Content uploaded by Mohan Paladugu
Author content
All content in this area was uploaded by Mohan Paladugu on Nov 22, 2021
Content may be subject to copyright.
Materialia 14 (2020) 100948
Contents lists available at ScienceDirect
Materialia
journal homepage: www.elsevier.com/locate/mtla
Full Length Article
Comparative micromechanics assessment of high-carbon martensite/bainite
bearing steel microstructures using in-situ synchrotron X-ray diraction
D. Foster
a , ∗
, M. Paladugu
b , ∗
, J. Hughes
a
, M. Kapousidou
a
, C. Barcellini
a
, D. Daisenberger
c
,
E. Jimenez-Melero
a
a
Materials Performance Centre, Department of Materials, The University of Manchester, M13 9PL, UK
b
The Timken Company World Headquarters (WHQ), North Canton, OH 44720, USA
c
Diamond Light Source Ltd, Diamond House, Harwell Science & Innovation Campus, Didcot OX11 0DE, UK
Keywords:
SAE 52100 steel
Crystal plasticity
Austenite stability
Martensitic transformation
Synchrotron X-ray diraction
Mechanical testing
The mechanical behavior of dierent microstructural constituents in SAE 52100 bearing steel has been studied
at room temperature in relation to quenching and partitioning (QP) and bainitization (B) process parameters,
and compared to the standard quenched and tempered (QT) microstructure using high-energy synchrotron X-ray
diraction in situ during tensile loading. Owing to a larger degree of carbon entrapment in its body-centered cubic
lattice and associated lattice distortion, martensite in the QT microstructure showed a larger lattice parameter
and broadened diraction peaks as compared to lower bainitic ferrite or partitioned martensite. A reduction in
diraction peak broadness in tempered martensite occurs at a true stress value of ~1800 MPa, and preserves its
peak broadness even after subsequent unloading. In contrast, an equivalent eect in peak broadness is detected
at ~1500 MPa in the lower bainitic ferrite or mixed bainitic/martensitic matrix characteristic of the B and
QP microstructures. In all studied microstructures, the metastable austenite phase transforms when a critical
stress is reached, the value of which increases with the bainitic ferrite/martensite fraction and with the carbon
content in austenite, but remains lower in the QP and B microstructures compared to the standard QT steel. These
results suggest that the carbon solid solution strengthening and associated lattice distortion in bainitic ferrite or
martensite are key in determining the mechanical performance of the constituent phases in the steel, with the
phase fraction and local carbon content playing an additional role on the austenite mechanical stability.
1. Introduction
Ball and roller bearings, especially rolling bearings, are key preci-
sion components that allow rotary motion of, or about, shafts in rela-
tively complex mechanisms operating in industrial machinery and com-
mercial devices. Rolling bearings are based on the rolling action of the
balls or rollers to minimize friction drag [1] . The low-alloy ( < 5 wt.%
total alloying elements), high carbon ( > 0.8 wt.%C) through-hardened
SAE 52100 steel grade was introduced decades ago in bearing technol-
ogy and still dominates bearing steel usage worldwide [ 2 , 3 ]. Its predom-
inance in bearing production (rolling elements and raceways) relies on
the high machinability found in its soft spheroidized-annealed condi-
tion, together with its good hardenability and the relatively high hard-
ness (HRC 61-65), strength and toughness found in the heat-treated state
of the nal product [1-3] . In the typical martensitic through-hardening
process, spheroidized steel is austenitized, and then followed by quench-
ing in oil and a low-temperature tempering, to yield a quenched and
tempered (QT) martensitic microstructure with < 15 vol.% metastable
∗ Corresponding authors.
E-mail addresses: daniel.foster-7@postgrad.manchester.ac.uk (D. Foster), mohan.paladugu@timken.com (M. Paladugu).
retained austenite and ~2-3 vol.% spheroid cementite ( 𝜃-Fe
3
C) phase
[ 2 , 4 , 5 ].
During operation, the bearing is subject to Hertzian contact stresses
of several GPa in magnitude together with subsurface shear stresses,
causing the phenomenon of rolling contact fatigue (RCF) [ 1 , 2 , 6 ]. The
resultant stress state and fatigue can trigger the development of surface
and/or subsurface damage to and cracking of the steel [7] . The con-
sequence is the appearance of macro surface fatigue spalls that cause
bearing damage and terminate bearing operation [8-10] . Therefore, the
fatigue resistance of steels is a key aspect in improving the reliability of
bearing lives.
Additionally, the retained austenite phase present in the steel, if not
thermo-mechanically stable enough, transforms into a body-centered
cubic phase, inducing an undesired increase in the bearing’s dimensions
over time [11-15] . But the additional work-hardening capability from
the transformation itself is thought to increase the fatigue resistance of
the bearing if the austenite stability is tailored for a progressive transfor-
mation [ 16 , 17 ], although recent results suggest that the solid solution
https://doi.org/10.1016/j.mtla.2020.100948
Received 5 August 2020; Accepted 3 November 2020
Available online 4 November 2020
2589-1529/© 2020 Acta Materialia Inc. Published by Elsevier B.V. All rights reserved.
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Table 1
Chemical composition (wt.%) of the SAE 52100 steel used in
this study.
