kth royal institute

of technology

Doctoral Thesis in Materials Science and Engineering

Creep Behavior of High

Temperature Cast Materials
for Exhaust Applications
CHRISTIAN ÖBERG

Stockholm, Sweden 2020

Creep Behavior of High
Temperature Cast Materials
for Exhaust Applications
CHRISTIAN ÖBERG

Academic Dissertation which, with due permission of the KTH Royal Institute of Technology,
is submitted for public defence for the Degree of Doctor of Philosophy on Friday the 18th of
december 2020, 09.00, Public defense by Zoom.

Doctoral Thesis in Materials Science and Engineering

KTH Royal Institute of Technology
Stockholm, Sweden 2020
© Christian Öberg

© Christian Öberg, Baohua Zhu, Stefan Jonsson, Paper 1.

© Christian Öberg, Ralf Rablbauer, Baohua Zhu, Stefan Jonsson, Paper 2.
© Christian Öberg, Baohua Zhu, Stefan Jonsson, Paper 3.
© Christian Öberg, Baohua Zhu, Stefan Jonsson, Paper 4.
© Christian Öberg, Baohua Zhu, Stefan Jonsson, Paper 5.

ISBN 978-91-7873-728-4
TRITA-ITM-AVL 2020:48

Printed by: Universitetsservice US-AB, Sweden 2020

Author’s contribution to appended papers
Appendix 1: “Plastic deformation and Creep of Two Ductile Cast irons, Simo51 and
Simo1000, during Thermal Cycling with Large Strains”.
Literature survey, mechanical testing, microscopy, wrote the paper with modifications from
supervisors.
Appendix 2: “Monotonic and Cyclic Creep of Cast Materials for Exhaust Manifolds”
Literature survey, mechanical testing associated with SRTC (stress relaxations with thermal
cycling), planning of STT (sequential tensile testing) and CL (constant load) creep tests, data
analyses associated with all test methods, SEM-microscopy, wrote the paper with
modifications from supervisors.
Appendix 3: “Creep, strain and oxidation damage in the ferritic ductile cast iron SiMo51 at
700°C”
Literature survey, tensile testing, oxidation testing, planning of CL tests, did data analyses
associated with all tests, microscopic imaging associated with LOM and regular SEM, did
EBSD microscopy together with Joacim Hagström, post-processing of EBSD-data, took part
in the discussion of the results of the thermodynamic calculations, wrote the paper, except
thermodynamic part, with modifications from supervisors.
Appendix 4: “Creep behaviour, creep damage and precipitation in the austenitic cast steel
HK30 at 750°C”
Literature survey, planning of the CL tests, data analyses, microscopic analyses associated
with LOM and regular SEM, did EBSD microscopy together with Joacim Hagström, post-
processing of EBSD-data, took part in the discussions of the results of the thermodynamic
Scheil simulation, wrote the paper, except Scheil simulation part, with modifications from
supervisors.
Appendix 5: “Creep behaviour, creep damage and precipitation in the austenitic ductile
cast iron D5S at 750°C”
Literature survey, planning of the CL tests, data analyses, LOM-imaging, EBSD microscopy
together with Joacim Hagström, post-processing of EBSD-data, wrote the paper with
modifications from supervisors.

1
Sammanfattning
Avhandlingen fokuserar på krypbeteende hos fyra gjutna material som används till avgasgrenrör i
lastbilsmotorer. Dessa material inbegriper två ferritiska segjärn, betecknade SiMo51 och SiMo1000,
ett austenitiskt segjärn, D5S, och ett austenitiskt gjutstål, HK30. Gjutjärnen är rika på kol och har en
mikrostruktur som utöver matrisen består av grafitnoduler och olika utskiljningar, de senare
huvudsakligen karbider och intermetalliska faser. Gjutstålet har ett lägre kolinnehåll och saknar därför
grafitnoduler, men innehåller andra utskiljningar. Under drift cyklas avgasgrenrören termiskt upp till
800 °C i ett låst tillstånd, eller mer specifikt fastskruvade till motorblocket. Lastfallet och miljön ger
upphov till krypdeformation, utmattning, oxidation och andra förändringar i mikrostrukturen.
Generellt sett har utvecklingen mot mer miljövänliga lastbilsmotorer lett till högre avgastemperaturer
vilket i sin tur innebär högre krav på materialen.
Det huvudsakliga målet med studien var att utreda krypbeteende och relaterade fenomen hos de
inkluderade materialen (se ovan). Ett sekundärt mål var även att jämföra resultat från tre olika typer
av kryptester: i) SRTC (stress relaxations with thermal cycling) är ett sätt att framkalla
spänningsrelaxationer i materialet genom att cykla det termiskt i ett låst tillstånd, ii) STT (sequential
tensile test) är ett dragprov vid konstant temperatur under vilket töjningshastigheten ändras vid givna
töjningsnivåer, iii) CL (constant-load creep test) är den traditionella krypprovningsmetoden, som
innebär konstant last och temperatur. SRTC och STT är avsedda som snabba och billiga
krypprovningsmetoder medan CL-metoden generellt betraktas som långsam och dyr. Resultat från de
tre metoderna jämfördes regelmässigt i s.k Norton plots, det vill säga logaritmiska krypdiagram med
spänning och töjningshastighet.
Resultat från i) SRTC (i kompression) och ii) STT (i drag) överensstämde generellt mycket väl vilket
indikerar att krypning i materialen inte beror av lastriktning. Utöver detta var töjningshastigheten som
registrerades i SRTC-testerna konstant med antalet cykler. Båda dessa upptäckter underlättar
modellering av cyklisk krypning i materialen, ett ämne som dock inte ingick i avhandlingen. Det finns
diskrepanser mellan den krypdata som erhölls med CL-metoden och den som erhölls med de snabbare
metoderna (STT och SRTC). Dessa kunde inte fullt ut förklaras, men olika alternativ diskuterades.
Hursomhelst är det givet att den tidsberoende krypskada som utvecklas under ett långsamt krypprov
(CL) inte kan uppstå efter snabba spänningsrelaxationer eller dragprov.
De mikrostrukturella förändringar som skedde under krypning dokumenterades genom LOM-, SEM-
och EBSD-mikroskopi, med flera olika tekniker för preparering och etsning.
I CL-provning vid 700 °C uppvisade SiMo51 primärkryp som mer eller mindre direkt följdes av
tertiärkryp. Tertiärkrypet var i sin tur uppdelat i två subregimer, av vilka den första assoceriades med
skada från kaviteter runt grafitnoduler och runt korngränser/karbider, och den andra med större
sprickor mellan grafitnodulerna. Den oxidation som skedde under krypning var påtaglig, men inte
tillräcklig för att förklara uppdelningen i två tertiärkrypregimer. Oxidationen på provytan och runt
nodulerna studerades explicit. Oxiden som helhet bestod av flera individuella lager som kunde
identiferas genom att kombinera EDX-data och termodynamiska beräkningar.
Både D5S och HK30 CL-testades vid 750 °C för att reflektera en högre driftstemperatur hos dessa
material jämfört med SiMo51. Efter långvarig krypexponering uppvisade HK30 typiska krypkaviteter
vid korngränserna, utskiljning av sigma-fas och G-fas, små oxidinträgningar och ett tunt
rekristalliserat lager på provytan. D5S å sin sida uppvisade olika typer av kaviteter/porer runt
grafitnodulerna, liknande den skada som noterades för SiMo51 vid 700 °C. Brottmekanismen bestod
av skjuvsprickor som propagerade mellan grafitnodulerna. Olika utskiljningar bildades under
krypexponeringen.

2
Abstract
This thesis focuses on creep of four cast materials intended for exhaust manifolds in heavy-duty truck
engines. Two of the materials are ferritic ductile cast irons, SiMo51 and SiMo1000, one is an
austenitic ductile cast iron, D5S, and another one is an austenitic cast steel, HK30. The ductile cast
irons, rich in carbon, have a microstructure with graphite nodules and precipitates, mainly carbides
and intermetallics. The cast steel, on the other hand, being meagre in carbon, has precipitates but lacks
graphite nodules. During service, the exhaust components are thermally cycled up to 800 °C in a
locked stated, bolted to an engine block. This gives rise to creep deformation, fatigue, oxidation and
microstructural changes. Driven by the development of environmental friendly engines of lower
emissions, the exhaust gas temperature is increasing, continuously leading to higher demands on the
materials.
The main aim was to investigate the creep behavior and related phenomena of the included materials.
A secondary aim was to compare results from three types of tests, i) SRTC (stress relaxations with
thermal cycling), provoking stress relaxations in a locked specimen subjected to thermal cycling, ii)
STT (sequential tensile test), changing the strain rate at selected strain levels during a tensile test at a
selected temperature, iii) CL (constant-load creep test), i.e. traditional creep testing, applying a
constant load at a given temperature. SRTC and STT are intended as quick and cheap methods while
CL is generally considered slow and associated with high costs. Results of the three methods were
regularly compared in Norton plots, i.e. double logarithmic plots of stress and strain rate.
Results of i) SRTC (in compression) and ii) STT (in tension) were generally in very close agreement
which indicates that creep of the included materials is independent of loading direction. In addition,
the creep rates obtained by SRTC were also constant with number of cycles. Both findings facilitate
modeling of cyclic creep, although this was not in the scope of the present thesis. There were
discrepancies between data sets of CL and SRTC/STT which could not be explained, although several
reasons were discussed. In addition, the time-dependent creep damage which develops during a slow
CL test is always missed in quick stress relaxation tests or tensile tests.
The microstructural events taking place during creep were documented using LOM, SEM and EBSD
microscopy techniques, with various etching and sample preparation procedures.
When CL tested at 700 °C, SiMo51 showed primary creep, more or less directly followed by tertiary
creep. The tertiary creep regime was in turn divided into two stages of which the first was associated
with the formation of typical creep cavities around the graphite nodules and at the grain boundaries,
and the second associated with larger cracks between the graphite nodules. Oxidation was significant
but not enough to be held responsible for the tertiary creep stages. The oxidation on the surface and
around the graphite nodules was explicitly studied. Layered oxides were identified by combining
EDX data with thermodynamic calculations.
Both D5S and HK30 were CL tested at 750 °C, reflecting a higher service temperature of these
materials compared with SiMo51. After prolonged creep exposure, HK30 exhibited typical creep
cavitation at the grain boundaries, precipitation of sigma phase and G phase, oxide intrusions and
recrystallization in a thin layer at the specimen surface. D5S exhibited various types of cavities/voids
around the graphite nodules (like SiMo51 at 700 °C) and fracture occurred by shear cracks growing
nodule-to-nodule. Various precipitates developed during creep.
Keywords: Creep, creep behaviour, creep damage, ductile cast iron, high-alloyed austenitic cast
steel, thermal cycling, tensile testing, stress relaxation, oxidation, precipitation.

3
Contents
1. Introduction ........................................................................................................................................................ 5
1.1. Development of exhaust manifold materials ................................................................................................... 6
1.1.1. SiMo51 ......................................................................................................................................................... 7
1.1.2. SiMo1000 ..................................................................................................................................................... 7
1.1.3. D5S ............................................................................................................................................................... 8
1.1.4. HK30 ............................................................................................................................................................ 9
2. Objectives ......................................................................................................................................................... 10
3. Theory............................................................................................................................................................... 11
3.1 The origin of creep ......................................................................................................................................... 11
3.2. Constant-load creep tests and creep curves ................................................................................................... 11
3.3. Five-power-law creep .................................................................................................................................... 13
3.4. Creep mechanisms, cavitation and rupture .................................................................................................... 14
3.5 Cyclic creep .................................................................................................................................................... 16
3.6. Creep and tensile testing of ductile cast iron ................................................................................................. 16
3.7. Oxidation of Fe, Fe-Si and Fe-C-Si-Mo (SiMo51) ........................................................................................ 17
3.8. SRTC (stress relaxations with thermal cycling) ............................................................................................ 18
3.9. STT (sequential tensile testing) ..................................................................................................................... 20
4. Experiments ...................................................................................................................................................... 21
4.1. Specimen preparation .................................................................................................................................... 21
4.2. SRTC (stress relaxations with thermal cycling) ............................................................................................ 22
4.2.1. Temperature Gradient ................................................................................................................................. 23
4.2.2. Extensometer measurements....................................................................................................................... 24
4.3. Sequential tensile test (STT) .......................................................................................................................... 27
4.4. Constant-load (CL) creep test ........................................................................................................................ 28
4.5. Regular, “non-sequential”, tensile tests ......................................................................................................... 29
4.6. Oxidation tests ............................................................................................................................................... 29
4.7 EBSD .............................................................................................................................................................. 29
5. Results and discussion ...................................................................................................................................... 30
5.1. Mechanical response to thermal cycling ........................................................................................................ 30
5.2. Comparing stress relaxations with thermal cycling (SRTC), sequential tensile tests (STT) and constant-load
(CL) creep tests..................................................................................................................................................... 31
5.3. Creep of SiMo51 at 700 °C, focusing on “Two-stage” tertiary creep............................................................ 34
5.4. Creep behaviour and damage of HK30 at 750 °C.......................................................................................... 37
5.5. Creep behaviour and damage of D5S at 750 °C ............................................................................................ 39
6. Summary........................................................................................................................................................... 43
7. Conclusions ...................................................................................................................................................... 44
8. Sustainability statement .................................................................................................................................... 44
9. Acknowledgements .......................................................................................................................................... 45
10. References ...................................................................................................................................................... 46

