MatCalc modelling software was used
to calculate this information.
163
8.3 Results and discussion
8.3.1 Microscopy
TEM imaging of alloys A, B and C (Figure 8.2) showed that the phase fraction, number
density and mean radius of γ′ precipitates were very similar for each alloy, validating the
kinetic modelling. The mean radius of the precipitates in the samples, demonstrated
clearly in Figure 8.2(d), were all approximately 15 nm, which again, was in good agree-
ment with the thermokinetic modelling prediction in Table 8.2. This result means that
any difference in trapping capacity between alloys A, B and C may be attributed to the
lattice mismatch between the γ and γ′ phases.
164
Figure 8.2: Bright field TEM images of Alloys a) A, b) B and c) C and d) Higher magnification
image of Alloy B (Table 8.1), subject to the heat treatment in Figure 8.1 to produce γ′
precipitates. The phase fraction, number density and precipitate size is very similar in all
three alloys as required and concur with the modelling predictions shown in Table 8.2.
8.3.2 X-ray diffraction
The lattice mismatch of alloys A, B and C was measured using a combination of a Bruker
D8 DAVINCI diffractometer (step size 0.02◦, step time 20 s and angular 2θ range 23 -
123◦) and a Phillips PW1820 (step size 0.01◦, step time 5 s and angular 2θ range 115 -
150◦) using Ni filtered Cu Kα radiation.
Thermokinetic modelling predicted the lattice mismatch between the γ and γ′ phases
165
of the three alloys as explained previously. XRD was used to quantify the true lattice
mismatch. Since both the γ and γ′ phase have fcc crystal structure, you would expect
the same {hkl} reflections to be present. The position of these peaks is of course
dependent upon the lattice parameter of each phase. If two peaks are formed within a
small angular range of the scan then there is a lattice mismatch between the two phases.
Perfect coherency would lead to the formation of only one peak as both phases have the
same crystal structure and lattice parameter.
This shifting of peaks is more apparent at larger 2θ angles. However, beyond a 2θ angle
of 134◦, a different diffractometer was required, producing lower intensity signals. A
noticeable skewing in the peak at a 2θ angle of around 140◦ for Alloy C (Figure 8.3)
suggested mismatch as described above. Rietveld analysis was performed on the spectra
using known crystallographic information files (CIFs) for the γ [325] and γ′ phase [326].
The refined lattice parameters for the γ and γ′ phases were 3.571 ± 0.001 and 3.576
± 0.001 respectively. This lattice misfit is significantly lower than predicted in Table
8.1. However, the skewing in this peak could be due to instrumental broadening and so
there is low confidence in this result.
The scans of all three alloys at lower angles showed small peaks at 2θ angles of 25.0◦
and 35.5◦, seen clearly in Figure 8.4 (c). This was found to fit well with existing γ′ CIFs
and importantly was not present in the γ CIF file. Reitveld refinement of the γ CIF
was performed first as the γ phase was the primary contributor to the signal due to
its larger volume fraction. The γ′ CIF was refined to match the 25.0◦ and 35.5◦ peaks.
This combination fit the shape of higher order peaks well, as seen in Figure 8.4 (b).
However, the intensity of some of the fitted peaks did not fit the measured data, which
is likely due to preferential orientation of grain growth. Similar analysis was carried out
for Alloy B and C as seen in Figures 8.5 and 8.6 respectively.
Table 8.3 presents the calculated lattice parameters for all phases in the studied samples.
These results have significantly lower error than the analysis from the higher angle
spectra, with the fitted spectra appearing to have a better visual match. Based on this,
the results of low angle analysis are taken to be a better prediction of the true lattice
mismatch. Thermokinetic modelling (Table 8.1) predicted sample A to have zero misfit
which was upheld experimentally. The measured misfit of Alloy B was approximately
half of the predicted misfit. Alloy C was measured to have no misfit (within experimental
error). This result means that the relationship between misfit and hydrogen trapping
capacity will be difficult to investigate, since only Alloy B displayed any misfit. However,
all three of the alloys have different chemical compositions and so could lead to differing
hydrogen trapping capacities due to chemistry-dependent hydrogen interactions.
166
(a)
(b)
Figure 8.3: High angle XRD analysis of Alloy C. (a) displays the entire spectrum. The black
and grey lines are the γ and γ′ phases respectively. A distinct skewing is observed in the peak
in (b), which may be due to instrumental broadening rather than lattice misfit.
Alloy Measured
γ′ Lattice
[A˚]
Measured γ Lattice
[A˚]
Predicted Misfit
[A˚]
Measured
Misfit [A˚]
A 3.5682 ±
0.0001
3.56822 ± 0.00001 0.000 0.00003
B 3.5731 ±
0.0002
3.56782 ± 0.00004 0.010 0.005
C 3.5712 ±
0.0001
3.57144 ± 0.00003 0.023 0.0002
Table 8.3: The predicted and measured γ/γ′ lattice misfit for alloys A, B and C. The lattice
parameters were measured using XRD.
167
(a)
(b)
(c)
Figure 8.4: XRD analysis of Alloy A. (a) displays the entire spectrum. The region from 38◦
to 44◦ was not analysed as the signal in this region was badly affected by the Ni filter used.
The Reitveld analysis was performed using the predicted phases matches the observed signal
well in (b). The green and grey peaks represent the contributions from the γ and γ′ phases
respectively. The small peaks at low angles seen in (c) are due to the γ′ phase only.
168
(a)
(b)
Figure 8.5: XRD analysis of Alloy B. (a) displays the entire spectrum. The green and grey
hatched peaks represent the contributions from the γ and γ′ phases respectively. The small
peaks at low angles seen in (a) are due to the γ′ phase only. The slight lattice misfit can be
seen in the markers at the top of the spectra in (b).
169
(a)
(b)
Figure 8.6: XRD analysis of Alloy C. (a) displays the entire spectra. The green and grey solid
peaks represent the contributions from the γ and γ′ phases respectively. The small peaks
at low angles seen in (a) are due to the γ′ phase only. No lattice misfit can be seen in the
markers at the top of the spectra in (b).
170
8.3.3 Hydrogen charging and thermal desorption analysis
Heat treated samples of Alloy A, B and C and a solutionised Alloy C sample underwent
TDA. All three heat treated samples trapped more hydrogen than the solutionised
sample (Table 8.4), suggesting that the γ/γ′ interface does indeed behave as a dominant
hydrogen trapping site, in comparison to the dislocation and grain boundary traps that
will dominate the solutionised γ phase.
The higher test temperatures revealed the presence of several higher temperature peaks
in the TDA graphs (Figure 8.7). This suggests that there are multiple types of trapping
sites in the alloys, which requires further investigation to identify. However, the heat
treatment led to a clear increase in hydrogen trapping capacity, indicating that the
γ/γ′ interface does indeed behave as a dominant hydrogen trapping site. The variation
between the three alloys is likely to be a chemical effect due to the varying composition
of each. The binding energy of the peak desorption rate temperature appears to increase
from Alloy C through A, suggesting that γ′ traps are the deepest in these alloys.
Sample Temperature of Peak
Desorption Rate [◦C]
Hydrogen Trapped
[ppm]
Alloy A (Heat Treated) 162 4.22
Alloy B (Heat Treated) 161 3.59
Alloy C (Heat Treated) 155 3.51
Alloy C (Solutionised) 153 1.98
Table 8.4: TDA results for Alloy A, B and C samples. The temperature of peak desorption
rate is related to the binding energy of any traps within the samples.
The Vickers hardness of Alloy A, B and C are displayed in Table 8.5. The results show
an increase in hardness with an increase in predicted lattice misfit (Table 8.1). However
as explained previously, XRD analysis revealed only Alloy B to have misfit and so the
relationship between lattice misfit and hardness is not observed experimentally. It is
likely that hardness has been affected by differences in the chemical nature of the alloys.
8.4 Summary
Three γ′ containing model nickel alloys were designed via thermodynamic and thermoki-
netic modelling to investigate how lattice misfit between the γ and γ′ phases affected
Alloy Vickers Hardness [MPa]
Alloy A 2620 ± 80
Alloy B 2730 ± 80
Alloy C 3200 ± 90
Table 8.5: The Vickers hardness of Alloy A, B and C samples.
171
0.00E+00 
1.00E-04 
2.00E-04 
3.00E-04 
4.00E-04 
5.00E-04 
6.00E-04 
7.00E-04 
0 50 100 150 200 250 300 350 400 450 500 
H
 d
es
or
pt
io
n 
ra
te
 / 
w
pp
m
 s
−1
 
Temperature / °C 
Alloy A 
Alloy B 
Alloy C 
Alloy C Solutionised 
Figure 8.7: The TDA plots for heat treated Alloys A, B and C and solutionised Alloy C. All
three heat treated samples retained more hydrogen than the solutionised sample, suggesting
that the γ/γ′ interface provided improved hydrogen trapping sites over γ phase alone. Data
was gathered with the assistance of Dr Tom Depover at Ghent University.
172
hydrogen trapping capacity. These alloys where characterised by transmission electron
microscopy, X-ray diffraction and thermal desorption analysis. Transmission electron
microscopy identified that the phase fraction, number density and mean size of γ′ pre-
cipitates were approximately equivalent for all three of the alloys. Any difference in
hydrogen trapping capacity could therefore be attributed to lattice misfit. X-ray diffrac-
tion found that Alloy C contained no significant lattice misfit when it was designed to
have the largest misfit. Alloy A showed no lattice misfit as intended. Alloy B had slight
misfit, but lower than intended. The heat treatment led to a clear increase in hydrogen
trapping capacity, affirming that γ/γ′ is the dominant hydrogen trapping site in these
alloys.
Although the purpose of Alloys A, B and C was to investigate the effect of misfit in
the γ/γ′ interface on hydrogen trapping capacity, only Alloy B displayed lattice misfit.
There was no clear evidence for the effect of lattice mismatch on trapping capacity.
New model alloys should be designed with a larger range of lattice misfit to assess the
relationship fully.
173
Part III
The design of novel hydrogen
embrittlement resistant alloys
174
Chapter 9
Nanocarbide containing alloys
Thermodynamic and kinetic modelling were carried out to develop hydrogen embrittle-
ment resistant steels and nickel-based alloys, incorporating the hydrogen trapping capa-
bilities of homogeneously distributed nano-carbides with the hardness and strength re-
quirements of current alloys. Six new steel compositions have been developed to achieve
such desired properties: two nanostructured superbainitic steels G5V and G5V(HC),
two high carbon (around 1 wt%) martensitic steels based upon 52100 steel, 100Cr6WV,
100Cr6+0.5V, one martensitic mild carbon case-hardened steel Grade159V, and one low
carbon offshore flexible pipe steel Flexpipe X, all with homogeneously distributed 10
nm sized vanadium carbides. In addition to the described steels, a novel nickel-based
alloy has been developed, Alloy 600V, again with homogeneously distributed around 10
nm sized vanadium carbides. The complete design and characterisation process through
which these alloys were developed is provided in the following sections.
As evidenced in the literature review, the presence of nano-sized titanium carbides,
niobium carbides, vanadium carbides and potentially tungsten carbides can provide an
increase in hydrogen trapping efficiency, both in terms of capacity and irreversibility
of such trapping. For a given ternary steel, varying both carbon and substitutional
element concentration, calculations for the carbides’ formation and dissolution tem-
peratures were carried out in order to identify the stability, and hence the industrial
practicality, of the carbides for use in commercial steels. The modern furnaces of in-
dustrial steel manufacturers operate with a maximum temperature of around 1300oC.
Thus, it would appear beneficial, both in regard to economy and practicality, to identify
which carbides dissolve below this temperature and the compositional dependence of the
dissolution temperature. In addition to the calculations for ternary systems, silicon ad-
ditions, given its effect on cementite formation, were studied to identify any influence on
carbide behaviour in equilibrium. The results of the thermodynamic equilibrium calcu-
lations for the ternary Fe-C-M and quaternary Fe-C-Si-M systems are shown in Figures
9.1 (a) through (e). These thermodynamic calculations were performed using the com-
175
mercially available software packages MatCalc v.5.60 and ThermoCalc v.4.0, with the
databases mc fe v2.021.tdb and TCFE v.6 respectively. ThermoCalc and MatCalc are
software packages for thermodynamic calculations and phase equilibrium calculations in
multicomponent systems that use the CALPHAD method to evaluate thermodynamic
databases and apply Gibbs free energy minimisation algorithms. For a given system,
the software computes the mass fractions and compositions of all phases at which the
free energy of the system is at a minimum. These calculations provide an indication of
the stability of the permissible phases for a given composition at equilibrium, allowing
one to establish the viability of managing these phases in the proceeding thermokinetic
calculations and corresponding real-life processing methods and heat treatments.
176
177
Figure 9.1: Thermodynamic equilibrium calculations for a quaternary steel’s carbon content
dependency of the dissolution temperature of (a) molybdenum carbide (b) titanium carbide
(c) niobium carbide (d) vanadium carbide and (e) tungsten carbide.
178
The dissolution temperature of titanium carbide reduces with carbon content. Silicon
additions tend to increase the dissolution temperature for carbon contents <0.5 wt%,
and reduce it for contents >0.5 wt%. As for titanium carbide, the dissolution temper-
ature of niobium carbide reduces with carbon content. However, Si additions tend to
increase the dissolution temperature at all carbon contents, except for the 1C-0.1Nb-
2Si variant. Molybdenum carbide has a significantly lower dissolution temperature in
comparison to that of titanium and niobium carbide. For molybdenum contents of 0.1
wt% and lower, the dissolution temperature reduces with carbon contents >0.01 wt%.
However, for 1 wt% molybdenum, carbon additions increase the dissolution tempera-
ture. Silicon additions increase the dissolution temperature for all variants. Tungsten
carbide behaves similar to that of molybdenum carbide, however with a reduction in
dissolution temperatures at carbon content >0.1 wt%. The dissolution temperature
of vanadium carbide increases with carbon content for vanadium contents of 0.1 wt%
and above. Silicon additions increase the dissolution temperature for all variants of
vanadium carbide.
Vanadium carbide, tungsten carbide and molybdenum carbide show great promise given
their lower dissolution temperatures (<1250oC for substitutional contents up to 1 wt%
and carbon contents up to 1 wt%) in comparison to that of titanium carbide and nio-
bium carbide. Titanium carbide and niobium carbide are of reduced practicality given
the lower carbon and/or substitutional contents necessary to lower the dissolution tem-
peratures into the industrially practical range. The influence of silicon on cementite is
predominantly during the initial cementite formation from epsilon carbides [327]. As
the results show, silicon additions provide little effect on the dissolution temperatures
of the selected carbides, however, as with cementite, it is permissible that silicon may
retard the nucleation and growth of the carbides.
Given that both high hardness and hydrogen resistance is desired, a higher carbon
content of the alloy, carbide phase fraction and finer distribution of carbides is desired
to achieve the optimal properties. It is thus more beneficial to use those carbide forming
elements for which a greater amount of both carbon and carbide forming element can
be added whilst remaining industrially practical in terms of their dissolution. For this
reason, tungsten carbide, vanadium carbide and molybdenum carbide appear the most
suitable. However, carbon segregation and/or case hardening affords the opportunity
to implement high alloying additions despite the resulting carbides’ stability.
179
9.1 Modelling and experimental methodology
9.1.1 Thermodynamic and kinetic modelling
All simulations were conducted using MatCalc version 5.6, with the thermodynamic
databases mc fe v2.021 (for steel), mc ni v2.033 (for nickel-based alloys) and diffusion
databases mc fe v2.006 (for steel) and mc ni v2.007 (for nickel-based alloys). For all
steels, V4C3 and cementite were restricted to nucleate on dislocations and homoge-
neously respectively. Cementite was allowed to form under paraequilibrium whereas
V4C3 was formed under orthoequilibrium. This choice was made on the basis that iron
carbides such as cementite form much quicker than alloy carbides due to the high mo-
bility of C. As with the simulations conducted in Part II, Chapter III, simulations were
carried out to generate an initial calculation state, replicating the microstructure of the
as-received commercial cast in its spheroidised state. The replicated microstructure was
used as the initial calculation state from which kinetic simulations would continue for
the primary heat treatment schedules. The key factors of influence on the proceeding
microstructural evolution were the carbide fractions formed from the liquid (primary
carbides), calculated using the Scheil-Gulliver method, the carbide fractions and distri-
bution formed from the solid solution (secondary carbides), and the resulting matrix
composition, both of which are calculated using kinetics. The dislocation density of
annealled austenite, ferrite and freshly formed martensite were taken as 1011 m−2, 1012
m−2 and 1015 m−2 respectively.
The alloys developed throughout this chapter are based upon established commercial
alloy compositions and their corresponding heat treatments. Such alloys, although the
most commercially successful in the case of 100Cr6, Grade 159, Flexpipe and Alloy 600
for their respective applications, have been shown to be highly susceptible to hydrogen
embrittlement. As such, minimal compositional variations were made to ensure limited
deviation from the commercial counterparts’ mechanical and corrosion properties with
the intention solely to improve hydrogen embrittlement resistance. In order to further
assist the application of the novel alloys, the heat treatments developed were designed to
use established industrial processes identical to that currently used for their commercial
counterparts.
For each alloy designed in Sections 9.2-9.6, an iterative optimisation methodology was
used. The initial composition was developed using thermodynamic equilibrium cal-
culations to ensure the intended nanocarbides could be dissolved under the relevant
industrial furnace temperature limitations (1200oC and below in most cases). The
composition was then experimentally cast and characterised to establish the initial mi-
crostructural conditions for preliminary inputs into the thermokinetic models as stated
previously. Heat treatments were then designed using the kinetic models described in
180
chapter 5. Samples of the cast material were subjected to the novel heat treatments and
characterised. Deviations from the predicted microstructures were accounted for and
the heat treatments revised to produce the desired microstructures. As such, the alloy
design was an iterative process requiring a number of heat treatment trials to acquire
the optimised conditions for proceeding hydrogen trapping property characterisation.
During these iterations, the kinetic modelling parameters were optimised to improve
accuracy. The optimised kinetic modelling parameters utilised are described for each
alloy in Sections 9.2-9.6. Unless specified otherwise, all other model parameters are set
to their default value in MatCalc.
9.1.2 Materials
Two 65 gram experimental casts of each alloy were produced using electric arc melting to
the respective compositions shown in Table 9.1. Larger industrial casts of 100Cr6+0.5V,
100Cr6 and Steel G were provided by Tata Steel in a spheroidised condition. Experimen-
tal casts were sealed in glass tubes backfilled with argon to prevent oxidation and decar-
burisation during the proceeding homogenisation at 1300oC for 48 hours and spheroidis-
ation treatment. Nickel-based alloy 600V was not spheroidised. The steels’ spheroidisa-
tion treatment involved holding the material above the austenitisation temperature (A1),
cooled to precipitate cementite and allow coarsening, forming spheroidised pearlite. The
material is then slow cooled to room temperature to ensure full ferritic transformation,
resulting in a hardness of approximately 190 Hv10. All casts were cold-swaged to rods
8 mm in diameter and sectioned into cylinders 10 mm in length for the proceeding
dilatometry tests. The heat treatments were conducted in a THERMECMASTOR-Z
dilatometer. Hardness tests were taken using a Vickers hardness testing machine with
a 30 kg load. The quoted resulting values represent the mean of 5 readings within one
standard deviation.
Heat treated dilatometry samples were hot mounted in Bakelite, ground and polished
to a 1 µm finish with diamond paste. The samples were etched in 2% nital solution. A
Zeiss Axioplan2 was used for optical microscopy. An FEI Nova NanoSEM was used for
scanning electron microscopy (SEM) and a FEI Tecnai F20 FEG TEM (200 keV) and
FEI Tecnai Osiris FEG-TEM (80 or 200 keV) were used for transmission electron mi-
croscopy (TEM). For TEM, 0.3 mm slices of the 8 mm diameter heat-treated dilatometry
samples were cut. These slices were ground using 800 grit SiC paper and then punched
or spark eroded to form 3 mm diameter foils. The foils were further ground using 1200
grit SiC paper and cleaned using ethanol and acetone. The foil samples were electropol-
ished using 5% perchloric acid, 25% glycerol and 70% ethanol by a Struers Tenupol 5
electropolisher. A voltage of 40 V at 16oC resulted in samples suitable for analysis by
TEM unless specified otherwise.
181
Table 9.1: Composition of the designed novel alloys in wt %. The compositions shaded are
those designed in this work.
Grade C Mn Si Cr Ni Mo Cu V Al W
100Cr6+0.5V 0.98 0.38 0.16 1.39 0.07 0.02 0.12 0.5 ... ...
100Cr6+0.2V 0.98 0.38 0.16 1.39 0.07 0.02 0.12 0.2 ... ...
100Cr6+0.08V 0.98 0.38 0.16 1.39 0.07 0.02 0.12 0.08 ... ...
100Cr6WV 0.98 0.38 0.16 1.39 0.07 0.02 0.12 0.7 ... 6
G5V 0.84 0.03 1.90 1.97 ... 0.28 ... 0.5 0.68 ...
G5V(HC) 0.9 0.03 1.90 1.97 ... 0.28 ... 0.5 0.68 ...
G 0.84 0.03 1.90 1.95 ... 0.27 ... ... 0.68 ...
Grade 159V 0.18 0.5 0.2 1.65 1.55 0.3 0.15 0.2 0.03 ...
Flexpipe X 0.54 0.71 1.47 0.66 0.04 ... 0.03 0.17 ... ...
9.1.3 Hydrogen charging and thermal desorption analysis
The charging and TDA equipment and parameters used on the novel alloys were identical
to those described in Part II, Chapter 7. All samples were charged electrochemically with
hydrogen using a 3% NaCl + 0.3% NH4SCN solution for 4 days. The applied current was
calculated based on sample dimensions in order to give a current density of 10 mAcm−2
unless specified otherwise. For each alloy, samples were charged simultaneously unless
specified otherwise. Following charging, samples were gently polished with 1200 grit
SiC paper and cleaned with ethanol and acetone and stored in ambient conditions for 5
days to allow diffusible and weakly trapped hydrogen to leave the sample prior to TDA.
9.2 100Cr6 steel variants
As previously discussed, 100Cr6 variants dominate bearing steel production, and were
first employed over a century ago. Such steels are typically supplied hot-rolled with a
pearlitic structure and will require annealing to spheroidise the cementite [86] for further
machining, after which heat treatments are applied to create the desired hard martensitic
microstructures required for improving rolling contact fatigue life. The salient concept
behind 100Cr6+0.5 V is that stark improvements in hydrogen trapping and hardness
can be achieved with minimal change in composition from 100Cr6, indicating that one
can improve on existing materials without compromising other properties. Given the
beneficial effects of vanadium carbide evidenced in the hydrogen trapping capacity of
vanadium containing 100Cr6+0.5V, a number of 100Cr6 variants have been designed and
characterised with respect to microstructure and hydrogen trapping efficiency, presented
in the following sections, in their entirety.
182
9.2.1 100Cr6+0.5V
As previously described, 100Cr6+0.5V is a vanadium containing variant of 100Cr6 with
a superior hydrogen trapping capacity over its parent material. Although of great
industrial appeal, the original heat treatment schedule designed by Szost [108] was
found to be unachievable using the patent holder’s established industrial production
line due to the complex, industrially demanding three stage treatment, as shown in
Figure 9.2. In the heat treatment, the steel is austenitised for 15 min at 860◦C. The
steel is then heated to 1200◦C for 1 min to dissolve the vanadium carbide and cementite
of the as-spheroidised microstructure (diameters are around 25 nm and 200 nm for the
as-spheroidised V4C3 and cementite respectively). V4C3 precipitates are then grown at
600◦C. However, cementite also grows at this temperature. Large cementite particles can
be detrimental to the RCF life of the bearing as they provide sites for crack nucleation.
Hence, another temperature spike to 860◦C is included to redissolve cementite, although,
as evidenced previously in this thesis, cementite does not dissolve at this temperature
in 100Cr6+0.5V. The steel is then tempered at 215◦C, promoting formation of a fine
cementite dispersion. Hereonafter, this treatment will be referred to as Szost’s heat
treatment. This heat treatment is not feasible using the current continuous production
line as it necessitates three seperate quenching steps and three separate furnaces. As
such, the focus of this section will be to produce a microstructure that can match
both the trapping capacity and mechanical performance of 100Cr6+0.5V as designed
by Szost, but using a more feasible heat treatment.
Figure 9.2: The Szost heat treatment [108] for bearing grade 100Cr6+0.5V
The novel heat treatment should produce an alloy which meets the microstructural
requirements within the constraints of the industrial processing conditions. The three
key limitations are the maximum temperature of 1200◦C, maximum heating rate of
10◦Cs−1 and the maximum of two furnaces within the continuous casting production
183
line. As such, the multiple quenching steps of Szost’s treatment are unfeasible and
should be replaced by a single treatment stage with a single quench, or two stages with
two quenches, with the prior being preferable.
Thermodynamic and kinetic modelling
In its spheroidised state, 30 nm and 200 nm diameter V4C3 and Fe3C particles respec-
tively are observed in 100Cr6+0.5V. These features were replicated in order to produce
an initial calculation state from which kinetic simulations would continue for the pri-
mary heat treatment schedules. The Ae3 temperature, above which the microstructure
is fully austenitic, is approximately 740◦C. The final quench to martensite must be
initiated from above this temperature to ensure the maximum martensite fraction. Ce-
mentite is seen to dissolve above 860◦C at equilibrium. Prior to this quench, cementite
precipitation should be avoided, as it limits the maximum phase fraction of vanadium
carbide, reduces the austenite carbon content, and thus the hardenability, and reduces
the number of potential temper carbides shown to improve trapping capacity. The
heat treatment must initially solutionise the coarse V4C3 of the as-cast microstructure.