C Cr Mn Si Cu Ni Al Fe
1.005 1.46 0.32 0.27 0.19 0.1 0.024 Bal.
strengthening eect of interstitial carbon atoms and substitutional so-
lutes in the martensitic matrix is the major contributor to the bearing’s
tolerance to RCF [18] . The presence of retained austenite is thought
to be benecial for local toughness and fatigue resistance, especially
in the context of contaminated or decient lubrication environments
[ 15 , 19 , 20 ]. The probability of micro-cracking also decreases with the
austenite grain size [ 19 , 20 ]. However, a thorough analysis of how these
microstructures have performed has not been conducted or reported in
literature.
Heat treatment modications to control austenite phase transforma-
tions and carbon diusion oer the opportunity to tailor the character-
istics of the constituent phases in the steel microstructure for enhanced
through-life bearing performance. One appealing option is to “austem-
per ”the material above but relatively close to the martensitic start tem-
perature ( M
S
) [21] , as a substitute to the quenching step in oil below M
S
.
Bainitic-type (B) bearing steel microstructures with nano-sized carbide
particles oer improved wear resistance compared to QT microstruc-
tures, but generally lower overall mechanical properties [ 22 , 23 ].
An alternative quench and partitioning (QP) heat treatment process
consists of adding a partitioning step above the M
S
temperature, after
the standard oil quenching, in order to promote the diusion of car-
bon from the supersaturated martensitic matrix into the neighboring
austenite grains, and potentially the formation of bainitic ferrite from
less stable austenite grains [ 24–26 ]. The QP process is usually applied to
steel grades high in Al or Si in order to hinder the formation of carbides
[27] , though the process promotes additional carbon partitioning into
remaining austenite. As such, the QP terminology has been applied in
this work to describe a two-step heat treatment process that promotes
carbon diusion into austenite. In this work, we have systematically as-
sessed the micro-mechanics of the constituent phases in B and QP-type
microstructures of SAE 52100 bearing steel, and compared their behav-
ior under deformation with that of standard QT microstructures of the
same steel. The collection of high-energy X-ray diraction data in situ
during the mechanical test has allowed us to probe the mechanical be-
havior & stability of the constituent phases in the bulk of the evolving
steel microstructure upon straining.
2. Experimental details
2.1. Material processing
The chemical composition of the steel used in this study is given in
Table 1 . The as-received material consisted of semi-cylindrical bars of ra-
dius ~12 mm in the spheroidized-annealed condition (see Fig. S1a). All
samples were machined from the same initial material. Two microstruc-
tural types were produced by a systematic variation in heat treatment
parameters after an initial austenitization step at 860°C, (see Fig. 1 b
& c): (i) bainitic “B-type ” microstructures were generated via austem-
pering and were salt quenched to a temperature of either 220°C, 240°C
or 260°C, and held for 30 min; (ii) mixed microstructure, quenched and
partitioned “QP-type ” microstructures were generated by salt quenching
to 170°C and held for 30 min in order to accommodate a small degree of
phase transformation (see Fig. S5) and allow the sample to reach tem-
perature uniformity in the salt bath. The quenching temperature was
selected to be ~30°C below M
S
for this steel grade, since this can ef-
fectively induce an acceleration of the bainitic transformation kinetics
in the subsequent step above M
S
and therefore shorten the processing
time at industrial scale. This is subsequently followed by a “partitioning ”
step of also 30 min at a temperature of either 220°C, 240°C or 260°C.
All samples were subsequently quenched to room temperature. The M
s
temperature of those samples was measured by dilatometry to be 203°C
(see Fig. S6). The intention of these two heat treatment types was to
generate a sample set of B-type microstructures, in which the matrix
is comprised primarily of lower bainite, with ~28 vol.% untempered
martensite formed during cooling to room temperature from 220°C (see
Fig. S2, 3 & 4), with retained austenite; and a sample set of QP-type mi-
crostructures, in which the matrix is comprised of a bainite/martensite
mix, with retained austenite. In high carbon bearing steel, this double
step (QP) treatment, i.e. quenching to ~30°C below M
S
followed by a
temperature rise into the bainite regime, aims to accelerate the austen-
ite transformation as compared to a one-step austempering treatment
without sacricing hardness [28] , and promote some additional carbon
diusion into austenite. This acceleration of the transformation kinet-
ics, potentially due to the dislocations introduced by the prior marten-
site formation, is reported to occur both in Si-containing and in Si-free
steel [29] . Transition carbides are expected to form in the strain elds of
the dislocations present in supersaturated martensite, as the transition
carbide formation is not retarded by Si [30] .