4
1. Introduction
Exhaust manifolds (see Fig. 1) are located in the beginning of the exhaust system of truck
engines. They collect the hot exhaust gas, flowing from the six or eight cylinders, and pass it
further along the exhaust system. Due to starts and stops of the engine, they become
subjected to thermal cycling during service. Since they are more or less locked to other
components, not allowed to expand or contract freely, stresses arise. To prevent failure, they
are usually designed in a way to reduce both stresses and strains in critical sections. In reality,
the highest stresses and strains usually arise at sharp transitions, i.e. sudden variations in
thickness (and thermal expansion). To avoid this, a “streamlined” design is generally desired.
The manifolds are cast in sand molds and regular materials are ductile cast irons and cast
stainless steels. The ferritic ductile cast irons are mainly used for diesel engines and the
austenitic cast steels for more demanding applications, e.g. the gasoline engine (otto-engine).
D5S may work for both types. The ferritic ductile cast irons are not traditional high
temperature materials and have to be spiced and processed in a special way to obtain the
required properties. For example, in the case of SiMo51, Si and Mo have been added to give
additional oxidation resistance and strength. In addition, the casting process is often
simulated beforehand, testing various geometries and configurations to get the desired
microstructure and properties.
Firstly, the materials are subjected to mechanical and thermal loads. The service condition
described above, involving thermal cycling of locked parts, is usually called thermo-
mechanical fatigue. It involves both thermal and mechanical loading – in cycles. Depending
on the location, the relationship between stress, strain and temperature differs. It is called
TMF-OP when the loading and temperature are Out of Phase and TMF-IP when they are In
Phase. For example, the inner surface of a manifold is compressed (-) when heated (+) since
thermal expansion is restricted by the cold area of the outer wall. Thus, there is a phase shift
of 180° between stress and temperature which is a clear case of TMF-OP. Due to the
frequency and lifetime of the thermal cycles in manifolds, this loading is usually associated
with LCF (low-cycle fatigue). HCF (high-cycle fatigue) is also present and originates from
vibrations from the road and other components. TMF-testing of ductile cast iron, and
sometimes with the effect of adding HCF signals, have been thoroughly described in several
studies ( [1], [2], [3], [4]).
Secondly, the manifolds are oxidized during service, both by hot exhaust gases on the inside
and by air on the outside. Most obviously, oxidation decreases the metallic area and may
therefore increase the stress. This is, however, not so trivial as the oxides are known to have
different load-carrying abilities. Another aspect is that of crack growth: it is well known that
an oxide may initiate for instance fatigue cracks but less well known that it may also have a
blunting effect on crack tips which may reduce stress concentrations. However, in spite of
these “dual” effects, oxidation is generally considered to do bad. In the context of exhaust
manifold materials, the effect of oxidation on fatigue has been studied in depth by Ekström
[5], and Xiang et al. ( [6], [7], [8], [9], [10]), among others.
Thirdly, and central for this thesis, is the creep deformation associated with the thermal
cycling. Creep is both a repair and damage process as it unloads the material but also leads to

5
strain (geometry change) and creep damage. The unloading is due to the transformation of
elastic strains into plastic strains, leading to a stress relaxation. Traditionally, creep damage is
associated with various sorts of cavitation at grain boundaries and their linkage into cracks.
As will be demonstrated, the situation is slightly different for the cast irons due to the
occurrence of big graphite nodules in the matrix. Besides the work of the present study, creep
of ductile cast irons has for instance been studied by Hug et al. [11].

Fig. 1. Exhaust manifold in service (in an engine test cell).

1.1. Development of exhaust manifold materials

As mentioned, the materials used for the exhaust manifolds are mainly ductile cast irons and
cast steels. Ductile cast irons are generally defined by a carbon content of minimum 2 wt%.
The ductile cast irons tested in the present study comprise a ferritic or austenitic matrix,
graphite nodules, carbides and/or intermetallics. The graphite phase has a lower density
compared to the metal phase and thus prevents shrinkage pores from developing during
solidification.
Compared with the cast steels, the ductile cast irons are generally cheaper and easier to cast.
The cast steels intended for the application, on the other hand, offer advantages such as
higher corrosion resistance and, according to Ekström [5], higher strength at elevated
temperatures. Ferritic grades (in general) have higher thermal conductivity and lower thermal
expansion, properties which are considered beneficial for the application. Regarding
fundamentals of creep, the activation energy of self-diffusion of iron in γ-austenite (310
kJ/mol) is higher compared to Fe in α-ferrite (280 kJ/mol), Cottrell [12]. For instance, Hug et
al. [11] referred to this when they observed more severe creep cavitation in a ferritic ductile
cast iron compared to an austenitic one.
The chemical compositions of the included materials are listed in Table 1. Since different
casting batches were used throughout the study, compositions are given in ranges according
to their respective standards. For the specific composition of each batch and the associated
experiments, see appendices 1-5.
Table 1. Chemical compositions (wt%, Fe bal.)

Alloy C Si Mn Ni Mg Cr P Cu Mo Al
SiMo51 3.0-3.5 4.2-4.8 <= 0.4 <= 0.1 0.02-0.08 <= 0.1 <= 0.05 <= 0.1 0.8-1.2 -
SiMo1000 3.0-3.9 2.0-3.2 <= 0.4 <= 1.0 - - - - 0.5-1.0 2.5-3.9
D5S <= 2.0 4.9-6.0 0.5-1.5 35.0-36.0 - 1.5-2.5 <= 0.08 <= 0.50 - -
HK30 0.25-0.35 0.5-2.0 - 19.0-22.0 - 23.0-27.0 <= 0.04 - <= 0.50 -

6
1.1.1. SiMo51
SiMo51 is probably the most common material for manifolds in diesel engines. It mainly
comprises a ferritic matrix, spherical graphite nodules and carbides. A micrograph of SiMo51
is shown in Fig 2, with one of the carbide segments of Fig 2a being highlighted in Fig 2b. All
parts of the segment are rich in Mo and some parts are additionally rich also in Si or Ti.
Besides these rather large structures, there are also small particles inside the grains, seen as
small black dots in Fig 2b. In creep theory, the larger grain boundary carbides are often
considered detrimental while small particles of appropriate structure and composition are
considered beneficial. The composition and crystal structure(s) of the carbides were not
completely verified but, searching elsewhere, Ekström et al. [13] list Mo- and Si-rich M6C
carbides.
The designation SiMo51 reflects the proportion of 5:1 between Si and Mo contents, see Table
1. Si provides increased “fluidity” of the melt during casting, solid solution strengthening and
oxidation resistance while Mo gives solid solution strengthening and carbide formation.
Oxidation resistance is derived from the formation of a thin, dense SiO2 barrier layer at
elevated temperatures [5], [14].

Fig 2. a) FSD (forward scattered diodes) image of the microstructure of SiMo51 and b)
EBSD band contrast image of the segment enclosed in a). (From appendix 3).

1.1.2. SiMo1000
SiMo1000 is a ferritic ductile cast iron which is similar to SiMo51 and is also used mainly for
diesel engines. It also contains Mo-rich carbides at the grain boundaries and smaller
precipitations inside the grains, see Fig 3. Differences are that it contains both spherical and
vermicular graphite and has a relatively high aluminium content of about 3 wt% (see Table 1)
which adds solid solution strengthening and changes the oxide barrier composition from SiO2
to Al2O3, see appendix 1. SiMo1000 is marketed as an improved version of earlier SiMo-
grades with better oxidation resistance, increased high-temperature strength and also a higher
ferrite-to-austenite transformation temperature [15]. It is also marketed as a cheaper
alternative to the austenitic D5S. The manufacturer and inventor, Georg Fischer Ltd,
recommends usage for applications such as turbocharger housings and exhaust manifolds
[15].

7
 Fig 3. Microstructure of SiMo1000.

1.1.3. D5S
Both regarding price and properties, the austenitic cast iron D5S is somewhere in between the
cheap, weak ferritic ductile cast irons (e.g. SiMo51, SiMo1000) and the expensive, strong,
oxidation-resistant, cast steels (e.g. HK30). The higher price is mainly due to the high Ni
content (≈34 wt%) which promotes formation of an austenitic matrix. D5S belongs to the Ni-
resist family which includes both flake and spheroidal (referring to graphite shape) cast iron.
The etched microstructure, taken from appendix 5, is shown in Fig 4. Casting shrinkage
porosities can be seen in a), graphite and mixed structures in b), and a close-up of a mixed
structure in c). The latter are “mixed” as they contain several phases, likely both
intermetallics and carbides. Unfortunately, they could not be fully identified using
quantitative EDX or EBSD, but EDX mapping showed that carbon rich areas are also rich in
Cr and that the intermetallics are rich in Si, Ni and, in some areas, also in Mn.

8
 Fig 4. a) Microstructure of D5S (shrinkage porosities are enclosed in red), b) higher
magnification, and c) a “mixed structure”. Electrochemically etched by applying 1.5V for
about 10s in 5 % HCl in ethanol. (From appendix 5).

1.1.4. HK30
As mentioned, HK30 can also be used for more demanding applications, e.g. gasoline
engines. A high Cr-content provides solid solution strengthening and precipitation
strengthening by carbide formation, and oxidation resistance through the formation of a thin
Cr-oxide film. A high Ni content promotes an austenitic matrix and also provides solid
solution strengthening. Compared with more common steels, such as austenitic 304 or 316
steels, the carbon content of HK30 is much higher (0.25-0.35 wt%). According to Piekarski et
al. [16], this is beneficial as all Cr is consumed in early formation of carbides and therefore
not available to form Cr-rich sigma phase. Sigma phase precipitation is generally considered
as detrimental for both mechanical and corrosive properties, but the true effect is complicated
and very much case-dependent (see literature review of appendix 4).
In the as-cast and subsequently heat-treated state, HK30 comprises an austenitic matrix and
M23C6 carbides, see Fig. 5 (from appendix 4). The carbides occur both as relatively thick
particles at the grain boundaries and as small dispersions inside the grains. The orientation
map of creep tested material in Fig 6 shows that one of the grains (i.e. dendrites) is at least 11
mm across. Since the creep specimens have a diameter of 10 mm, creep test results will
inevitably reflect properties close to those of a single crystal. There is very limited creep data
on HK30 available but there is more on HK40 (25Cr-20Ni-0.4C) which has a higher carbon
content of 0.4 wt%. For instance, HK40 has been studied with regard to creep by Whittaker et
al. [17] and Konosu et al. [18].

9
 Fig. 5. a) Microstructure of HK30. (1) points towards an area of shrinkage porosities. b)
segregation boundaries. (2) points towards M23C6 carbides and (3) towards a TiN-particle.
Images from appendix 4.

Fig 6. Orientation map of creep tested HK30 material. (59 MPa, 750 °C). Images from
appendix 4.

2. Objectives
The primary objective of the present thesis is to investigate creep behaviour of the included
cast materials, both in terms of creep parameters/curve characteristics and creep damage.
Since the ductile cast irons, SiMo51 and SiMo1000, have big graphite nodules and oxidize
rather heavily, the creep response is expected to be different from that of pure metals as well
as more traditional high-temperature materials, e.g. Ni-base superalloys, and highly alloyed
steels. HK30 is similar to such “super grades” in several respects but the casting defects and
the coarse-grained structure make it unpredictable. The austenitic ductile cast iron D5S is
interesting as it lies somewhere in between the SiMo-materials and HK30. It includes large
graphite nodules but also comprises a highly alloyed austenitic matrix.
The investigated grades are used in various types of exhaust truck engines all over the world
and are constantly exposed to creep – but have barely been creep tested. There may be
several reasons for this:

10
i) Service temperatures have steadily increased with engine development, implying a
gradually increasing contribution from creep deformation. Thus, creep may not have been
that critical in the past.
ii) Compared with regular tensile tests, constant-load creep tests are more complicated, much
more expensive, and generally more time-consuming (obviously depending on load and
temperature). Additionally, since they are not controlled by deformation rate, it is more
difficult to estimate times to rupture in constant-load creep tests and thus to select suitable
loads.
iii) Large steel companies have financed creep testing of processed steel. The foundries are
often operated by relatively small companies with less resources to finance such research.
iv) Cyclic creep, which is the case of manifolds, has perhaps been considered too complicated
to account for.
As a second objective of the thesis, connecting with ii) and iv), attempts have been made to
compare faster methods, more specifically cyclic stress relaxations and monotonic, sequential
tensile testing, with the monotonic constant-load creep test method.