Szost’s heat treatment indicated that a hold at 1200◦C for 1 minute is sufficient for
full solutionisation. However, Szost held the steel at 860◦C for 15 minutes prior to this
dissolution step, perhaps presuming it necessary for austenitisation. However, as will
be shown, this stage is redundant and full austenitisation can be achieved with solely
the 1 minute hold.
In order to nucleate and age vanadium carbides, without growing cementite, a slow cool
from the solutionisation temperature is utilised so that only one quenching step and
two furnaces are necessary. Figure 9.3 shows the results from kinetic simulations of two
novel heat treatments, HT3 and HT4, presented in more detail in Figure 9.4. Cementite
and V4C3 are shown to dissolve during the 1 minute hold at 1200
◦C. In HT3, the steel
is then cooled to 800◦C and held for 500 seconds to age V4C3 to a mean radius of 5
nm. The final V4C3 precipitate number density is of the order of 10
18 m−3. The final
cementite distribution has a mean radius of around 8 nm. This is not large enough to
have any detrimental effect on mechanical properties and may well improve the trapping
capacity in a similar way to temper cementite.
In HT4, a slow cool to 600◦C is conducted, followed by a reheat up to 900◦C where
it is held for 80 seconds. The monotonic cool of HT3 was changed to incite a burst
of V4C3 nucleation and hence improve the number density of the carbides at 10 nm
diameter. Ageing at 900◦C also improved the distribution of V4C3 radii around the
mean because the nucleation rate is so low at this temperature. Figures 9.5a and b
show histograms of V4C3 particle radii at the end of HT3 and HT4, respectively. In
HT3, there is a large number of very small precipitates (less than 3 nm in diameter)
184
Figure 9.3: Modelling results for heat treatment HT3 and HT4, showing: a) the treatment
schedule, b.) the mass fraction of precipitates c.) the number density of precipitates, and d.)
mean radii of precipitates. The final quench is not shown.
185
Figure 9.4: Heat treatment schedules for HT3 and HT4.
which, as shown previously, will not contribute significantly to the hydrogen trapping
capacity. HT4 shows little deviation away from the mean size of 5nm, resulting in a
significantly higher number density of carbides at their optimum size for trapping. The
final cementite number density is much lower in HT4 compared to HT3 (1013 compared
to 1019 m−3). All heat treatments use a mean cooling rate of 5◦Cs−1.
186
(a)
(b) VC
HT4
Figure 9.5: Histograms of V4C3 and cementite particle distributions for a) HT3 and b) HT4.
Microscopy
The spheroidised casts were sectioned into cylinders 8 mm in diameter and 6 mm in
length. The cylinders were heat treated under vacuum according to the schedules shown
in Figure 9.4. After quenching, samples were tempered for 2 hours at 210◦C. A temper-
ing treatment such as this is standard practice for 100Cr6, as the steel is too brittle to
be used in the as-quenched condition. Tempering anneals the highly strained martensite
and induces the formation of a variety of iron carbides, reducing the degree of carbon
saturation, further relieving the residual strain.
Samples were hot mounted in Bakelite, ground and polished to a 1 µm finish. They
were etched in 2% nital (2% nitric acid and 98% methanol). Figure 9.6 shows the
microstructure of the sample in the as-spheroidised condition.
SEM imaging was conducted using a FEI Nova NanoSEM at 5keV. Figure 9.7 shows
the martensitic microstructures of 100Cr6+0.5V after the Szost heat treatment, HT3
and HT4. The observed martensite is much coarser after HT3 and HT4 in comparison
to that after Szost’s, most likely due to the coarseness of the austenite grains prior to
quenching. Such grain growth is due to the increased period of time at high temper-
atures in comparison to Szost’s treatment. Samples were etched in nital to reveal the
187
Figure 9.6: The microstructure of 100Cr6+0.5V after spheroidisation.
prior austenite grain boundaries. A Zeiss Axioplan2 optical microscope was used to
characterise the grains.
The mean prior austenite grain size of around 200 µm after HT3 is shown in Figure 9.8.
For TEM, 3 mm diameter discs of HT3 were spark eroded from slices cut from the
centre of the treated dillatometry samples. Samples were polished with a 1200 grit SiC
paper and electropolished in 5% perchloric acid, 25% glycerol and 70% ethanol using a
Struers Tenupol 5 electropolisher. A FEI Tecnai Osiris FEG-TEM (200 keV) was used
for all TEM based techniques. Figure 9.9 shows a STEM bright field image of HT3 and
its corresponding EDX vanadium map, indicating the distribution of the small, around
10 nm diameter vanadium carbides.
Hardness was measured for all heat treatments using a Vickers hardness testing machine
with a 30 kg load. The quoted resulting values represents the mean of 5 readings within
one standard deviation. The hardnesses measurements were as follows: 785 Hv for
100Cr6, 804 Hv for Szost, 842 Hv for HT3, and 694 Hv for HT4. Indicating that the
hardnesses achieved HT3 are above that of Szost, but HT4 is below it. The reasons
behind these variations remain inconclusive but are most likely due to carbon variations
in the matrix and retained austenite content.
188
Figure 9.7: SEM micrographs of 100Cr6+0.5V after a) Szost’s treatment b) HT3 c) HT4.
189
Figure 9.8: Optical micrograph of 100Cr6+0.5V after HT3.
V
Figure 9.9: A STEM bright field image of HT3 alongside a corresponding vanadium EDX
map.
190
X-ray diffraction
Samples of HT3 and HT4 were scanned over a 2θ of 35 to 125◦ with a step size of 0.06◦
and a dwell time of 2 seconds using a Bruker D8 diffractometer with Cu Kα radiation.
A 0.18 mm thick nickel filter was used. Rietveld refinement was performed using the X-
pert High Score Plus package. The Volume fraction of retained austenite was estimated
to be 5.2% for HT3 and 10.4% for HT4. Figure 9.10 shows the XRD patterns for HT3
and HT4. Figure 9.11 shows the relative intensities of the first and second peaks due to
the reflections from the 111γ and 110α planes respectively in both samples. The volume
fractions are calculated through integrating the peaks.
191
Figure 9.10: XRD patterns for HT3 and HT4 with Rietveld difference plots.
192
0 
1 
2 
3 
4 
5 
6 
41 43 45 47 
(111)
(111)
↵(110)
↵(110)HT3 HT4 Relative intensity
1
HT3 HT4 Relative intensity
1
41 43 45 47 

(3
1
1
)
R
el
at
iv
e
in
te
n
si
ty
1
2✓
Figure 9.11: XRD patterns showing the relative intensities of the first and second peaks due
to the reflections from the {111}γ and {110}α planes respectively for HT3 and HT4.
193
Hydrogen charging and thermal desorption analysis
Figure 9.12 presents the results from the thermal desorption analysis of HT3, HT4 and
Szost’s treatment.
0 
0.001 
0.002 
0.003 
0.004 
0.005 
0.006 
0.007 
0.008 
0.009 
0.01 
0 25 50 75 100 125 150 175 200 225 250 275 300 
H
 d
es
or
pt
io
n 
ra
te
 / 
w
pp
m
 m
in
−1
 
Temperature / °C 
BS (2.52 wppm) 
HT3 (3.15 wppm) 
HT4 (2.98 wppm) 
Figure 9.12: Hydrogen desorption data for HT3, HT4 and Szost’s treatment (BS). A heating
rate of 25◦C hr−1 was used. The total trapped hydrogen content is shown in the legend.
As shown, similar results are observed to that identified in Part II, Chapter 7 with re-
spect to the two prominent peaks. HT3 is seen to trap more hydrogen than HT4. Based
on values of binding energy reported in the literature (21-29 kJ mol−1) for coherent
Fe3C [28] compared to 30 kJ mol
−1 for V4C3, shown in Part I, Chapter 2, Section 2,
it is speculated that both fine coherent cementite and V4C3 is responsible for trapping
hydrogen at these low energies in HT3 and HT4. HT3 and HT4 trapped more hydro-
gen than Szost’s treatment and therefore provide an improvement in hydrogen trapping
capacity both with respect to Szost’s treatment and 100Cr6. Two possibilities as to the
nature of traps giving rise to the higher temperature peak are considered: either that
hydrogen is trapped by retained austenite, or at the interface between the matrix and
incoherent cementite. The fact that the greatest peak area at high temperature is seen
in HT4 is evidence in support of the former since XRD identified HT4 as containing a
significantly higher proportion of retained austenite (10.4% compared to 5.2% in HT3).
Kinetic modelling predicted much less cementite in HT4. Despite more hydrogen being
released at high temperature in HT4 compared to HT3, the total trapping capacity
of HT3 is still 30% larger than HT4 (2.58 ppmw compared to 1.99 ppmw), suggesting
194
that HT3 contains a higher number density of V4C3 nanoprecipitates. As such, TEM
characterisation should be conducted on HT4 to affirm this.
Summary
Two novel, industrially feasible heat treatments, HT3 and HT4, have been developed
for 100Cr6+0.5V, optimised for hydrogen trapping capacity. Kinetic simulations were
conducted to design the heat treatments, validated by microscopy, indicating a homo-
geneous distribution of V4C3 with a mean diameter of 10 nm, the optimum size for
hydrogen trapping. TDA showed an improvement in trapping capacity for both heat
treatments in comparison to Szost’s treatment, and in turn commercial 100Cr6. HT4
had a slightly lower trapping capacity than HT3, likely due to a lower number den-
sity of vanadium carbides and/or the carbides growing beyond the optimum size. HT4
contained a higher proportion of retained austenite, likely due to the higher austenitisa-
tion temperature and lower fraction of carbides, further affirming HT3 as the preferable
treatment.
Both novel treatments produced a much coarser martensite structure than that of Szost’s
treatment, attributed to rapid prior austenite grain growth during HT3 and HT4. A
final grain size of 200 µm was achieved for both HT3 and HT4. A grain size following
the Szost heat treatment is seen to be significantly less than this, around 40 µm. This
accelerated grain growth occurs during the slow cool from 1200◦C. Once carbides begin
to nucleate, these will likely pin grain boundaries and inhibit continued growth. For
bearing applications, steels should have a fine or ultra-fine grain structure of around
30 µm. As such, recrystallisation is often relied upon to achieve such grain refinement.
A remarkably high hardness of 842 Hv was achieved in HT3. 694 Hv in HT4 is still
acceptable for bearing steel applications, which normally have hardnesses upwards of
700 Hv. The high hardness in HT3 is due to the high number density of fine carbides,
as expected by kinetics, and validated via TEM. The lower hardness of HT4, again,
is likely due to a reduced number density of carbides, and these carbides being larger
in size than HT3. Retained austenite volume fraction was found to be significantly
higher after HT4 compared to HT3 (10.4% and 5.2%, respectively). Retained austenite
is not neccesarily a negative, having been shown to both improve and reduce RCF life
depending on the operating conditions [63]. Typically, martensitic bearing steels have a
retained austenite content of 2-6%. Retained austenite content is affected most notably
by carbon content, affecting the martensitic start temperature. There is a positive
correlation between retained austenite content and austenitisation temperature however,
which would explain the retained austenite variability as HT4 and HT3 were quenched
from 900◦C and 800◦C respectively. A higher austenitisation temperature reduces the
phase fraction of carbides, increasing matrix carbon content, and as a consequence,
195
retained austenite fraction.
9.2.2 100Cr6WV
Thermodynamic and kinetic modelling
100Cr6WV, a steel composition, based on the 52100 steel, was developed with the
maximum tungsten and vanadium content permissible for the resulting carbides to be
dissolved at 1300oC under reasonable times. The composition is shown in Table 9.1,
its phase stablility with temperature is shown in Figure 9.13. A martensitic steel was
chosen due to the high strength requirements of bearings and of structural steels such as
those used in nuclear fusion blanket designs [328]. Tungsten and vanadium content was
governed by the dissolution temperature of the resultant carbides and the temperature
limit of industrial furnaces by increasing tungsten and vanadium content until a disso-
lution temperature of 1200oC was reached under equilibrium for both carbides. 1200oC
was chosen to provide a buffer of 100oC from industrial furnaces’ upper temperature
limit of 1300oC to ensure full dissolution is achieved under reasonable times at 1300oC
(< 5 minutes to prevent excessive grain growth).
Figure 9.13: Phase equilibrium diagram for 100Cr6WV
Using the Scheil-Gulliver model, 100Cr6WV’s solidification temperature and primary
carbide fraction and composition at solidification are quantified. The results of which
are shown in Figure 9.14. The primary carbides, vanadium carbide (fcc) and tungsten
carbide (hcp), reduce the solute content of the matrix and so alter the thermodynamic
196
behaviour of the steel. As such, the new matrix composition is quantified for use in the
proceeding kinetic calculations.
austenite liquid
VC
WC
M6C
Figure 9.14: Scheil-Gulliver Calculation for 100Cr6WV
Thermokinetic calculations were performed with an initial grain size of 60 µm. The
nucleation coefficient (1.1) and nucleation sites (dislocations for cementite and vana-
dium carbide, and homogeneous nucleation for tungsten carbide (WC)) resulted in the
most accurate predictions. Large secondary carbides form upon casting, the nucleation
and growth of which are modelled using kinetic simulations. To reproduce the phase
fractions and precipitate sizes of the secondary carbides, virtual heat treatments were
produced in three separate calculation states with secondary carbides of mean radius
500 nm, 1000 nm and 2000 nm, respectively, through manipulation of nucleation coeffi-
cients, nucleation sites, and the minimum permitted precipitate radius within classical
nucleation theory. Although 500 nm primary carbides are typically seen in similar steels,
larger carbides were induced to study the effects of the carbides on alloy behaviour dur-
ing forthcoming heat treatments. Each calculation state was used for proceeding kinetic
simulations and the development of the optimum heat treatment schedule for the de-
sired microstructure. It is typically from experience that one establishes the fraction,
size and distribution of carbides formed upon casting. Thus, by covering the typical
ranges observed in similar steels one ensures that proceeding kinetic calculations cover
all possible scenarios. The microstructure of each of the calculation states is then used as
the initial microstructure for the proceeding primary heat treatment simulations. Two
primary heat treatments were developed to produce an optimised microstructure for
197
100Cr6WV, having the maximum phase fraction of homogeneously distributed around
10 nm sized vanadium carbides and tungsten carbides. Figures 9.15 and 9.16 show the
results of the kinetic modelling for 100Cr6WV with initial secondary carbides 1000 nm
in radius for primary heat treatment options A and B respectively.
Figure 9.15: Thermokinetic simulations for 100Cr6WV showing volume fraction, radii and
the number density of precipitates evolving during the primary heat treatment schedule for
heat treatment option A, indicating the primary objectives during the heat treatment.
Heat treatment option A is the ideal optimised treatment for the steel, although per-
haps not suitable for producing larger homogeneous casts given the intermediate rapid
quenching stage. The primary heat treatment B, with its low heat up and cooling rates
was developed to be applicable to both small and large casts where the permissible cool-
ing rates are limited. This heat treatment, not requiring the partial quench and reheat of
its counterpart, indicates that although a wider distribution of vanadium and tungsten
carbide particle sizes are developed, an effective microstructure, albeit un-optimised,
can be produced by using a slow-cool method suitable for large casts.
To precipitate the desired around 10 nm tungsten and vanadium carbides and fine ce-
mentite structure, the coarse secondary carbides formed in prior manufacturing stages
need to be dissolved. As shown in Figure 9.13, the equilibrium dissolution temperature
is just below 800 oC and 1200 oC for cementite and tungsten/vanadium carbides respec-
tively. Precipitation kinetics simulations indicate that a 1 minute hold at 1300 oC is
sufficient to dissolve vanadium and tungsten carbides 1 µm in radius. The primary heat
treatment is also designed to fully austenitise the steel prior to quenching to martensite.
Given the difficulties in accurately modelling austenitisation, the industrially practised
198
Figure 9.16: Thermokinetic simulations for 100Cr6WV showing volume fraction, radii and
the number density of precipitates evolving during the primary heat treatment schedule for
heat treatment option B, indicating the primary objectives during the heat treatment.
austenitisation treatment of around 15 minutes at 860oC used for 100Cr6 is used as
a lower limit to ensure full austenitisation of 100CrWV. Using an industrially viable
heat-up rate of 0.3oC/s from room temperature to the 1300oC 1 minute hold, the alloy
is held above 860oC for 20 minutes, thus ensuring full austenitisation.
The partial quench from 1300oC to 730oC at -25oC/s, promotes the rapid nucleation
of vanadium and tungsten carbides, reaching a stable number density of around 1020
m-3. As the temperature is increased at 0.03oC/s, vanadium and tungsten carbides
begin to grow to an equilibrium mean radius of around 9 nm and 11 nm respectively
upon reaching 850oC, at which point all cementite has been dissolved. The steel is
then quenched from 850oC to room temperature to form martensite, as discussed in the
following section.
The final stage of heat treatment, used to temper the martensite and precipitate fine
cementite (not shown in Figure 9.15), is a hold at 215oC for 2 hours, no change is seen
in the structure of vanadium and tungsten carbides during this treatment. As such,
the finished microstructure will be that of tempered martensite, nano-structured with
vanadium and tungsten carbides of around 10 nm mean radius, with homogeneously
distributed fine cementite particles around 1 nm in size surrounding the martensite.
100Cr6WV’s heat treatment option B, shown in Figure 9.16, produces a very similar
microstructure except a greater size distribution of vanadium and tungsten carbides
is observed. Assuming the optimal size for hydrogen trapping is 10 nm, this greater
199
size distribution implies a reduced hydrogen trapping efficiency despite the mean radius
of around 10 nm remaining the same as for option A. An additional concern of heat
treatment option B is that of large austenite grain growth due to the increased time
at higher temperatures seen in comparison to option A. The benefit however, is that
option B eliminates the need for the middle quenching step of option A and replaces it
with a slow cool. Thus option B is more appropriate for larger casts where quench rates
may be limited and slow cooling is preferred.
The strengthening contribution of the microstructure’s tungsten carbide, vanadium car-
bide and cementite for the given heat treatment schedule should be significant. In
comparison to 100Cr6, absent of vanadium and tungsten carbides, 100Cr6WV will pro-
vide an increase in strength both from solid solution strengthening and precipitate
hardening. As shown in Figure 9.17, calculations carried out using MatCalc to com-
pute the total yield strength of the finished microstructure, incorporating solid solution,
dislocation, grain and precipitate strengthening indicate the contributions to the total
strength. MatCalc has the ability to model Orowan stress and dislocation shearing
mechanisms and anti-phase boundary hardening and incorporate these into the calcula-
tion for the total yield strength of the material. However, martensite is not included in
the commercially available databases used here, and thus its strengthening contribution
is not accounted for, but will be the most significant [5]. These calculated strengthening
contributions should only be used to provide an indication of the significance of these
contributions and not for accurate calculations. For precise quantification of strength,
physical data must be acquired first, such as dislocation density, grain size and marten-
site packet size from an experimental cast, which in turn can be used to produce the
appropriate parameters for quantifying strength.
Figure 9.17: Modelled strengthening contributions of carbides and matrix for 100Cr6WV
200
Figure 9.18 shows the time-temperature-transformation (TTT) diagram for both 100Cr6
and 100Cr6WV. The data for the TTT curves were computed using the software pack-
age MUCG83 and were used to establish suitable quench rates to avoid deleterious
transformations. The martensite start temperature (Ms) was calculated using Ishida’s
model [25], using the composition of the austenite matrix prior to quenching as calcu-
lated from kinetics:
Ms = 545-330C+2Al+7Co-14Cr-13Cu-233Mn-5Mo-13Ni-7Si+3Ti+4V (wt%) (9.1)
The calculated Ms was found to be 330oC for 100Cr6WV and 184oC for 100Cr6. With a
known martensite start temperature and final end-of-quench temperature, the marten-
site content can be estimated using the Koistinen-Marburger model:
1 - Vα′ = exp -[0.011(Ms - TQ)] (9.2)
where TQ, the temperature below the Ms to which the steel is quenched, was taken as
20oC for both 100Cr6WV and 100Cr6. The martensite volume fraction was calculated
as 0.84 for 100Cr6 and 0.97 for 100Cr6WV. As a consequence, the retained austenite
content is 0.16 for 100Cr6 and 0.03 for 100Cr6WV. Such a small fraction of retained
austenite as that seen for 100Cr6WV can be beneficial with respect to improving rolling
contact fatigue life [5].
Figure 9.18: TTT diagram calculated with MUCG83, with Ms calculated using Oshida’s
model, for 100Cr6 and 100Cr6WV.
Using the Fischer model incorporated into MatCalc, described in Part II, Chapter 1
201
of this thesis, the simultaneous modelling of microstructural and hydrogen trapping
evolution during the primary heat treatment of 100Cr6WV was carried out, quantifying
the hydrogen concentrations trapped at each trapping site and the resulting total free
diffusible hydrogen during the corresponding microstructural evolution.
To understand the flux of hydrogen atoms within a certain matrix, we must first under-
stand the influence of the microstructural features present, including the distribution of
lattice defects, as well as their interactions with hydrogen. The initial hydrogen content
is assumed to be 0.01 ppmw. The mean dislocation density immediately after austeni-
tisation and after final quenching to martensite is assumed to be 1011 m-2 and 1015 m-2
respectively. The grain size is assumed to be 600 µm, as found for the experimental cast
of 100Cr6WV.
As the trapping mechanisms and sites are not clearly established, we use the charac-
teristic trapping energies of hydrogen ∆EH to dislocations of 20.6 kJ/mol and to grain
boundaries of 58.6 kJ/mol as suggested by Gaude-Fugarolas ( [87] after [253]). These
energies were found to describe experimental results with exceptional accuracy [87].
The calculations are performed with a single trapping energy value for vanadium car-
bide and tungsten carbide of 30 kJ/mol [25]. In reality, it is unlikely that tungsten
carbide is equal in its trapping energy to vanadium carbide, however, given the lack of
any trapping energy in the literature for tungsten carbide, this assumption was taken.
To take the size-dependence of trapping energy for precipitates into account, the mean
radius of the precipitates is related to the binding energies, formulated in accordance
with the size-dependence found for titanium carbides [87]:
∆EH−carbide = 30 + 7 · 107 · rmean(carbide) (9.3)
Figures 9.19 and 9.20 show the modelled redistribution of hydrogen into the various
trapping sites during the heat treatment schedule shown in Figure 9.15. During heat
treatment, the hydrogen atoms are initially trapped on the coarse carbides, dislocations
and grain boundaries present after spheroidisation. As the steel is heated up to 1300oC,
the diffusible hydrogen content increases as hydrogen is liberated from dislocations and
coarse carbides as they are dissolved. Upon quenching, the hydrogen partitions from
the free interstitial sites towards the dislocations nucleated during quenching and the
fine vanadium and tungsten carbides precipitated. As expected, the fraction of dif-
fusible hydrogen reaches a minimum when the maximum phase fraction of vanadium
and tungsten carbides is achieved. At the end of the heat treatment, the majority of the
remaining hydrogen is trapped on dislocations created upon quenching to martensite.
The binding enthalpies for tungsten carbides of varying size in iron will be both modelled
using density-functional-theory and experimentally measured using thermal-desorption
analysis of a Fe-C-W system in future work. However, despite the trapping energy
202
4000 50001000 2000 3000
Figure 9.19: Free diffusible hydrogen evolution during the primary heat treatment of
100Cr6WV, as shown in Figure 9.16, with an initial hydrogen content of 0.01 ppmw (5.55×10-7
molar fraction)
Figure 9.20: Total trapped hydrogen evolution during primary heat treatment of 100Cr6WV,
as shown in Figure 9.16 with an initial hydrogen content of 0.01 ppmw (5.55×10-7 molar
fraction)
assumptions for tungsten carbides, the modelling indicates the significant benefit of the
vanadium carbides with respect to hydrogen trapping capacity in comparison to the
contributions of grain boundaries and dislocations.
203
Microscopy
As shown in Figures 9.21 and 9.22, large around 20 µm primary carbides were iden-
tified post-homogenisation, the EDX analyses of which are shown in 9.22 and Figure
9.23. These MC grain boundary primary carbides appear semi-continuous, and are of
the composition shown in Figure 9.23. Rod shaped M2C carbides are also seen dis-
tributed intermittently across grain boundaries, as shown in EDX spot 2 of Figure 9.22.
The compositions of these primary carbides contradict the modelled compositions from
Scheil-Gulliver calculations. Such large carbides are detrimental in regard to rolling con-
tact fatigue and so it would be preferable to dissolve/break these up during proceeding
hot-working. In addition to these primary carbides, secondary carbides can be seen dis-
tributed throughout the interior of the grains. Post homogenisation, the 600 µm grains
contain mostly ferritic centres surrounded by pearlite. Such large grains are detrimen-
tal to rolling contact fatigue life [5], and so it would be desirable to recystallise during
the proceeding hot-working process typical of industrial bearing production. Hardness
was evaluated using ten 30 kg indentations radially across the sample. Hardness was
measured to be 535 Hv within one standard deviation. No significant difference was
observed in hardness across the sample’s radius.
Figure 9.21: Optical micrographs showing the as-homogenised structure of 100Cr6WV.
The spheroidised microstructure of 100Cr6WV is shown in Figures 9.24, 9.25, 9.26 and
9.27. Post-spheroidisation, the hardness was found to be 290 Hv using 20 kg indents with
a standard deviation of 3.1, and shows a mix of widmanstatten ferrite, allotriomorphic
ferrite and pearlite. Large homogeneously distributed un-spheroidised MC particles
with smaller spherical particles surrounding the particles are seen. Given the size of the
204
Figure 9.22: SEM Micrographs of as-homogenised 100Cr6WV, showing a large primary car-
bides along a grain boundary, with spots used for EDX analysis on: 1 - MC carbide, 2 - M2C,
3 - Bulk with small homogeneously distributed carbides.