As a reference, a QT-type microstructure was obtained by austeni-
tization at 840°C for 30 min, based on industrial practice to minimize
distortions or cracking eects during manufacturing, followed by oil
quenching to 80°C, rinsing in water and nally tempering at 180°C for
90 min (see Fig. 1 a). The austenitization temperature of 860°C for the B
and QP –type samples was selected in order to further increase the car-
bon content in austenite by enhancing Fe
3
C dissolution, consequently
increasing retained austenite stability in subsequent steps of the heat
treatment [31] . The average grain size of the prior austenite was ~16
µm at 840°C (see Fig. S1b), with a standard deviation of 2.1 µm, and
~15 µm at 860°C, with a standard deviation of 1.7 µm. The resultant
microstructures were characterized using a Zeiss Ultra 55 scanning elec-
tron microscope (SEM) operating at an accelerating voltage of 10 kV
and a beam current of 2.3 nA. The Vickers micro-hardness of each mi-
crostructure was determined from the average of 5 indents using a 10
N load (HV
1.0
). Dog-bone tensile specimens with a total length of 38
mm, a gauge length of 4 mm and a cross section in the gauge length of
1.5 ×0.7 mm
2
were machined from each microstructure (see Fig. 1 d for
full sample dimensions), mechanically polished down to a 0.4 𝜇m col-
loidal silica nish, and then electropolished at room temperature using
a solution of 5 vol.% perchloric acid in 95 vol.% glacial acetic acid.
2.2. In-situ synchrotron X-ray diffraction during tensile deformation
The dog-bone samples were tested sequentially at room temperature
using a 2kN micro-tensile rig placed at the high-energy I15 beamline
of the UK Diamond Light Source. A schematic diagram of the in-situ
synchrotron X-ray diraction (SXRD) experiment is shown in Fig. 2 a.
The center of the sample gauge length was illuminated with a 72 keV
( 𝜆= 0.1722 Å) X-ray beam with a 70 µm-diameter round beam, and the
sample was deformed step-wise at a strain rate of 0.1 mm/min during
loading up to the maximum load of 2 kN. At each deformation step, the
sample was kept under load controlled and rotated in steps of Δ𝜔 = 5°
along the loading direction, covering a total angular range of 𝜔 = ± 45°.
During each rotation step, the diracted intensity was recorded contin-
uously using a 2D PerkinElmer 1621EN detector placed at 1000 mm
behind the sample. That sample to detector distance was selected to
be able to distinguish the {111} austenite and {110} ferrite diraction
rings at zero deformation in this experiment. Beyond those measure-
ments, changes in gauge length and width were monitored by recording
the transmission of the X-ray beam as it scanned the sample prior to the
collection of the diraction signal. After maximum load of ≥ 2000 MPa
was reached, the sample was unloaded and the diracted intensity was
also recorded. Due to the micro-tensile sample geometry adopted and
the low strain rate, no sharp macroscopic yielding was observed within
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Fig. 1. Schematic representation of the heat treatment to produce (a) quenched & tempered (QT), (b) bainitic (B) and (c) quenched & partitioned (QP) microstructures.
(d) Dimensions (in mm) of the dog-bone samples machined from those microstructures for tensile testing during the in-situ synchrotron X-ray diraction experiment.
OQ = Oil Quenching; SQ = Salt Quenching, WQ = Water Quenching.
the loaded range of this experiment. LaB
6
(NIST Standard Reference
Material 660c) was used as calibrant.
The 2D diraction patterns recorded as a function of applied stress
and sample were summed for all 𝜔 -rotations and processed into 1D pat-
terns using the DAWN software [32] . The resultant 1D patterns were
subsequently analyzed using the Rietveld method and the Caglioti’s for-
mula [33] implemented in the FullProf Suite software package [34] .
The diraction patterns showed the presence of three crystallographic
phases: face centred-cubic phase (fcc austenite), body-centred cubic (bcc
bainitic ferrite and/or martensite) and orthorhombic cementite. The Ri-
etveld analysis of each 1D pattern included the renement of the scale
factors, phase fractions, FWHM (full width at half maximum) and lat-
tice parameters. A signicant reduction in austenite fraction beyond the
experimental uncertainty marked the critical stress for austenite to com-
mence to transform into martensite upon loading. During deformation,
the load is distributed between the phases present in the evolving mi-
crostructure. The observed shifts in peak positions during deformation
were translated into the corresponding average lattice strain for a given
phase ( ⟨𝜀
𝑝ℎ
⟩) according to:
⟨𝜀
𝑝ℎ
⟩=
𝑎
𝑝ℎ
− 𝑎
0
𝑝ℎ
𝑎
0
𝑝ℎ
(1)
where 𝑎
𝑝ℎ
and 𝑎
0
𝑝ℎ
correspond to the lattice parameter of a constituent
phase in the microstructure at a given deformation step and at zero de-
formation, respectively. Additionally, we have carried out a single-peak
t of selected (weakly overlapping or non-overlapping) reections to a
pseudo-Voigt prole function, in order to obtain the plane strains ( ⟨𝜀
ℎ𝑘𝑙
⟩)
containing information about the load partitioning amongst specic hkl
planes of each phase, using the expression:
⟨𝜀
ℎ𝑘𝑙
⟩=
𝑑
ℎ𝑘𝑙
− 𝑑
0
ℎ𝑘𝑙
𝑑
0
ℎ𝑘𝑙
(2)
where 𝑑
ℎ𝑘𝑙
and 𝑑
0
ℎ𝑘𝑙
correspond to the lattice plane spacing at a given de-
formation step and at zero deformation, respectively. Furthermore, the
2D-to-1D integration and Rietveld analysis were also performed using
the diracted intensity collected in the 2D detector at ± 7.5° with respect
to the loading direction and also its perpendicular direction. The grains
with their plane normal oriented parallel to the loading direction will be
in a tensile state upon sample loading, whereas grains with their plane
normal perpendicular to the loading direction will be in compression.