3. Theory
3.1 The origin of creep
Creep refers to slow, time-dependent plastic deformation at high temperature, usually starting
at around half of the melting temperature. As stated by Kassner [19], creep is enabled due to
a decrease in yield stress, derived both from high temperature and low deformation rate.
The temperature dependence arises since atomic motion is assisted by vibrations and because
the atoms are able to climb and diffuse around obstacles, Manson and Halford [20]. Another
effect is that the atomic bonds are longer and weaker at higher temperatures.
The strain-rate dependence can be explained by the smaller force required to move atoms
slowly [21], explaining why creep is considered a “below the yield stress” phenomenon. As
exemplified by Manson and Halford [20], the yield stress obtained in a slow tensile test of,
for instance, 10-7 s-1, is often much lower compared to the “conventional” yield stress which
is obtained in a regular tensile test (≈10-4 s-1).

3.2. Constant-load creep tests and creep curves

Constant-load creep testing is the conventional way of testing creep. Regarding control
parameters, it is quite simple since the load is kept constant at constant temperature,
throughout the whole test. The general trends with stress are illustrated by three curves in Fig
7. The four main segments of a creep curve (1-4) have been indicated in the case of the
lowest stress, 𝜎1 .
(1) marks the initial loading, seen as a vertical shift of each creep curve.
(2) marks the primary creep regime, characterized by an increase in dislocation density and a
resulting decrease in creep rate.

11
In the secondary creep regime (3), the deformation hardening in (2) is balanced by
dislocation annihilation, i.e. recovery, resulting in a constant strain rate. For this regime to
develop, it is also necessary that the microstructure is stable. The secondary creep regime is,
however, an ideal case. In real life creep testing, the strain rate is seldom, or in very rare
cases, completely constant as some type of “damage” sets in and give at least a slight increase
in strain rate [22].
Finally, (4) marks the tertiary regime in which accumulated creep damage, crack growth and
a reduced load-carrying area accelerate deformation to the point of creep rupture, often
indicated by an ‘x’.
A lower stress (𝜎1 ) usually leads to a longer secondary creep (in time), a longer time to
rupture and a low strain to rupture [19]. The time and strain to rupture are naturally related:
longer times usually mean more evolved creep cavitation and a more brittle fracture, with
only small elongation upon rupture. Correspondingly, relatively high stresses (𝜎3 ) mean
negligible or shorter secondary creep, less evolved creep cavitation and a more ductile
fracture. Regular characteristics are then: large strain-induced voids, shear cracks, necking
and high strain to rupture [23].
In order to better understand the sequence of events during a creep test, creep strain rates are
usually plotted, either as a function of time or of strain, see illustration in Fig 8. The creep
strain-rate curves observed in the present study can be explained by general dislocation
and/or crack theories. The trends are again illustrated by three curves, one for each stress, see
Fig 8. For each curve, the red marker illustrates the lowest value of strain rate. If the strain
rate stays on this level, it is called the “steady-state” creep rate. If it instead shows an
inflection point, i.e. a direct transition, the rate is simply referred to as the “minimum creep
rate”. A direct transition is made possible by, for instance, cavitation, early crack growth,
oxidation, or in any other way an unstable microstructure. Since complete steady state is
more of a theoretical or “ideal” state, the minimum creep rate is usually used when presenting
results, for instance in a Norton plot which includes the minimum creep rate and applied
stress [22].
The trends illustrated in Fig 8 are, however, valid both in the case of a steady state and in the
case of a direct transition. Real plots of strain rate usually have log-10 scale (semi-log or log-
log). Regarding vocabulary, “hardening” and “softening” are sometimes used to point
towards segments of decreasing and increasing creep rates, respectively.
In the ε̇ -t-curve (Fig 8a), the steady state/minimum creep rate is reached earlier for higher
stresses. This is quite natural as a high stress generates many dislocations which quickly start
annihilating – a requirement for steady state. Additionally, it is natural to have early crack
growth, leading to a direct transition. In the case of a lower stress, there is instead a slow
deformation hardening, and it takes a long time to build the dislocation density needed for
wide-spread annihilation. It also takes longer time for cracking (causing a direct transition) to
take place since the creep damage accumulates more slowly.
In the ε̇ -ε-curve (Fig 8b), the situation is reversed: the steady state/minimum creep rates are
generally reached at lower strains for lower stresses. A high strain is simply not needed to
reach either steady state or cause early cracking – if things are allowed to develop slowly.
“Things” include a high dislocation density, leading to a steady state, or cracks, leading to a

12
direct transition. The general trend that a higher strain is needed to reach steady state if the
stress is high is also emphasized by Kassner, see Fig. 5 in [19] (originally published in [24]).

Fig 7. Typical creep curves obtained from a constant-load creep test.

Fig 8. Typical scenario for three stress levels with: Creep strain rate as a function of a) time
and b) strain. (x = rupture).

3.3. Five-power-law creep

For most pure metals and many alloys, the “steady-state” creep rate is linked to the applied
stress through Norton’s law [25]:
𝜀̇𝑠𝑠 = 𝐴0 exp⁡[−𝑄𝐶 /kT] (𝜎𝑆𝑆 /𝐸)𝑛 (1)
𝜀̇𝑆𝑆 = secondary creep rate, 𝐴0 = constant, 𝑄𝐶 = activation energy for creep, k = the
Boltzmann constant, E = Young’s modulus, n = Norton exponent.
As outlined, in real life creep testing we usually present a minimum creep rate as the creep
rate is practically never completely constant [22]. Because of the power-law relationship in
eq. 1, minimum creep rates are often plotted as a function of stress in double logarithmic,

13
Norton plots. The slope (n) is basically a measure of how dislocations are created and
annihilated in the material. The minimum creep rate is selected as that is the point where
dislocations are created and annihilated with the most even rate [22]. For pure metals and
many alloys, the slope (n) is then usually around 5. Because of these circumstances,
expressions like “five-power-law creep” or simply “power-law creep” are often used. The
limit is not very distinct and n-values of about 4-7 fall within five power-law creep [19]. If
the n-value is the same over the tested stress range, it indicates that the creep mechanism is
constant. A change in n (in any direction) is usually referred to as “power-law breakdown”
and indicates that a new creep mechanism becomes predominant.
The microstructural events taking place during creep, referring to the “dislocation” length
scale, are well-documented within the range of “five-power-law creep”. When plastic
deformation commences, the dislocation density increases and subgrains form. They arise as
a low energy geometry of the introduced dislocations and typically give rise to
misorientations of around 1°. Inside these subgrains, so-called Frank networks form. These in
turn consist of single dislocations, arranged in “criss-cross” substructures [19].
The subgrain boundaries and their movement are often linked to the creep strength of a metal.
It is preferred to have a lot of small dispersions inside the grains which “pin” the subgrains,
thereby lowering creep rates [26].

3.4. Creep mechanisms, cavitation and rupture

Creep deformation is controlled by various mechanisms which may be more or less active,
depending on stress and temperature. The relationships are sometimes presented in Ashby
deformation maps, named after their inventor. An illustration is shown in Fig 9. Each of the
four regions are associated with a deformation mechanism.
Dislocation creep (2) consists of both dislocation glide and climb where the climb step, often
called “recovery”, is a form of diffusion by which entire dislocation segments annihilate each
other. Even though the glide step creates almost all of the strain, the climb step sets the rate,
and as a result the mechanism is often termed “climb-controlled” [27]. Within five power law
creep (see paragraph 3.3) the activation energy of creep (QC) has been observed to closely
follow that of lattice self-diffusion (QSD). This has led to the nowadays widely accepted
conclusion that creep, within this regime, is climb-controlled (since climb is a form of
diffusion) [19].
“Diffusional flow” is usually divided into Coble and Nabarro-Herring creep (3-4). In the case
of Coble creep, it refers to a stress-induced flow of matter along grain boundaries, while in
Nabarro-Herring creep, it refers to flow by lattice diffusion. Coble creep is more easily
activated and occurs at lower temperatures and stresses while Nabarro-Herring is activated at
higher temperatures and stresses [27]. As mentioned, creep tests are rarely conducted in these
regions as strain and damage evolve too slowly. In this context, it can be pointed out that
clear documentations of diffusional flow are rare, and it has only been documented in a few
creep studies on steel. This includes work by Nilsson et al. [28] who did constant-load creep
tests on austenitic (20Cr, 30Ni, Ti stabilized) steel at 800 °C. They found a distinct transition
with a change in Norton exponent from n=1 to n=2.7 at 5 MPa, which was related to a change
from Coble creep to power-law (dislocation based) creep. By TEM-imaging, they could see

14
that no matrix or twin-boundary dislocations occurred in the case of 3 MPa whereas
significant amounts could be found in specimens tested above 5 MPa.

Fig 9. Ashby deformation map

In an extensive review by Ashby et al. [23], the mechanisms of Fig 9 are related to the
formation of creep cavities and the subsequent fracture. They point out that ductile fracture at
lower temperature is basically controlled by the formation of large, strain induced ovals
around inclusions, often followed by intragranular fracture. However, as the temperature is
raised, a new type of fracture is viable. Ashby et al. [23] call it “intergranular, creep-
controlled fractured”. Under these conditions, grain boundaries start sliding, wedges are
initiated, and creep cavities grow under mysterious circumstances. Since grain boundaries
typically concentrate stress, they are likely to be found as initiation spots for cavitation.
The mechanism of creep cavity nucleation is widely debated and to this day not verified.
However, there are several theories, of which Kassner [19] lists the following: i) grain
boundary sliding, ii) vacancy condensation at a high stress interface, iii) dislocation pile-up.
Grain boundary sliding, i), is by some believed to nucleate cavities due to the stress
concentrations arising particularly at triple points, although it is debated whether the stresses
really are big enough to nucleate cavities. Vacancy condensation, ii), refers to the diffusion-
based accumulation of vacancies into cavity “nuclei”. Dislocation pile-up, iii), may readily
occur against hard particles and grain boundaries, following a so-called Zener-Stroh
mechanism. The theory, reviewed by Kassner [19], suggests that the stress concentrations
arising from the pile-ups are sufficient to nucleate cavities.
In the review paper by Ashby et al. [23] another, perhaps related explanation, is also listed.
This proposes that cavities are nucleated by pure diffusional flow at stress-concentrating
surfaces but later grow in a coupled state. In this mode, pure diffusional flow is active on a
local scale around a single cavity but coupled to global areas of power-law “creep cages”
which set, or at least influence, the local diffusion rate around the cavity.

15
3.5 Cyclic creep
Few studies have involved how creep behaviour or creep rates are affected by a reversal of
the load. One reason for this may be that most traditional creep test rigs cannot do the
reversal from tensile to compressive loading as they use dead weights to load the specimen.
In that case, other machines, likely designed for fatigue experiments, are needed. Another
may be that compressive creep is not considered an issue since any cracks remain closed.
However, for the application of manifolds, creep rates in compression are also of interest to
get the net strain after many cycles.
On this topic, Swindeman [29] tested Inconel at 816 °C, comparing monotonic creep tests in
tension to cyclic creep tests, using the same (absolute) load values. In the cyclic tests, the
load was reversed when the strain reached a desired level, meaning that the test was load-
controlled but strain-triggered. Quite remarkably, the obtained times to rupture in monotonic
tensile loading were up to 4 times shorter at higher stresses and 2 times shorter at lower
stresses compared with the cyclic creep test. From this, Swindeman concluded that the time
under tensile loading is critical. He also pointed out the quite obvious: the stress increases in
a monotonic tensile test due to a decrease in load-carrying area but essentially remains the
same in a cyclic test (if the net strain is controlled to zero). A bit surprising, the measured
strain rates were higher in the cyclic creep test (although primary creep accounted for most of
it). Manson and Halford [20] explain the matter further by assuming a creep mechanism of
grain boundary sliding along the 45°-direction (with respect to load). They suggest that strain
rates are higher in tension due to less “friction” between the atoms compared with shear in
compression (along the same boundary).