Figure 9.23: EDX results for spots shown in Figure 9.22.
particles with respect to the limitations of EDX, X-ray diffraction was carried out on
the spheroidised microstructure to provide an accurate phase identification, the results
of which are shown in Figure 9.28. A continuous scan was taken from 25o to 85o with
a step Size of 0.04◦ and time of 8 seconds using Cu-Kα radiation. The generator was
set to 40 mA at 40 kV. A weighted profile fitting factor of 0.3 was achieved, indicating
a good fit of the data to the reference datasets.
The larger particles shown in Figure 9.26 appear to be that of M2C with the smaller
particles that of MC carbides, small enough in phase fraction to be undetectable by
XRD. This is contrary to that predicted from the modelling with respect to the M2C
carbide and as such, the stability of the carbide is unknown.
Post-spheroidisation, dilatometry was carried out on 4 mm diameter by 12 mm length
samples according to the heat treatment schedule shown in Figure 9.16. The resulting
microstructure is shown in Figures 9.29 and 9.30. The microstructure is fully marten-
sitic with no identifiable retained austenite, concurring with the modelling. The M2C
205
Figure 9.24: Optical Micrograph of as-spheroidised 100Cr6WV, showing large 600µm grains.
Figure 9.25: Optical Micrograph of as-spheroidised 100Cr6WV, showing areas of pearlite
carbides have not been fully dissolved during heat treatment unlike the smaller MC
carbides which were not observed by SEM and thus appear to have been dissolved or at
least partially dissolved and are consequently of a nanometre size. However, due to the
detrimental presence of continous eutectic M2C and MC carbides decorating the grain
boundaries, the steel is annapropriate for application for bearings [71]. Nonetheless, the
eutectic carbides have been reduced in volume by a mean of 30% during the swaging
and heat treatment proceeding spheroidisation, suggesting that full dissolution could be
achieved, perhaps with a longer hold at 1300◦C. on either side of the grain boundary
206
Figure 9.26: SEM Micrograph of as-spheroidised 100Cr6WV, showing the large grain bound-
ary eutectic carbides and smaller homogeneously distributed secondary carbides.
Figure 9.27: SEM Micrograph of as-spheroidised 100Cr6WV, showing the larger iron and
tungsten-rich MC carbides (darker) and smaller homogeneously distributed vanadium-rich
carbides (lighter).
eutectic carbides, is a carbide-depleted, tungsten-rich region around 5 µm in width,
beyond which, iron and tungsten-rich MC carbides alongside nanosized vanadium-rich
secondary carbides have precipitated, smaller than those observed in the spheroidised
structure, indicating partial dissolution throughout the structure. The more complete
207
ferite
ferite
2 2M CM C
Figure 9.28: X-ray diffraction data of as-spheroidised 100Cr6WV
dissolution seen in the regions surrounding the grain boundary eutectic carbides indi-
cates carbon-depletion, which would encourage the dissolution of carbides in this region
as evidenced in the carbon depedence of carbide stability shown in Figure 9.1. Such
carbon depletion is expected given the significant volume of carbon-rich grain boundary
eutectic carbides.
The hardness of the heat treated and un-tempered 100Cr6WV was found to be 843 HV.
This is typical for a bearing steel. However, given the large grain size and only partially
dissolved carbides, a more refined grain structure (permissible through recrystallisation
during hot-working) with the proposed around 10 nm sized VC and WC carbides (per-
missible with a longer dissolution hold time) would result in a greater hardness and
strength than that observed here. As such, a longer/higher temperature dissolution
step should be attempted in future work to study whether full carbide dissolution can
be achieved. For such optimised materials, hydrogen charging and thermal desorption
analysis should be conducted to assess the trapping efficiency of the observed secondary
phases. In its current state, 100Cr6WV showed little to no significant trapping carac-
teristics.
208
Figure 9.29: Optical micrograph of heat-treated 100Cr6WV
Figure 9.30: Micrographs of the final heat-treated structure of 100Cr6WV showing large pri-
mary tungsten-rich carbides surrounded by a martensitic structure containing small partially-
dissolved M6C3 carbides.
Summary
The parameters for the thermokinetic modelling of 100Cr6WV proved to be ineffective in
accurately modelling the microstructural evolution during heat treatment and as such,
209
further dilatometry and characterisation work is required to optimise the heat treatment
to achieve the desired microstructure.
100Cr6WV forms grain boundary MC and M2C eutectic carbides with smaller secondary
solid-state carbides precipitated in the grains. The final heat treated material was of high
strength, with a hardness upwards of 800 Hv, however, the continuous network of grain
boundary carbides are expected to be of detriment to RCF life and as such, an optimised
thermomechanical treatment, breaking up these carbides during forging, is necessary to
attempt to fully dissolve or limit any large carbides and reprecipitate nanosized carbides
for hydrogen trapping. The observed M2C carbides have a fishbone like structure,
but with a ragged boundary that does not clearly outline the interface between the
matrix and the eutectic pool, suggesting a limited coupling between eutectic phases.
It has been observed that MC carbides favour iron and tungsten, whilst M2C favours
tungsten. Vanadium is suspected to dominate the composition of M4C3, precipitated as
nanocarbides.
It is hypothesised that during solidification, dendrites form surrounded by a more or
less continuous interdendritic network of eutectic carbides. The residual interdendritic
liquid, enriched with carbon and carbide former elements, decomposes through differ-
ent eutectic reactions as it moves down a eutectic trough, leading to the formation of
M2C and MC. As such, the resultant microstructure contains a continuous network of
interdendritic carbides.
The heat treatment schedule of 100Cr6WV should be optimised in future work to ac-
quire the desired microstructure, A cost-benefit analysis will need to be carried out
for 100Cr6WV, once hydrogen trapping behaviour of tungsten carbides has been estab-
lished, to develop an optimised composition for use as a bearing steel. Alternatively,
a new composition based solely on the replacement of vanadium with tungsten by its
equivalent cost in 100Cr6+0.5V could be developed. This will better affirm the appeal
of tungsten carbide both with respect to its contributions to strengthening and hydro-
gen trapping. It was the decision of the industrial sponsor to discontinue the work on
100Cr6WV to prioritise more appealing steel concepts.
9.3 Offshore Flexpipe X steel
A novel nanostructured martensitic steel has been designed, Flexpipe X, the composition
of which is shown in Table 9.1, consisting of fine tempered martensite decorated with
nanosized vanadium carbides optimised for hydrogen trapping, providing a suitable
hydrogen embrittlement resistant steel for application in offshore flexible pipe systems.
This section compares the hydrogen trapping capacity of two heat treatment variants
of Flexpipe X against a solutionised benchmark sample.
210
Flexpipe steel is utilised within the flexible piping system of offshore oil rigs. Given
its required high strength and exposure to hydrogen-rich environments, hydrogen em-
brittlement resistant high strength steels are of great desire for application in these
components. Such steels can demand a range of strengths depending on the specific
application. As such, a number of heat treatment options are presented in this thesis
for various yield strengths by utilising untempered to highly tempered variants.
9.3.1 Thermodynamic and kinetic modelling
The specifications for the primary heat treatment, due to the current industrial pro-
duction line of the supplier, was to have a maximum of two furnaces, in series, with a
maximum total treatment time of 6 minutes between these furnaces. As shown in Figure
9.31 and Table 9.2, heat treatment options A, B and C consist of an initial heat up to
1100oC, for which the heating rate is 100oC·s-1 as used in the established production
line. The steel is held for two minutes at 1100oC to dissolve all secondary carbides. This
step also ensures the steel is fully austenitised. The steel is then quenched at -20oC·s-1
to room temperature to form martensite. The second step consists of a heat up at
100oC·s-1 to a tempering temperature of 680oC and held for 4 minutes in option A, 3
minutes in option B and 1 minute in option C. This tempering temperature is just be-
low the transition temperature for ferrite to austenite, see Figure 9.32, chosen to retain
the martensitic microstructure but provide the maximum growth rate of vanadium car-
bides during the hold without reaustenitisation. Flexpipe X shows the most significant
dependence of simulated vanadium carbide evolution on the most basic precipitation
parameters. As such, this will be used as an example of the most basic optimisation
methods available for kinetic modelling of carbide evolution in mid to high carbon steels
using MatCalc. Figure 9.33 presents the kinetic simulation results of the primary heat
treatment, assuming homogeneous precipitation of all carbides with nucleation coeffi-
cients of 1.0, incubation time constants of 1.0, and a precipitation domain of austenite
and ferrite above and below 800◦C respectively. Dislocation density was assumed to be
1011 m−2 for both ferrite and austenite. All precipitate species were assumed to have
an ortho-equilibrium composition.
Table 9.2: Hold times used to precipitate different sizes of vanadium carbides (cf. Fig. 9.31)
Option Hold / s
A 240
B 180
C 60
As shown in Figure 9.33, the maximum 4 minute hold permissible for the production line
211
680
120s1100
Figure 9.31: Heat treatment steps for Flexpipe X
Figure 9.32: Equilibrium diagram for Flexpipe X.
appears to be insufficient to grow the vanadium carbides to the desired 10 nm diameter
for hydrogen trapping. However, TEM for this heat treatment indicated that vanadium
carbides had grown much larger, around 20 nm in diameter, as shown in Figure 9.35.
A heat treatment of 3 minutes at 680◦C produced carbides 15 nm in diameter, as
shown in 9.36. This data is used to optimise kinetics parameters in order to replicate
212
the microstructure observed. Figure 9.34 presents the results of the optimised kinetic
simulations. All that was changed was the number of nucleation sites by altering the
dislocation density in ferrite/martensite to 1015 m−2.
Figure 9.33: Unoptimised kinetic simulations for the microstructural evolution of Flexpipe
X, showing the evolution of phase fractions, mean radius and number density of precipitates
during primary heat treatment.
213
Figure 9.34: Optimised kinetic simulations for the microstructural evolution of Flexpipe X,
showing the evolution of phase fractions, mean radius and number density of precipitates
during primary heat treatment.
9.3.2 Microscopy
TEM foil samples were electropolished using 5% perchloric acid, 25% glycerol and 70%
ethanol by a Struers Tenupol 5 electropolisher. A voltage of 35 V at 8oC resulted in
samples suitable for analysis by TEM. The hardness for treatment option A was found
to be 413 Hv20. Option B provided a hardness of 472 Hv20. The solutionised and
quenched state of Flexpipe X had a hardness of 679 Hv20.
TEM and scanning transmission electron microscopy (STEM) techniques were used to
affirm the presence and distribution of vanadium carbide and cementite, and quantify the
refinement of the martensite structure. Figure 9.35 shows a bright and dark field image
of the same area of the sample treated to option A, indicating the martensitic structure
and high number density of carbides. The carbides are all larger than the optimum
10 nm desired for hydrogen trapping. This was also the case for option B despite its
lower hold time, as shown in Figure 9.36. As such, further work is needed to optimise
Flexpipe for hydrogen trapping using a lower temperature and or hold time. A higher
dislocation density is preferable with respect to both hydrogen trapping and strength,
as the strength of martensite is proportional to the dislocation density. As such, the
hold times and temperatures should be optimised for each individual application to
ensure maximum dislocation and vanadium carbide number density permissible for the
respective strength limits.
214
Figure 9.35: 200keV Transmission electron microscopy bright field image of Flexpipe X option
A.
Figure 9.36: 200keV Transmission electron microscopy dark field image of Flexpipe X option
B.
A solutionised sample of Flexpipe X, treated at 1100◦C for 2 minutes, as for the other
Flexpipe X samples tested prior to the 680◦C hold, evidencing full dissolution of any
carbides present from casting. This sample was produced to represent a benchmark
against which the efficiency of vanadium carbides with respect to hydrogen trapping
efficiency could be compared for the carbide containing variants.
215
9.3.3 Hydrogen charging and thermal desorption analysis
All three states of Flexpipe X were hydrogen charged and subjected to TDA. Each
sample was hydrogen charged individually for 72 hours at a surface current density
of 1 mA cm-2. A Carbolite tubular furnace heated up the specimens at a rate of 25
K/hour during TDA analysis. The obtained TDA results are shown in Figure 9.37.
As shown, The trapped hydrogen contents are very similar for all treatments, with
only small shifts in peak desorption rate temperatures. The solutionised and quenched
treatment has trapped the most hydrogen, on its higher number of dislocations. Option
A’s peak desorption temperature is shifted to a lower temperature in comparison to
both solutionised and Option B, and trapped the least hydrogen, most likely due to the
reduced dislocation density and greater growth of vanadium carbides to lesser efficient
sizes for trapping, an effect evidenced in Part II of this thesis, showing the reduction
in efficiency beyond 10 nm diameter. The shift in peak desorption rate to the left is
likely due to the change in coherency for the cementite particles. For the same reason,
it is believed that option B has a reduced total capacity due to a lower dislocation
density, but a higher peak desorption temperature in comparison to option A due to
more coherent carbides. Further work is required to justify this, however.
Figure 9.37: Hydrogen desorption rate for 3 steel samples after charging for 72 hours at a
current density of 10 mA cm-2. Samples were left to desorb any diffusible hydrogen at room
temperature for 4 days. A heating rate of 25◦C·s−1 was used.
216
9.3.4 Summary
A vanadium nanocarbide strengthened flexpipe steel has been developed for improved
hydrogen trapping capacity within the limitations of the supplier’s established pro-
duction line. Thermokinetic modelling, validated by microscopy, was used to design
a heat treatment in order to produce a tempered martensitic matrix with homoge-
neously distributed vanadium carbides of 10 nm mean radius. The initial prescribed
heat treatments were unsuccessful with respect to producing the size of carbides de-
sired. However, kinetics have been optimised in order to continue the work. TDA found
that such carbides may trap hydrogen but to such an insignificant degree that the re-
duced dislocation density, resulting from tempering during treatment, in comparison
to as-quenched, resulted in a reduced total trapping capacity. The optimised kinetics
suggest that a treatment similar to those prescribed but with a hold of 50 seconds at
680◦C will provide the optimum carbide distribution.
9.4 Case carburised Grade 159V steel
A novel nanostructured case hardened martensitic steel has been designed, Grade 159V,
the composition of which is shown in Table 9.1, based on an established commercial
material, Grade 159. Grade 159V consists of a softer low carbon martensite core and
a hard high carbon martensite case, both decorated with nanosized vanadium carbides
optimised for hydrogen trapping.
Case carburised steels afford a bearing manufacturer the enhanced ductility and tough-
ness of a low carbon core with a high strength, wear resistant high carbon case. As
such, bearings are commonly final machined prior to carburisation. Such carburisa-
tion treatments are conducted at temperatures at which the entirety of the steel is
solutionised, including the surface, which when exposed to the carbon containing atmo-
sphere, such as carbon monoxide, will inherit carbon and increase the driving force for
carbide formation. Grade 159V and its accompanying heat treatment were designed for
manufacture within the limits of the supplier’s established production line. Thus, the
desired microstructure and corresponding heat treatment must ensure no large carbides
are formed in the bulk or case, which could act as crack initiation sites during operation
but vanadium nanocarbides are precipitated homogeneously throughout the structure.
The proposed heat treatment follows a three stage schedule, shown in Figure 9.38. The
initial carburisation stage is conducted at 1000◦C for a minimum of 24 hours, which is
depth dependent. At this stage, the steel is a solid solution. Following carburisation,
the steel is quenched to martensite in oil at a rate of 5◦C·s−1 or greater and reheated at
0.1◦C·s−1 to the austenitisation temperature of 800◦C, where it is held for 30 minutes
to recystallise, reaustenitise and precipitate the desired vanadium nanocarbides. The
217
steel is then quenched as before, and tempered at 200◦C for 2 hours to grow temper
cementite and increase ductility.
1000
0.1 °C/s
Figure 9.38: Heat treatment schedule of Grade 159V
As discussed in the literature review, bearing failures commonly initiate close to the
surface, and the initiation time of such failures are drastically reduced by hydrogen.
The carburised surface of a case hardened bearing, where this failure will occur, is
commonly very shallow, depending on carburisation parameters. As such, it would be
ideal to trap hydrogen both in the core of the bearing and the case. The benefit of
trapping in the core is hydrogen is drawn away from the damage region, preventing any
detrimental interactions. Additionally, given the shallow, lesser volume of the carburised
layer, it is more advantageous to optimise hydrogen trapping in the core rather than
the case of the bearing, to achieve the highest trapping capacity. As such, the design of
Grade 159 was conducted to optimise trapping capacity through trapping both in the
case and core, but optimised for the core with respect to vanadium carbide distribution
and size.
9.4.1 Thermodynamic and kinetic modelling
As evidenced by the equilibrium diagrams of Figures 9.39 and 9.40, the carburisation
temperature of 1000◦C is sufficient for full dissolution both in the core and the surface.
Noting that the microstructure post-carburisation is that of a carbide-free martensite,
Figures 9.41 9.42 present the microstructural evolution of the core and carburised surface
respectively during the second stage austenitisation. The final microstructure of the core
and case consists of a martensite matrix with 1022m−3 vanadium carbides of 9 and 12
nm mean diameter respectively.
218
Figure 9.39: Equilibrium diagram for the low carbon core of Grade 159V.
Figure 9.40: Equilibrium diagram for carburised Grade 159V.
219
Figure 9.41: Kinetic simulations for the microstructural evolution of the low carbon core of
Grade 159V, showing the evolution of (a) phase fractions (b) mean radius of precipitates and
(c) number density of precipitates, during primary austenitisation (d).
220
Figure 9.42: Kinetic simulations for the microstructural evolution of the carburised surface of
Grade 159V, showing the evolution of (a) phase fractions (b) mean radius of precipitates and
(c) number density of precipitates, during primary austenitisation (d).
221
9.4.2 Materials
Two 65 gram experimental casts of Grade 159V were produced using electric arc melting
to the composition shown in Table 9.1, but with one cast at a carbon content of 1.1
wt%, the other at 0.18 wt%. The higher carbon content cast was produced in order to
represent the outer surface/case of a bearing subjected to the designed carburisation,
as carburisation facilities were unavailable. Following homogenisation, the dillatometry
samples were sealed in glass tubes backfilled with argon and subjected to the simulated
carburisation hold at 1000◦C in a Carbolite furnace. The second stage austenitisation
was conducted in a THERMECMASTOR-Z dilatometer.
9.4.3 Microscopy
The TEM foil samples were electropolished using 5% perchloric acid, 25% glycerol and
70% ethanol by a Struers Tenupol 5 electropolisher. A voltage of 40 V at 10oC resulted
in samples suitable for analysis by TEM. The hardness for the core was found to be
723 HV30. The surface/case provided a hardness of 855 HV30. TEM and scanning
transmission electron microscopy (STEM) techniques were used to affirm the presence
and distribution of vanadium carbide. Figures 9.43 and 9.44 show two bright field
images of the heat treated core and case samples respectively. Both show a martensitic
structure and high number density of vanadium carbides around 10 and 15 nm diameter
for the core and case respectively. The EDX results shown in Figure 9.43 validate the
presence of vanadium carbide.
Figure 9.43: 200keV Transmission electron microscopy bright field image of Grade 159V core,
showing an EDX map of the same area.
222
Figure 9.44: 200keV Transmission electron microscopy bright field image of Grade 159V
surface/case.
9.4.4 Hydrogen charging and thermal desorption analysis
An 8 mm diameter, 10 mm length cylinder of the heat treated low carbon core of Grade
159V was hydrogen charged and subjected to TDA along with a sample of heat treated
martensitic 100Cr6, austenitised at 830◦C for 30 minutes and then quenched to room
temperature. The Grade 159V and 100Cr6 samples were hydrogen charged individually
for 72 hours at a surface current density of 1 mA cm-2 and 10 mA cm-2 respectively. The
lower current density used for Grade 159V was chosen as greater currents were observed
to produce corrosion pitting on the sample surface. A Carbolite tubular furnace heated
up the specimens at a rate of 100 K/hour during analysis.
The TDA results are shown in Figure 9.45. Grade 159V is seen to trap more hydrogen
than 100Cr6. Although the difference is not significant, due to the lower dislocation den-
sity of the Grade 159V one would expect a reduction in trapping capacity. As such, the
increased trapping capacity seen is evidence of the contribution of vanadium carbides.
Notably, it is unlikely that the Grade 159V sample is saturated given the low current
used during charging. The 100Cr6 sample is considered saturated as charging beyond
72 hours at 10 mA cm-2 was found to not result in any change with respect to total hy-
drogen content after 5 days at room temperature. Further work is required to establish
the full saturated trapping capacity of Grade 159V core and case microstructures.
223
Figure 9.45: Hydrogen desorption rate for Grade 159V and 100Cr6 samples after charging for
72 hours at a current density of 1 and 10 mA·cm−2 respectively. Samples were left to desorb
any diffusible hydrogen at room temperature for 5 days. A heating rate of 100◦C·s−1 was
used during TDA.
9.4.5 Summary
A vanadium nanocarbide strengthened case carburised steel has been developed for im-
proved hydrogen trapping capacity. Thermokinetic modelling was conducted to simulate
the microstructural evolution of both the core and case of the bearing during carburisa-
tion and the proceeding austenitisation. The prescribed heat treatments were successful
in producing the modelled microstructure of homogeneously distributed vanadium car-
bides, with a mean radius close to 10 nm, within a martensitic matrix. TDA found
that the core structure produced improved hydrogen trapping in comparison to com-
mercial through hardened 100Cr6 despite its reduced dislocation density. TDA should
be conducted to assess the trapping capacity of the bearing case also. However, given
the virtually identical microstructure for both case and core, it is likely to show similar
trapping improvements.
9.5 Steel G5V
A novel nanostructured superbainitic steel has been designed, G5V, the composition of
which is shown in Table 9.1, consisting of fine platelets of ferrite within a carbon rich
austenite matrix, decorated with nanosized vanadium carbides optimised for hydrogen
trapping, providing a suitable hydrogen embrittlement resistant alternative to current
commercial bearing steels. This section systematically compares the hydrogen trap-
224
ping capacity of the novel nano-vanadium-carbide hardened superbainitic Steel G5V, a
carbide free superbainitic Steel G, 100Cr6 and 100Cr6+0.5V.
The trapping efficiency of vanadium carbides, and the resulting hydrogen embrittlement
resistance has been established previously in martensitic bearing steels [108]. However,
the combination of such trapping sites, coupled with a fine bainitic microstructure,
presents a promising means to enhance hydrogen embrittlement resistance in steels. Al-
though not currently implemented commercially, a nanostructured superbainitic bear-
ing steel has been proposed as an alternative to the market dominating martensitic
steels [90]. Superbainitic steels are of significant appeal with respect to hydrogen em-
brittlement resistance. It has been speculated that a softer lower-bainitic structure
of 100Cr6 bearing steel could provide a longer RCF life, in comparison to martensitic
counterparts when in the presence of hydrogen, due to the increased ductility and tough-
ness [5, 91]. The austenite content, once exceeding a percolation threshold, allows the
bearing to resist the penetration of hydrogen [329]. As such, engineering the austenite
content is critical to enhancing hydrogen embrittlement resistance.
Applying degassing treatments to hydrogenated steels containing high austenite contents
must be done with caution. There have been cases where typical degassing treatments
have failed to ensure total degassing in bearing steels of high retained austenite content,
due to the hydrogen diffusivity in austenite being up to a factor of 104 less than that
in ferrite and the high solubility of hydrogen in the austenite [330]. As such, retained
austenite can behave as a reservoir for hydrogen, gradually diffusing its contained hydro-
gen into the ferrite over time, enhanced by stress, thus increasing the mobile hydrogen
concentration within the steel [56].
Despite retained austenite’s apparent behaviour as a reservoir for hydrogen, thermal
desorption analysis (TDA) suggests that retained austenite might actually trap hydro-
gen either reversibly or irreversibly. It has been shown that for hydrogen trapped on
retained austenite, the TDA peak can vary from 873.15 K (600oC) with an activation
energy of around 55 kJ/mol in super duplex stainless steel and ferritic welds [44, 49]
to 583.15 K (310oC) with around 45 kJ/mol in 0.82C-0.48Mn-0.02Cr (wt%) steel [48].
Moreover, Perez-Escobar et al. found that retained austenite displayed a hydrogen des-
orption peak of 773.15 K (500◦C) in undeformed and uncharged 0.17C-1.6Mn-0.4Si-2.0Al
TRIP steel, claiming that the hydrogen was introduced during processing [331]. It has
been postulates that typical outgassing treatments are not able to egress all hydrogen
stored in bulky retained austenite due to the reservoir effect. Outgassing treatments at
higher temperatures could ensure total egress, however, higher outgassing temperatures
could lead to detrimental phase transformations. As such, retained austenite is of signif-
cant concern with respect to hydrogen embrittlement. Andreone and Murut’s research
on 0.42C-0.78Mn-0.35Si-0.80Cr-1.79Ni-0.33Mo (wt%) steel with varying amounts of re-
tained austenite showed that increases in retained austenite content leads to a higher
225
susceptibility to hydrogen embrittlement [56]. Notably, there is no literature on the
possibility of trapping hydrogen irreversibly within retained austenite, be it through an
incoherent interface or through microstructural traps precipitated within the austenite
itself.
Nanostructured superbainitic steel, containing significant levels of retained austenite, are
typically cementite-free, nano-structured steels with a fine dispersion of bainitic-ferrite
and carbon-enriched retained austenite which cluster in the form of sheaves. The ferrite
platelets are 20-40 nm in thickness and it is their small size which contributes to the
high strength of these steels. Notably, such semi-coherent ferrite-austenite interfaces
have been shown to trap hydrogen [314]. The high interface to volume ratio causes
interference with slip/twinning mechanisms, giving ultimate tensile strengths of up to
2.5 GPa, toughnesses of 30-40 MPa m1/2 and long fatigue lives [332, 333]. However,
there is very little literature on the response of superbainite to hydrogen.