This procedure allowed us to decouple the parallel ( 𝜀
∥
ℎ𝑘𝑙
) and perpen-
dicular ( 𝜀
⊥
ℎ𝑘𝑙
) lattice plane strains from the isotropic changes in lattice
parameter due to variations in average carbon content in austenite [ 35 ].
The values of the FWHM discussed later in this work for dierent { hkl }
reections correspond to the diracted intensity collected in the 2D de-
tector at a ± 7.5° azimuthal angle with respect to the loading direction.
3. Results
3.1. Initial microstructures
Fig. 3 shows secondary electron SEM images of representative mi-
crostructures. Bainite can be visually identied with its sheaf-like struc-
ture in all samples shown; the morphology of martensite can also be
identied with its long needle-like structure. Further, large, globular
cementite is present across all samples. Fig. 2 b shows the Rietveld anal-
ysis of a 1D diraction pattern before mechanical loading. The experi-
mental values of the austenite fraction and the lattice parameter of the
main phases present in the unloaded state, together with the Vickers
micro-hardness values, are shown in Table 2 . The carbon content in the
retained austenite phase in the unloaded state can be estimated using
the expression [ 14 , 36 , 37 ]:
𝑎
𝛾= 3 . 556 + 0 . 0453 𝑥
𝐶
+ 0 . 00095 𝑥
𝑀𝑛
+ 0 . 056 𝑥
𝐴𝑙
+ 0 . 0006 𝑥
𝐶𝑟
+0 . 0015 𝑥
𝐶𝑢
− 0 . 0002 𝑥
𝑁𝑖
(3)
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Fig. 2. (a) Schematic diagram of the high-energy synchrotron X-ray diraction experiment performed in situ upon straining. An X-ray beam with a size of 70 ×70
µm
2
and an energy of a 72keV ( 𝜆= 0.1722Å) illuminated the center of the gauge length of the dog-bone tensile specimen. The diracted intensity was recorded
continuously during step-wise 𝜔 - rotations using a 2D PerkinElmer 1621EN detector placed at 1000 mm behind the sample. The Rietveld analysis of a representative
1D diraction pattern is shown in (b) at zero load and in (c) at the true stress of 2087 MPa. The three green rows of vertical lines correspond to (top row) body-centered
cubic (bainitic ferrite and/or martensite), (middle row) face centered-cubic phase (austenite) and (bottom row) orthorhombic cementite.
Fig. 3. Secondary electron SEM micrographs of the bainitic (B) and quenched and partitioned (QP) microstructures produced using the austempering/partitioning
temperature of either 220°C or 260°C.
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Table 2
Process parameters used to produce the seven microstructures under study, namely the quenching/austempering temperature ( T
q
) and time at that temperature ( t
q
), together with the
subsequent partitioning temperature ( T
p
) and time ( t
p
) where relevant. The table also contains the average Vickers hardness (HV
1.0
), the martensite/bainitic ferrite lattice parameter
( a
𝛼) and carbon content ( 𝑥
𝛼
𝐶
), the austenite lattice parameter ( a
𝛾), carbon content ( 𝑥
𝛾
𝐶
) and volume fraction ( f
𝛾) before mechanical testing, and also the critical stress for the austenite
transformation into martensite upon straining ( 𝜎𝛾→𝛼
𝑦 ). A Fe
3
C fraction of ~2-3 vol.% was detected in all samples. The values of carbon content given in the table were derived taking
into account only the eect of alloying elements on the lattice parameter (see text). Estimated microstructural composition of each sample prior to straining is presented as calculated
from dilatometry and XRD, where RA stands for retained austenite.
Material T
q
(°C) t
q
(min) T
p
(°C) t
p
(min) HV
1.0
a
𝛼(Å) 𝑥
𝛼
𝐶
(wt.%) a
𝛾(Å) 𝑥
𝛾
𝐶
(wt.%) f
𝛾(vol.%) 𝜎𝛾→𝛼
𝑐 (MPa) Microstructure
QT RT - 180 60 780(9) 2.86807(13) 0.074(3) 3.5862(2) 0.593(5) 13.3(3) 1140 ~87 % martensite, RA
B-220C 220 30 - - 725(10) 2.86741(14) 0.042(3) 3.5883(2) 0.640(5) 16.0(3) 700 ~55 % bainite, ~28 % martensite, RA
B-240C 240 30 - - 689(7) 2.86726(11) 0.034(2) 3.5931(2) 0.746(4) 13.6(2) 900 ~86% bainite, RA
B-260C 260 30 - - 661(5) 2.86656(11) ~0 3.5974(2) 0.843(5) 8.5(2) 1070 ~91 % Bainite, RA
QP-220C 170 30 220 30 690(6) 2.86778(17) 0.060(4) 3.5917(2) 0.715(5) 19.6(4) 730 ~30 % martensite, 50 % bainite, RA
QP-240C 170 30 240 30 682(9) 2.86734(4) 0.038(1) 3.5954(2) 0.796(3) 12.7(2) 830 ~30 % martensite, ~57
% bainite, RA
QP-260C 170 30 260 30 664(5) 2.86669(11) 0.006(2) 3.5976(5) 0.848(11) 4.8(3) - ~30 % martensite, ~65 % bainite, RA
where 𝑎
𝛾is the measured lattice parameter of austenite in Åand
𝑥
𝐶
, 𝑥
𝑀𝑛
, 𝑥
𝐴𝑙
, 𝑥
𝐶𝑟
, 𝑥
𝐶𝑢
and 𝑥
𝑁𝑖
are the concentrations of the respective
alloying elements in wt.%. The presence of silicon does not inuence
the austenite lattice parameter within the experimental accuracy [37] .