3.6. Creep and tensile testing of ductile cast iron

Creep studies on cast irons are quite scarce, but a few were found. For example, Martinsson
et al. [30] tested a ductile cast iron at 100 and 125° up to 41,000 h. The material was tested
for the purpose of nuclear waste storage, or more specifically the supporting frame of the
copper canisters used to store the waste fuel. Results showed logarithmic creep, i.e. creep
rates which after some time barely could be measured. A closer microscopical investigation
showed no creep cavitation.
Being more relevant for the present study, Hug et al. [11] constant load tested three ductile
cast irons, two ferritic and one austenitic, between 650 and 900 °C. The presented creep
curves showed small primary creep followed by small or negligible secondary creep and
dominating tertiary creep. Quite remarkably, all data could be fitted to one Monkman Grant
line, indicating one common creep mechanism for ferritic and austenitic cast irons. The
damage was reported as plastic flow of the matrix in the high stress/short-time specimens and
diffusion-based cavitation in the low stress/long-term specimens. Oxidation was also
observed, but the layers were reportedly too thin to have a significant effect on creep.
The deformation behaviour of ductile cast iron has been extensively tested - using high
deformation rates. In a study by Hervas et al. [31], ferritic SiMo cast iron was tested under
tensile and compressive loading at temperatures up to 800 °C. Up to 400 °C, the tensile test
curves showed deformation hardening which was, however, replaced by efficient recovery
above a test temperature of 500 °C. In tension, at 400 °C and below, damage was reported as
plastic deformation and cracking of the nodules. The nodules were also damaged through an

16
“onion-like” mechanism. In this case, the outer shell, obtained by solid state diffusion during
solidification, was peeled off from the inner core (which is formed directly from the melt).
At higher temperatures, above 400 °C, there was significant plastic flow of the ferritic matrix,
even more pronounced void formation around nodules and a higher strain to rupture. In
compression, below 400°C, damage was reportedly plastic flow of the ferritic matrix and
fractured nodules at high strains. Above 400 °C, plastic flow was reported, but neither cracks
nor voids could be observed.
The significance of graphite nodules for plastic properties has been discussed in several
studies. For example, Sjögren et al. [32] tested flake and compacted cast iron and suggest
micro-yielding in the vicinity of graphite particles. They similarly suggest that higher stresses
lead to void formation at the graphite-matrix interface.
Furthermore, Dong et al. [33] did in-situ observations as they tensile tested a ferritic ductile
cast iron directly inside the SEM. They could see the following stages of deformation: 1) slip-
lines in the elastic regime, 2) small decohesion at the graphite-matrix interface right after
exceeding the macroscopic yield stress, 3) large ellipsoidal voids around graphite nodules, 4)
facture by shear cracks, linking the voids together, 5). In their modelling of the tensile test
curves, which reportedly was successful, they assumed a porous material in which the
graphite nodules were simply replaced by voids. The assumption was supported by the
observation that voids develop at the matrix-graphite interface immediately upon yielding (2)
without any “resistance” from the nodules.

3.7. Oxidation of Fe, Fe-Si and Fe-C-Si-Mo (SiMo51)

If oxidation is severe, it may influence creep both regarding crack growth and decrease in
load-carrying area. In any of these cases, the effect of oxidation on the registered creep strain
values can be considered and compared with other effects such as creep damage growth and
reduction of load-carrying area due to pure straining. Of the materials included in the present
study, the three cast irons SiMo51, SiMo1000 and D5S were prone to oxidize heavily
compared to the cast steel, HK30. Especially in CL-testing of SiMo51 at 700 °C, oxidation
was pronounced with a mean thickness of about 500 µm (10 % of the original specimen
radius) along the gauge of the most prolonged specimen (339 days at 12 MPa). (However, as
described more detailed in appendix 3, only half of the oxide was estimated to reduce the
metallic area).
The theory surrounding oxidation of metals states that oxidation occurs outwards from the
base metal due to migration of metal ions towards air, and inwards due to migration of
oxygen ions towards the base metal [5]. For stoichiometric reasons, scales of FeO, Fe3O4 and
Fe2O3 form when pure iron oxidises (ordered from the base metal to air).
The addition of Si to Fe is interesting as SiMo51 is included in the present study. Although
addition of Si also has other effects, for instance on casting and strength properties, the main
reason is often to obtain a barrier of SiO2 during service at high temperatures. As it is
generally dense and uniform, such a scale has the ability to effectively prevent further
oxidation. Since oxide growth generally slows down by time due to protection from existing
scales, growth rates tend to follow a logarithmic relationship with time. By adding Si, this
effect becomes even stronger. Another situation, not desired, is when the oxides spall off,
continuously refreshing the exposure of metallic surface to air. This is common, in for

17
instance TMF-cycling, were the difference in thermal expansion and plastic properties
between oxide and metal sometimes leads to spallation, i.e. oxides “jumping” off the surface.
Regarding pure Fe-Si alloys, Liu et al. [34] oxidation tested three grades of 2.75, 4.97 and
6.96 wt% Si. As the 4.97 wt% Si grade (close to SiMo51 in Si content) was exposed at 700°C
for 24h, a thin, covering Si-oxide formed together with “oxide nodules”. The nodules
included Fe2O3, growing outwards from the base metal, and Fe3O4, growing inwards, mixed
iron and Si oxides and, innermost, a thick Si-rich internal oxidation layer, also containing
SiO2 particles.
In the even more specific case of SiMo cast iron, Ekström et al. [13] did oxidation tests of
SiMo51 at 700 and 800 °C, observing Fe-oxides growing around graphite nodules and
carbides followed by “island growth” (similar to “oxide nodules”). The more uniform scale,
obtained after “steady-state”, consisted of layers of Fe2O3 and Fe3O4, both growing outwards,
and layers of Fe3O4, spinel and SiO2, growing inwards.
Ebel et al. [14] also did oxidation tests of a SiMo ductile cast iron, at 700 and 800 °C.
Similarly, two outwards growing layers of Fe2O3 and Fe3O4, and two inwards growing layers
of Fe3O4 and Fe2SiO4 were observed. The outwards growing layers were reportedly filled
with porosities while the inward growing oxides contained trace marks of old spheroids.

3.8. SRTC (stress relaxations with thermal cycling)

This type of test starts by heating of a typical tensile test specimen under zero load
(continuous unloading) to a desired start temperature, also acting as the lower limit of the
thermal cycling test. From then on, the machine is instructed to keep the position of the
movable grip completely still, thereby locking the specimen in its full length (in the present
case the lower grip was locked as the upper grip is always fixed). This condition is active
throughout the whole test without interruption.
Following this instruction, the specimen is thermally cycled between the lower and upper
temperature limits. Since the specimen is locked, heating results in a compressive stress while
cooling leads to a tensile stress. Between heating and cooling, there are hold times of constant
temperature. If the hold time temperatures and stresses are high enough, there are stress
relaxations arising from creep. See example of HK30, cycled between 700 and 800 °C (Fig
10). In each reversal, the stress builds up until the yield stress is reached. Since the yield
stress was often reached during the ramps, before the hold times, the hold time was generally
preceded by some pre-deformation. Only this hold time data, with accompanying stress
relaxation, were compared to data obtained by the two other methods (STT and constant-load
creep-testing).
Interpretation of results is complicated as the measured strain includes thermal, elastic and
plastic contributions. Thermal and elastic strains directly follow temperature and stress.
Plastic strains include plastic flow and creep strain, the former by pure dislocation glide and
the latter by some creep mechanism (e.g. dislocation creep). The contributions from elastic,
thermal and plastic strains could be varied by changing the settings of the thermal cycle, i.e.
the temperature limits and/or the heating/cooling rates or hold times.

18
The creep rates during hold times could be obtained by first calculating the true strain, this is
𝜎
plotted in Fig 10, and then removing the elastic strain contribution, ( ), to obtain the plastic
𝐸
strain.
After the true plastic strain had been obtained, the data was standardized in order to facilitate
curve fitting of the strain evolution during the hold times, a procedure which is described
more detailed in appendix 2.
The following function, combining a linear and exponential relationship, was used for curve
fitting of (standardized) strain evolution with time:
𝑦𝑆 = 𝑝1 + 𝑝2 ∙ 𝑥𝑆 + 𝑝3 ∙ 𝑒𝑥𝑝(𝑝4 ∙ ⁡ 𝑥𝑆 )
The standardized experimental time, 𝑡𝑆 , and standardized experimental strain, 𝜀𝑆 , were used
to evaluate 𝑝1 to 𝑝4 by minimizing the error of:
(𝑥𝑆 , 𝑦𝑆 ) = (𝑡𝑆 , 𝜀𝑆 )
during a relaxation. The modelled strain vs. time was obtained by reverting the
standardization procedure:
𝑥 − 𝑥̅
𝑥𝑆 =
𝑠𝑡𝑑(𝑥)
𝑦 − 𝑦̅
𝑦𝑆 =
𝑠𝑡𝑑(𝑦)
where 𝑥̅ and 𝑦̅ represent the average values and std defines the standard deviation.
The corresponding strain rate during relaxation was calculated from:
𝑑𝜀 𝑑𝑦 𝑠𝑡𝑑(𝑦) 𝑑𝑦𝑆 𝑠𝑡𝑑(𝑦)
𝜀̇ = = = ∙ = {𝑝 + 𝑝3 𝑝4 ∙ 𝑒𝑥𝑝(𝑝4 ∙ 𝑥𝑆 )}
𝑑𝑡 𝑑𝑥 𝑠𝑡𝑑(𝑥) 𝑑𝑥𝑆 𝑠𝑡𝑑(𝑥) 2
The same function of 𝑦𝑆 was used to fit the stress relaxation data (during hold times) which,
as can be seen in Fig 10, follow a similar evolution with time. This curve fitting was
necessary for removing scattering in the stress evolution during relaxation, mainly arising
from temperature fluctuations which are directly converted to stress fluctuations in the locked
specimen.

19
 Fig 10. Cycling of HK30 between 700 and 800 °C by
heating/cooling rates of 200 °C/min and hold times of 3 min. The
temperature is indicated on the
stress axis and has been vertically shifted by 600 °C.

3.9. STT (sequential tensile testing)

In the present study, sequential tensile tests (STT:s) were carried out to see if the results
agreed with, or in any way could be related to those obtained from traditional constant-load
creep tests. The meaning of “sequential” is that the strain rate is changed multiple times in
sequences over the strain range of one single tensile test. Each time, the applied strain rate
results in a “saturation stress”, ideally also corresponding to “steady-state”. Thus, this is a
direct way of testing strain-rate sensitivity. A fundamental difference between the methods is
the control parameter: tensile tests are controlled by strain rate and creep tests by load. But is
there any difference in the output? Kassner [19], who instead calls the test a “strain-rate
increase” test, points out that the saturation stress obtained in a tensile test can be treated as
steady state if it has been reached due to a balance between hardening and dynamic recovery
processes. However, Kassner adds that the stress at which there is constant dislocation
structure is often “ambiguous”. Another complication, discussed in appendix 2, is that
deformation hardening taking place in sequences of high strain rate affect the following
sequences of the test due to deformation hardening. To sum up, there is a lot to consider
when comparing the constant-load creep tests and tensile tests.
The same procedure was used for all STTs in the present study. As described, the test is
controlled by strain rate but there are trigger points at certain strain levels where the strain
rate is changed. In the first sequence (0-2 % strain), a high strain rate of 10-4 s-1 is set. The
strain rate is then lowered in steps (2-4) down to 10-7 s-1 followed by a new increase each 0.5
% of new strain up to 3.3·10-3 s-1. See Table 2 and Fig 11. It can be noted that the test routine
is designed to evenly distribute data in a logarithmic diagram. The resulting Norton plots of
each test may include more or fewer than eight points, for instance due to early fracture
(below 5.5 % strain) or oscillations in stress or strain rate leading to discarding points or
letting two values from the same segment indicate the spread. In the past, Appelt [35] has
used the very same STT method (and equipment).

20
Table 2. Order of sequences in STT.
Sequence 1 2 3 4 5 6 7 8
ℇ [-] 0.000- 0.020- 0.025- 0.030- 0.035- 0.040- 0.045- 0.050-
0.020 0.025 0.030 0.035 0.040 0.045 0.050 0.055
ℇ̇ [s-1] 10-4 10-5 10-6 10-7 3.3·10-6 3.3·10-5 3.3·10-4 3.3·10-3

Fig 11. STT test of SiMo51 at 700 °C.

4. Experiments
4.1. Specimen preparation
Specimens were manufactured from cast plates, see Fig 12a. In order to avoid casting defects,
the plates were cast in connection to large cylinders which were cut off and thrown away.
Only the mid-section was used for manufacturing of test specimens, as indicated by the
schematic test specimen added to the illustrated in a). Various foundries were involved:
HK30 material was cast by Smålands gjuteri AB, Eksjö, Sweden, SiMo51 material by
Castings P.L.C, West Midlands, England, SiMo1000 by Georg Fischer Eisenguss GmbH,
Herzogenburg, Austria and D5S by Josef Brechmann GmbH & Co, Germany.
Specimen geometries are shown in Fig 12: b) CL, c) STT and d) tensile test specimens (used
for both SRTC and regular tensile testing).

21
 a) b)

c) d)
Fig 12. Geometries of a) Cast plates and specimens used for b) CL, c) STT and d) SRTC and
tensile tests.