Superbainite forms by a displacive mechanism through quenching from austenite to a
temperature below conventional bainite transformation temperatures and holding at
this temperature for a period of time necessary to induce the desired ferrite content.
The resulting microstructure is dependent upon both composition and thermomechan-
ical history. A lower temperature of transformation will give finer platelets of bainitic
ferrite and higher strength as a consequence [334]. In order to reach a low enough tem-
perature to form superbainite (ferrite lathes with an average thickness of <150 nm),
the martensite transformation start temperature must be significantly lower than that
typical of common martensitic bearing steels. This is achieved by adding additional
alloying elements, in particular carbon. The formation of vanadium carbides will re-
duce the matrix carbon content and as a consequence, increase both the bainite and
martensite start temperatures. As such, the bainite hold temperature is calculated
for the matrix composition proceeding the vanadium carbide formation desired in the
austenite. Notably, vanadium additions are observed to reduce both the martensite and
bainite start temperatures, as shown in table 9.3 for solid solutions of G5V with various
vanadium contents. These temperatures were calculated using the freeware programme
MUCG83 using an altered Ishida model [335]. As such, it is preferable to increase the
carbon content to strengthen the austenite, refining the proceeding bainite structure,
and provide enough carbon to form sufficient vanadium carbide volume fractions for
trapping hydrogen. To increase the driving force for vanadium carbide formation, one
could increase the vanadium content. However, the practical additions of carbon and
vanadium are limited by the stability of resulting carbides, which must be dissolved
during primary heat treatment so to allow reprecipitation to the optimum size of 10
nm for hydrogen trapping [62]. As such, a vanadium and carbon content was chosen to
maximise the carbon content in the austenite matrix prior to quenching and maximise
the volume fraction of vanadium carbides whilst still permitting their dissolution at
226
Table 9.3: Martensite and Bainite formation start temperatures of G5V as a function of
vanadium content.
V content (wt%) Ms (oC) Bs (oC)
0 130 386
0.4 110 366
0.5 105 361
0.85 87 344
temperatures below the melting point of the alloy. The melting point of G5V, 1350oC,
was calculated using the Scheil-Gulliver method. For Steel G5V, a carbon and vana-
dium content of 0.84 and 0.5 wt% respectively were chosen to meet these requirements,
taking consideration for the consequences to the relevant transformation temperatures,
as discussed below.
A lower transformation temperature, although providing a more refined bainite struc-
ture, will require longer transformation times [336]. Given the significant cost of long
hold times at temperature, a designer must find a compromise between lowering the tem-
perature to refine the microstructure and increasing the temperature to give a faster,
more economical transformation.
To understand the chosen composition of G5V, it is necessary to understand the fun-
damental steps in its heat treatment schedule and the corresponding objectives with
respect to microstructural evolution. Figure 9.46 presents the primary heat treatment
schedule for G5V and the microstructural evolution during the treatment, modelled us-
ing thermokinetics, as is discussed in the proceeding section. As shown, the vanadium
carbides are formed in the initial austenite matrix prior to quenching to the bainite
transformation temperature. This ensures a homogeneous distribution of vanadium
carbides thoughout the intended bainitic microstructure whilst potentially enhancing
the refinement of the bainitic structure through the nucleation and pinning of bainite
plates by the carbides. As previously discussed, once the carbides have been precipi-
tated to the desired size and distribution, the carbon content in the austenite matrix
must be sufficiently high to ensure the austenite is strong enough to form a superbainitic
structure upon quenching and holding at the chosen transformation temperature [337].
The composition of the austenite at this quench point was obtained from the kinetic cal-
culations and was used to calculate the proceeding transformation temperatures using
the freeware programme MUCG83, as is discussed later.
Silicon was added to G5V to retard the precipitation of cementite, as validated in earlier
work
227
9.5.1 Thermodynamic and kinetic modelling
As shown in Figure 9.46, the primary heat treatment consists of an initial heat up to
1300oC, for which the heating rate is not strict and should be carried out at the maximum
rate to limit grain growth. For these calculations, a heat up rate of 1oC·s-1 was used.
The steel is held for two minutes at 1300oC to dissolve all secondary carbides. This
step also ensures the steel is fully austenitised. The steel is then quenched at -10oC·s-1
to 830oC and held for 0.5 hours to precipitate a homogeneous distribution of 10 nm
vanadium carbides. Vanadium carbides were assumed to nucleate on both dislocations
and grain boundaries. Dislocation density post-quenching was taken as 1015 m-2 with
a grain size of 600 µm. The final treatment step is a quench at -10oC·s-1 to the bainite
transformation temperature of 250oC and held for 6 hours.
The partial quench from 1300oC to 830oC is done so to increase the dislocation density,
creating more nucleation sites for the proceeding precipitation of vanadium carbides in
the austenite grains, and to increase the driving force for the carbide formation, a result,
among others, of the supersaturation of carbon. Kinetic simulations, and validatory
microscopy, as is discussed later, indicate that G5V has 1.4 x 1021 m-3 carbides with
an average diameter of 10 nm, homogeneously distributed throughout the austenite at
the end of this hold. The martensite start temperature was found to be approximately
240 to 250oC, and as such, a bainite transformation temperature of 250oC was chosen.
Transformation temperatures were calculated using MUCG83 software. The resulting
TTT diagrams for G5V and G (a vanadium free variant) are shown in Figure 9.47. The
equilibrium phase diagram for G5V is shown in Figure 9.48.
Figure 9.46: Kinetic simulations for the microstructural evolution of G5V, showing the evo-
lution of (a) phase fractions (b) mean radius of precipitates and (c) number density of pre-
cipitates, during primary heat treatment (d).
If vanadium carbides were formed during the bainite transformation holding step, it is
thought that the carbides will form preferentially within the ferrite due to the reduced
228
Figure 9.47: TTT diagrams for G and G5V, including (a) the shear transformation curves
for bainite with varying vanadium content, indicating the delay in transformation initiation
with increasing content, and (b) the transformation temperatures of bainite and martensite
around the region chosen for G5V’s bainite transformation hold temperature.
Figure 9.48: Equilibrium diagram for G5V.
carbon solubility and high dislocation density. However, after quenching to the bainite
transformation temperature, the high density of dislocations in the austenite will provide
nucleation sites for vanadium carbide formation. Notably, the increased carbon content
within Cottrell atmospheres surrounding the dislocations will enhance carbide forma-
tion. As such, it is feasible that vanadium carbide could form in both phases despite the
higher carbon solubility in austenite. Through vanadium additions, one can increase the
supersaturation of vanadium within the phases and engineer an alloy to form vanadium
carbides above or below the austenitisation temperature depending on the desired fi-
nal microstructure and/or maximum temperature limitations during processing. Higher
229
vanadium contents result in higher dissolution temperatures of the resulting carbides.
To ensure that hydrogen trapping efficiency was homogeneous throughout the alloy, it
is preferrable to precipitate vanadium carbides prior to the bainitic transformation, and
as such, these carbides are formed within the austenite.
9.5.2 X-ray diffraction
The amount of retained austenite in the examined specimens was measured by X-ray
diffraction (XRD). 8 mm diameter and 3 mm long cylindrical samples were cut and
ground to a 0.5 µm finish then etched with a combination of (Butyl-)ethanol and Per-
chloric Acid for 20 s in order to relieve surface stress. The samples were step scanned
from a 2θ of 5 to 80o with a Philips PW1820 diffractometer using unfiltered Cu Kα
radiation. The step size was 0.015o, with a scan step time of 7.5 s. The diffractometer
was operated with generator setting of 40 mA current with 40 kV voltage. A divergence
slit of 0.5o, an anti-scatter slit of 0.5o, and a receiving slit of 0.2 mm were used to restrict
the beam size and the counts obtained. The volume fractions of ferrite and austenite
were obtained using the software High Score plus and the Rietveld refinement method
based on the ratio method for the intensities of four peaks, ferrite (200) and (211), and
austenite (220) and (311).
The bainite transformation hold time dictates the extent to which austenite transforms
to bainite and will thus affect the mechanical properties of the steel, such as hardness,
and the hydrogen diffusion rate, both of which increase with ferrite content [329].
Several different bainite transformation hold times were tested for G5V to evaluate the
rate of austenite to bainite transformation and its effects on hardness. The results of
XRD, as shown in Figure 9.49, indicate a hold time of 16 hours at 250oC resulted in an
austenite and ferrite content of 16 and 84% respectively. Based on the 111 austenite and
110 ferrite peaks, the austenite and ferrite lattice parameters are 3.61498 and 2.87414
A˚respectively. The austenite and ferrite was found to have a Nishiyama-Wasserman (N-
W) orientation relationship of (110)α//(111)γ, [001]α//[011¯]γ. The resulting hardness
was found to be 666 HV30. A hold time of 72 hours resulted in a hardness of 697 HV30.
This increase was attributed to the increased volume fraction of bainitic ferrite.
Vanadium carbide is of stoichiometric VC composition with a rocksalt-type B2 face
centred cubic structure [108, 314] with each octahedral site occupied by carbon atoms.
The lattice parameter a is around 0.416 nm. A cube to cube orientational relation in
austenite has been reported elsewhere [338]: (001)γ —— (001)VC - [100]γ —— [100]VC.
A Baker and Nutting orientation relationship has also been found in ferrite {100} ——
{100}VC, {011}α —— {001}VC. The molar phase fraction of vanadium carbide was
calculated to be 0.096.The lattice mismatch for the carbides within austenite and ferrite
was found to be 11.5 % and 14.5 % respectively.
230
Figure 9.49: X-ray diffraction data of G5V for 2θ of between 5o and 80o, showing a ferrite
and austenite content of 83.9% and 16.1% respectively. Based on the 111 austenite peak, the
austenite lattice parameter is 3.6153 angstroms. Based on the 110 ferrite peak, the ferrite
lattice parameter is 2.8748 angstroms.
9.5.3 Microscopy
The TEM foil samples were electropolished using 5% perchloric acid, 25% glycerol and
70% ethanol by a Struers Tenupol 5 electropolisher. A voltage of 40 V at 16oC resulted
in samples suitable for analysis by TEM.
The optical micrographs in Figure 9.50 show the spheroidised microstructure of G5V,
which showed large 600 µm grains of bainite with homogeneously distributed spheroidised
cementite particles. As shown in the SEM image of Figure 9.50, the cementite particles
have a mean diameter of 360 nm with an inter-carbide spacing of 700 nm. In addition,
small vanadium-rich carbides of less than 30 nm in diameter can be observed distributed
amongst the larger cementite particles. This data was used to optimise the thermoki-
netic simulations for the proceeding primary heat treatment with respect to the initial
microstructural state. The bulk hardness was found to be 232 HV20.
The spheroidised G5V samples proceeded to be heat treated to the primary heat treat-
ment schedule shown in Figure 9.46. As shown in Figures 9.51, 9.52 and 9.53, a fully
bainitic microstructure is achieved. Little or no prior austenite grain growth was ob-
served.
Figure 9.52 shows an SEM image of the same sample, affirming that the large spheroidised
carbides, present prior to heat treatment, have been dissolved during the initial temper-
231
Figure 9.50: Spheroidised Steel G5V, with (a) the spheroidisation treatment schedule used
for 100Cr6, 100Cr6+0.5V, G, G5V and G5V (HC), (b) optical micrograph showing the large
bainitic ferrite grain structure and (c) spheroidised cementite particles of spheroidised G5V,
etched in 2% Nital, and (d) back scattered diffraction SEM micrograph, taken at 20keV at
a working distance of 10 mm, showing the large homogeneously distributed cementite and
smaller vanadium carbides in spheroidised G5V, etched in 2 % Nital.
ature spike. The large dark blocks in Figure 9.52 are that of retained austenite. Such
retained austenite could be beneficial with respect to increasing toughness due to the
TRIP effect [339, 340]. However, large blocks may also behave as crack initiation sites
due to the formation of brittle martensite upon deformation [189].
TEM and scanning transmission electron microscopy (STEM) techniques were used to
affirm the presence and morphology of vanadium carbide and quantify the refinement
of the bainite structure. Figure 9.53 shows a bright and dark field image of the same
area of G5V, indicating the 114 ± 9 nm mean thickness of the austenite and bainite
plates. Figure 9.53 provides an additional bright field image for the quantification. A
finer structure is preferable with respect to both hydrogen trapping and strength, as the
strength of bainite is inversely proportional to the thickness of the ferrite plates. Figure
9.53 shows a STEM image and corresponding Energy-dispersive X-ray spectroscopy
(EDX) map and diffraction patterns for G5V, showing ferrite and austenite plates, and
the fine V4C3 particles precipitated throughout the microstructure.
232
Figure 9.51: Optical micrographs showing a) the fine bainite structure and large prior austen-
ite boundaries and b) absence of large cementite particles seen in spheroidised G5V, etched
in 2 % Nital.
The homogenous distribution of vanadium carbide in G5V is important so to ensure
homogeneous strength, toughness and hydrogen embrittlement resistance. The cuboidal
shape of the precipitates is likely due to the minimisation of interfacial energy during
precipitation in the austenite. The cubic carbides precipitate in the fcc austenite lattice
and so form an orientation relationship, ensuring increased coherency.
Steel G, having been heat treated identically to that of G5V, shows a significantly
more refined structure. This indicates that it is the vanadium addition which results
in the loss of refinement. Given that the vanadium carbides will form prior to bainite
transformation, the austenite matrix will have a reduced carbon and vanadium content.
According to thermokinetic calculations, at the austenitisation temperature post vana-
dium carbide formation, G5V’s austenite contains around 0.8 wt% carbon. Prior to
formation, all carbon is in solution, as it is in Steel G. Although this reduced carbon
content is small, it is suggested that this is the cause of the loss in refinement.
As such, a higher carbon variant of G5V was cast, G5V(HC), and heat treated identically
to that of G5V to validate this hypothesis, the microstructure for which is shown in
Figure 9.54. The carbon addition of 0.08 wt% was chosen to result in an identical
233
Figure 9.52: Back scattered diffraction SEM micrograph, taken at 20keV at a working distance
of 10 mm, showing the fine bainite structure and large prior austenite boundaries, absent of
the large cementite particles seen in spheroidised G5V, etched in 2 % Nital.
austenite carbon content to that of Steel G at the point of quenching to form bainite.
By doing so, the effects of austenite carbon content or vanadium carbides can be isolated
with respect to the refinement and strength of the steel. G5V(HC), in addition to a
higher carbon content, has a reduced vanadium content to replicate G5V’s vanadium
carbide stability, ensuring they are dissolved during the initial temperature spike of the
primary heat treatment. The volume fraction and distribution of vanadium carbides are
similar to that of G5V but the mean diameter is 8 nm. As the finer bainite structure seen
in G5V(HC) confirms, the carbon content is crucial in the refinement of superbainite,
with a mean bainite plate spacing of around 40 nm. The hardness of G5V and G5V(HC)
was found to be 666 Hv and 760 Hv respectively.
234
Figure 9.53: 200 keV Transmission electron microscopy images of G5V, showing (a) a bright
field image and (b) a dark field image of the bainitic structure, (c) a STEM image with an
EDX map of iron and vanadium, (d) a bright field image of vanadium carbides within a lath,
and diffraction patterns for the (e) bainitic ferrite and (f) austenite.
235
Figure 9.54: 200 keV bright field Transmission electron microscopy image of G5V (HC),
showing the refined superbainitic structure.
236
9.5.4 Hydrogen charging and thermal desorption analysis
Four materials were hydrogen charged and analysed; Steels G5V and G, heat treated
identically, martensitic 100Cr6 and martensitic 100Cr6+0.5V. Each sample was hydro-
gen charged individually for 72 hours at a surface current density of 1 mA cm-2. A
Carbolite tubular furnace heated up the specimens at a rate of 25 K/hour during anal-
ysis.
The obtained TDA results are shown in Figures 9.55 and 9.56. Figure 9.55 shows the
room temperature diffusible hydrogen egressing shortly after charging for both G5V
and 100Cr6. As shown, 100Cr6 contained more diffusible hydrogen then G5V, indi-
cating a higher fraction of trapped hydrogen in G5V. Provided the ingress rates are
equal amongst samples, the sum of diffusible and trapped hydrogen should be equal
for amongst all samples. It is necessary to differentiate between trapped and diffusible
hydrogen during charging, especially with respect to steels with significant austenite
contents due to the reservoir effect discussed previously. Hence, the samples were left
until all diffusible hydrogen had egressed at room temperature. Worthy of note, despite
the reduced diffusible hydrogen content in the G5V, the time to egress all of the dif-
fusible hydrogen is almost equal to that of 100Cr6. This is due to the described reservoir
effect.
Figure 9.55 shows the trapped hydrogen desorption rate as a function of temperature
for all samples tested, indicating that G5V (HC) traps the most hydrogen, followed
by G5V, both with a notable increase over 100Cr6 and Steel G, indicating that the
vanadium carbides provide successful hydrogen trapping sites in superbainite, increasing
the trapping capacity of the Steel. The fact the total hydrogen trapped for G5V was
only slightly smaller than G5V(HC) indicates that the refinement of the ferrite-austenite
interfaces has little influence on hydrogen trapping in comparison to that of vanadium
carbides. The refinement of these interfaces appears to broaden the desorption peak
however, likely providing deeper traps than that of vanadium carbide, but with a reduced
capacity. The higher temperature peak seen in the 100Cr6 desorption profile is that of
cementite, which is absent in all bainitic samples tested.
The peaks in desorption rate were attributed in each material to: trapping by vanadium
carbides, ferrite-austenite interfaces and dislocations in G5V (which, as shown in Table
3.2 are very similar), and trapping by dislocations and ferrite-austenite inrefaces in Steel
G.
237
Figure 9.55: Hydrogen desorption rate for Steel G5V at room temperature after charging for
72 hours at a current density of 1 mA·cm−2.
238
Figure 9.56: Hydrogen desorption rate for 3 steel samples after charging for 72 hours at a
current density of 1 mA·cm−2. Samples were left to desorb any diffusible hydrogen at room
temperature for 4 days.
239
9.5.5 Mechanical testing
Given the excellent hydrogen trapping characteristics of the superbainitic steels stud-
ied. Further investigations were conducted to establish the strength of the baseline
superbainitic Steel G. Dogbone tensile samples of heat treated Steel G were machined
according to ASTM E8/E8M, with a gauge width of 14.956 mm and sample thickness
of 4.999 mm. Before testing, the samples were instrumented with Cu-Ni alloy foil strain
gauges/sensors. The strain sensors were fixed to the centre of the gauge section of the
sample with a cyanoacrylate base cement after grinding the surface with 400 grit SiC
paper. The wires of the strain sensor were soldered to a small terminal, glued outside
the gauge section, and to the two 0.5 mm diameter tinned copper conductors of a sensor
cable that connected the strain sensor to the amplifier. The results are shown in Table
9.4.
Tensile Strength (GPa) Young’s Modulus (GPa) 0.2% Proof Stress (GPa)
1.87 175 1.62
Table 9.4: Tensile test data for Steel G
0 
200 
400 
600 
800 
1000 
1200 
1400 
1600 
1800 
2000 
0.00 0.50 1.00 1.50 2.00 2.50 3.00 
St
re
ss
 (M
Pa
) 
Strain (%) 
Figure 9.57: The mean tensile test data for three dogbone specimens of Steel G, indicating
the 0.2 % proof stress. Tests were conducted at Intertek - Manchester.
As shown, the material is of the high strength indicative of superbainite, with a 0.2%
proof stress and ultimate tensile strength of 1619 and 1867 MPa respectively. Given
the ductility, high strength and hydrogen embrittlement resistance, this steel and its
vanadium containing variants are currently under consideration for application as a
possible alternative to Flexpipe X, described previously.
240
9.5.6 Summary
Two vanadium nanocarbide strengthened superbainitic steels and their accompanying
heat treatments have been developed for high strength bearings with improved hydro-
gen trapping capacity. The prescribed heat treatments have significantly shorter bainitic
transformation times (6 hours) than typical of established bainitic steels with the same
degree of transformation, attributed to the reduced manganese content. Optimised
thermokinetic modelling, validated by microscopy, was used to design these heat treat-
ments in order to produce a refined superbainitic matrix with homogeneously distributed
vanadium carbides of 10 nm mean radius. TDA found that such carbides are the dom-
inant hydrogen trap, trapping in both the asutenite and the ferrite. All superbainitic
steels evaluated provided an improved hydrogen trapping capacity in comparison to the
commercially leading martensitic 100Cr6. It has been shown that the bainite struc-
ture can be refined by increasing the carbon content of the austenitic matrix prior to
quenching to the bainite transformation temperature. As such, the formation of carbides
prior to transformation will reduce the refinement, as such, a trade-off is established be-
tween refinement and carbide volume fraction. However, the refinement of the structure
results in only a slight increase in the trapping capacity of the steel, with the vana-
dium carbides being the predominant trapping species. Its contributions to strength are
significant however. Such carbide reinforced superbainitic steels could afford bearing
manufacturers a ductile yet strong steel with high hydrogen trapping capacity, which
during deformation, will strengthen through the TRIP effect. The work affirms that: it
is possible to conceive a steel, via carefully engineered trapping species, that combines
both high hardness and hydrogen trapping capacity, resulting in a grade suitable for
bearing applications, that thermodynamic and kinetic modelling can be optimised to
accurately predict the microstructural evolution of such steels and that in carbide free
nanostructured bainite, hydrogen trapping occurs at the ferrite/retained austenite in-
terfaces and at dislocations, but this is insignificant in comparison to the contribution
of vanadium carbides.
9.6 Nickel-based Alloy 600V
As stated in the literature review, nickel-based alloys have become a prominent mate-
rial in Gen-IV reactor designs. However, their through-life structural integrity suffers
due to the significant neutron radiation doses in such reactors. Neutron radiation pro-
duces two primary damage mechanisms: the displacement of atoms by neutron-atom
collisions and the creation of impurity atoms (primarily hydrogen and helium), leading
to embrittlement. Nano-sized vanadium carbides (VCs) have recently been shown to
reduce hydrogen embrittlement, via hydrogen trapping, in martensitic bearing steels. It
241
is theorised that these could be applied to nickel-based alloys, with VC providing resis-
tance to both radiation embrittlement mechanisms. In this work, a neutron radiation
resistant variant of Alloy 600 (a well established nickel-based nuclear alloy) has been
designed using thermodynamic and thermokinetic modelling. This alloy, and its par-
ent material, were cast, heat treated, and characterised with respect to microstructure
and hydrogen trapping capacity. As will be discussed, for the proposed nanostructured
alloy, the susceptibility to hydrogen embrittlement can be correlated with helium em-
brittlement susceptibility and matrix hardening rates. Thus providing an insight into
the overall irradiation resistance. Alloy 600 was selected as a base composition due to
its common use, both historically and in future reactor designs, and the wide range of
data available for it. In the proceeding two sections, the design of the novel vanadium
containing Alloy 600 variant, and its characterisation, both with respect to microscopy
and trapping efficiency, are presented respectively.
Alloy Ni Fe Cr Mo Co Ti Al Mnˆ Cˆ Nb+Ta V
X-750 70* 7 16 - 1.0ˆ 2.26 1.0ˆ 1.0 0.08 1.0 -
600 72* 8 16 - - - - 1.0 0.15 - -
718 53 Bal. 19 3 1.0ˆ 0.9 0.2ˆ 0.35 0.08 5.2 -
600V Bal. 8 14 - - - - - 0.14 - 1
625 Bal. 4.3 20.5 8.2 0.1 0.1 0.1 0.1 0.03 3.6 -
Table 9.5: Nominal compositions of commercial nickel-based alloys (with the exception of
Alloy 600V) in wt% (*min , ˆmax).
9.6.1 Thermodynamic and kinetic modelling
The first step of the proposed heat treatment is intended to dissolve any pre-existing
VCs. As previously stated, Industrial furnaces are often limited to ∼1300 ◦C [341].
As such, the chosen vanadium content must be such that complete dissolution of vana-
dium carbides is achievable below the maximum temperature of 1300◦C. Experience has
shown that a ∼100 ◦C buffer above the equilibrium dissolution temperature is sufficient
for industrially feasible dissolution times of the order of minutes. Hence, a maximum
vanadium content is chosen for which an equilibrium dissolution temperature of 1200
◦C is acquired.
Vanadium carbides were set to nucleate homogeneously within the austenitic matrix,
with full equilibrium assumed which in practice is rarely achieved. The optimum vana-
dium content was found by recording the dissolution temperature of the VC phase as a
function of the modified Alloy 600 vanadium content (Figure 9.58). A vanadium con-
tent of 1 wt% produced close to the ideal equilibrium dissolution temperature of 1200
◦C, and so was selected for further investigation. For the remainder of the thesis, this
vanadium containing variant is defined as Alloy 600V (Table 9.5).
242
Name of Treatment Treatment Conditions Predicted VC Radius
[nm]
600TT [342] 704 ◦C for 15 hours N/A
Solutionised 1300 ◦C for 5 minutes 0
Underaged 700 ◦C for 30 minutes ∼1
Aged 840 ◦C for 10 hours ∼5
Overaged 860 ◦C for 10 hours ∼6
Table 9.6: Alloy 600V heat treatments designed using MatCalc to produce a range of VC
precipitate sizes.
Having established the composition of Alloy 600V, heat treatments were developed in
order to control the size of the precipitating VCs. Heat treatments consisted of two
stages, as shown in Figure 9.59. The first is a dissolution hold to dissolve any VC
formed during casting (Table 9.6). This is proceeded by a quench to room temperature,
followed by heating to the precipitation temperature, and held for a given time, allowing
VCs to grow (Table 9.6).