We did not observe a clear separation of the {002} reection from
the {200}/{020} doublet, characteristic of the tetragonal distortion of
the body-centered cubic unit cell with a c/a ratio higher than one and
increasing with carbon content in the lattice [38] , for the measured
carbon content in either bainitic ferrite or martensite, or evidence of
martensite tetragonality in the 2 𝜃-range covered in this experiment.
The average lattice parameter of tetragonal martensite with 1.0 wt.%C
only diers 0.3 % from the lattice parameter of bcc ferrite [36] . There-
fore the bcc ferritic phase, termed ‘ 𝛼phase’, included in the Rietveld
analysis corresponds to bainite ferrite and/or martensite. The latter cor-
responds in this study to ‘tempered martensite’ in QT sample, ‘parti-
tioned martensite’ in QP samples, ‘untempered martensite’ in B-220C
sample (discussed later on in the results section), and ‘mechanically-
induced’ martensite in all samples upon loading. This is in line with re-
cent experimental observations of bcc martensite induced mechanically
at room temperature in high-carbon Fe-8Ni-1.26C (wt.%), revealing that
the bcc martensite with a random distribution of carbon atoms amongst
all octahedral interstices through short-range diusion, rather than the
body-centred tetragonal (bct) martensite with ordering of carbon atoms,
would be stable in high-carbon steels [39] . The carbon content in the
surrounding bcc matrix was derived using the following relationship,
based on the average carbon eect on the martensite unit cell in steels
containing ≤ 1.51 wt.%C [38] and on model alloys with ≤ 40 at.%Mn
[40] :
𝑎
𝛼= 2 . 8664 + 0 . 00055 𝑥
𝑀𝑛
+ 0 . 020 𝑥
𝐶
(4)
where 𝑎
𝛼is the measured lattice parameter of the surrounding bainitic
ferrite/martensite (namely ‘ 𝛼-phase‘) in Å, and 𝑥
𝐶
and 𝑥
𝑀𝑛
are the con-
centrations of the respective alloying elements in wt.%. Substitutional
solute elements such as Cr or Mn are reported not to partition or in-
uence carbide precipitation signicantly at the relatively low trans-
formation temperatures ≤ 250°C for the formation of lower bainite in
high-carbon steels [ 21 , 41 ]. The eect of substitutional solutes on the
bcc lattice parameter is the same for bainitic ferrite and martensite. The
values of the carbon content in the austenite and in its surrounding bcc
matrix using Eqs. (3) and (4) are also assembled in Table 2 , and they
are plotted as a function of austempering/partitioning temperature in
Fig. 4 a and 4 b.
The carbon content in the “𝛼-phase ” reduces progressively with the
increase in temperature in both the B and QP microstructures, with
somewhat higher values in the case of the QP microstructures. In all
cases, the bcc lattice parameter and carbon content remain lower than
in the tempered martensite phase present in the QT sample. Concomi-
tantly, the micro-hardness values in the B and QP samples decrease at
higher temperatures and remain lower than the micro-hardness in the
QT microstructure. In contrast, the lattice parameter and carbon content
in the austenite phase in the QT microstructure are lower than in any
of the B or QP microstructures. The austenite carbon content in the B
samples remains consistently lower than in the QP samples, with the ex-
ception of the 260°C partitioning/austempering temperature where the
values of the austenite carbon content in both microstructures lie within
the experimental uncertainty.
The value of the martensite start temperature experimentally deter-
mined in this steel by dilatometry was M
S
= 203°C (Fig. S6). The marten-
site volume fraction formed at a given quench temperature can be es-
timated using the Koistinen-Marburger expression, based on data from
Fe-C alloys, carbon steels and SAE 52100 steel [42] :
𝑓
𝛼= 1 − exp
(−0 . 011
(𝑀
𝑆
− 𝑇
𝑄
)) (5)
where 𝑇
𝑄
denotes the quench temperature.
In the QP microstructures, the steel was initially quenched to
𝑇
𝑄
= 170°C from the austenitization temperature of 860°C. Conse-
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Fig. 4. Variation in carbon content: (a) in the martensite/bainite matrix ( 𝛼), and (b) in the austenite phase ( 𝛾), together with (c) the initial austenite fraction before
straining, with the partitioning/austempering temperature. (d) Critical stress of the austenite transformation into martensite as a function of the initial carbon content
in austenite. QP: quenched and partitioned, B: bainitic and QT: quenched and tempered microstructures.