4.2. SRTC (stress relaxations with thermal cycling)

A test rig with a 100 kN alignment fixture, a 25 kN Instron load cell, water-cooled grips and
an EDC-580 V Doli control system were attached to a load frame which was originally
intended for an Instron 8561 machine, see Fig 13. An induction system of 5kW (by
Teknoheat AB) and an induction coil of copper is used for heating. There are two possible
options of measuring strain: by a laser extensometer, relying upon speckle pattern tracking
(Laser Xtens compact TZ, Zwick Roell), or by a contact extensometer with aluminum-oxide
pins (MTS 632.53F-14). A few tests were also conducted in inert atmosphere, see appendix 1.
For that practice, a gas-tight chamber setup, developed by Ekström [5], was used, see Fig
13b. Thermocouples of S-type were spot-welded to the surface, following the TMF validated
code of practice of a small distance (<1 mm) between the two weld points [36].

22
 a) b)
Fig 13. a) Test rig used for SRTC and b) air-proof chamber.

4.2.1. Temperature Gradient

Compared with a furnace, a clear disadvantage of using induction heating is the inevitable
temperature gradient developed in the test specimens. The gradient is for instance affected by
convection, thermal and electrical conductance and radiation, as well as any magnetic
transition. Since experiments involving thermal cycling were conducted (SRTC), the gradient
and its possible effects on the results had to be elucidated. Fig 14 shows results from a
measurement of the gradient using the ferritic ductile cast iron SiMo51, involving three
thermocouples which were welded 5 mm below, right at, and 5 mm above the middle of the
gauge length. The middle thermocouple was used for control and thus shows the set
temperature in distinct ramps.
As can be seen in Fig 14, the gradient increases steadily as the temperature is raised. This is
related to heat conductance; the temperature difference between the hot zone and the
surroundings becomes larger and larger with the grips of the test rig acting as heating sinks,
Verma et al. [37].
At about 700 °C, the Curie temperature is reached. This marks a magnetic transition from
ferromagnetic (always magnetic with internal magnetic domains) to paramagnetic (magnetic
only in the presence of an external field). At this point, the gradient instantly becomes almost
zero. Verma et al. [37] explains that in the ferromagnetic range, both joule and hysteresis
heating are active. Joule heating is caused by the eddy currents generated by the alternating
magnetic field and “hysteresis” refers to heat generated from magnetic hysteresis losses. In
the paramagnetic state, hysteresis heating vanishes, and an extra energy input is needed to
continue heating the specimen at the same rate. From this, see Fig 14, it appears very likely
that the point of the middle thermocouple first becomes paramagnetic, leading to an increase
in effect (the energy input) which increases the temperatures in the surrounding, still
ferromagnetic parts of the gauge, quickly pushing the gradient to almost zero. With this result
in mind, the effect on thermal cycling tests (e.g. SRTC) is that the strains measured by the
contact extensometer (L0=12 mm) are less accurate at lower hold temperatures, below the
Curie temperature. That is, when the extensometer pins are pressing against colder areas than
what is given by the control thermocouple, strains in the very middle will be underestimated
as the measurement length of 12 mm in that case is too long. At temperatures above the Curie

23
temperature, the more homogenous temperature leads to more homogenous deformation
along the measurement zone, and measurements are more precise.

Fig 14. Temperature recorded 5 mm above, at, and 5 mm below the middle of the gauge
length.

4.2.2. Extensometer measurements

As mentioned, it was possible to do the SRTC tests using either a contact or laser
extensometer. In appendix 1, the laser extensometer was used, while in appendix 2, the
contact extensometer was used. It was found that the two devices generally show the same
strain evolution, i.e. the same strain curve characteristics, but that the signals differ in strain
level (by several factors).
Fig 15a shows a specimen lying horizontally, with the two ends of the thermocouple (TC)
having a weld point below the middle winding of the induction coil (Ind. coil). L0, the
measurement length, is set by changing the distance between the two green boxes of master
and slave. The strain measurement relies upon surface pattern tracking; as the laser light hits
the specimen surface, a speckle pattern forms based on the scattering angles of the laser light.
During a test, the individual speckle patterns in the master and slave boxes are tracked by a
video camera and the distance between them is converted to a strain.
Clear advantages using this quite novel measurement technique are non-existing preparation
times and a low risk of damaging the specimen or extensometer device.
The contact extensometer, see Fig 15b, measures the strain as the distance between two pins
of aluminium oxide, pressed onto the specimen surface. It has a fixed measurement length, in
this case of L0 = 12 mm. An advantage is the matter of trust: this type of device has been used
for a long time and has a good track record. A clear disadvantage is the physical contact point
between the pins and the specimen which may initiate cracks.

24
 a) Tracking of speckle patterns using the laser extensometer. The green tracking boxes,
thermocouple (TC), induction coil and specimen are shown in a horizontal view. Arrows
indicate the axial direction of the specimen. The weld point of the TC is below the middle
winding of the induction coil marked Ind. coil (where the two TC wires meet).

b) Strain measurement using the contact- and laser extensometers simultaneously.

Fig 15.
To validate the performance and compare results between the two extensometers, a test was
carried out using both of them simultaneously, see experimental setup in Fig 15b, above.
HK30 was SRTC-tested with a lower and an upper limit of 500 and 800 °C, respectively. In
Fig 16a, the strain-time evolutions of the two devices are shown with (a) temperature and (b)
stress also plotted in dashed lines using the second y-axis. For pedagogical reasons, the strain
is given in mm instead of the usual [-] or [%]. There are three segments of the routine:
heating to the lower (start) temperature (1), pre-cycling in zero load control (2) and load by
thermal cycling in a locked state (3).
As can be seen in steps (1) and (2), the two devices show exactly the same strain when the
load is zero. Then, as the specimen is locked and loading by thermal cycling starts (3), the
extensometers register strains of about -0.1 mm (contact extensometer) and -0.3 mm (laser
extensometer) at the end of the first compression. It can be commented that Lo=12 mm for
both devices, fixed for the contact extensometer and manually set for the laser extensometer.
Thus, the results show that there is an actual difference in measured strain, a difference which
grows with the number of cycles.
So, which device shows the correct strain? A way of estimating is to look at the thermal strain
during pre-cycling at zero load (2), which is basically a dilatometry test between the lower
and upper temperature limits. This shows the strain contribution originating purely from
thermal expansion. During thermal cycling in a locked state, this thermal strain, along with a

25
small elastic strain, is transformed into plastic strain. Thus, the measured strain is not
expected to be much higher than the thermal strain during pre-cycling. Since it is slightly
higher in the case of the contact extensometer, and several times higher in the case of the
laser extensometer, the laser extensometer appears to show unreasonably high strain.
One reason for this potentially large error is that the laser extensometer cannot keep track of
the speckle patterns during deformation. The speckle pattern depends on the scattering angle
of the diffracted laser light which depends on surface topology, for example including the
orientation of surface-exposed grains. With this in mind, the tracked speckle patterns in the
green boxes are probably changed because of strain itself (leading to a changed diffraction).
The strained grains give a change in speckle pattern making it move which is then
misinterpreted as a strain. (This idea was suggested by Joakim Lindblom, see
acknowledgements). However, since the two devices show the same character of the strain
curve, they can both be used to investigate tendencies in strain. For the present study, this
means that the strains presented in appendix 1 are most likely too high but, nevertheless,
show the correct behaviour.
It can be commented that a similar measurement was made for SiMo51 which showed the
same result.

26
 a)

b)
Fig 16. Strains registered during thermal cycling (SRTC) of HK30 using the contact and laser
extensometers. Temperature (a) and stress (b) were also plotted using the second (right) y-
axes. (1) Heating to the lower (start) temperature. (2) Pre-cycling at zero load, (3) Load by
thermal cycling in a locked state.

4.3. Sequential tensile test (STT)

STT was carried out at Volkswagen, Wolfsburg, Germany. A Zwick/Roell Kappa 050DS test
rig was used, see Fig 17. The machine allows testing up to 1400 °C using a radiation furnace.
Six thermocouples (the middle used for control) were tied to the specimen. The temperature
gradient was about ±2 °C over the measuring length. Strains were measured using two
ceramic contact extensometers (see Fig 17b) which were fastened to the knife edges of the
specimens.

27
 a) b)
Fig 17. a) Zwick/Roell Kappa test rig used “Sequential tensile tests”. b) Ceramic
extensometers setup.

4.4. Constant-load (CL) creep test

These tests were carried out at the research institute of Swerim, located in Kista, Sweden.
Tests were conducted following the international creep standard SS-EN ISO 204:2018 [38].
Instead of the traditional way of applying the load through dead weights, a step motor was
used. The step motor loads the load cell and specimen and uses the feedback signal from the
load cell to control the load to a constant value, obtaining ±3 N from the set-point (well
within the requirement of 1 % of the load). Strain was measured with two contact
extensometers which were fastened to the knife edges of a specimen, obtaining an accuracy
of ±100 nm. Three thermocouples were tied to the gauge length, with the middle one used for
control. A temperature gradient of ±1 °C was obtained which is also well within the
requirement of ±4 °C between the set and obtained temperature.
An illustration of the setup of the creep test rig is shown in Fig 18a and an image of the
specimen, the two extensometers and two of the thermocouples (a third upper one is missing
in the image) is shown in Fig 18b.

28
 a) b)

Fig 18. Setup of constant-load creep tests. a) Schematics of the rig, b) Specimen,
extensometers and two thermocouples (a third is yet to be tied above these two).

4.5. Regular, “non-sequential”, tensile tests

Some regular, non-sequential tensile tests were carried out, see appendix 3. In those tests, an
MTS-810 servo-hydraulic test rig, with axial loading capacity of 100 kN, was used. The rig is
equipped with an induction heating system identical to the one used in the test rig for SRTC.
The temperature gradient is about ±10 °C over the measuring length (as told by an operator of
the machine). A similar contact extensometer, attached with aluminium oxide pins (L0=12
mm), was used for measuring the strain. The test rig is located at Scania CV AB, Södertälje,
Sweden.

4.6. Oxidation tests

In the third study, see appendix 3, oxidation testing was also conducted in air at 700 °C using
a Carbolite CST 12/70 furnace. The furnace, originally intended for a tensile test rig, was
flipped and placed horizontally on a desk. It was opened through hinges and specimens were
placed directly on the inner ceramic lining. The area where the specimens were placed was
checked by manually measuring the temperature using a thermocouple (type K). In this area,
there were deviations of ±5 °C from the set temperature (700 °C).

4.7 EBSD
In all EBSD-measurements, specimens were polished using oxide particles (Buehler
Colloidal Silica polishing suspension). A LEO 1530 FEG-SEM, with the EBSD detector
“Symmetry”, was used. Various step lengths were used depending on the accuracy needed.
29
For post-processing of data, the software analysis platform Aztek was used together with the
“Tango” software, included in the “Channel 5” package. Local misorientations were
calculated in each pixel as the average of the misorientations to the eight surrounding pixels.