The optimum VC size is likely to be of similar size to that found previously in this thesis
for bearing steels. As such, a heat treatment was developed to produce precipitates of 10
nm mean diameter. Two other heat treatments were created so a range of carbide sizes
could be investigated with respect to trapping efficiency to validate this (Table 9.6).
Figure 9.60 shows the results of the kinetic modelling for the aged heat treatment
(Table 9.6). The size, distribution and volume fraction of VC and M23C6 precipitates
are shown. The plots 9.60a and 9.60b display how the phase fraction and number density
of precipitates vary with time respectively. Both of these plots show a drop to zero for
the VC during the initial high temperature hold, predicting full solutionisation. The
phase fraction of M23C6 remains at zero for the rest of the heat treatment, as desired.
Plot 9.60d shows that the radius of VC steadily increases during the heat treatment to
a size of ∼5 nm.
243
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
400
600
800
1,000
1,200
1,400
Vanadium Content [wt%]
D
is
so
lu
ti
o
n
T
em
p
er
at
u
re
[◦
C
]
Figure 9.58: VC equilibrium dissolution temperature for Alloy 600V. The red line represents
the industrial oven temperature limit. The dotted red line represents the thermodynamic
temperature limit, to allow industrially feasible dissolution hold times.
0 500 1,000 1,500 2,000 2,500
1,300
700
25
(a)
(b)
Time [s]
T
em
p
er
at
u
re
[◦
C
]
Figure 9.59: An example heat treatment profile for Alloy 600V. Stage (a) is a dissolution hold
to remove any carbides formed during the Alloy’s thermomechanical history. Stage (b) is the
precipitation hold to allow VCs to nucleate and grow.
244
(a) (b)
(c) (d)
Figure 9.60: Thermokinetic results for the aged treatment (Table 9.6) of Alloy 600V. Data
for VC and M23C6 is plotted showing, (a) the temperature, (b) the phase fraction, (c) the
number density and (d) the mean radius of precipitates as a function of treatment time.
245
9.6.2 Microscopy
A solutionisation treatment (Table 9.6) was run to confirm the solutionisation step was
adequate to remove any carbides formed during casting. The aged and underaged treat-
ments (Table 9.6) were carried out, with quenching provided by helium. For SEM, a
FEI Nova NanoSEM working at 20 keV was used. An SEM image of Alloy 600TT (Ta-
ble 9.6) is shown in Figure 9.61. Semi-continuous micron-sized Cr-rich M23C6 carbides
are found at the grain boundaries, as evidenced by EDX (Figure 9.61).
Grain boundary carbides were not seen in Alloy 600V after the solutionised treatment,
indicating the treatment had successfully dissolved all M23C6 particles. This was also
found for the underaged sample. However, the aged sample displayed similar grain
boundary carbides to the Alloy 600TT sample, although rich in V as well as Cr (Fig-
ure 9.62). Although thermokinetic modelling did not predict M23C6 precipitate forma-
tion in any treatment, this result is potentially very useful. As discussed, Alloy 600
relies on intergranular carbides to provide SCC resistance [126]. An alloy with both
intragranular carbides for H and He trapping, and defect recombination, in addition to
intergranular carbides for SCC resistance could show significant improvements with re-
spect to commercial Alloy 600TT in corrosion susceptibility and radiation embrittlement
resistance.
For TEM, 0.3 mm slices were removed from the centre of the heat-treated dillatometry
samples. Following grinding on 800 grit SiC paper to 150 μm, these were punched to
form 3 mm diameter cylindrical foils. Further grinding was carried out using 1200 grit
SiC paper and cleaning with ethanol and acetone. The foil samples were electropolished
using 5 % perchloric acid, 25 % glycerol and 70% ethanol by a Struers Tenupol 5
electropolisher. A voltage of 22 V at 16 ◦C resulted in samples suitable for analysis by
TEM. An FEI Tecnai Osiris FEG-TEM at 200 keV was used for all TEM. EDX was
performed on select Alloy 600V samples in STEM mode to confirm the presence of VC.
TEM imaging of the solutionised sample (Figure 9.63a) showed no evidence of nano-sized
carbide formation. The precipitates formed by the underaged treatment (Figure 9.63b)
had a radius of approximately 2.5 nm, slightly larger than that predicted by thermoki-
netics (Table 9.6). These precipitates are distinctly closer in size and shape to those
found most effective in bearing steels [62]. The predicted size of VC precipitates for
the aged heat treatment was ∼5 nm (Table 9.6), however TEM imaging of the sample
(Figure 9.63c) showed that distinctly larger needle shaped precipitates formed of the
order of 100-1000 nm.
The size of the precipitates produced by the underaged sample were too small to resolve
with STEM EDX and so was carried out on the aged sample only. Figure 9.63d shows
that the rod-like precipitates contained increased levels of V. This region was also found
to have a slight depletion in Ni compared to the bulk. This suggests that some form
246
Figure 9.61: SEM analysis of Alloy 600TT after its industry standard thermal treatment. (a)
Micron sized carbides have developed at the grain boundaries. (b) EDX imaging showing a
high Cr concentration in these carbides.
of Ni(Cr,V) phase is developing. This precipitate morphology was seen by Ferrer et
al. [343] when observing the effects of heat treatments above 800◦C on Alloy 625 (Table
9.5), which has a close compositional similarity to Alloy 600V, with increased Nb content
instead of V. Ageing of Alloy 625 at 820 ◦C for 100 hours formed similarly long rod-like
precipitates of the order of a few microns. These were characterised as δ phase Ni3Nb.
The rods observed in Figure 9.63c are therefore likely to be δ phase Ni3V due to the
similarities of Nb and V with respect to kinetic behaviour.
A different region of the same sample displayed in Figure 9.64 was analysed via EDX.
Small spherical regions of V are observed which coincide with increased C concentration.
This provides good evidence that spherical VCs have formed in the aged Alloy 600V
sample. Their mean radius is approximately 20 nm, significantly larger than predicted
by thermokinetic modelling (Table 9.6). This finding shows similarities once again to
the work of Ferrer et al. [343] on Alloy 625. A heat treatment of 820 ◦C for 100 hours
produced δ phase Ni3Nb rods growing from NbCs. A similar effect could be occurring
in Alloy 600V with V precipitates instead of Nb. Further work is necessary to affirm
this.
Figure 9.65 shows the compositional map produced by a distinctly shorter rod-like pre-
cipitate found in the aged sample. The rod-like precipitate appeared to have a noticeable
increase in V and Cr composition but a distinct lack of C and Ni, and so are unlikely
to be VC. Several smaller spherical regions show slight increases in both V and Cr with
no depletion in Ni, but a lack of C. Again, these are possibly δ phase Ni3(Cr,V). The
area of increased C is within the vicinity of a small spherical region of very high V
247
Figure 9.62: EDX analysis of the aged Alloy 600V sample (Table 9.6). Chromium-rich pre-
cipitates have formed at the grain boundaries, similar to Alloy 600TT.
248
Figure 9.63: Bright field TEM images of Alloy 600V subject to heat treatments according to
Table 9.6. (a) Several dislocations are visible. Possible carbides may be present, however they
are few in number and very small in size. (b) Nano-sized VC have formed homogeneously.
Large VC rods are visible in (c), which contains increased V content to the bulk (d).
249
Figure 9.64: EDX mapping of the region shown in Figure 9.63 (c) of the aged Alloy 600V
sample. The C regions correlates very well with the V regions, suggesting the presence of
approximately 40 nm VCs.
concentration. This could be indicating that small spherical VC have actually formed
but were undetectable through TEM.
The data from the STEM EDX analysis in Figure 9.65 was analysed further using
Hyperspy analytical software. Principal component analysis is used to deconstruct the
original dataset and generate a much more compact set of components compared to the
initial reference, which expressed each spectrum independently. The first components of
this compact representation will contain all of the useful signal, allowing for the removal
of higher order noise components. The components extracted also contain information
on the correlation between different features in the spectra. This allows mapping based
on compounds to be generated as opposed to elemental mapping from conventional
EDX [344].
The Hyperspy analysis can be seen in Figure 9.66. The analysis suggested that Figure
9.66 (a) primarily consisted of Ti. Given that there is no Ti in the composition of
Alloy 600V (Table 9.5) this is likely to be the misinterpretation of a V signal, as the
characteristic Lα X-rays for Ti and V are 4.508 and 4.949 keV respectively. This same
region was identified as V in conventional EDX (Figure 9.65), supporting this claim.
Interestingly a diffuse region consisting of mainly C in Figure 9.66 (b) coincides with
the V rich region in Figure 9.66 (a), suggesting the formation of a spherical VC.
Figure 9.66 (d) identifies that the rod-like precipitate consists primarily of Fe, with small
contributions of Cr and V. This finding was rather unexpected as an Fe rich precipitate
was not predicted to form from thermokinetic modelling.
A Ni(Cr,V) phase was identified in Figure 9.66 (e), which is consistent with the findings
of conventional EDX analysis in Figure 9.65. Figure 9.66 (c) was identified as the
austenite phase.
250
Figure 9.65: STEM EDX of Alloy 600V subject to the overaged treatment. It can be seen
from this image that there are various regions with both an increase in V and Cr but depletion
in C, suggesting a Ni(Cr,V) phase has formed. The region with increased carbon corresponds
to a region of very high V concentration.
251
(a) (b)
(c) (d) (e)
Figure 9.66: Hyperspy analysis of Figure 9.65. (a) is a phase that is apparently rich in Ti,
however given that there is no Ti in the alloy it is likely that the analysis is misinterpreting
a V signal which is similar to Ti. (b) is a C rich phase and appears to coincide with the V
rich region in (a). The austenite phase is shown in (c). The rod-like precipitate (d) that was
believed to be VC was found to be largely comprised of Fe, with some Cr and V. A Ni(Cr,V)
phase is shown in (e).
252
9.6.3 X-ray diffraction
XRD was carried out on overaged Alloy 600V to confirm the presence of carbides. A
Bruker D8 DAVINCI diffractometer with unfiltered Cu Kα radiation was used, with a
step size of 0.02◦, step time of 20 s and angular 2θ range of 20 - 70◦. It was hoped
that XRD could be carried out on the underaged Alloy 600V sample to establish the
morphology of the precipitates found by TEM. However, the volume fraction of these
precipitates was too small for any detectable deviation in spectral signal compared to
solutionised 600V. STEM EDX had already confirmed the presence of vanadium in the
large rods found in the overaged Alloy 600V sample, however the morphology is yet to
be confirmed.
No diffraction data was found with respect to any morphology of VC in nickel. As
such, the solution would be to fit known data for other carbides, for example tungsten
carbide, to the spectra by modifying the lattice parameter of the data file. However,
no detectable signal was seen and so the structure of the precipitates was unable to be
investigated.
9.6.4 Thermal desorption analysis
Four cylinders (Table 9.7), 8 mm in width and 12 mm in length, were produced for solu-
tionised Alloy 600, Alloy 600TT, Alloy 600V aged and Alloy 600V underaged conditions.
All samples were charged at the same time with equal electrode lengths from the power
source to the sample to ensure equal charge distribution. Charging was carried out
at room temperature. TDA was carried out at the Ghent University Material Science
and Engineering Department to quantify the hydrogen trapping capacity of the studied
samples. The hydrogen content was detected by a mass spectrometer, which is under
constant vacuum. Calibration was done by a standard gaseous calibration method. The
carrier gas used during testing was high purity helium. The hydrogen is detected by
a mass spectrometer that measures current versus cycles. While calculating the total
trapped hydrogen, the background hydrogen reading is subtracted. The regular cycles
are converted to temperature by the data generated through a thermocouple in contact
with the sample. Each sample was held at room temperature for 3 days to allow all
diffusible hydrogen to egress prior to testing. Proceeding egress at room temperature,
and prior to the TDA of each sample, samples were stored in liquid nitrogen to ensure
all samples had identical egress times at room temperature. During TDA, the samples
were subject to a heating rate of 100 ◦C per hour up to a maximum of 900 ◦C to asses
the egress of trapped hydrogen.
The results are shown in Figure 9.67. Alloy 600TT provides hydrogen trapping sites
at its grain boundary chromium carbides and its dislocations. The solutionised Alloy
253
600 contained no grain boundary chromium carbides and trapped far less hydrogen,
suggesting that the chromium carbides are strong hydrogen trapping locations. This
is expected, as Alloy 600TT is highly susceptible to hydrogen-induced grain boundary
decohesion, which necessitates hydrogen conglomeration on the grain boundaries. The
dominant traps of the solution annealed Alloy 600 are that of its high density of dis-
locations, which are likely to be far greater than that of any other samples due to the
rapid quenching from its high temperature solid solution.
The underaged treatment of Alloy 600V trapped the most hydrogen of all samples tested.
This finding suggests that the nano-sized VCs, which are significantly smaller than for
the aged treatment, provide greatly improved trapping capacity. This sample contained
no grain boundary chromium carbides and would have a reduced dislocation density
in comparison to that of solution annealed Alloy 600. As such, the dominant trapping
mechanism is that of nano-sized VCs. This sample also had the highest temperature of
peak desorption rate. Given peak desoprtion rate temperatures are directly related to
the binding energy of the dominant trap, it can be inferred that VCs are the deepest
traps observed for the alloy conditions tested.
The aged treatment of Alloy 600V had the lowest hydrogen trapping capacity. Noting
that this treatment produced similar grain boundary chromium carbides to the Alloy
600TT sample and the grain size was fairly consistent amongst all the conditions, the
likely reasons behind this reduced capacity are as follows: a reduced dislocation den-
sity in comparison to solution annealed Alloy 600, that the large V-rich delta phase
Ni3(Cr,V) needles do not trap hydrogen and finally, that the V addition alters the
trapping behaviour of the grain boundary Cr-V rich M23C6 carbides, perhaps due to
a coherency change or a chemical effect. It is unknown whether hydrogen is trapped
within such carbides or on the interfaces. As such, further work is necessary to es-
tablish the mechanism behind this effect. Regardless, if V additions do reduce the
trapping capacity and binding energy of hydrogen on grain boundary Cr-rich carbides,
this could greatly inhibit the susceptibility to impurity conglomeration on these grain
boundaries, reducing the likelihood of hydrogen-induced grain boundary embrittlement.
As such, the combination of V-rich grain boundary carbides and intragranular vanadium
nanocarbides could significantly outperform Alloy 600TT with respect to hydrogen em-
brittlement resistance, and most likely He embrittlement, by inhibiting such impurities
accumulating on the grain boundaries.
254
Figure 9.67: The TDA plots for Alloy 600 and Alloy 600V. The underaged heat treatments
of Alloy 600V retained the most hydrogen, whereas the aged treatment retained the least.
255
Sample Temperature of Peak
Desorption Rate [◦C]
Hydrogen Trapped
[ppm]
600TT 171 3.01
600 (Solutionised) 168 2.20
600V (Underaged) 179 3.74
600V (Aged) 144 1.34
Table 9.7: TDA results for Alloy 600 and Alloy 600v samples. The temperature of peak
desorption rate is related to the binding energy of any traps within the samples.
9.6.5 Summary
A vanadium containing variant of Alloy 600 has been designed via thermodynamic and
thermokinetic modelling. The modelling of which has been validated by microscopy,
XRD and TDA. The underaged heat treatment produced what appeared to be homo-
geneously dispersed vanadium carbides with a radius of 2.5 nm. These were too small
to accurately characterise by EDX, but they have been identified as vanadium and car-
bon rich. Longer heat treatments at higher temperatures (the aged heat treatment)
produced needle shaped precipitates on the micron length scale. EDX identified that
these were composed of nickel, chromium and vanadium, a form of δ-phase Ni3(Cr,V).
Spherical vanadium carbides, approximately 20 nm in radius, were also observed. XRD
was unsuccessful in identifying the morphology of the precipitates produced by the un-
deraged and aged samples due to their low abundance and small size. TDA concluded
that the underaged treatment, with its nanosized vanadium carbides of Alloy 600V
provided improved hydrogen trapping capacity compared to all other tested variants
of Alloy 600 and 600V, including commercial Alloy 600TT. The results indicate that
vanadium carbides are the dominant traps with a higher binding energy than any other
traps in the samples tested, showing, for the first time, that vanadium carbides can
be utilised to improve the intragranular trapping capacity of nickel-based alloys such
as Alloy 600. Conversely, but of equal promise, is that of the aged treatment of Alloy
600V, containing larger nano-sized vanadium-rich δ phase Ni3(Cr,V) and grain bound-
ary vanadium polluted chromium-rich carbides, displayed a lower hydrogen trapping
capacity than Alloy 600TT. It is suggested that this is a result of vanadium pollution
of grain boundary chromium carbides inhibiting the trapping capacity of these car-
bides. Grain boundary carbides are utilised in Alloy 600TT to inhibit SCC. Hydrogen
conglomeration on grain boundaries is well known to induce decohesion and result in
premature failure. A reduced trapping capacity of these carbides would be beneficial in
resisting hydrogen-enhanced embrittlement by reducing the hydrogen concentration on
the grain boundaries. Finally, the results for the aged sample also indicates that δ-phase
Ni3(Cr,V), in its described state, does not trap hydrogen.
The aim of the VC precipitates was to provide impurity trapping and defect recombina-
256
tion sites. It has been shown that nano-sized vanadium carbides, 5 nm in diameter, in
Alloy 600V, leads to improved hydrogen trapping capacity compared to Alloy 600TT.
However, the heat treatment did not produce grain boundary chromium carbides, which
are important for providing stress corrosion cracking resistance. The next step would
be to develop a heat treatment to replicate these nano-sized vanadium carbides but also
produce grain boundary chromium carbides. The hydrogen trapping capacity and stress
corrosion cracking resistance will need to be assessed in comparison to Alloy 600TT.
It is also important to investigate the effects of other embrittlement mechanisms. He-
lium charging and desorption analysis can be carried out in a similar manner to hy-
drogen charging. An irradiation programme should be conducted, which could assess
the susceptibility to overall irradiation embrittlement of Alloy 600V. Although the more
essential work is to irradiate the alloys with neutrons to elucidate the overall effects of
trapping species, using irradiation techniques that avoid transmutation, such as proton
radiation, could isolate the role of secondary phases with respect to matrix hardening.
257
Part IV
Conclusions and future work
258
Chapter 10
Conclusions and future work
10.1 Microstructure-hydrogen interactions
The wide range of mechanisms reviewed and the somewhat conflicting evidence in the
literature, make it challenging to give an opinion on how the various microscopic mech-
anisms (HID, HELP, HESIV, etc.) combine to produce a macroscopic brittle fracture
response. There is sufficient evidence, although not proven experimentally, that in many
situations a combination of two or more mechanisms is required to explain the embrit-
tlement. Most frequently observed is the combination of HELP and HID, or HELP and
HESIV. It is understood that hydrogen trapped in steels does not contribute to the
degradation of the material properties with early experiments having verified that dif-
fusible hydrogen is responsible for the embrittlement process. The ratio between trapped
and diffusible hydrogen is highly dependent upon the material’s microstructure, includ-
ing key properties such as solubility, diffusivity and trapping behaviour. During the
manufacturing process, there are several options for hydrogen control. In the molten
state, a vacuum treatment can encourage degassing of otherwise solubilised gasses, in-
cluding gaseous oxides, nitrogen and indeed hydrogen. This can be enhanced through
the use of an inert gas-bubbling method, in order to further reduce the total hydro-
gen content in the melt. Once cast, materials can then be annealed at sufficiently low
temperatures to avoid side transformations whilst enhancing diffusion and thus out-gas
hydrogen. However, whilst such an approach can limit the total hydrogen content, it
does require long heat treatments at low temperatures, which are dependent on the dif-
fusion path, and are thus specific to part geometries. This approach is designed to avoid
so-called “intrinsic hydrogen”, which, whilst important, does not control hydrogen that
is introduced from other sources (sometimes referred to as “environmental hydrogen”).
In general, it is important to control the quantity of diffusible hydrogen within the mi-
crostructure in order to ensure its resistance to embrittlement, either through preventing
hydrogen ingress or ensuring an intrinsic resistance within the microstructure.
259
Whilst there has been considerable progress in recent years, stimulated both by ex-
perimental insight and by modelling at the atomic and continuum scales, there are a
number of key open questions that remain. One of the main issues is that there is still
considerable disagreement in the scientific literature concerning the underlying processes
that are responsible for hydrogen embrittlement, even in simple material systems. A
comprehensive understanding of hydrogen embrittlement is essential to underpin design
strategies for the next generation of ultra-high-strength steels and nickel-based alloys.
This will require the integration of a wide range of computational modelling schemes and
experimental techniques at different scales. The new understanding derived from these
studies might guide the development of new procedures for the design of microstructures
that are resistant to hydrogen embrittlement. Significant contributions have been made
in the field of imaging techniques for deuterium at the atomic scale with new advanced
APT techniques for imaging of deuterium in steels providing important insight in the
characterisation of traps. Such improved understanding could be extremely important
in order to identify microstructures that are resistant to hydrogen embrittlement and
thus to develop new procedures for alloy design that take into account the influence
of hydrogen on mechanical performance, leading to ultra-high-strength steels that are
resistant to hydrogen embrittlement.
In Chapter 5, a thermokinetic model for hydrogen redistribution in pure iron and steel
has been developed in combination with the thermokinetic software package MatCalc,
allowing the synergistic modelling of microstructural and hydrogen trapping evolution.
The model is based on the reduction of the Gibbs free energy for a multi-component
system incorporating hydrogen trapping sites, the evolution of which is governed by
classical nucleation growth and coarsening models. This synergistic combination is the
first of its kind, providing a validated method [87] to evaluate novel and established alloys
with respect to internal hydrogen behaviour for a range of environments including those
under mechanical loading, including both processing and operational environments. The
thermokinetic mean-field approach is important as it provides the link between statistical
distributions of microstructural defects, such as precipitate size distributions, dislocation
densities and the presence of grain boundaries, with hydrogen redistribution during heat
treatment. Although this model has been applied solely to iron and steel, it is equally
applicable to other alloys for which the thermodynamic and diffusion databases exist.
DFT based calculations have been conducted to evaluate the effects of hydrogen on
the stability and mechanical properties of cementite to establish the mechanisms of
hydrogen embrittlement in steels. The results show that in the presence of hydrogen,
cementite’s stability is notably reduced and its shearability increased, resulting in the
enhanced dissolution of cementite under fatigue, as evidenced in hydrogen-charged RCF
tests. The shearing of a cementite particle destabilises the particle. As hydrogen enters
the carbide, in addition to a reduction in its overall stability, such shearing becomes
260
more favourable, further increasing the likelihood of dissolution. Regardless of whether
cementite particles are dissolved, the enhanced plasticity in the presence of hydrogen
results in a reduction in strength, and a reduction in the fatigue life of the steel as a
result. RUS and DSC results on hydrogen charged cementite containing 100Cr6 bearing
steel validate the findings with respect to hydrogen’s effects on shearability and overall
stability respectively. However, given the limited cementite phase fraction in the steels
tested and the minimal hydrogen contents, the degree of such effects with respect to
operational conditions on the shearability of cementite could not be accurately quantified
by RUS. Similarly, DSC could not accurately quantify the reduced stability of cementite
in the presence of hydrogen due to hydrogen egress from the sample during testing.
Nonetheless, the preliminary results suggest the hypothesised effects are at least in-part
correct, though nanopillar testing is necessary to affirm this. One method by which the
changes in mechanical properties and stability could be more accurately quantified, is
through the use of bulk cementite for RUS testing. Additionally, conducting DSC with
hydrogen as a carrier gas could reduce the effects of hydrogen egress from the sample
during testing. Work is ongoing to conduct RUS testing on bulk cementite, as is work on
in situ electrochemically charged nanopillar compression testing for both bulk cementite
and cementite containing 100Cr6.
The role of γ′ precipitates on the hydrogen trapping characteristics of nickel-based alloys
has been investigated. Three γ′ containing model nickel alloys have been designed via
thermodynamic and thermokinetic modelling to investigate how lattice misfit between
the γ and γ′ phases affects hydrogen trapping behaviour. Although the calculated lat-
tice mismatches between the three alloys did not correlate with that found via XRD,
the results suggest that the γ/γ′ interface does indeed behave as a dominant hydrogen
trapping site and misfit may have an effect, although the variations observed in this
work could not be affirmed as being either a chemical effect or misfit effect. The TDA
profiles of each heat treated alloy revealed what appears to be three trapping species -
dislocations, grain boundaries and γ′. The results indicate that γ′ could indeed be im-
plemented to reduce hydrogen embrittlement susceptibility and that spherical nanosized
γ′ are deep, high capacity traps.
The optimised condition of vanadium carbide containing steels, and in part nickel-based
alloys, has been established with respect to hydrogen trapping capacity. In steels, 10
nm diameter carbides have been confirmed to be the optimum size. In nickel-based
alloys, the optimum size appears to be of the same scale, with 5 nm diameter carbides
providing significant trapping. Vanadium carbides in nickel-based alloys do not appear
to be stable at high temperatures (greater than 800◦C) over long periods of time however,
with a vanadium-rich delta phase dominating. The large needles of delta phase found
to form are not efficient hydrogen traps. Vanadium pollution of chromium-rich grain
boundary carbides is found to inhibit trapping upon these carbides. Such an effect could
261
be utilised to inhibit hydrogen conglomeration on the grain boundaries and reduce the
susceptibility of such alloys to grain boundary embrittlement.