quently, a martensite volume fraction of 𝑓
𝛼~30 vol.% is expected
to form, which is later tempered at either 220°C, 240°C or 260°C dur-
ing the partition step; a portion of the remaining austenite after the
quench is expected to transform into bainitic ferrite during partition-
ing, based on the values measured for the austenite fraction after parti-
tioning ( Table 2 ). It is expected that carbon diuses towards the strain
elds of the dislocations present in supersaturated martensite, and the
subsequent nucleation and growth of nano-sized 𝜀 / 𝜂carbides within
the martensite crystals [30] . In the B microstructures, the steel was
quenched to an austempering temperature of either 220°C, 240°C or
260°C. Based on dilatometry data (see Fig. S2), a bainite volume frac-
tion of 𝑓
𝛼~55 vol.% is formed after austempering at 220°C for 30 min,
generating a microstructure after quenching to room temperature that
is comprised of 55 vol.% lower bainite, 16 vol.% retained austenite, 1
vol.% Fe
3
C and 28 vol.% untempered martensite formed during nal
cooling from the salt bath to room temperature. There was no marten-
site formed during cooling from austempering after 30 min at 240°C
and 260°C (see Fig. S4), and though increased formation of nano-scale
carbides is expected with increasing austempering temperature, this was
not detected within the experimental resolution. For example, according
to dilatometry 86 vol.% bainite is formed after 30min at 240°C (see Fig.
S3); in agreement to this amount of transformation, we have measure
an austenite volume fraction of 13.6 % in the B240C sample. In the ref-
erence QT microstructure, the steel was quenched in oil to 80°C, which
forms martensite as major phase with some retained austenite ( 𝑓
𝛾~ 13.3
vol.%). Fig. 4 c shows the variation of the austenite volume fraction in
the as-heat-treated state with the partitioning/austempering tempera-
ture for the microstructures under examination. The austenite fraction
reduces with increasing temperature in both the B and QP microstruc-
tures, with a crossover at a temperature of 240°C. At that temperature,
the austenite volume fraction in both types of microstructures remains
very close to the austenite fraction detected in the QT microstructure,
despite the fact that the austenite carbon content in the B-240C and QP-
240C microstructures is signicantly higher than in the QT sample (see
Fig. 4 b).
3.2. Mechanical behavior of the QT microstructure
Fig. 5 summarizes the change in austenite fraction (5a), in the
austenite (5a) and martensite (5b) lattice parameters, the lattice plane
strains (5c and 5d), and the full width at half maximum (FWHM) of the
probed { hkl } reections (5e and 5f) with the increase in true stress for
the QT microstructure. The retained austenite remains untransformed
until a true stress value of 1140 MPa (5a), where the start of the austen-
ite transformation into martensite is triggered. The transformation pro-
gresses gradually up to a true stress close to 2000 MPa, where the re-
maining austenite fraction of ~7 vol.% remains untransformed. The
austenite lattice parameter value upon unloading is lower than its value
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Fig. 5. Mechanical response of the quenched and tempered (QT) microstructure during uniaxial tensile testing at room temperature: (a) austenite ( 𝛾) volume fraction
(red) and average lattice parameter (blue) as a function of true stress. (b) Average lattice parameter of martensite ( 𝛼) as a function of true stress. Both (a) and (b) show
the lattice parameter in the post-loading state plotted as a single date point at zero stress. (c) and (d) show individual elastic plane strains along and perpendicular
to the applied load as a function of true stress in the 𝛾-phase and 𝛼-phase, respectively; error bars in all lattice plane strains are smaller than the symbols. (e) and (f)
show full width at half maximum (FWHM) of the austenite and martensite {hkl} reections as a function of true stress. The grey area depicts the region of austenite
mechanical stability, whereas the dashed line marks the critical stress for the austenite transformation into martensite upon straining. (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of this article.)
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Table 3
Diraction elastic constants in GPa for austenite ( 𝛾) and its surrounding martensite/bainite matrix ( 𝛼),
together with the carbon content (wt.%) in 𝛾and in 𝛼, based on the { hkl } reections probed in this
synchrotron X-ray diraction experiment.
Sample 𝑥
𝛾
𝐶
(wt.%) E
{200} 𝛾E
{311} 𝛾E
{220} 𝛾𝑥
𝛼
𝐶
(wt.%) E
{200} 𝛼E
{310} 𝛼E
{110} 𝛼E
{211} 𝛼
QT 0.593(5) 180.4 240.8 293.4 0.074(3) 159.8 177.7 222.1 214.5
B220 0.640(5) 214.2 272.6 297.1 0.042(3) 157.9 175.1 215.7 214.0
B240 0.746(4) 236.9 319.7 396.6 0.034(2) 174.6 198.6 238.0 237.6
B260 0.843(5) 213.1 295.7 394.5 ~0 160.1 178.8 207.7 210.4
QP220 0.715(5) 213.6 276.7 349.5 0.060(4) 161.8 174.6 211.6 213.8
QP240 0.796(3) 216.4 287.1 367.4 0.038(1) 161.3 181.1 222.7 222.7
QP260 0.848(11) 221.2 360.7 368.0 0.006(2) 159.3 178.5 218.6 217.4
Avera ge 213.7 293.3 352.4 162.1 180.6 219.5 218.6
before the mechanical test. Furthermore, the lattice plane strains for
austenite (5c) vary linearly with true stress up to the aforementioned
critical stress for austenite transformation. Beyond that critical stress,
the plane strain behavior of {200}, {220} and {311} deviates simul-
taneously from linearity (5c), and there is concomitantly an increase
in the FWHM of those austenite reections (5e). The error in the sin-
gle peaking tting propagated through to values obtained for the plane
strains is between 1.9E-5 in zero load condition with maximum volume
of retained austenite, and 7.2E-5 at maximum load with minimum retain
austenite. The rate of FWHM increase in austenite becomes less severe
≥ 2000 MPa.