5. Results and discussion

5.1. Mechanical response to thermal cycling
In the first study (see appendix 1), thermally induced, cyclic deformation was studied through
SRTC (stress relaxations with thermal cycling). The study included SiMo51 and SiMo1000.
The method, described in previous sections, is similar to TMF-OP (out of phase) but differs
regarding the control parameter. By standard [4], TMF-OP is carried out using strain control
while SRTC is instead controlled by specimen length. More precisely, in TMF-OP the strain
signal, measured by an extensometer attached to the specimen gauge section and calibrated
for a specified measurement length, is used to control the movement of one of the specimen
grips, producing distinct ramps in strain (strain amplitudes). In SRTC, the positions of both
grips are instead kept constant, meaning that the specimen is locked in its full length. The
idea is that even if the total specimen length is constant, local strains can be registered by the
extensometer within the measurement length. Thus, by having a high temperature and high
stress along the gauge, it is possible to provoke and measure local plastic deformation, still
keeping the full specimen length constant. Due to the condition of constant specimen length,
such a deformation, for instance a heavy compression right in the middle of the gauge section
appearing during heating, must be balanced by elastic specimen extension in surrounding,
slightly colder areas of the gauge section, working as loading springs. This deformation
produces a thicker mid-section which during cooling, and corresponding reversion of stress
into tension, resists plastic deformation better than the surroundings. Hence, new local
deformations in tension may occur in the specimen gauge section but slightly (or partly)
outside the measurement length. This process may be repeated every cycle, changing the
specimen geometry considerably. Indeed, this was directly observed: in some cases, the
middle part of the gauge was very thick while areas a bit further away were considerable
thinner compared with the original diameter.
A possible advantage of using this procedure compared to TMF-OP is possibly that the
condition of a locked specimen length and spontaneous strain evolution is more realistic with
regard to the real conditions of exhaust manifolds (compared to strain control). It is also
beneficial that the method gives valuable output without strain-control, which is sometimes
difficult to achieve. Even with the most reliable extensometers, the PID-parameters of the
system need to be trimmed in, which is not trivial.
Most importantly, the study in appendix 1 led to an understanding of the mechanical response
during thermal cycling and also showed the effect of changing the various parameters. The
illustration from appendix 1 is shown below in Fig 19 and is representative of the behaviour
of both SiMo51 and SiMo1000. Hypothetical yield stress curves were added as grey, dashed
lines and the segments of the loop were coloured in blue or red for elastic and plastic events,
respectively.
The specimen is locked at (1), and as it is heated a compressive stress develops (1-2), see Fig
19a. Looking at the corresponding elastic strain change, b), it is practically zero. At (2), the

30
yield stress is exceeded followed by a stress relaxation (2-3), see a), and heavy compressive
straining, see b). In reality, deformation hardening was small and is exaggerated in (2-3). (3-
4) mark the high temperature hold which was introduced to allow further creep deformation
during constant temperature. For later investigations, the second paper, appendix 2, this was
the most important segment. During (4-5), the load is reversed by cooling of the specimen. At
(5), the yield stress is exceeded on the tensile side, giving rise to a strain increase in b). Then,
as the yield stress becomes higher at lower temperatures, another segment of elastic
deformation (6-7) is followed through. (7-8) is the low temperature hold which, if this lower
temperature is high enough, allows strain increase due to creep deformation, see b). In (8-9),
the load is again reversed by heating. Since the stress is insufficient to cause yielding, (8-9)
only gives small elastic straining, see b). Due to the curvature of the hypothetical yield curve,
plastic deformation occurs again but only temporarily at 9-10, leading to substantial increase
in strain. Elastic deformation (10-11) proceeds until the yield stress is reached, creating a new
compression. Since the cycles did not overlap, the specimen “climbed” towards positive (as
illustrated) or negative strains. Climb towards negative strains was more common and
resulted in “barrelling” which could be readily observed after many cycles.

Fig 19. Illustration of SRTC (1.5 thermal cycles) with stress a) and strain b).
Blue = elastic segments, Red = plastic segments.

5.2. Comparing stress relaxations with thermal cycling (SRTC),

sequential tensile tests (STT) and constant-load (CL) creep tests
In the second study (see appendix 2), the method SRTC was dedicated to the purpose of
obtaining creep data which was compared with results of STT and CL. All four materials
were tested: SiMo51, SiMo1000, D5S and HK30. SiMo51 was tested by all three methods
while the other three were only tested by SRTC and STT. SRTC and STT are generally quick
methods, or at least they are intended as such, while the CLs are costly and associated with
longer times. Obviously, the time to acquire data depends on settings of stress and
temperature; for instance, temperature interval and heating/cooling rates (SRTC), selected
strain rates (STT) and set load (CL).
In Fig 20, a comparison of SRTC and STT results from testing of HK30 is shown in a Norton
plot. The SRTC test was conducted at 300-800 °C but only the stress relaxation and strain

31
rate evolutions during the high temperature holds (800 °C) of cycles 1-91 have been included
in the plot. For STT, data at 800 °C, including obtained strain rates and saturation stresses,
are included with the associated regression line. As can be seen, the data sets agree almost
perfectly. From this and similar plots including the other materials, it was concluded that
creep in tension, by tensile testing, and compression, by stress relaxations, are the same for
these materials. It can also be noticed that creep rates remain unchanged with the number of
cycles. This is valuable to know in various fatigue lifetime simulations which need input of
cyclic creep parameters, which in this particular case are the same as the monotonic ones. For
instance, the effect of adding creep models to fatigue lifetime simulations of exhaust
manifolds have been studied by Seifert et al. [39], [40].
The same tendencies, with small modifications, could be seen for the other three materials
SiMo51, SiMo1000 and D5S. See more details in appendix 2.
Comparing all three methods, applied to SiMo51 at the upper limit test temperature of 700
°C, things became more complicated. In the case of SRTC, an extra-long hold time of 8.3 h
was introduced to really enter the creep regime in compression. Just as for other comparisons,
only the high temperature (700 °C) hold time data was evaluated. As can be seen in Fig 21,
the SRTC and STT results agree almost perfectly at higher stresses, just as for HK30 in Fig
20. At lower stresses, where the results of the CLs can be found, the strain rate evolution of
the SRTC test abruptly goes down to practically zero, just as if there was a critical stress for
creep deformation. The CLs fall about one decade lower in strain rate compared with STT
and SRTC (found at higher stresses), but the data sets have about the same slope, indicating a
similar creep mechanism. The cause of the shift of one decade in strain rate could not be fully
verified. However, in appendix 2, oxidation is briefly discussed as a possible reason. The
thought behind this is that the growing oxides increase the specimen diameter during primary
creep and thus lower the stresses/minimum creep rates in relation to those of SRTC and STT
(where times for oxidation were much shorter). Quite naturally, the credibility of this theory
depends on the load-carrying ability of the oxide scale as a whole, which is not established.
To avoid confusion, it can be commented that in estimations of the effect of oxidation on
creep strain rates (appendix 3) the load-carrying ability was instead set to zero by assuming
that the inwards-growing oxide layer did not carry load. See paragraph 5.3, or most detailed,
appendix 3. Another reason could be that different casting batches were used for SiMo51 in
the CL tests (one batch) compared with STT/SRTC (another batch), which could give rise to
differences in microstructure (even if such could not be observed). The chemical
compositions of the batches were generally very similar, but one difference was that the batch
used for the CL test of SiMo51 contained 0.20 wt% Ni whereas that used for STT/SRTC
contained 0.03 wt% Ni which could give a difference in solution strengthening.
The critical stress for creep deformation, as mentioned above, was only observed in the
SRTC test. A possible explanation can be traced to the level of pre-deformation associated
with each test. To begin with, the SRTC stress relaxation included in Fig 21 was preceded by
deformation (compression) of 2.1 % during the heating ramp prior to the hold time. In
addition, 1.2 % strain accumulated during the actual hold time. With the STT tests, the
present settings (see Table 2) always result in a pre-deformation of 2 % strain (by tensile
stress). In the CL test of 14 MPa, on the other hand, pre-deformation was only 0.4 % at the
point of reaching the minimum creep rate. Thus, pre-deformations were much larger for the
two faster methods compared with the slow CL test. This could explain the creep stop, i.e.

32
why the strain rates of the SRTC test go down to zero at about 14 MPa whereas the CL test of
14 MPa still gives a measurable minimum creep rate of about 10-9 s-1, see Fig 21.
Regarding prolonged “steady state” or a direct transition to tertiary creep, the compressive
SRTC sequences probably only allow steady state since cracks generally do not grow in
compression. However, a comparison is meaningful in the case of STTs and CLs. The STT
tests, due to the large pre-deformation (2 %), probably exhibit a steady state rather quickly.
This is confirmed by the plateau of the first sequence (10-4 s-1) of the 700 °C STT test of
SiMo51, previously shown in Fig 11. Additionally, the following sequences in Fig 11 also
show plateaus (except the second), indicating steady state. The CLs, on the other hand,
quickly go into tertiary creep (although the curves level out before, see paragraph 5.3). An
explanation of this difference is the aspect of time: in a creep test, damage is allowed to grow
more slowly and cause a direct transition. In the tensile tests, dislocations are instead quickly
pumped into the material and a steady state is reached just as fast – but there is not enough
time for damage to grow.
Another viable cause for a quick transition is oxidation, see paragraph 5.3.

Fig 20. HK30, cycled between 300-800 °C. Results at 800 °C, including the stress-strain rate
evolution of the hold time (SRTC) and strain rates-saturation stresses (STT).

33
 Fig 21. Norton plot of SiMo51 including STT, CL and SRTC (hold time of 8.3 hours in
compression) at 700 °C.
The comparison between CL and STT could also be briefly commented for the materials
HK30 and D5S, which were both tested at 750 °C in STT and CL.
For HK30, STT at 750 °C was not included in appendix 2, since the deformation hardening
observed in the STTs at 750 °C was too heavy and no “plateau”, i.e. saturation stress, was
reached in any of the sequences. Thus, no sound comparison could be made with the CL tests
at 750 °C in that case. A suggestion for future work could be to test HK30 by avoiding high
deformation rates (e.g. 10-4-10-3 s-1).
In the case of D5S, the CL data at 750 °C was not shifted to a lower level compared to the
STT data at 750 °C, but in that case the slopes of the two Norton plots (the n-values) differed
instead. In summary, there is no consistency in the difference in results between the STT and
CL for the materials. Hence, the discrepancies could simply be related to differences in
microstructure and chemical composition between the various casting batches which were
used for each test.

5.3. Creep of SiMo51 at 700 °C, focusing on “Two-stage” tertiary

creep
In the third paper, see appendix 3, the same CL specimens as in appendix 2 were investigated
regarding the microstructural damage resulting from the creep tests. Additionally, this
damage was compared to that of regular tensile tests and also to oxidation tests. All tests were
carried out at 700 °C since that was the temperature of the CL tests.
The true strain rate curves of the four CL tests are shown in Fig. 22 with strain rate as a
function of strain (log-log). As can be seen, primary creep regimes are followed by tertiary
creep. However, by plotting the very early parts of the creep curves (primary creep) with a
linear time scale instead of a logarithmic strain scale, see Fig. 23, it can also be observed that
the curves more or less “level out” before entering tertiary creep, indicating that they for a
short while were in, or close to, a steady state. Fig. 23 is a close-up of Fig. 8 in appendix 3

34
and the additional points (black squares) were in that case used to evaluate the true strain
rate-time dependence.

Fig. 22. True strain rate (creep rate) as a function of true strain. SiMo51, tested at 700 °C.
Numbers beside points indicate two hypothetical stress evolutions of the 12 MPa sample
taking area reduction into consideration with respect to volume conservation (Strain Comp)
and with respect to surface oxidation (Ox. Comp), respectively. (From appendix 3).

Fig. 23. The first part of the constant-load creep tests of SiMo51 with a linear time scale.
Referring to Fig. 22 again, primary creep was followed by a “first” tertiary creep stage which
is seen as a linear segment in the log-log plot of true strain rate as a function of strain. This is
in turn followed by a “second” tertiary stage, being more accelerated in nature. A question
being raised was how the tertiary stage could be subdivided into two stages, i.e. what caused
the first stage. Three possible reasons were considered: i) reduced load-carrying area due to
pure straining (volume conservation), ii) reduced load-carrying area due to oxidation and iii)
softening from creep damage. In calculations of the various contributions (Fig. 22), the 12
MPa specimen served as an example.

35
The contribution from pure straining, i), was obtained by calculating the true stresses at
certain true strains, using the technological stress of 12 MPa and the constant-volume
approach. The stress was then inserted into Norton’s creep law (obtained by regression
analyses of the minimum creep rates) to obtain a “strain-compensated” strain rate. See the
black line, labelled “Strain Comp”, in Fig. 22. As can be seen, the effect of pure straining on
stress and resulting creep strain rate is rudimental during the “first” tertiary stage. Therefore,
it cannot be held responsible for the measured increase in strain rate.
Secondly, the effect of oxidation, ii), was estimated by plotting the oxide thickness obtained
from one (separate) oxidation test and the four creep tests. For more details, see appendix 3.
In Fig. 22, each strain corresponds to a time which, from the oxidation curve, was translated
into an oxide thickness. Since about half of the oxide grew inwards and decreased the
metallic area, the total oxide thickness, plotted in the oxidation thickness curve, was divided
by 2 and removed from the original specimen radius. The remaining area was finally
translated into “oxidation-compensated” stresses which, by Norton’s law, in turn were
translated into strain rates, see “Ox. Comp” in Fig. 22. As a last check, the purely metallic
area of each fractured specimen was measured with the light optical microscope. These areas
were translated to stresses (since the load is constant) and strain rates (again through Norton’s
law) and are indicated by filled, coloured circles. As can be seen, the oxidation compensated
strain rate is not high enough to account for the measured rate. Neither is the strain rate
calculated from the purely metallic area. It is also interesting to note that in the “first” tertiary
stage, the last strain rate of “Ox. Comp” (marked 13.24) agrees perfectly with the strain rate
calculated from the purely (measured) metallic area (filled blue circle). This confirms that
practically all area reduction was from oxidation and that the original radius was correctly
reduced by only half of the total oxide thickness.
Rejecting the options listed above, the increasing strain rate in the “first” tertiary stage was
attributed to accumulating creep damage, iii). The existence of such creep damage was
confirmed through microscopical investigations. In Fig 24, the various types of creep damage
are indicated, this time with the 14 MPa specimen serving as an example. (1) is a typical form
of creep cavitation, located at grain boundary carbides. (2) is cavitation around the graphite
nodules. In quick tensile testing, such damage also occurred but in the shape of large ovals
around the nodules. In the creep tests, such ovals are clear signs of locally elevated
deformation rates. (3) is the linkage of ovals/cavities/nodules into larger cracks. When
connected to air, this type of “dimple” crack is oxidized, see (4). (5) indicates the outer oxide
scale.
Connecting this to the discussion on the “two-stage” tertiary creep, (1-2) are related to the
first stage and (3) to the second stage of tertiary creep.