10.2 The design of novel hydrogen embrittlement resistant
microstructures
Six novel alloy compositions and their corresponding heat treatments have been de-
veloped: 100Cr6+0.5V, 100Cr6WV, G5V, Flexpipe X, Grade 159V and Alloy 600V,
designed via thermodynamic and kinetic modelling. SEM and TEM investigations have
validated the modelling of all alloys, except 100Cr6WV, for which modelling parameters
require further work to be optimised. TEM has validated that the desired distribution
of nanoscale vanadium carbides has been achieved, although optimisation is necessary
for Flexpipe X. With the except of Flexpipe X, where the vanadium carbides were too
large to sufficiently trap hydrogen, TDA indicates that the novel microstructures have a
significantly increased hydrogen trapping capacity in comparison with their benchmark
counterparts. Notably, for the first time, the precipitation of vanadium carbides and its
benefits to hydrogen trapping have been established for nickel-based alloys and super-
bainitic steels. The novel alloys developed in this thesis have hardnesses equivalent to
their leading commercial counterparts but with improved hydrogen embrittlement resis-
tance. This confirms that novel high strength hydrogen embrittlement resistant steels
and nickel-based alloys can be developed using the prescribed methodologies established
in this work and as such, these methods can be utilised in any future work. A number
of the designed alloys are currently under review for patenting.
10.3 Future work
The composition and heat treatment schedule of 100Cr6WV should be optimised to
acquire the desired microstructure, this should be done through the methods presented
in this thesis. Similarly, Flexpipe X should be optimised with respect to vanadium
carbide distribution. By means of TDA, the hydrogen binding energy and trapping
capacity of tungsten carbide should be calculated, establishing the optimal size of these
carbides for trapping. The novel alloys designed in this thesis should be produced
using established industrial scale facilities. These larger casts should be used to assess
the proposed treatments on an industrial scale, affirm the required process routes, and
establish the mechanical properties, notably rolling contact fatigue life in the case of
bearing steels. With respect to the role of cementite on hydrogen induced failure, work
is currently ongoing to establish the effects of hydrogen on the mechanical properties of
cementite using in situ electrochemically charged nanopillar compression testing.
262
Many other areas of work would be of benefit to our understanding of hydrogen em-
brittlement. The development of an in situ hydrogen charging RCF test would be of
great benefit, validating the role hydrogen plays in reducing rolling contact fatigue life
and elucidating the effects of different ingress rates, and the relationship between load
and hydrogen content with respect to fatigue life. The role of retained austenite in hy-
drogen trapping during processing and operation should also be established. Retained
austenite is considered to behave as a hydrogen reservoir for diffusive hydrogen, liber-
ating the contained hydrogen under low stresses. This behaviour needs to be validated.
With respect to processing, the presence of retained austenite could result in high hy-
drogen contents when entering service. However, provided any retained austenite is free
of hydrogen, the austenite could provide the means of trapping hydrogen or providing
barriers to its diffusion, permissibly inreasing the hydrogen embrittlement resistance of
the corresponding steel. In addition, the effects of elastic strain on hydrogen migration
and trapping in bulk materials should be established, providing a more in-depth under-
standing of hydrogen interactions during mechanical loading such as under RCF. This
could be investigated through electrochemical hydrogen permeation tests of samples un-
der elastic and/or plastic strains. Such data could be implemented into a finite element
model, combining the kinetics of hydrogen-microstructure interactions modelled in this
thesis, with the strain evolution already modelled in the literature for bearings under
RCF. This could provide a comprehensive modelling package for evaluating bearings of
established and novel microstructures under RCF in the presence of hydrogen, providing
designers the currently absent technique of modelling the effects of hydrogen containing
microstructures with respect to RCF for specific operational conditions.
The work conducted on γ′ hydrogen trapping should be repeated with new compositions
that provide greater lattice misfits than those tested in this thesis to establish the role of
misfit on hydrogen trapping capacity and binding energies. Similar to the optimisation
of vanadium carbides for hydrogen trapping conducted in this thesis, the optimisation
of such secondary phases in nickel-based alloys is necessary to design better resistant
microstructures. Additionally, such optimisation work should be conducted with re-
spect to both helium trapping and radiation induced defect recombination. Such work
would require helium charging systems and ideally helium desorption analysis, although
helium bubbles can be visible in TEM at high concentrations. With respect to defect
recombination, the alloys could be neutron irradiated to establish the overall effects
of trapping species, but using irradiation techniques that avoid transmutation, such as
proton radiation, could isolate the role of secondary phases with respect to matrix hard-
ening. In addition to nickel-based alloys, the role of misfit/coherency strain surrounding
vanadium carbides in steel would also be useful to establish, using high resolution TEM
and TDA on vanadium carbides of differing compositions.
It is a pity that hydrogen-charged RCF tests for some of the steels designed here were not
263
completed in time for the publication of this thesis. Such data is critical to affirming the
benefits of trapping species, such as vanadium carbides, to resisting hydrogen-enhanced
RCF failure. Such tests are more complex than they may first appear however. It has
been presented elsewhere through tensile tests of hydrogen-charged materials contain-
ing such trapping species, that such traps can in fact enhance embrittlement in the
presence of hydrogen. However, naively absent in much of this work is the definition
of trap saturation, whether such traps are saturated, whether benchmark samples had
comparative contents of hydrogen during testing and the relevance of such tests to com-
mercial applications. RCF tests on hydrogen charged nanostructured steels must ensure
that trap-free variants contain equal hydrogen content throughout testing alongside the
trapping variants, and that this content is equal to that observed in industrial opera-
tion, such as that at the end of a particular bearing application’s life. If traps are fully
saturated during testing, they become redundant and such species can in fact enhance
embrittlement by providing sources of hydrogen and/or stress concentrators at which
hydrogen enhanced failure can initiate. As such, a steel designed principally to resist
hydrogen through trapping must at least trap the hydrogen content expected at the
corresponding component’s end-of-life to see an improvement. As such, in situ hydro-
gen charging during testing provides the most appealing means to assess embrittlment
resistance.
264
Bibliography
[1] A.G. Varias and A.R. Massih. Simulation of hydrogen embrittlement in zirco-
nium alloys under stress and temperature gradients. Journal of Nuclear Materials,
279(2-3):273–285, 2000.
[2] W.H. Johnson. On some remarkable changes produced in iron and steel by the
action of hydrogen and acids. Proceedings of the Royal Society, 23:168–179, 1875.
[3] M. Nagumo. Hydrogen related failure of steels - a new aspect. Materials Science
and Technology, 20(8):940–950, 2004.
[4] A. Barnoush. Hydrogen embrittlement revisited by in situ electrochemical nanoin-
dentation. PhD thesis, Universita¨t Des Saarlandes, December 2011.
[5] H.K.D.H. Bhadeshia. Steels for bearings. Progress in Materials Science, 57:268–
435, 2012.
[6] H.K. Birnbaum and P. Sofronis. Hydrogen-enhanced localized plasticity - a mecha-
nism for hydrogen-related fracture. Materials Science and Engineering: A, 176(1-
2):191–202, 1994.
[7] R.A. Oriani, W.W. Gerberich, M. Lii, X. Chen, and T. Foecke. Necessity of both
plasticity and brittleness in the fracture thresholds of iron. Philosophical Magazine
A: Physics of Condensed Matter, Structure, Defects and Mechanical Properties,
63(2):363–376, 1991.
[8] J.B. Bai, C. Prioul, and D. Francois. Hydride embrittlement in ZIRCALOY-
4 plate: Part I. influence of microstructure on the hydride embrittlement in
ZIRCALOY-4 at 20◦C and 350◦C. Metallurgical and Materials Transactions A,
25(6):1185–1197, 1994.
[9] I.M. Robertson and D. Teter. Controlled environment transmission electron mi-
croscopy. Journal of Microscopy Research and Technique, 42(4):260–269, 1998.
[10] D. Teter, I.M. Robertson, and H.K. Birnbaum. The effects of hydrogen on the
deformation and fracture of beta titanium. Acta Materialia, 49(20):4313–4323,
2001.
265
[11] International Atomic Energy Agency (IAEA). Delayed hydride cracking in zirco-
nium alloys in pressure tube nuclear reactors. Technical Report IAEA-TECDOC-
1410, Vienna, 2004.
[12] B. Shan, Y.Y. Zong, Y. Lv, and B. Guo. The effect of hydrogen on the strength-
ening and softening of Ti-6Al-4V alloy. Scripta Materialia, 58(6):449–452, 2008.
[13] H. Jagodzinski and Y. Hanninen. Interaction of hydrogen with dislocation pile-ups
and hydrogen induced softening of pure iron. Scripta Materialia, 43(3):245–251,
1997.
[14] P.J. Ferreira, I.M. Robertson, and H.K. Birnbaum. Hydrogen effects on the inter-
action between dislocations. Acta Materialia, 46(5):1749–1757, 1998.
[15] R.A. Oriani. Whitney award lecture—1987: Hydrogen—the versatile embrittler.
Corrosion, 43(7):390–397, 1987.
[16] M.S. Daw and M.I Baskes. Application of the embedded atom method to hydrogen
embrittlement. Chemistry and Physics of Fracture, 130:196–218, 1987.
[17] H. Vehoff and W. Rothe. Gaseous hydrogen embrittlement in FeSi and Ni single
crystals. Acta Metallurgica, 31(11):1781–1793, 1983.
[18] J.P. Carrasco, J.M.A. Barbosa, A.A. Silva, and M.A.D.S. Irma˜o. Application of
elastic fracture and damage mechanics models for numerical simulation of hydro-
gen embrittlement in steels. Engineering Computations, 29(6):596–604, 2012.
[19] K. Takai, H. Shoda, H. Suzuki, and M. Nagumo. Lattice defects dominating
hydrogen-related failure of metals. Acta Metallurgica, 56:5158–5167, 2008.
[20] M. Nagumo, H. Shimura, T. Chaya, H. Hayashi, and I. Ochiai. Fatigue damage
and its interaction with hydrogen in martensitic steels. Materials Science and
Engineering A, 348:192–200, 2003.
[21] A.M. Cuitino and M. Ortiz. Ductile fracture by vacancy condensation in fcc single
crystals. Acta Materialia, 44(2):427–436, 1996.
[22] K. Sakaki, T. Kawase, M. Hirato, M. Mizuno, H. Araki, Y. Shirai, and M. Nagumo.
The effect of hydrogen on vacancy generation in iron by plastic deformation.
Scripta Materialia, 55:1031–1034, 2006.
[23] R.B. McLellan and Z.R. Xu. Hydrogen-induced vacancies in the iron lattice.
Scripta Materialia, 36:1201–1205, 1997.
[24] N. Kino and K. Otani. The influence of hydrogen on rolling contact fatigue life
and its improvement. JSAE Review 24, pages 289–294, 2003.
[25] B.A. Szost. Hydrogen trapping in bearing steels: mechanisms and alloy design.
PhD thesis, Department of Materials Science and Metallurgy, University of Cam-
bridge, 2012.
266
[26] J.A. Ciruna and H.J. Szieleit. The effect of hydrogen on the rolling contact fatigue
life of AISI 52100 and 440C steel balls. Wear, 24(1):107–118, 1973.
[27] R. Lu. Investigation of decomposition of hydrocarbon oil on the nascent surface
of steel. Tribology Letters, 27(1):25–30, 2007.
[28] F.M. Al-Faqeer and H.W. Pickering. Analytical methods for studying cathodic
hydrogen charging into metals under competitive adsorption conditions, effects
of hydrogen on materials. In Proceedings of the 2008 International Hydrogen
Conference, pages 248–291, 2009.
[29] F.W.H. Dean. Measurement of hydrogen permeation through structural steel sec-
tions of varying thickness at 19oC. Materials Science and Technology, 21(3):347–
351, 2005.
[30] R. Dekkers, B. Blanpain, J. Plessers, and P. Wollants. Steel cleanliness and hy-
drogen in liquid steel. In VII International Conference on Molten Slags Fluxes
and Salts, pages 753–762. The South African Institute of Mining and Metallurgy,
2004.
[31] T.A. Harris, M.A. Ragen, and R.F. Spitzer. The effect of hoop and material resid-
ual stresses on the fatigue life of high speed rolling bearing. Tribology Transactions,
35:194–198, 1992.
[32] G.M. Pressouyre. A classification of hydrogen traps in steel. Metallurgical Trans-
actions A, 10(10):1571–1573, 1979.
[33] P.M. Scott. A review of environment-sensitive fracture in water reactor materials.
Corrosion Science, 583(8-9):583–606, 1985.
[34] T. Kushida. Hydrogen entry into steel by atmospheric corrosion. ISIJ Interna-
tional, 43(4):470–474, 2003.
[35] F. Ropital. Corrosion and degradation of metallic materials: Understanding of
the phenomena and applications in petroleum and process industries. IFP Publi-
cations, pages 58–60, 2005.
[36] R.A. Oriani. The diffusion and trapping of hydrogen in steel. Acta Metallurgica,
18:147–157, 1970.
[37] A. Turnbull, M.W. Carroll, and J.H. Ferriss. Analysis of hydrogen diffusion
and trapping in a 13 % chromium martensitic stainless steel. Acta Metallurgica,
37(7):2039–2046, 1989.
[38] E.M.K. Hillier. The effect of zinc-cobalt electroplating on the hydrogen embrittle-
ment of high strength steel. PhD thesis, Cranfield University, 2001.
[39] R.J. Hyun. Hydrogen Embrittlement in TRIP and TWIP Steels. PhD thesis, De-
partment of Ferrous Technology (Computational Metallurgy), Graduate Institute
of Ferrous Technology, Pohang University of Science and Technology, 2012.
267
[40] G.M. Pressouyre and F.M. Faure. Quantitative analysis of critical concentrations
for hydrogen-induced cracking. Hydrogen Embrittlement: Prevention and Control
(STP 962), pages 353–371, 1988.
[41] S. Hinotani, Y. Ohmori, and F. Terasaki. Effect of nickel on hydride formation
and hydrogen embrittlement in NiCrFe alloys. Materials Science and Engineering,
74(2):119–131, 1985.
[42] F. Gehmann, H.J. Grabke, and E. Riecke. Effects of the nitrides and carbides
of V, Nb and Ti on the diffusion and dissolution of hydrogen in iron. Hydrogen
Transport and Cracking in Metals, pages 216–226, 1995.
[43] G.M. Pressouyre and I.M. Bernstein. An example of the effect of hydrogen trap-
ping on hydrogen embrittlement. Metallurgical Transactions A, 12A:825–844,
1981.
[44] I. Maroef. Hydrogen trapping in ferritic steel weld metal. International Materials
Reviews, 4(47):191–223, 2002.
[45] G. W. Hong and J. Y. Lee. The measurement of the trap binding energy by the
thermal analysis technique. Scripta Metallurgica, 17(7):823–826, 1983.
[46] J. P. Hirth. Effects of hydrogen on the properties of iron and steel. Metallurgical
Transactions A, 11A:861–890, 1980.
[47] S.M. Lee and J. Lee. Hydrogen trapping phenomena in metals with B.C.C. and
F.C.C. structure by the thermal desorption analysis technique. Surface and Coat-
ings Technology, 28:301–314, 1986.
[48] D. Enomoto, M. Hirakami, and T. Tarui. Thermal desorption analysis of hydrogen
in high strength martensitic steels. Metallurgical and Materials Transactions A,
43(2):572–581, 2012.
[49] Y.D. Park, I.S. Maroef, A. Landau, and D.L. Olson. Retained austenite as a
hydrogen trap in steel welds. Welding Journal, 81(2):27–35, 2002.
[50] G. W. Hong and J. Y. Lee. The interaction of hydrogen with dislocation in iron.
Acta Metallurgica, 32(10):1581–1589, 1984.
[51] Y. Mine and Z. Horita. Hydrogen effects on ultrafine-grained steels processed by
high-pressure torsion. Materials Transactions, 53(5):773–785, 2012.
[52] P. Castano Rivera, V.P. Ramunni, and P. Bruzzoni. Hydrogen trapping in an API
5L X60 steel. Corrosion Science, 54:106–118, 2012.
[53] R.L.S. Thomas, D. Li, R.P. Gangloff, and J.R. Scully. Trap-governed hydrogen
diffusivity and uptake capacity in ultrahigh-strength AERMET 100 steel. Metal-
lurgical and Materials Transactions A, 33(7):1991–2004, 2002.
[54] S. Frappart, A. Oudriss, X. Feaugas, J. Creus, J. Bouhattate, F. Thebault, L. De-
lattre, and H. Marchebois. Hydrogen trapping in martensitic steel investigated
268
using electrochemical permeation and thermal desorption spectroscopy. Scripta
Materialia, 65(10):859–862, 2011.
[55] W.Y. Choo and J.Y. Lee. Thermal analysis of trapped hydrogen in pure iron.
Metallurgical Transactions A, 13(1):135–140, 1982.
[56] C. Andreone and A. Marut. Influence of the retained austenite in the hydrogen
embrittlement in AISI 4340. Scripta Metallurgica, 24:1453–1458, 1990.
[57] M.J. Robinson and R.M. Sharp. The effect of post-exposure heat treatment on the
hydrogen embrittlement of high carbon steel. Corrosion, 41(10):582–586, 1985.
[58] T. Tarui and S. Yamasaki. Evaluation method of delayed fracture property of
overcoming techniques of delayed fracture of high strength steels. Journal of the
Iron and Steel Institute of Japan, 88:612–619, 2002.
[59] G. Chalaftris. Evaluation of Aluminium-Based Coatings for Cadmium Replace-
ment. PhD thesis, School of Industrial and Manufacturing Science, Cranfield
University, 2003.
[60] S.M. Myers, M.L. Baskes, H.K. Brinbaum, J.W. Corbett, and G.G. Deleo. Hy-
drogen interactions with defects in crystalline solids. Reviews of Modern Physics,
64:559–617, 1992.
[61] K. Takahashi and J. Kawakami. The first direct observation of hydrogen trapping
sites in tic precipitation-hardening steel through atom probe tomography. Scripta
Materialia, 63:261–264, 2010.
[62] S. Yamasaki and H.K.D.H. Bhadeshia. M4C3 precipitation in Fe-C-Mo-V steels
and relationship to hydrogen trapping. Proceedings of the Royal Society A,
462:2315–2330, 2006.
[63] A.T.W. Barrow, J.H. Kang, and P.E.J. Rivera-Diaz-del-Castillo. ε→ η → θ tran-
sition in 100Cr6 and its effect on mechanical properties. Acta Materialia, 60:2805–
2815, 2012.
[64] P. Tra¨skelin. Molecular dynamics simulations of hydrogen bombardment of tung-
sten carbide surfaces. Physics Review B, 75:174113/1–174113/7, 2007.
[65] M. Iino. Evaluation of hydrogen-trap binding enthalpy I. Metallurgical Transac-
tions A, 18A:1559–1564, 1987.
[66] Y. Mine, T. Tsumagari, and Z. Horita. Hydrogen trapping on lattice defects
produced by high-pressure torsion in Fe-0.01 mass% C alloy. Scripta Materialia,
63(5):552–555, 2010.
[67] FD Fischer, J Svoboda, and E Kozeschnik. Interstitial diffusion in systems with
multiple sorts of traps. Modelling and Simulation in Materials Science and Engi-
neering, 21(2):1–13, 2013.
269
[68] A. McNabb and P.K. Foster. A new analysis for the diffusion of hydrogen in iron
and ferritic steels. Transactions AIME, 227:618–627, 1963.
[69] M. Koiwa. Trapping effect in diffusion of interstitial impurity atoms in bcc lattices.
Acta Metallurgica, 22:1259–1268, 1974.
[70] M.I. Luppo and J. Ovejero-Garcia. The influence of microstructure on the trapping
and diffusion of hydrogen in a low carbon steel. Corrosion Science, 32(10):1132–
1136, 1991.
[71] G.L.W. Hart, B.M. Klein, and S. Begay. Vacancy ordering and non-stoichiometry
in TiC and TiN. Complex Inorganic Solids, pages 99–109, 2005.
[72] T. Imran, B. Jacobson, A. Shariff. Quantifying diffused hydrogen in AISI-52100
bearing steel and in silver steel under tribo-mechanical action: Pure rotating
bending, sliding-rotating bending, rolling-rotating bending and uni-axial tensile
loading. Wear, 261:86–95, 2006.
[73] W. Beck, J.O.M. Bockris, J. McBreen, and L. Nanis. Hydrogen permeation in
metals as a function of stress, temperature and dissolved hydrogen concentration.
Proceedings of the Royal Society, London, 290-A:220–235, 1966.
[74] C. Beachem. New model for hydrogen-assisted cracking. Metallurgical Transac-
tions, 3(2):441–455, 1972.
[75] G.M. Pressouyre and I.M. Bernstein. A kinetic trapping model for hydrogen-
induced cracking. Acta Metallurgica, 27(1):89–100, 1979.
[76] H. Matsui, H. Kimura, and S. Moriya. The effect of hydrogen on the mechanical
properties of high-purity iron - softening and hardening of high-purity iron by
hydrogen charging during tensile deformation. Materials Science Engineering,
40:207–216, 1979.
[77] G. Bergmann. Studies on Hydrogen Embrittlement of NiAl, TiAl and CMSX-6.
PhD thesis, Max-Planck Institute for Iron Research, 1995.
[78] D. Shih, I. Robertson, and H. Birnbaum. Hydrogen embrittlement of α titanium
in situ TEM studies. Acta Metallurgica, 36:111–124, 1988.
[79] G.M. Bond, I.M. Robertson, and H.K. Birnbaum. On the determination of the
hydrogen fugacity in an environmental cell TEM. Scripta Metallurgica, 20:653–
658, 1986.
[80] D. Caillard. TEM in situ straining experiments in Fe at low temperature. Journal
of Physics: Conference Series, 240(1):1–4, 2010.
[81] A. Barnoush and H. Vehoff. Recent developments in the study of hydrogen embrit-
tlement: Hydrogen effect on dislocation nucleation. Acta Materialia, 58(16):5274–
5285, 2010.
270
[82] A. Barnoush and H. Vehoff. Electrochemical nanoindentation: A new approach to
probe hydrogen/deformation interaction. Scripta Materialia, 55(2):195–198, 2006.
[83] A. Barnoush, M. Asgari, and R. Johnsen. Resolving the hydrogen effect on dislo-
cation nucleation and mobility by electrochemical nanoindentation. Scripta Ma-
terialia, 66(6):414–417, 2012.
[84] C.F. Jatczak. Hardenability in high carbon steels. Metallurgical Transactions,
4(10):2267–2277, 1973.
[85] W.S. Owen. The effect of silicon on the kinetics of tempering. Transaction of the
ASM, 46:812–829, 1954.
[86] S. Chattopadhyay and C. M. Sellars. Kinetics of pearlite spheroidisation during
static annealing and during hot deformation. Acta Metallurgica, 30(1):157–170,
1982.
[87] M.A. Stopher, P. Lang, E. Kozeschnik, and P.E.J. Rivera-Diaz-del-Castillo. Mod-
elling hydrogen migration and trapping in steels. Materials and Design, 106:205–
215, 2016.
[88] Y.M. Fimenko, A.B. Kuslitskii, D.V. Chaban, G.V. Karpenko, and B.A. Movchan.
Effect of electron-beam melting upon the properties of grade ShKh15 ball-bearing
steel. Soviet Materials Science, 1:333–335, 1966.
[89] R. Schlatter and J.P. Stroup. Improved M50 aircraft bearing steel through ad-
vanced vacuum melting process. Journal of Vacuum Science and Technology,
9:1326–1333, 1972.
[90] W. Solano-Alvarez, E.J. Pickering, and H.K.D.H. Bhadeshia. Degradation of
nanostructured bainitic steel under rolling contact fatigue. Materials Science and
Engineering A, 617:156–164, 2014.
[91] J.M. Hampshire, J.V. Nash, and G.E. Hollox. Materials evaluation by flat washer
testing. Rolling contact fatigue testing of bearing steels, ASTM STP771, pages
46–66, 1982.
[92] P.V. Krishna, R.R. Srikant, M. Iqbal, and N. Sriram. Effect of austempering and
martempering on the properties of AISI 52100 steel. ISRN Tribology, 2013:1–6,
2013.
[93] V. Bhargava, G.T. Hahn, and C.A. Rubin. Rolling contact deformation, etching
effects, and failure of high-strength bearing steel. Metallurgical Transactions A,
21(7):1921–1931, 1990.
[94] V. Smel¸ova, A. Schwedt, L. Wang, W. Holweger, and J. Mayer. Microstructural
changes in White Etching Cracks (WECs) and their relationship with those in
Dark Etching Region (DER) and White Etching Bands (WEBs) due to Rolling
Contact Fatigue (RCF). International Journal of Fatigue, 100:148–158, 2017.
271
[95] Y.J. Li, M. Herbig, S. Goto, and D. Raabe. Atomic scale characterization of
white etching area and its adjacent matrix in a martensitic 100Cr6 bearing steel.
Materials Characterization, 123:349–353, 2017.
[96] JeeHyun Kang. Mechanisms of Microstructural Damage during Rolling Contact
Fatigue of Bearing Steels. PhD thesis, Darwin College, University of Cambridge,
2013.
[97] N.P. Suh. Martensite decay during rolling contact fatigue in ball bearings. Met-
allurgical Transactions A, 7A:1099–1110, 1976.
[98] A. Grabulov. Fundamentals of rolling contact fatigue. PhD thesis, University of
Delft, 2010.
[99] Y. Murakami, T. Kanezaki, and S. Matsuoka. Hydrogen embrittlement mechanism
in fatigue of austenitic stainless steels. Metallurgical and Materials Transactions
A, 39(6):1327–1339, 2008.
[100] G. Popescu, A. Gaeblli, G.M. Espejel, and B. Wemekamp. Micro-plastic material
model and residual fields in rolling contacts. Journal of ASTM International,
3(5):1–12, 2006.
[101] R.B. McLellan and L. Yang. Hydrogen-vacancy interaction in palladium. Acta
Metallurgica Materialia, 43(6):2463–1467, 1995.