An increased density of lattice defects due to crystal or trans-
formation plasticity causes additional lattice strain and consequently
diraction-line broadening [43] . In contrast, the lattice parameter (5b)
and lattice plane strains (5d) from tempered martensite show a linear
dependence on the true stress up to 1800 MPa, whereupon small devi-
ations from linearity can be observed in the average lattice parameter
and in the four { hkl } reections probed for the martensitic matrix, to-
gether with a small decrease in FWHM (5f). For a cubic single crystal,
the orientation-dependent variation in elastic strain caused by a tensile
stress 𝜎is characterized by the cubic elastic anisotropy factor A
hkl
de-
ned as [ 44 , 45 ]:
𝐴
ℎ𝑘𝑙
=
ℎ
2
𝑘
2
+ 𝑘
2
𝑙
2
+ 𝑙
2
ℎ
2
(ℎ
2
+ 𝑘
2
+ 𝑙
2
)2
(6)
where h, k , and l are the Miller indices of the diracting plane. For
planes aligned with the plane normal approximation along the loading
axis (i.e. generating diracted intensity in the 2D detector at ± 7.5° with
respect to the loading direction), a greater value of A
hkl
implies a greater
stiness:
𝐸
ℎ𝑘𝑙
=
𝜎
𝜀
∥
ℎ𝑘𝑙
(7)
where 𝜀
∥
ℎ𝑘𝑙
corresponds to the parallel hkl plane strain. For austenite
A
200
< A
311
< A
220
(i.e. A
200
= 0.00, A
311
~ 0.16, A
220
= 0.25), and
therefore E
200
< E
311
< E
220
; whereas for the martensite phase A
200
<
A
310
< A
110
= A
211
( A
200
= 0.00, A
310
~ 0.15, A
110
= A
211
= 0.25), and
therefore E
200
< E
310
< E
110
= E
211
. The experimental data in Figs. 5 c
and 5 d, and also the resulting values of E
hkl
in Table 3 , conrm the
order predicted by the cubic elastic anisotropy factor. In both phases, the
{200} direction is the most compliant crystallographic direction along
the axial loading direction. All elastic modulus calculations were limited
to the linear range of hkl -dependent strain [46] .
3.3. Alternative microstructures upon straining
The variation of the austenite volume fraction and average lattice
parameter with true stress is shown in Fig. 6 for the B and QP mi-
crostructures under study. In all cases apart from the QP-260C sample,
the austenite fraction remains constant until a critical value of true stress
is reached. Once that critical stress value is surpassed, austenite trans-
forms progressively up to the highest true stress reached in each tensile
test. In each sample, there remains ~4-7 vol.% untransformed austen-
ite. QP-260C showed a negligible amount of austenite transformation
over the full tensile test. The critical stress for austenite transformation
increases with the austempering/partitioning temperature and as the
𝛼-phase volume fraction increases.
Fig. 4 d displays the critical stress for transformation as a function
of the austenite carbon content prior to the commencement of the ten-
sile test for the three B and the two transforming QP microstructures,
together with the QT microstructure as reference. In both alternative mi-
crostructural types, the critical stress increases with the carbon content
in austenite, but remains below the critical stress characteristic of the
QT microstructure, despite the fact that the lattice parameter and car-
bon content in austenite presents its lowest value in the QT microstruc-
ture. It is of interest to note that although in the QP microstructures the
retained austenite has a higher carbon content than in the B microstruc-
tures, the critical stress for austenite-to-martensite transformation is, in
general, lower than that for bainite. The only exception is the critical
stress in the QP-260C microstructure in which no signicant transfor-
mation occurred, which is in contrast to the B-260C microstructure, in
which the austenite transformation was triggered at 1070 MPa, despite
the fact that the carbon content in austenite is the same in the QP-260C
and B-260C microstructures within the experimental uncertainty. The
QP-260C microstructure presents the lowest initial austenite fraction of
4.8(3) vol.%. Furthermore, the dierence between the pre- and post-
tested austenite lattice parameters decreases with increasing austem-
pering/partitioning temperature, with a return to the initial state and a
slight increase measured in the B-260C and QP-260C microstructures,
respectively.
The lattice plane strains for austenite ( 𝛾) and bainitic fer-
rite/martensite ( 𝛼)are shown in Figs. 7 and 8 , respectively, whereas the
simultaneous variation in FWHM of the {310}
𝛼and {311}
𝛾reections
is displayed in Fig. 9 together with the relative change in austenite frac-
tion. An equivalent variation in FWHM of the other probed 𝛼and 𝛾
{hkl} reections to those shown in Fig. 9 was observed as a function
of true stress. The initial linear behavior of the plane strains in both
phases follows the behavior predicted by the cubic elastic anisotropy
factor described in the previous section (see Table 3 ). The critical stress
for austenite transformation into martensite marks the onset of devia-
tions from linear behavior in all austenite plane strains ( Fig. 7 ), and also
a gradual increase in FWHM of the austenite {311} reection ( Figs. 9 c
and 9 d). This is not the case in the QP-260C microstructure, where the
austenite plane strains do not follow a linear behavior at stresses above
1000 MPa —an observation that also corresponds with a relatively large
increase in FWHM. The austenite reections in QP-260C also undergo
a smaller rst increase in FWHM at stresses above 500 MPa ( Fig. 9 d).