36
Fig 24. Various types of damage in SiMo51, CL tested by 14 MPa at 700 °C (tr= 114 days).

5.4. Creep behaviour and damage of HK30 at 750 °C

In the fourth paper, the austenitic cast steel HK30 was creep tested at 750 °C, applying loads
corresponding to initial stresses of 50, 53, 59, 70 and 95 MPa. The higher temperature of 750
°C, compared to 700 °C for SiMo51, was selected as HK30 is regularly used for more
demanding applications, e.g. manifolds in gasoline engines. Compared with SiMo51, this
austenitic steel showed a more conventional creep behaviour. At higher stresses, there were
direct transitions between primary and tertiary creep whereas at lower stresses, more
prolonged secondary creep was accomplished, see Fig 25.
It can be commented that the creep curve of the 70 MPa specimen initially exhibited an
unexpectedly low creep rate relative to the others which, however, instantly increased to a
new level. Since the new level agreed better with the evolution of the rest of the creep curve
and also with the rates of the other tests (by the Norton plot), it was selected as the “real”
minimum creep rate. It can be noted that without this adjustment, the data would not follow
Norton’s law. The phenomenon was possibly attributed to a sudden activation of creep in one
of the large grains by a chain-reaction with another grain (a redistribution of the stress). See
appendix 4 for more details.
In addition to the conventional creep curves, conventional creep cavitation was observed,
especially in the most long-lasting specimens tested at 50 MPa and 53 MPa. Since

37
documentation of the 53 MPa specimen was not included in the paper, appendix 4, it is added
here instead. As can be seen in Fig. 26, showing unetched material, creep cavitation grows as
isolated segments which, in other spots, see b), have connected into larger cracks.
Other effects during creep were surface re-crystallization and particle precipitation. Both are
featured in Fig. 27. In Fig. 27a, an orientation map of a cross-section including a grain-
boundary creep-crack along with the re-crystallized specimen surface (bottom) is shown. As
can be seen, the crack is located at a grain boundary of the original, un-recrystallized material
whereas only a hint of a crack can be observed in the recrystallized layer. This indicates an
effective grain boundary sliding mechanism in the recrystallized layer, preventing the crack
from entering this region. Furthermore, much smaller recrystallized grains can be found near
the crack which is indicative of a dynamic recrystallization process.
Regarding the creep testing artefact from recrystallization, it was considered to not
fundamentally influence creep as the layer was too thin. Altogether, it seems as though the
recrystallized layer was like a ductile skin of the material, but without adding any apparent
strength. The original material cracked along the original grain boundaries, seemingly with
no relation to the recrystallized layer. As a result, the crack may be difficult to see on the
specimen surface, although being very clear inside the material.
In the original, untested HK30 material, only M23C6 precipitations could be found. During
creep exposure, two new phases precipitated: sigma phase and, likely, G phase. The
distribution of the three phases is shown in Fig. 27b, a highlighted area of Fig. 27a. As can be
seen, the sigma phase was bulkier compared with the others and was preferentially located at
original, un-recrystallized grain boundaries and also recrystallized grain boundaries. The
cracks do not grow through the precipitations but around them, as evidenced by Fig. 27b.
Thus, the particle-matrix interface appears to be the weak link of HK30.
As previously shown, Fig 6, HK30 has a very big grain size of several mm (up to cm size).
From a creep perspective, this may be an advantage as the grain boundary area is relatively
small. However, only a very few, large grains allow less rotation as a response to
deformation, generally leading to increased brittleness.

Fig 25. HK30, creep tested at 750 °C. Creep strain rate as a function of time (semi-log axes).

38
 Fig. 26. a), b) Creep cavitation/cracks in HK30 specimen tested at 53 MPa and 750 °C,
tr= 1,557 h.

Fig. 27. a) Orientation map featuring a grain boundary crack and the recrystallized surface
layer (bottom), b) phase precipitations in the recrystallized layer and along the crack. The
specimen surface is just below the bottom in the images.

5.5. Creep behaviour and damage of D5S at 750 °C

Since it is used for similar conditions as HK30, D5S was also creep tested at 750 °C. The
creep strain rate curves are shown in Fig 28. This material shows the same trend as HK30
with quick transitions between primary and tertiary creep at higher stresses and more or less
pronounced secondary creep at lower stresses.

39
 Fig 28. D5S, creep tested at 750°C. Creep strain rate as a function of time (semi-log axes).
In order to see the damage originating purely from secondary creep, the 18.5 MPa specimen
was interrupted after 6494 h. For this specimen, and for the shorter lasting specimen of 20
MPa, the documented cavitation and damage is shown in Fig 29. As can be seen, small
cavities have emerged around graphite nodules in both materials. Since 20 MPa was tested to
rupture, it also contains clear traces of tertiary creep in the form of shear cracks (see Fig 29d).

40
Fig 29. Microstructure of creep specimens of D5S tested at 750 °C. a) Creep cavitation (18.5
MPa), b) same as a) in lower mag. c) creep cavitation (20 MPa), d) shear cracks (20 MPa).
The 20 MPa specimen was also investigated using EDX/EBSD and it was revealed that
significant amounts of Cr7C3 carbides along with larger particles, including Fe2Ni2Si and an
unidentified carbide, had precipitated during creep exposure. The effect on creep rate is hard
to account for as there are at least two opposing mechanisms: decreased solution
strengthening and increased precipitation hardening (depending on the particles). Cr7C3
carbides precipitated both in the shape of plates and small spheres while the other particles
were large and irregularly shaped. Images of the various precipitations and their
corresponding EDX maps can be found in appendix 5.
In local misorientation measurements, sub-grain boundaries of about 1°, i.e. very small
misorientations from deformation, were measured. The corresponding misorientation images
(maps) show the local imbalance in dislocation density and thus only indicate the degree of
plastic deformation. It could be emphasized that gliding dislocations produce plastic
deformation whereas locally stored dislocations, with an imbalance in the sum of Burgers
vector produce a local misorientation, which is being observed. At higher temperatures
recovery and annihilation often restores the lattice leaving no apparent trace of the strain
created by gliding dislocations while shaping more well-defined cell walls and sub-grains
with clearer orientation differences. Thus, the local misorientation technique is useful to show
tendencies in the dislocation microstructure being formed, rather than measuring local
deformations. Compared with a regular misorientation measurement, which gives the
orientation from one pixel to the next, the local misorientation in each pixel is the average of

41
the misorientations between this pixel and the surrounding pixels (3*3 matrix). This noise-
reducing moving average filter is sometimes critical for observing the very small
strains/misorientations due to creep.
Local misorientations were measured on untested and creep tested (20 MPa) materials, see
Fig 30. In Fig 30a, it can be seen that even the untested material contains low angle
misorientations/subgrain boundaries of about 1°. In Fig 30b, the effect of creep deformation
is seen as a significant increase in subgrain boundaries of around 1°.
In addition, the same areas are shown as “regular” orientation maps in Fig. 31. There, it can
be noted that both untested and creep tested materials contain additional subgrain boundaries
(>2°) but there is no apparent increase. Hence, creep adds a lot of 1° subgrain boundaries to
the material, in accordance with Kassner et al. [19].

Fig 30. Local misorientation (0-2°) in D5S, a) untested and b) creep tested (20 MPa)
materials.

Fig. 31. “Regular” orientation maps of the areas in Fig 30 with a) untested and b) creep tested
(20 MPa) materials. High angle grain boundaries (>15°) are shown as thick black lines while
low-angle boundaries (>2°) are shown as subtle gray lines. Red = {001}, Green = {101},
Blue = {111}. Reference: Y-axis (in the specimen plane).

42
6. Summary
Four cast materials, including three ductile cast irons and one cast steel, were tested mainly
with regard to slow and fast plasticity. The first two papers (see appendices 1 and 2) treat the
method development and data output analyses associated with the three methods SRTC, STT
and CL. The third, fourth and fifth papers are more focused on damage, mainly from creep
but, in the third paper, also from oxidation and tensile testing.
The first paper (appendix 1) mainly describes the mechanical response to thermal cycling and
the method of SRTC as an alternative to the more traditional TMF testing. In general, SRTC
is easier to conduct as it is not strain-controlled. The thermal cycling loops were described in
terms of elastic and plastic deformation segments. It was shown that they carry essential
information about mechanical properties, such as evolution in flow stress over the selected
temperature interval. Several test parameters, such as temperature interval, heating/cooling
rates and hold times were changed to see the various effects. The contributions in strain
during each part of the thermal cycle were compared.
In the second paper (appendix 2), the hold-time stress-relaxation-data, generated from
thermal cycling using SRTC, was compared with sequential tensile tests and constant-load
creep tests. The results were inserted into “Norton plots”, double logarithmic stress-strain rate
diagrams which are regularly used to compare creep data. Results obtained by the two fast
methods, SRTC and STT, in many cases showed perfect agreement. Since SRTC results were
often evaluated with regard to hold time compression at the upper temperature of thermal
cycling while STT includes tensile loading, the results showed that creep rates are
independent of load direction. “Cyclic” creep rates were more or less unchanged with the
number of cycles. At 700 °C, the CL data set ended up one order of magnitude lower (in
strain rate) compared with SRTC and STT but had the same slope which indicates one
common creep mechanism.
In the third paper (appendix 3), the damage from the CL tests of SiMo51 (appendix 2) was
investigated. Tensile tests and oxidation tests were also carried out for comparison. As
described above, primary creep was more or less directly followed by a “two-stage” tertiary
creep. Out of these two stages, the first accounted for the largest part of the creep life and was
therefore of biggest interest. After considering pure straining and oxidation, rejecting both, it
was attributed to cavitation around graphite nodules and at the grain boundaries. The second
tertiary stage was attributed to the usual characteristics of tertiary creep, namely coalescence
of smaller creep cracks into larger ones. Tensile test damage was associated with necking and
large oval voids around the graphite nodules.
In the fourth paper (appendix 4) the cast steel HK30 was investigated in constant-load creep
testing at 750 °C. During creep, sigma phase and presumed G-phase precipitated. The phase
identities were verified using EBSD and Thermo-Calc. As a specimen preparation artefact,
the surface of the specimens recrystallized during creep testing. However, since the
recrystallized zone was so thin and apparently did not crack, it is not considered to have had
any real effect. Instead, cracks propagated in grain boundaries, predominantly following grain
boundary particle interfaces of the original, “un-recrystallized” material. Ultimately, the large
grains, sometimes exceeding the specimen diameter, were considered the big weakness of
HK30.

43
In the fifth paper (appendix 5), D5S was CL tested, also at 750°C due to similar service
conditions as HK30. The creep damage reminded of the one observed for SiMo51: “delicate”
cavitation around graphite nodules at low stresses and long times, and large oval voids
around nodules at high stresses and short times. For both SiMo51 and D5S, the delicate
cavities were considered a time-dependent creep-phenomenon, likely following one of the
many theories on creep cavitation. The large ovals were attributed to strain-dependent
dislocation glide. Cr7C3 particles precipitated together with larger particles, including
Fe2Ni2Si and an unidentified carbide. The effect from the observed precipitation is
complicated to predict as there is both decreased solution strengthening and increased
precipitation hardening. Creep deformation created a lot of additional subgrain boundaries of
about 1° in the material.

7. Conclusions
In the present study, two ferritic ductile cast irons (SiMo51, SiMo1000), one austenitic cast
iron (D5S) and one austenitic cast steel (HK30), were included. The materials were mainly
tested using three methods: stress relaxations with thermal cycling (SRTC), sequential tensile
testing (STT) and the constant-load creep test (CL). Data sets generated by STT and SRTC
generally agreed very well, but there were discrepancies compared with the CL tests.
When subjected to constant-load creep testing at 700 °C, the ferritic ductile cast iron SiMo51
showed more or less direct transitions between primary and tertiary creep. Creep cavities
could be found around the graphite nodules and at the grain boundaries. Highly deformed
areas showed large elliptical pores around the graphite nodules, a damage which was also
found in the neck of tensile tested specimens.
The austenitic cast steel HK30 and the austenitic ductile cast iron D5S were constant-load
creep tested at 750 °C. Both materials showed direct transition between primary and tertiary
creep at higher stresses and more or less prolonged secondary creep at lower stresses. HK30
showed typical grain boundary creep-cavitation after long-term creep exposure. Sigma phase
and G phase precipitated during creep. The cracks were intergranular and followed particle-
matrix interfaces.
D5S showed large elliptical pores around graphite nodules when tested at higher stresses and
delicate creep cavities around the nodules at lower stresses (similar to SiMo51). Various
particles precipitated during creep exposure. Subgrain boundaries of 1° formed due to creep
deformation. Final rupture occurred by shear cracks which propagated nodule-to-nodule.