[102] M.H. Evans, A.D. Richardson, L. Wang, and R.J.K. Wood. Effect of hydrogen on
butterfly and white etching crack (WEC) formation under rolling contact fatigue
(RCF). Wear 306, pages 226–241, 2013.
[103] R. H. Vegter and J. T. Slycke. The role of hydrogen on rolling contact fatigue
response of rolling element bearings. Journal of ASTM International, 7(2):1–12,
2010.
[104] H.K. Birnbaum et al. Mechanisms of hydrogen related fracture - a review. Second
International Conference on Corrosion-Deformation Interactions. CDI’96, 1996.
[105] R. K. Dayal, N. Parvathavarthini. Hydrogen embrittlement in power plant steels.
Frontiers in Materials Science, pages 431–453, 2005.
[106] K.N. Jha, M.K. Sardar, N.N. Jha, and S. Chakraborty. Hydrogen control during
steel making for medium carbon wheels. Scandinavian Journal of Metallurgy,
32(6):296–300, 2003.
[107] D. Gaude-Fugarolas. Understanding hydrogen redistribution and designing a new
hydrogen extraction method. Materials Today: Proceedings, 2:S393–S400, 2015.
[108] B.A. Szost, R.H. Vegter, and P.E.J. Rivera-Dı´az-del Castillo. Developing bearing
steels combining hydrogen resistance and improved hardness. Materials & Design,
43:499–506, 2013.
272
[109] B.G. Pound. A comparison of hydrogen ingress behavior in Alloys 625 and 716.
Scripta Metallurgica et Materialia, 29(11):1433–1438, 1993.
[110] B.G Pound. The effect of aging on hydrogen trapping in precipitation-hardened
alloys. Corrosion Science, 42(11):1941–1956, 2000.
[111] A.F. Rowcliffe, L.K. Mansur, D.T. Hoelzer, and R.K. Nanstad. Perspectives on
radiation effects in nickel-base alloys for applications in advanced reactors. Journal
of Nuclear Materials, 392(2):341–352, 2009.
[112] T. Abram and S. Ion. Generation-IV nuclear power: a review of the state of the
science. Energy Policy, 36(12):4323–4330, 2008.
[113] G.O. Hayner et al. Next generation nuclear plant materials selection and qualifi-
cation program plan, report no. INEEL/EXT-03-001128. Technical report, Idaho
National Engineering Laboratory, November 2003.
[114] T. Schulenberg, L.K.H. Leung, D. Brady, Y. Oka, K. Yamada, Y. Bae, and
G. Willermoz. IAEA Report: IAEA-CN-164-5S06: Supercritical Water-Cooled
Reactor (SCWR) Development through GIF Collaboration. Technical report,
IAEA, 2009.
[115] C.D. Judge, N. Gauquelin, L. Walters, M. Wright, J.I. Cole, J.Madden, G.A. Bot-
ton, and M. Griffiths. Intergranular fracture in irradiated Inconel X-750 containing
very high concentrations of helium and hydrogen. Journal of Nuclear Materials,
457(2):165–172, 2015.
[116] D. Chersola, G. Lomonaco, and R. Marotta. The VHTR and GFR and their use
in innovative symbiotic fuel cycles. Progress in Nuclear Energy, 83:443–459, 2015.
[117] S. Bortot, P.C. Camprini, G. Grasso, and C. Artioli. Conceptual core design study
for a high-flux LFR demonstrator. Progress in Nuclear Energy, 54(1):56–63, 2012.
[118] C. Cabet and F. Rouillard. Corrosion of high temperature metallic materials in
VHTR. Journal of Nuclear Materials, 392(2):235–242, 2009.
[119] O.D. Kazachkovskii. Radiation swelling of materials. Atomic Energy, 81(2):605–
610, 1996.
[120] F.A. Comprelli, H. J. Busboom, and C. N. Spalaris. Comparison of radiation
damage studies and fuel cladding performance for Incoloy-800. Irradiation Effects
on Structural Alloys for Thermal and Fast Reactors, ASTM STP 457, pages 400–
413, 1969.
[121] T.M. Pollock and S. Tin. Nickel-based superalloys for advanced turbine engines:
Chemistry, microstructure, and properties. Journal of Propulsion and Power,
22(2):361–374, 2006.
[122] C. Judge, M. Griffiths, L. Walters, M. Wright, G. Bickel, O. Woo, M. Stewart,
S. Douglas, and F. Garner. Embrittlement of nickel alloys in a CANDU reactor
273
environment. In Effects of Radiation on Nuclear Materials: STP 1547: 25th
Volume, pages 161–175. ASTM International, 2013.
[123] M. Griffiths, G.A. Bickel, and S.R. Douglas. Irradiation-induced embrittlement
of Inconel 600 flux detectors in CANDU reactors. Journal of Energy and Power
Engineering, 6:188–194, 2012.
[124] S.J. Zinkle and G.S. Was. Materials challenges in nuclear energy. Acta Materialia,
61(3):735–758, 2013.
[125] K.L. Murty and I. Charit. Structural materials for Gen-IV nuclear reactors: Chal-
lenges and opportunities. Journal of Nuclear Materials, 383(1-2):189–195, 2008.
[126] P. Ford, P. Combrade, and P. Scott. Degradation of Alloy 600 SG Tubes in Sec-
ondary PWR coolant, Environmentally-assisted degradation of nickel-base alloys
in LWRs. Advanced Nuclear Technology International, September 2011.
[127] P. Scott, P. Combrade, and P. Ford. Environmentally-Assisted Degradation of
Nickel-Base Alloys in LWRs (Chapter 2.2.3.1). Advanced Nuclear Technology
International, September 2011.
[128] H. Zhang., Z. Yao, M.A. Kirk, and M.R. Daymond. Stability of Ni3(Al, Ti)
Gamma Prime Precipitates in a Nickel-Based Superalloy Inconel X-750 Under
Heavy Ion Irradiation. Metallurgical and Materials Transactions A, 45(8):3422–
3428, 2014.
[129] J.X. Dong, X.B. Liu, X.S. Xie, and R.G. Thompson. The segregation of sulfur and
phosphorus in nickel-base Alloy 718. Acta Metallurgica Sinica (English Letters),
10(6):510–514, 2009.
[130] A.L. Ward and J. M. Steichen and R.L. Knecht. Irradiation and thermal effects
on the tensile properties of Inconel-718. Irradiation Effects on Microstructure and
Properties of Metals, ASTM STP 611, page 156, 1976.
[131] T.T. Claudson and H.J. Pessl. Irradiation effects on high-temperature reactor
structural metals. Flow and Fracture of Metals and Alloys in Nuclear Environ-
ments, ASTM STP 380, pages 156–170, 1965.
[132] J Moteff, FC Robertshaw, and FD Kingsbury. Effects of neutron irradiation on
the stress-rupture properties of high-temperature precipitation hardening alloys.
Journal of Nuclear Materials, 17(3):245–258, 1965.
[133] T.T. Claudson. Effects of neutron irradiation on the elevated temperature me-
chanical properties of nickel-base and refractory metal alloy. Effects of Irradiation
on Structural Metals, ASTM STP 426, pages 67–94, 1967.
[134] M.L. Bleiberg and J.W. Bennett. Radiation effects in breeder reactor structural
materials. Technical report, American Institute of Mining, Metallurgical, and
Petroleum Engineers, Inc., New York, 1977.
274
[135] C. Cawthorne and E.J. Fulton. Voids in irradiated stainless steel. Nature, 216:575–
576, 1967.
[136] WG Johnston, JH Rosolowski, AM Turkalo, and T Lauritzen. An experimental
survey of swelling in commercial Fe-Cr-Ni alloys bombarded with 5 MeV Ni Ions.
Journal of Nuclear Materials, 54(1):24–40, 1974.
[137] J.F. Bates and W.G. Johnston. Proceedings of the international conference on
radiation effects in breeder reactor structural materials. The Metallurgical Society
of AIME, New York, page 625, 1977.
[138] R.L. Klueh and A.T. Nelson. Ferritic/martensitic steels for next-generation reac-
tors. Journal of Nuclear Materials, 371(1-3):37–52, 2007.
[139] E.E Bloom, J.O Stiegler, A.F Rowcliffe, and J.M Leitnaker. Austenitic stainless
steels with improved resistance to radiation-induced swelling. Scripta Metallurgica,
10(4):303–308, 1976.
[140] P. Yvon and F. Carre. Structural materials challenges for advanced reactor sys-
tems. Journal of Nuclear Materials, 385(2):217–222, 2009.
[141] J.A. Horak et al. U.s. department of energy national cladding/duct materials
development quarterly progress report, TC-160-27. Technical report, US Dep. of
Energy, 1980.
[142] G.R. Gessel et al. DOE Cladding/Duct Materials Development Quarterly Progress
Report, TC-160-18. Technical report, US Dep. of Energy, 1978.
[143] H. Ullmaier. Radiation damage in metallic materials. MRS Bulletin, 22(4):14–21,
1997.
[144] L.R. Greenwood and F.A. Garner. Hydrogen generation arising from the 59Ni
(n, p) reaction and its impact on fission-fusion correlations. Journal of Nuclear
Materials, 233-237:1530–1534, 1996.
[145] F.A. Garner, L.R. Greenwood, and B.M. Oliver. A re-evaluation of helium/dpa
and hydrogen/dpa ratios for fast reactor and thermal reactor data used in fission-
fusion correlations. Technical report, Proceeding of a Technical Committee Meet-
ing, Obnisk, Russian Federation, IAEA-TECDOC-1039, 16-19 June,1997.
[146] L.R. Greenwood, F.A. Garner, B.M. Oliver, M.L. Grosseck, and W.G. Wolfer.
Surprisingly large generation and retention of helium and hydrogen in pure nickel
irradiated at high temperatures and high neutron exposures. Effects of Radiation
on Materials: 21st International Symposium, STP11254S, 1(4):529–539, 2004.
[147] S.J. Zinkle, W.G. Wolfer, G.L. Kulcinski, and L.E. Seitzman. Stability of vacancy
clusters in metals. Part II. Effect of oxygen and helium on void formation in metals.
Philosophical Magazine A, 55(1):127–140, 1987.
275
[148] H. Schroeder and H. Ullmaier. Helium and hydrogen effects on the embrittlement
of iron- and nickel-based alloys. Journal of Nuclear Materials, 179-181(1):118–124,
1991.
[149] F.A. Garner, B.M. Oliver, L.R. Greenwood, D.J. Edwards, and S.M. Bruemmer.
Generation and retention of helium and hydrogen in austenitic steels irradiated in
a variety of LWR and test reactor spectral environments. Technical report, Pacific
Northwest National Laboratory (PNNL), Richland, WA (US), 2002.
[150] C.H. Lund. Physical metallurgy of nickel based superalloys. Technical report,
Battelle Memorial Institute, Defense Metals Information Center, Columbus, Ohio,
1961.
[151] D.J. Tillack. Selection of nickel, nickel-copper, nickel-chromium, and nickel-
chromium-iron Alloys. ASM International, part of Handbook Volume 6, Welding,
Brazing, and Soldering, 1993.
[152] Y. Yin, R.G. Faulkner, P. Moreton, I. Armson, and P. Coyle. Grain bound-
ary chromium depletion in austenitic alloys. Journal of Materials Science,
45(21):5872–5882, 2010.
[153] Y. Yin, F. Zhu, E. Miller R. Faulkner, P. Moreton, I. Armson, and B. Borradaile.
Predicting chromium depletion of nickel base alloys. In 15th International Confer-
ence on Environmental Degradation of Materials in Nuclear Power Systems-Water
Reactors, pages 343–354. John Wiley & Sons, Inc., 2012.
[154] R. Bajaj, W.J. Mills, M.R. Lebo, B.Z. Hyatt, and M.G. Burke. Irradiation-
assisted stress corrosion cracking of HTH Alloy X-750 and Alloy 625. In Seventh
International Symposium on Environmental Degradation of Materials in Nuclear
Power Systems–Water Reactors, volume 2, pages 1093–1107, 1995.
[155] R.C. Baumann and T.Z. Hossain. Electronic device and process achieving a re-
duction in alpha particle emissions from boron-based compounds essentially free
of boron-10, March 7 1995. US Patent 5,395,783.
[156] M. Kangilaski. The effects of neutron radiation on structural materials, NASA-
CR-93128, REIC-45. Technical report, Battelle Memorial Inst., Columbus, Ohio.
Radiation Effects Information Center, 1967.
[157] F.W. Wiffen. Response of Inconel 600 to simulated fusion reactor irradiation.
Effects of Radiation on Structural Materials, ASTM STP 683, pages 88–106, 1979.
[158] T.M. Angeliu, J.T. Ward, and J.K. Witter. Assessing the effects of radiation
damage on Ni-base alloys for the Prometheus space reactor system. Journal of
Nuclear Materials, 366(1):223–237, 2007.
[159] M.I. Baskes. Recent advances in understanding helium embrittlement in metals.
MRS Bulletin, 11(04):14–18, 1986.
276
[160] W.J. Mills. Fracture toughness of two Ni-Fe-Cr alloys. Engineering Fracture
Mechanics, 26(2):223–238, 1987.
[161] C.W. Hunter, R.L. Fish, and J.J. Holmes. Channel fracture in irradiated EBR-II
type 304 stainless steel. Transactions of the American Nuclear Society, 15:254–
255, 1972.
[162] R.L. Fish. Notch effect on the tensile properties of fast-reactor-irradiated type
304 stainless steel. Nuclear Technology, 31(1):85–95, 1976.
[163] L.A. James. Effect of fast neutron irradiation on fatigue-crack growth behavior of
three nickel-base alloys. Nuclear Technology, 53(1):64–68, 1981.
[164] J.J. Kai and R.D. Lee. Effects of proton irradiation on the microstructural and mi-
crochemical evolution of Inconel 600 alloy. Journal of Nuclear Materials, 207:286–
294, 1993.
[165] S.I. Porollo, Y.V. Konobeev, A.M. Dvoriashin, M. Alexander, V.M. Krigan, and
F.A. Garner. Determination of the lower temperature limit of void swelling of
stainless steels at relatively low displacement rates. Technical report, Pacific
Northwest National Laboratory (PNNL), Richland, WA (US), 2002.
[166] L.K. Mansur. Effects of point defect trapping and solute segregation on irradiation-
induced swelling and creep. Journal of Nuclear Materials, 83(1):109–127, 1979.
[167] M. Kurban, U. Erb, and K.T. Aust. A grain boundary characterization study of
boron segregation and carbide precipitation in Alloy 304 austenitic stainless steel.
Scripta Materialia, 54(6):1053–1058, 2006.
[168] S.I. Golubov, R.E. Stoller, S.J. Zinkle, and A.M. Ovcharenko. Kinetics of coars-
ening of helium bubbles during implantation and post-implantation annealing.
Journal of nuclear materials, 361(2):149–159, 2007.
[169] L.K. Mansur and K. Farrell. On mechanisms by which a soft neutron spectrum
may induce accelerated embrittlement. Journal of Nuclear Materials, 170(3):236–
245, 1990.
[170] P.L. Andresen, F.P. Ford, S.M. Murphy, and J.M. Perks. Environmental degra-
dation of materials in nuclear power systems - water reactors. Proceedings of the
4th International Symposium, pages 1–83, 1990.
[171] A.F. Rowcliffe and J.A. Horak. Tensile properties and fracture behavior of irra-
diated nickel alloys. Transactions of the American Nuclear Society, 38:266–267,
1981.
[172] F.H. Huang and R.L. Fish. Effects of fast neutron irradiation on the fracture
behavior of stainless steel. Effects of Radiation on Materials, ASTM STP 782,
pages 701–719, 1982.
277
[173] W.J.S. Yang and B.J. Makenas. Microstructure of irradiated Inconel 706 fuel pin
cladding. Effects of Radiation on Materials, ASTM STP 870, 2:127–138, 1985.
[174] S. Vaidyanathan, W.L. Bell, and T. Lauritzen. Irradiation embrittiement in some
austenitic superalloys. Effects of Irradiation on Structural Materials, ASTM STP
782, pages 619–635, 1982.
[175] R. Bajaj and R.P. Shogan and C. DeFlitch and R.L. Fish and M.M. Paxton and
M.L. Bleiberg. Tensile properties of neutron-irradiated nimonic PE16. Effects of
Radiation on Materials, ASTM STP 725, pages 326–351, 1981.
[176] A.L. Chang. Microstructural basis and crack growth theories for post-irradiation
ductility loss in Nimonic PE16. Micro and Macro Mechanics of Crack Growth,
pages 223–239, 1982.
[177] R.M. Boothby and D.R. Harries. Embrittlement and precipitation in irradiated
Nimonic PE16. Proceedings of the International Conference on Mechanical Be-
haviour and Nuclear Applications of Stainless Steel at Elevated Temperatures,
pages 157–164, 1982.
[178] R.M. Boothby. The microstructure of fast neutron irradiated Nimonic PE16.
Journal of Nuclear Materials, 230(2):148–157, 1996.
[179] G.H. Broomfield. High-temperature tensile properties of unirradiated and ther-
mal reactor irradiated Nimonic PE16. Irradiation Effects in Structural Alloys for
Thermal and Fast Reactors, ASTM STP 457, pages 38–66, 1969.
[180] A.F. Rowcliffe, G.J.C Carpenter, H.F. Merrick, and R.B. Nicholson. An electron
microscope investigation of high-temperature embrittlement of irradiated stainless
steel. Effects of Radiation on Structural Metals, ASTM STP 426, pages 161–199,
1967.
[181] R.D. Nicholson and R.B. Jones. Fracture toughness of irradiated Nimonic PE16
at high temperatures. Effects of Radiation on Materials, ASTM STP 683, pages
529–545, 1979.
[182] D.M. Symons. A comparison of internal hydrogen embrittlement and hydrogen
environment embrittlement of X-750. Engineering Fracture Mechanics, 68(6):751–
771, 2001.
[183] J.M. Walsh and B.H. Rear. Direct evidence for boron segregation to grain bound-
aries in a nickel-base alloy by secondary ion mass spectrometry. Metallurgical and
Materials Transactions A, 6(1):226–229, 1975.
[184] T.R. Allen and G.S. Was. The effect of ordering on radiation-induced segregation
in austenitic iron- and nickel-base alloys. Effects of Radiation on Materials, ASTM
STP 1325, pages 619–634, 1999.
278
[185] J.M. Perks, A.D. Marwick, and C.A. English. Computer code to calculate
radiation-induced segregation in concentrated ternary alloys, AERE-R-12121.
Technical report, UK Atomic Energy Agency (UKAEA), Harwell, 1986.
[186] Y. Yang, K. Field, and J. Busby T. Allen. Development of a robust modeling
tool for radiation-induced segregation in austenitic stainless steels. Technical Re-
port ORNL/TM-2015/479, Oak Ridge National Laboratory (ORNL), Oak Ridge,
Tenessee, USA, 2015.
[187] L. Fournier, D. Delafosse, and T. Magnin. Cathodic hydrogen embrittlement in
Alloy 718. Materials Science and Engineering: A, 269(1-2):111–119, 1999.
[188] P.D. Hicks and C.J. Altstetter. Hydrogen-enhanced cracking of superalloys. Met-
allurgical Transactions A, 23(1):237–249, 1992.
[189] M.A. Stopher and P.E.J. Rivera-Diaz-del-Castillo. Hydrogen embrittlement in
bearing steels. Materials Science and Technology, 32(11):1184–1193, 2016.
[190] E.V. Kornelsen and A.A. Van Gorkum. A study of bubble nucleation in tungsten
using thermal desorption spectroscopy: clusters of 2 to 100 helium atoms. Journal
of Nuclear Materials, 92:79–88, 1980.
[191] G. Hultquist, M.J. Graham, J.L. Smialek, and B. Johnsson. Hydrogen in metals
studied by thermal desorption spectroscopy (TDS). Corrosion Science, 93:324–
326, 2015.
[192] G.A. Young, N. Lewis, M. Hanson, W. Matuszyk, B. Wiersma, and S.C. Gonzalez.
Microstructural and Microchemical Characterization of Dual Step Aged Alloy X-
750 and its relationship to environmentally assisted cracking, KAPL Report LM-
01K034. Technical report, 2001.
[193] H. Zhang, Z. Yao, M.R. Daymond, and M.A. Kirk. Cavity morphology in a
Ni based superalloy under heavy ion irradiation with cold pre-injected helium.
Journal of Applied Physics, 115(10):1–7, 2014.
[194] F.A. Garner and D.S. Gelles. Neutron-induced swelling of commercial alloys at
very high exposures. Effects of Radiation on Materials, ASTM STP 1046, 2:673–
683, 1990.
[195] C. San Marchi, B.P. Somerday, and S.L. Robinson. Permeability, solubility and
diffusivity of hydrogen isotopes in stainless steels at high gas pressures. Interna-
tional Journal of Hydrogen Energy, 32(1):100–116, 2007.
[196] J. Kameda and C.J. McMahon. Solute segregation and hydrogen-induced inter-
granular fracture in an alloy steel. Metallurgical Transactions A, 14(4):903–911,
1983.
[197] M.I. Baskes and V. Vitek. Trapping of hydrogen and helium at grain boundaries in
nickel: An atomistic study. Metallurgical Transactions A, 16(9):1625–1631, 1985.
279
[198] R. Sugano, K. Morishita, H. Iwakiri, and N. Yoshida. Effects of dislocation on
thermal helium desorption from iron and ferritic steel. Journal of Nuclear Mate-
rials, 307-311:941–945, 2002.
[199] Y.M. Zakaria. Thermal Desorption Analysis of Helium Trapping in Ion-Implanted
Beryllium. PhD thesis, McMaster University, 1996.
[200] G.J. Thomas, W.A. Swansiger, and M.I. Baskes. Low-temperature helium release
in nickel. Journal of Applied Physics, 50(11):6942–6947, 1979.
[201] J. Ehrenberg, B.M.U. Scherzer, and R. Behrisch. Thermal desorption spectroscopy
of He from Ni at and below saturation. Radiation effects, 78(1-4):405–416, 1983.
[202] J. Barnaby, P.J. Barton, R.M. Boothby, A.S. Fraser, and G.F. Slattery. Post-
irradiation mechanical properties of AISI type 316 steel and Nimonic PE16 alloy.
In Radiation effects in breeder reactor structural materials, pages 159–174. Metal-
lurgical Society of AIME, 1977.
[203] H.G. Harman. Post-irradiation tensile and creep-rupture properties of several
experimental heats of Incoloy 800 at 700 and 760◦C, ORNL-TM-2305. Technical
report, Oak Ridge National Lab., Tenn., 1968.
[204] L.K. Mansur. Theoretical evaluation of a mechanism of precipitate-enhanced cav-
ity swelling during irradiation. Philosophical Magazine A, 44(4):867–877, 1981.
[205] F.A. Garner. Irradiation performance of cladding and structural steels in liquid
metal reactors. Materials Science and Technology, Nuclear Materials, 10A(part
1):419–543, 1994.
[206] J.C. Zhao, M. Larsen, and V. Ravikumar. Phase precipitation and time–
temperature-transformation diagram of Hastelloy X. Materials Science and Engi-
neering: A, 293(1):112–119, 2000.
[207] K. Bhanu et al. On massive carbide precipitation during high temperature low
cycle fatigue in alloy 800H. Scripta Metallurgica et Materialia, 31(4):381–386,
1994.
[208] Y. Yin, R.G. Faulkner, and Z. Lu. Irradiation-induced precipitation modelling of
ferritic steels. Journal of Nuclear Materials, 389(2):225–232, 2009.
[209] K.C. Russell. Phase stability under irradiation. Progress in Materials Science,
28(3-4):229–434, 1984.
[210] J.R. Lindgren. Irradiation effects on high-temperature gas-cooled reactor struc-
tural materials. Nuclear technology, 66(3):607–618, 1984.
[211] P. Caceras, B. Ralph, G.C. Allen, and R.K. Wild. Grain boundary segregation in
Inconel 600. Surface and Interface Analysis, 12(3):191–197, 1988.
[212] J.K. Heuer, P.R. Okamoto, N.Q. Lam, and J.F. Stubbins. Disorder-induced melt-
ing in nickel: implication to intergranular sulfur embrittlement. Journal of Nuclear
Materials, 301(2):129–141, 2002.
280
[213] W.C. Johnson, J.E. Doherty, B.H. Kear, and A.F. Giamei. Confirmation of sulfur
embrittlement in nickel alloys. Scripta Metallurgica, 8:971–974, 1974.
[214] B.J. Berkowitz and R.D. Kane. The effect of impurity segregation on the hydrogen
embrittlement of a high strength nickel base alloy in H2S environments. Corrosion,
36(1):24–29, 1980.
[215] E.H. Lee and L.K. Mansur. Mechanisms of swelling suppression in cold-worked
phosphorous-modified Fe-Ni-Cr alloys. Philosophical Magazine A, 61(5):733–749,
1990.
[216] T. Yamamoto, G.R. Odette, P. Miao, D.T. Hoelzer, J. Bentley, N. Hashimoto,
H. Tanigawa, and R.J. Kurtz. The transport and fate of helium in nanostructured
ferritic alloys at fusion relevant He/dpa ratios and dpa rates. Journal of Nuclear
Materials, 367:399–410, 2007.
[217] Y. Gan, H. Zhao, D.T. Hoelzer, and D. Yun. Energetic study of helium cluster
nucleation and growth in 14YWT through first principles. Materials, 9(1):1–12,
2016.
[218] D.T. Hoelzer, J. Bentley, M.K. Sokolov, M.K. Miller, G.R. Odette, and M.J.
Alinger. Influence of particle dispersions on the high-temperature strength of
ferritic alloys. Journal of Nuclear Materials, 367:166–172, 2007.
[219] P.D. Edmondson, C.M. Parish, Q. Li, and M.K. Miller. Thermal stability of
nanoscale helium bubbles in a 14YWT nanostructured ferritic alloy. Journal of
Nuclear Materials, 445(1-3):84–90, 2014.
[220] D.T. Hoelzer, J.P. Shingledecker, R.L. Klueh, M.K. Miller, and J. Bentley. The
microstructural stability of a ruptured thermal creep specimen of MA957. Tech-
nical report, Oak Ridge National Laboratory, Tennessee, USA, June 30, 2008.
[221] H. Kronmller and P. Vargas. Binding energy of hydrogen-impurity complexes in
nickel. Philosophical Magazine A, 51(1):59–70, 1985.
[222] S. Lee and J. Lee. The trapping and transport phenomena of hydrogen in nickel.
Metallurgical Transactions A, 17(2):181–187.
[223] F. Jambon, L. Marchetti, M. Sennour, F. Jomard, and J. Chene. SIMS and
TEM investigation of hydrogen trapping on implantation defects in a nickel-based
superalloy. Journal of Nuclear Materials, 466:120–133, 2015.
[224] J.E. Angelo, N.R. Moody, and M.I. Baskes. Trapping of hydrogen to lattice de-
fects in nickel. Modelling and Simulation in Materials Science and Engineering,
3(3):289, 1995.
[225] A. Turnbull, R.G. Ballinger, I.S. Hwang, M.M. Morra, M. Psaila-Dombrowski,
and R.M. Gates. Hydrogen Transport in Nickel-Base Alloys. Metallurgical Trans-
actions A, 23(12):3231–3244, 1992.
281
[226] D.M. Symons. An investigation into the effects of hydrogen on the fracture and
deformation behavior of Alloy X-750, PhD Thesis. PhD thesis, Carnegie-Mellon
Univ., Pittsburgh, PA (United States), 1994.
[227] N.R. Moody, S.L. Robinson, S.M. Myers, and F.A. Greulich. Deuterium concen-
tration profiles in Fe-Ni-Co alloys electrochemically charged at room temperature.
Acta Metallurgica, 37(1):281–290, 1989.
[228] B.G. Pound. The effect of aging on hydrogen trapping in precipitation-hardened
alloys. Corrosion Science, 42(11):1941–1956, 2000.
[229] D.H. Lassila and H.K. Birnbaum. Intergranular fracture of nickel: the effect of
hydrogen-sulfur co-segregation. Acta Metallurgica, 35(7):1815–1822, 1987.
[230] A.W. Thompson and I.M. Bernstein. The role of metallurgical variables in
hydrogen-assisted environmental fracture. Advances in Corrosion Science and
Technology, 7:53–175, 1979.
[231] B.G. Pound. A comparison of hydrogen ingress behavior in Alloys 625 and 716.
Scripta Metallurgica et Materialia, 29(11):1433–1438, 1993.
[232] L.K. Mansur, E.H. Lee, P.J. Maziasz, and A.P. Rowcliffe. Control of helium
effects in irradiated materials based on theory and experiment. Journal of Nuclear
Materials, 141:633–646, 1986.
[233] E.O. Ezugwu, Z.M. Wang, and A.R. Machado. The machinability of nickel-based
alloys: a review. Journal of Materials Processing Technology, 86(1-3):1–16, 1999.
[234] S. Zamberger and E. Kozeschnik. Carbo-nitride precipitation in tempered
martensite-computer simulation and experiment. Material Science Forum,
706:1586–1591, 2012.
[235] P. Lang, Y.V. Shan, and E. Kozeschnik. The life-time of structural vacancies in
the presence of solute trapping. Material Science Forum, 794:963–970, 2014.
[236] J. Svoboda, F.D. Fischer, P. Fratzl, and E. Kozeschnik. Modelling of kinetics
in multi-component multi-phase systems with spherical precipitates: I: Theory.
Materials Science Engineering A, 385(1):166–174, 2004.
[237] E. Kozeschnik, J. Svoboda, P. Fratzl, and F.D. Fischer. Modelling of kinetics in
multi-component multi-phase systems with spherical precipitates: II: numerical
solution and application. Materials Science Engineering A, 385(1):157–165, 2004.
[238] K. Russell. Nucleation in solids: the induction and steady state effects. Advanced
Colloidal Interfaces, 13:205–318, 1980.
[239] B. Sonderegger B and E. Kozeschnik. Generalized nearest-neighbor broken-bond
analysis of randomly oriented coherent interfaces in multicomponent fcc and bcc
structures. Metallurgical Material Transactions A, 40:499–510, 2009.
282
[240] B. Sonderegger and E. Kozeschnik. Size dependence of the interfacial energy
in the generalized nearest-neighbor broken-bond approach. Scripta Materialia,
60:635–638, 2009.
[241] Available from: http://matcalc.tuwien.ac.at.
[242] R. Kampmann and R. Wagner. Kinetics of precipitation in metastable binary
alloys - theory and application to Cu-1.9 at % Ti and Ni-14 at % Al. In P. Haasen,
V. Gerold, R. Wagner, and M.F. Ashby, editors, Decomposition of Alloys: the
Early Stages, pages 91–103. 1984.
[243] U.F. Kocks and H. Mecking. Physics and phenomenology of strain hardening: the
fcc case. Progress in Materials Science, 48:171–273, 2003.
[244] L.E. Lindgren, K. Domkin, and S. Hansson. Dislocations, vacancies and solute
diffusion in physical based plasticity model for aisi 316l. Mechanics of Materials,
40:907–919, 2008.
[245] F. Roters, D. Raabe, and G. Gottstein. Work hardening in heterogeneous alloysa
microstructural approach based on three internal state variables. Acta Materialia,
48:4181–4189, 2008.
[246] S. Brinckmann, R. Sivanesapillai, and A. Hartmaier. On the formation of vacancies
by edge dislocation dipole annihilation in fatigued copper. International Journal
of Fatigue, 33:1369–1375, 2011.
[247] M.S. Soliman, E.A. El-Danaf, and A.A. Almajid. Effect of heat treatment condi-
tions on the high temperature deformation of 6082-AL alloy. Advanced Materials
Research, 83:407–414, 2009.
[248] R. Radis and E. Kozeschnik. Kinetics of Aln precipitation in microalloyed steel.
Modelling and Simulation in Materials Science and Engineering, 18(5):1–10, 2010.
[249] P. Warczok and G. Stechauner. Matcalc technical papers [matcalc documentation]
from http://matcalc.tuwien.ac.at/wiki/doku.php?id=techpapers. retrieved March
2016.
[250] H. Shoda K. Takai. Dynamic behavior of hydrogen desorption from pure iron and
inconel 625 during elastic and plastic deformations. Revista Materia, 15:267–274,
2010.
[251] T. Awane, Y. Fukushima, T. Matsuo, Y. Murakami, and S. Miwa. Highly sen-
sitive secondary ion mass spectrometric analysis of time variation of hydrogen
spatial distribution in austenitic stainless steel at room temperature in vacuum.
International Journal of Hydrogen Energy, 39(2):1164–1172, 2014.
[252] K. Takai and R. Watanuki. Hydrogen in trapping states innocuous to environmen-
tal degradation of high-strength steels. ISIJ International, 43(4):520–526, 2003.
283
[253] D. Gaude-Fugarolas. Effect of microstructure and trap typology on hydrogen
redistribution in steel. Metal 2013, pages 1–5, 2013.
[254] F.G. Wei and K. Tsuzaki. Quantitative analysis on hydrogen trapping of TiC
particles in steel. Metallurgical Transactions A, 37:331–353, 2006.
[255] J.R.G. da Silva and R.B. McLellan. The solubility of hydrogen in super-pure-iron
single crystals. Journal of Less-common Metals, 50:1–5, 1976.
[256] K. Kiuchi and R.B. McLellan. The solubility and diffusivity of hydrogen in well-
annealed and deformed iron. Acta Metallurgica, 31(7):961–984, 1983.
[257] J.R.G. da Silva, S.W. Stafford, and R.B. McLellan. The thermodynamics of the
hydrogen-iron system. Journal of Less-common Metals, 49:407–420, 1976.
[258] P. Sofronis and R.M. McMeeking. Numerical analysis of hydrogen transport near a
blunting crack tip. Journal of the Mechanics and Physics of Solids, 37(3):317–350,
1989.
[259] E.J. Song, D-W. Suh, and H.K.D.H. Bhadeshia. Theory for hydrogen desorption
in ferritic steel. Computational Materials Science, 79:36–44, 2013.
[260] E. Wallaert, T. Depover, M. Arafin, and K. Verbeken. Thermal desorption spec-
troscopy evaluation of the hydrogen-trapping capacity of NbC and NbN precipi-
tates. Metallurgical and Materials Transactions A, 45(5):2412–2420, 2014.
[261] S. Zhang et al. Effect of Nb on hydrogen-induced delayed fracture in high strength
hot stamping steels. Materials Science and Engineering A, 626:136–143, 2015.
[262] F-G. Wei and K. Tsuzaki. Hydrogen trapping character of nano-sized NbC pre-
cipitates in tempered martensite. Proceedings of the International Hydrogen Con-
ference, Wyoming, USA, pages 456–463, 2008.
[263] M. Ohnuma, J. Suzuki, F.G. Wei, and K. Tsuzaki. Direct observation of hydrogen
trapped by nbc in steel using small-angle neutron scattering. Scripta Materialia,
58(2):142–145, 2008.
[264] H.G. Lee and J-Y. Lee. Hydrogen trapping by TiC particles in iron. Acta Metal-
lurgica Materalia, 32(1):131–136, 1984.
[265] H.G. Lee and J-Y. Lee. The effect of the interface character of TiC particles on
hydrogen trapping in steel. Acta Metallurgica Materalia, 35(11):2695–2700, 1987.
[266] N. Winzer et al. Hydrogen diffusion and trapping in Ti-modified advanced high
strength steels. Materials and Design, 92:450–461, 2016.
[267] E. Wallaert, L. Duprez, A. Atrens, and K. Verbeken. Thermal desorption spec-
troscopy study of the interaction of hydrogen with TiC precipitates. Metals and
Materials International, 19(4):741–748, 2013.
284
[268] D. Perez-Escobar, L. Duprez, A. Atrens, and K. Verbeken. Thermal desorption
spectroscopy study of experimental Ti/S containing steels. Materials Science and
Technology, 29(3):261–267, 2013.
[269] G.M. Pressouyre and I.M. Bernstein. A quantitative analysis of hydrogen trapping.
Metallurgical Transactions A, 9(11):1571–1580, 1978.
[270] M.A. Stopher. The effects of neutron radiation on nickel-based alloys. Materials
Science and Technology, 33(5):518–536, 2017.
[271] A. Turnbull et al. Hydrogen transport in nickel-base alloys. Metallurgical Trans-
actions A, 23(12):3231–3244, 1992.
[272] B.G. Pound. The effect of ageing on hydrogen trapping in precipitation-hardened
alloys. Corrosion Science, 42(11):1941–1956, 2000.
[273] W.J. Qi et al. Hydrogen embrittlement susceptibility and hydrogen-induced addi-
tive stress of 7050 aluminum alloy under various aging states. Journal of Materials
Engineering and Performance, 24(9):1941–1956, 2015.
[274] O. Verners, G. Psofogiannakis, and A.C.T. van Duin. Comparative molecular
dynamics study of fcc-Al hydrogen embrittlement. Corrosion Science, 98:40–49,
2015.
[275] N.D. Alexopoulos, Z. Velonaki, C.I. Stergiou, and S.K. Kourkoulis. The effect of
artificial ageing heat treatments on the corrosion-induced hydrogen embrittlement
of 2024 (AlCu) aluminium alloy. Corrosion Science, 102:413–424, 2016.
[276] A. Laurino, E. Andrieu, J.P. Harouard, G. Odemer, J.C. Salabura, and C. Blanc.
Effect of corrosion on the fatigue life and fracture mechanisms of 6101 aluminum
alloy wires for car manufacturing applications. Materials Design, 53:236–249,
2014.
[277] J. Languillaume, G. Kapelski, and B. Baudelet. Cementite dissolution in heavily
cold drawn pearlitic steel wires. Acta Materialia, 45(3):1201–1212, 1997.
[278] J.-H. Kang and P. E. J. Rivera-Dı´az-del Castillo. Carbide dissolution in bearing
steels. Computational Materials Science, 67:364–372, 2013.
[279] G. Kresse and J. Furthmu¨ller. Efficiency of ab initio total energy calculations for
metals and semiconductors using a plane-wave basis set. Computational Materials
Science, 6(1):15–50, 1996.
[280] G. Kresse and D. Joubert. From ultrasoft pseudopotentials to the projector
augmented-wave method. Phys. Rev. B, 59:1758–1775, 1999.
[281] J.P. Perdew, K. Burke, and M. Ernzerhof. Generalized gradient approximation
made simple. Phys. Rev. Lett., 77:3865–3868, 1996.
[282] H.J. Monkhorst and J.D. Pack. Special points for brillouin-zone integrations.
Phys. Rev. B, 13:5188–5192, 1976.
285
[283] M. Methfessel and A.T. Paxton. High-precision sampling for brillouin-zone inte-
gration in metals. Phys. Rev. B, 40:3616–3621, 1989.
[284] P.E. Blo¨chl, O. Jepsen, and O.K. Andersen. Improved tetrahedron method for
brillouin-zone integrations. Phys. Rev. B, 49:16223–16233, 1994.
[285] D.E. Jiang and E.A. Carter. Diffusion of interstitial hydrogen into and through
bcc fe from first principles. Phys. Rev. B, 70:064102, 2004.
[286] E.-J. Song, H.K.D.H. Bhadeshia, and D.-W. Suh. Effect of hydrogen on the surface
energy of ferrite and austenite. Corrosion Science, 77:379–384, 2013.
[287] X. Zhang, T. Hickel, J. Rogal, S. Fa¨hler, R. Drautz, and J. Neugebauer. Structural
transformations among austenite, ferrite and cementite in Fe-C alloys: A unified
theory based on ab initio simulations. Acta Materialia, 99:281–289, 2015.
[288] Y. Gao, Z. Lv, S. Sun, M. Qu, Z. Shi, R. Zhang, and W. Fu. First principles study
on surface structure and stability of alloyed cementite doped with Cr. Materials
Letters, 100:170–172, 2013.
[289] A. Dick, F. Ko¨rmann, T. Hickel, and J. Neugebauer. Ab initio based determina-
tion of thermodynamic properties of cementite including vibronic, magnetic, and
electronic excitations. Phys. Rev. B, 84:125101, 2011.
[290] C.M. Fang, M.A. van Huis, and H.W. Zandbergen. Structural, electronic, and
magnetic properties of iron carbide Fe7C3 phases from first-principles theory. Phys.
Rev. B, 80:224108, 2009.
[291] I.R. Shein, N.I. Medvedeva, and A.L. Ivanovskii. Electronic and structural prop-
erties of cementite-type M3X (M=Fe, Co, Ni; X=C or B) by first principles cal-
culations. Physica B: Condensed Matter, 371(1):126–132, 2006.
[292] H.I. Faraoun, Y.D. Zhang, C. Esling, and H. Aourag. Crystalline, electronic, and
magnetic structures of θ-Fe3C, χ-Fe5C2, and η-Fe2C from first principle calcula-
tion. Journal of Applied Physics, 99(9):1–8, 2006.
[293] I.G. Wood, L. Vocˇadlo, K.S. Knight, D.P. Dobson, W.G. Marshall, G.D. Price, and
J. Brodholt. Thermal expansion and crystal structure of cementite, Fe3C, between
4 and 600K determined by time-of-flight neutron powder diffraction. Journal of
Applied Crystallography, 37(1):82–90, 2004.
[294] M. Umemoto, Z. G. Liu, K. Masuyama, and K. Tsuchiya. Influence of alloy
additions on production and properties of bulk cementite. Scripta Materialia,
45(4):391–397, 2001.
[295] H.P. Scott, Q. Williams, and E. Knittle. Stability and equation of state of Fe3C
to 73 GPa: Implications for carbon in the Earth’s core. Geophysical Research
Letters, 28(9):1875–1878, 2001.
286
[296] E. Duman, M. Acet, E. F. Wassermann, J. P. Itie´, F. Baudelet, O. Mathon, and
S. Pascarelli. Magnetic instabilities in Fe3C cementite particles observed with Fe
K-Edge X-Ray circular dichroism under pressure. Phys. Rev. Lett., 94:075502,
2005.
[297] R.C. Reed and J.H. Root. Determination of the temperature dependence of the lat-
tice parameters of cementite by neutron diffraction. Scripta Materialia, 38(1):95–
99, 1997.
[298] E. J. Fasiska and G. A. Jeffrey. On the cementite structure. Acta Crystallographica,
19(3):463–471, 1965.
[299] Y. Le Page and P. Saxe. Symmetry-general least-squares extraction of elastic
data for strained materials from ab initio calculations of stress. Phys. Rev. B,
65:104104, 2002.
[300] F. Mouhat and F-X. Coudert. Necessary and sufficient elastic stability conditions
in various crystal systems. Phys. Rev. B, 90:224104, 2014.
[301] R. Hill. The Elastic Behaviour of a Crystalline Aggregate. Proceedings of the
Physical Society. Section A, 65(5):349–354, 1952.
[302] W. Voigt. Lehrbuch der Kristallphysik. Teubner, Leipzig, 1928.
[303] A. Reuss. Berechnung der Fließgrenze von Mischkristallen auf Grund der Plas-
tizita¨tsbedingung fu¨r Einkristalle. ZAMM - Journal of Applied Mathematics and
Mechanics / Zeitschrift fu¨r Angewandte Mathematik und Mechanik, 9(1):49–58,
1929.
[304] S.F. Pugh. Relations between the elastic moduli and the plastic properties of
polycrystalline pure metals. The London, Edinburgh, and Dublin Philosophical
Magazine and Journal of Science, 45(367):823–843, 1954.
[305] D.H. Chung and W.R. Buessem. The Elastic Anisotropy of Crystals. Journal of
Applied Physics, 38(5):2010–2012, 1967.
[306] S.I. Ranganathan and M. Ostoja-Starzewski. Universal elastic anisotropy index.
Phys. Rev. Lett., 101(5):055504, 2008.
[307] J.F. Nye. Physical properties of crystals: their representation by tensors and
matrices. Oxford University Press, 1985.
[308] A. Migliori and J.L. Sarrao. Resonant ultrasound spectroscopy: applications to
physics, materials measurements, and nondestructive evaluation. Wiley, 1997.
[309] M. Umemoto, Z.G. Liu, H. Takaoka, M. Sawakami, K. Tsuchiya, and K. Ma-
suyama. Production of bulk cementite and its characterization. Metallurgical and
Materials Transactions A, 32(8):2127–2131, 2001.
[310] S.M. Lee and J.Y. Lee. The effect of the interface character of TiC particles on
hydrogen trapping in steel. Acta Metallurgica, 35(11):2695–2700, 1987.
287
[311] M. Peet and H.K.D.H. Bhadeshia. MUCG83 - a programme for the calculation of
TTT diagrams in steels, 2015.
[312] L. Lutterotti, S. Matthies, H.-R. Wenk, A. S. Schultz, and J. W. Richardson Jr.
Combined texture and structure analysis of deformed limestone from time-of-flight
neutron diffraction spectra. Journal of Applied Physics, 81(2):594–600, 1997.
[313] L. Lutterotti. A rietveld extended program: Materials Analysis Using Diffraction
(MAUD). http://maud.radiographema.eu, retrieved: 2016.
[314] B.A. Szost, R.H. Vegter, and Pedro E.J. Rivera-Dı´az-del Castillo. Hydrogen-
trapping mechanisms in nanostructured steels. Metallurgical and Materials Trans-
actions A, 44(10):4542–4550, 2013.
[315] T.N. Baker. Processes, microstructure and properties of vanadium microalloyed
steels. Materials Science and Technology, 25(9):1083–1107, 2009.
[316] R.G. Baker and J. Nutting. Precipitation processes in steels. ISI Special Report,
64:1–12, 1959.
[317] G. Kurdjumow and G. Sachs. Walz- und Rekristallisationstextur von Eisenblech.
Zeitschrift fu¨r Physik, 62(9):592–599, 1930.
[318] I. J. Langford and D. Loude¨r. Powder diffraction. Reports on Progress in Physics,
59(2):131–234, 1996.
[319] J. Takahashi, K. Kawakami, and T. Tarui. Direct observation of hydrogen-trapping
sites in vanadium carbide precipitation steel by atom probe tomography. Scripta
Materialia, 67(2):213–216, 2012.
[320] T. Epicier, D. Acevedo, and M. Perez. Crystallographic structure of vanadium
carbide precipitates in a model Fe-C-V steel. Philosophical Magazine, 88(1):31–45,
2008.
[321] Y. Oba, S. Koppoju, M. Ohnuma, T. Murakami, H. Hatano, K. Sasakawa,
A. Kitahara, and J. Suzuki. Quantitative Analysis of Precipitate in Vanadium-
microalloyed Medium Carbon Steels Using Small-angle X-ray and Neutron Scat-
tering Methods. ISIJ International, 51(11):1852–1858, 2011.
[322] Y.-S. Chen, D. Haley, S.S.A. Gerstl, A.J. London, F. Sweeney, R.A. Wepf, W.M.
Rainforth, P.A.J. Bagot, and M.P. Moody. Direct observation of individual hy-
drogen atoms at trapping sites in a ferritic steel. Science, 355(6330):1196–1199,
2017.
[323] T. Yokota and T. Shiraga. Evaluation of hydrogen content trapped by vanadium
precipitates in a steel. ISIJ International, 43(4):534–538, 2003.
[324] A.W. Thompson and J.A. Brooks. Hydrogen Performance of Precipitation-
Strengthened Stainless Steels Based on A-286. Matellurgical Transactions A,
6A(July):1431–1442, 1975.
288
[325] H.E Swanson. Standard x-ray diffraction powder patterns. US Department of
Commerce, 1953.
[326] A. Taylor and R. Floyd. The constitution of nickel-rich alloys of the Ni-Ti-Al
system. Journal of the Institute of Metals, 81:25–32, 1952.
[327] B. Kim, C. Celada, D. San Martin, T. Sourmail, and P.E.J. Rivera-Diaz-del-
Castillo. The effect of silicon on the nanoprecipitation of cementite. Acta Mate-
rialia, 61(7):6983–6992, 2013.
[328] J. Knott. Nuclear power plants: types, components and material requirements.
Structural Alloys for Power Plants, pages 69–101, 2014.
[329] L. Fielding, E.J. Sung, D.K. Han, H.K.D.H. Bhadeshia and D.W. Suh. Hydro-
gen diffusion and the percolation of austenite in nanostructured bainitic steel.
Proceedings of the Royal Society A, 470:20140108, 2014.
[330] K.G. Solheim, J.K. Solberg, J. Walmsley, F. Rosenqvist, and T.H. Bjrn. The
role of retained austenite in hydrogen embrittlement of supermartensitic stainless
steel. Engineering Failure Analysis, 34:140–149, 2013.
[331] D. Prez Escobar, T. Depover, L. Duprez, K. Verbeken, and M. Verhaege. Com-
bined thermal desorption spectroscopy, differential scanning calorimetry, scanning
electron microscopy and x-ray diffraction study of hydrogen trapping in cold de-
formed TRIP steel. Acta Materialia, 60(67):2593–2605, 2012.
[332] C. Garc´ıa-Mateo and F.G. Caballero. The role of retained austenite on ten-
sile properties of steels with bainitic microstructures. Materials Transactions,
46(8):1839–1846, 2005.
[333] M.J. Peet, P. Hill, M. Rawson, S. Wood, and H.K.D.H. Bhadeshia. Fatigue of
extremely fine bainite. Materials Science and Technology, 27(1):119–123, 2011.
[334] H.K.D.H. Bhadeshia. The first bulk nanostructured metal. Science and Technology
of Advanced Materials, 14(1):014202, 2013.
[335] H.K.D.H. Bhadeshia and M. Peet. Program: MAP Steel MUCG83.
http://www.msm.cam.ac.uk/map/steel/programs/mucg83.html, retrieved: 2016.
[336] H.K.D.H. Bhadeshia. Bainite in steels. Inst. of Metals, 1992.
[337] M.J. Peet. Transformation and tempering of low-temperature bainite. PhD thesis,
University of Cambridge, 2010.
[338] S. Allain J.P. Chateau, A. Dumay and A. Jacques. Precipitation hardening of a
FeMnC TWIP steel by vanadium carbides. Journal of Physics: Conference Series
(ICSMA-15), 240, 2010.
[339] J. Chiang, B. Lawrence, J.D. Boyd, and A.K. Pilkey. Effect of microstructure
on retained austenite stability and work hardening of {TRIP} steels. Materials
Science and Engineering: A, 528(1314):4516–4521, 2011.
289
[340] H.K.D.H. Bhadeshia. Nanostructured bainite. In Proceedings of the Royal Society
of London A: Mathematical, Physical and Engineering Sciences, volume 466, 2113,
pages 3–18. The Royal Society, 2010.
[341] Caltherm-UK. Industrial Furnaces: 650oC to 1300oC, 2016.
[342] Corrosion evaluation of thermally treated Alloy 600 tubing in primary and faulted
secondary water environments, EPRI NP-6721s-M. Technical report, Electric
Power Research Institute, 2008.
[343] L. Ferrer, B. Pieraggi, and J.F. Uginet. Microstructural Evolution During Ther-
momechanical Processing of Alloy 625. In Proceeding of the International Sympo-
sium on Superalloys 718, 625, 706 and Various Derivatives, pages 217–228. The
Minerals, Metals and Materials Society, 1991.
[344] HyperSpy: multi-dimensional data analysis toolbox, available at
http://hyperspy.org, retrieved: January 2017.
290