8. Sustainability statement
The strive towards lower emissions in truck diesel- and gasoline engines has led to better
engines in term of efficiency but also to higher temperatures of the exhaust gases. This
development inevitably puts higher demands on the cast materials which are used for the
application. Since creep deformation is one of the main degradation processes of the manifold
materials, it is also essential to test the materials regarding their resistance against such
exposure.
Referring to the 17 sustainable development goals set by the European Commission, this
relates to: 7) affordable and clean energy, 9) industry, innovation and infrastructure, 11)

44
sustainable cities and communities, 12) responsible consumption and production and 13)
climate action.

9. Acknowledgements
Regarding financing, the first part of the work (up to the licentiate degree) was completely
financed by Scania CV AB. I am very grateful for this contribution.
The second part of the work (post-licentiate) was financed by a VINNOVA project, no. 2018-
04019, “Inverkan av korrosion och krypning på utmattning av värmebeständiga material till
avgaskomponenter” within the research program “Fordonsstrategisk Forskning och
Innovation, FFI”, at VINNOVA, Swedish Governmental Agency for Innovation Systems. I
am also very grateful for this governmental support and for the time the supervisors spent on
writing the application.
Stefan Jonsson (main supervisor), is acknowledged for his contributions to appendices 1-5,
for being a big inspiration as a researcher and supervisor, for teaching me a lot of new skills,
for always being very careful and thorough in his review of my work, and for highly
stimulating discussions.
Baohua Zhu (co-supervisor), is acknowledged for her contributions to appendices 1-5, for
being highly supportive as a project leader, for teaching me about various sample preparation
techniques, and for writing funding applications.
Joakim Lindblom, is acknowledged for teaching me a lot about materials science and testing
methodology in general, for suggesting me to start spot welding the thermocouples, and for
his theories regarding the speckle pattern tracking (4.2.2). Also, for taking me out sailing with
the Wayfarer.
Henrik Östling, Taina Vuoristo and Mats Larsson are acknowledged for carrying out the
creep tests. Henrik is especially acknowledged for providing a lot of insightful questions as
opponent during my licentiate defense, for teaching me a lot about creep, and for being very
helpful.
Joacim Hagström, is acknowledged for teaching me about EBSD, very beautiful imaging and
for a very good collaboration.
Jan Linder, is generally aknowledged for fruitful discussions and for providing valuable
suggestions.
Ralf Rablbauer, is acknowledged for arranging the STT tests, for arranging our visits to
Volkswagen and for his contribution to appendix 2.
Marc Imiela is acknowledged for carrying out the STT tests at Volkswagen and for teaching
me how to do the tests for a second round of tests.
Madeleine Ekström is acknowledged for fruitful discussions, for valuable previous
contributions to the research field, and for introducing me to the topic.
Anders Tjernberg is aknowledged for a good collaboration (even though we could not
publish) and for teaching me about a lot about creep and stress relaxations.

45
Peter Skoglund is acknowledged for lending me his test rig and for teaching me about testing
methodology and TMF.
Jessica Elfsberg is acknowledged for helping with questions related to casting and nodule
measurements.
Nulifer Ipek is acknowledged for being inspirational and highly supportive.
Lasse Miettunen is acknowledged for arranging his course in test programming and for being
highly forthcoming in adjustments of the test rig used for SRTC.
Shengmei Xiang is acknowledged for co-operation in lab development.
Peter Szakalos is acknowledged for his valuable comments regarding the oxidation of
SiMo51.
Rolf Sandström is acknowledged for teaching me about creep.
The group at Scania (YTMB) is acknowledged for being supportive, friendly and helpful.
The Workshop of Scania is acknowledged for doing parts of the sample preparation, and for
being very helpful in this matter.
My colleagues at KTH, including other PhD students, are acknowledged for good
comradeship.
My mother Heléne, my father Mikael and my brother Eric are acknowledged for unlimited
support and help with everything outside of the PhD life.
Temporary colleagues at Adaptus Rehab AB, including Petra, Kristin, Fanny, Andrea and
Andreas, are acknowledged for being a good company during my last time as a PhD student.
Close friends, including Petter, Carl and Wictor, are acknowledged for good times and good
support.

10. References
[1] V. Norman, Fatigue of Heavy-Vehicle Engine Materials, Doctoral thesis, Linköping:
Linköping University, 2018.

[2] F. Bruch, Thermo-mechanical fatigue of ductile iron, Masters thesis, Kassel: Institut fur
Werkstofftechnik, Metallische Werkstoffe, 2015.

[3] N. Collin, Thermo-mechanical fatigue of cast iron for engine application, Masters
thesis, Stockholm: KTH (Royal Institute of Technology), 2014.

[4] H.-J. Kuhn, R. B and B. Skrotzki, "Thermomechanical fatigue of heat-resistant

austenitic cast iron EN-GJSA-XNiSiCr35-5-2 (Ni-Resist D-5S)," International Journal
of Fatigue, vol. 99, pp. 295-302, 2017.

[5] M. Ekström, Oxidation and corrosion fatigue aspects of cast exhaust manifolds,
Doctoral thesis, Stockholm: KTH (Royal Institute of Technology), 2015.

46
[6] S. Xiang, S. Jonsson, B. Zhu and J. Odqvist, "Corrosion fatigue of austenitic cast iron
Ni-resist D5S and austenitic cast steel HK30 in argon and synthetic diesel exhaust at
800°C," International Journal of Fatigue, vol. 132, pp. 1-11, 2020.

[7] S. Xiang, High-Temperature Corrosion-Fatigue of Cast Alloys for Exhaust Manifolds,

Licentiate thesis, Stockholm: KTH (Royal Institute of Technology), 2018.

[8] S. Xiang, B. Zhu and S. Jonsson, "HIgh-Temperature Corrosion-Fatigue Behavior of

Ductile Cast Irons for Exhaust Manifolds Applications," Materials Science Forum, vol.
925, pp. 369-376, 2018.

[9] S. Xiang, P. Hedström, B. Zhu, S. Jonsson and J. Odqvist, "Influence of graphite

morphology on the corrosion-fatigue properties of the ferritc Si-Mo-Al cast iron
SiMo1000," International Journal of Fatigue, vol. 140, 2020.

[10] S. Xiang, S. Jonsson, P. Hedström, B. Zhu and J. Odqvist, "Influence of ferritic

nitrocarburizing on the high-temperature corrosion-fatigue properties of the Si-Mo-Al
cast iron SiMo1000," International Journal of Fatigue, vol. 143, 2021.

[11] E. Hug, C. Keller, J. Favergon and K. Dawi, "Application of the Monkman-Grant law
to the creep fracture of nodular cast irons with various matrix composites and
structures," Materials Science and Engineering A, vol. 518, pp. 65-75, 2009.

[12] A. Cottrell, An Introduction to Metallurgy, London: The Institute of Materials, 1975.

[13] M. Ekström, P. Szakalos and S. Jonsson, "Influence of Cr and Ni on High-Temperature

Corrosion Behavior of Ferritic Ductile Cast Iron in Air and Exhaust Gases," Oxidation
of Metals, vol. 80, pp. 455-466, 2013.

[14] A. Ebel, S. Yapi Brou, B. Malard, J. Lacaze, D. Monceau and L. Vaissiere, "High-
Temperature Oxidation of a High Silicon SiMo Spheroidal Cast Iron in Air with In Situ
Change in H20-content," Materials Science Forum, vol. 925, pp. 353-360, 2017.

[15] Georg Fischer, "SiMo1000 Materials Data Sheet," Georg Fischer, 2014.

[16] B. Piekarski and J. Kubicki, "Creep-resistant austenitic cast steel," Archives of Foundry
Engineering, vol. 8, pp. 115-120, 2008.

[17] M. Whittaker, B. Wilshire and J. Brear, "Creep fracture of the centrifugally-cast

superaustenitic steels, HK40 and HP40," Materials Science and Engineering A, vol.
580, pp. 391-396, 2013.

[18] S. Konosu, T. Koshimizu, T. Lijima and K. Maeda, "Evaluation of Creep-Fatigue

Damage Interaction in HK40 Alloy," Journal of Mechanical Design, vol. 115, pp. 41-
46, 1993.

[19] M. Kassner, Fundamentals of Creep in Metals and Alloys, Butterworth Heinemann,

2015.

47
[20] S. Manson and G. Halford, Fatigue and Durability of Metals at High Temperatures,
ASM International, 2009.

[21] S. Jonsson, Mechanical Properties of Metals and Dislocation Theory from an

Engineer’s Perspective, Stockholm: KTH (Royal Institute of Technology), 2010.

[22] H. Östling, Interviewee, Swerim, Private communication. [Interview]. October 2020.

[23] M. Ashby, C. Gandhi and D. Taplin, "Fracture-mechanism maps and their construction
for F.C.C metals and alloys," Acta Metallurgica, vol. 27, pp. 699-729, 1979.

[24] T. Tozera, O. Sherby and J. Dorn, Trans. ASM , vol. 49, p. 173, 1957.

[25] F. Norton, The Creep of Steels at high Temperatures, New York: McGraw-Hill, 1929.

[26] R. Sandström, "Creep Strength of austenitic stainless steels for boiler applications,"
Coal Power Plant Materials and Life Assessment, pp. 127-145, 2014.

[27] H. Frost and M. Ashby, "Deformation-mechanism maps for pure iron, two austenitic
stainless steels, and a low-alloy ferritic steel," Fundamental Aspects of Structural Alloy
Design, 1977.

[28] J.-O. Nilsson, P. Howell and G.L. Dunlop, "Interfacial microstructure and low stress,
high temperature creep of an austenitic stainless steel.," Acta Metallurgica, vol. 27, pp.
179-186, 1979.

[29] R. Swindeman, "The interrelation of Cyclic and Monotonic creep Rupture," in

International conference on creep, London, Paper 32, pp. 71-76, 1963.

[30] Å. Martinsson, H. Andersson-Östling, F. Setisleam, R. Wui and R. Sandström, "Creep

testing of nodular iron at ambient and elevated temperatures (nbr: R-10-64)," Swerea
Kimab AB, Kista, 2010.

[31] I. Hervas, A. Thuault and E. Hug, "Damage Analysis of a Ferritic SiMo Ductile Cast
Iron Submitted to Tension and Compression Loadings in Temperature," Metals, vol. 5,
pp. 2351-2369, 2015.

[32] T. Sjögren, P. Vomacka and I. Svensson, "Comparison of mechanical properties in

flake graphite and compacted graphite cast irons for piston rings," International Journal
of Cast Metals Research, vol. 17, pp. 65-71, 2004.

[33] M. Dong, C. Prioul and D. Franscois, "Damage Effect on the Fracture Toughness of
Nodular Cast Iron: Part 1. Damage Characterization and Plastic Flow Stress Modeling,"
Metallurgical and Materials Transactions A, vol. 28, pp. 1997-2245, 1996.

[34] L. L. Liu, Q. Guo and Y. Niu, "Transition Between Different Oxidation Modes of
Binary Fe-Si Alloys at 600-800C in Pure O2," Oxidation of Metals, vol. 79, pp. 201-
224, 2013.

48
[35] M. Appelt, Eisen-Aluminium-Legierung fu ̈r thermomechanisch belastete
Komponenten im Abgasstrang (Doctoral thesis), Karlsruher institut fur technologie,
2013.

[36] P. Hähner, E. Affeldt, T. Beck, H. Klingelhöffer, M. Loveday and M. Rinaldi,

"Validated Code-of-Practice for Strain-Controlled Thermo-Mechanical Fatigue
Testing," European commission, Directorate General, Joint Research Centre, 2006.

[37] A. Verma, M. Sundararaman, J. Singh and S. Nalawade, "A new method for
determining the Curie temperature using a dilatometer," Measurement Science and
Technology, vol. 10, pp. 105-106, 2010.

[38] Swedish Institute of Standards, "Metallic materials - Uniaxial creep testing in tension -
Method of test (ISO 204:2018)," Swedish Institute of Standards, 2018.

[39] T. Seifert, R. Hazime and S. Dropps, "TMF Life Prediction of High Temperature
Components Made of Cast Iron HiSiMo: Part II: Multiaxial Implementation and
Component Assessment," SAE International Journal of Materials and Manufacturing,
vol. 7, pp. 421-431, 2014.

[40] T. Seifert, R. Hazime and S. Dropps, "TMF life prediction of high temperature
components made of cast iron HiSiMo: Part I: uniaxial tests and fatigue life model.,"
SAE International Journal of Materials and Manufacturing, vol. 7, pp. 439-445, 2014.

49