In this microstructure, the 𝛼plane strains deviate from linear behavior
above 1500 MPa, and the austenite shows only a total change of ~1
vol.%.
A true stress of ~1500 MPa is also required to observe a non-linear
behavior in the 𝛼plane strains in the other microstructures ( Fig. 8 ).
At that stress in the B-220C microstructure, there is a signicant in-
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Fig. 6. Austenite ( 𝛾) volume fraction (red)
and average lattice parameter (blue) as
a function of true stress for the studied
bainitic (B) and quenched and partitioned
(QP) microstructures. The lattice parame-
ter in the post-loading state is plotted as
a single date point at zero stress in each
graph. The grey area depicts the region of
austenite mechanical stability, whereas the
dashed line marks the critical stress for the
austenite transformation into martensite
upon straining in each microstructure. (For
interpretation of the references to color in
this gure legend, the reader is referred to
the web version of this article.)
crease in FWHM in the austenite reections ( Fig. 9 c), even though a
more gradual change in FWHM is also observed from the start of the
austenite transformation at the lowest critical stress of 700 MPa. The
FWHM of the 𝛼-phase {310} reection remains linear until a decrease
is seen at applied stress values of 1500 MPa or higher ( Figs. 9 a and 9 b).
The magnitude of the FWHM in the 𝛼-phase decreases with increasing
partitioning/austempering temperature due to the reduction in carbon
content in solution in the 𝛼-phase.
4. Discussion
4.1. Microstructures before tensile deformation
The as-spheroidized samples were austenitized for 30 min at 840–
860°C. Spheroidized microstructures of high-carbon steels take up to
180 min to reach equilibrium during austenitization at temperatures
below 880–900°C [41] . In the QT sample, the subsequent oil quenching
to 80°C and water rinse and nal tempering at 180°C for 90 min led to
an austenite fraction of 13.3(3) vol.% with a carbon content of 0.593(5)
wt.%. The tempering at 180°C induces the loss of carbon from the su-
persaturated martensite, and the formation of nano-sized 𝜀 / 𝜂transition
carbides [47] . In contrast, the retained austenite would not transform
into a bcc structure during tempering at temperatures lower than 200°C
[14] .
Three B-type microstructures were produced by austempering for
30 min at either 220°C, 240°C or 260°C after the austenitization step
at 860°C. SAE 52100 steel undergoes an isothermal transformation into
bainite at temperatures lower than B
S
= 450°C after the microstruc-
ture has been quenched in a molten salt bath to the austempering tem-
perature, with lower bainite being the transformation product at tem-
peratures below 350°C [48] . Plate-like nano-scale carbides form within
supersaturated ferrite in lower bainite in this steel under paraequilib-
rium conditions [ 49 , 50 ], with an equal probability of forming 𝜀 -carbide
(Fe
2.4
C) and 𝜃-cementite (Fe
3
C) from the thermodynamics standpoint
and a potential transition reaction with time and higher temperatures
[50] :
𝜀 -Fe
2.4
C + 0.6Fe ↔𝜃-Fe
3
C (9)
During lower bainite formation, the carbon demand for carbide for-
mation competes with carbon partitioning into the remaining austenite
[51] , so that the untransformed austenite becomes further stabilized and
its M
S
temperature decreases [52] . Our results indicate that the carbon
content in solid solution in lower bainitic ferrite is reduced as compared
to that of tempered martensite, at least with austempering temperatures
of 240°C and 260°C in which no martensite is formed from the remain-
ing austenite during cooling down to room temperature (see Fig. S4),
and decreases with increasing austempering temperature for a holding
time of 30 min (see Table 2 ). The carbon content in the microstruc-
D. Foster, M. Paladugu, J. Hughes et al. Materialia 14 (2020) 100948
Fig. 7. Individual elastic plane strains along and per-
pendicular to the applied load in the austenite ( 𝛾)
phase as a function of true stress for the studied bainitic
(B) and quenched and partitioned (QP) microstruc-
tures. The grey area depicts the region of austenite
mechanical stability, whereas the dashed line marks
the critical stress for the austenite transformation into
martensite upon straining in each microstructure.
ture generated in B-220C is lower in both the martensite/bainitic ferrite
matrix and austenite as compared to its QP contemporary, indicating a
greater loss of carbon to nano-scale carbides formed as a result of the
austempering temperature. In this B-220C sample, we expected a 55
vol.% lower bainitic ferrite based on dilatometry data (see Fig. S2), 2-3
vol.% Fe
3
C and the remaining fraction would correspond to untrans-
formed austenite. However, the measured austenite fraction at room
temperature is only 16 vol.% and is also lower than in the QP-220C,
revealing the formation of untempered martensite during cooling from
220°C to room temperature. Furthermore, the austenite carbon content
increases with temperature despite the relatively low content of carbide
inhibitors such as Si or Al in this steel, but at the expense of a lower
austenite fraction remaining untransformed at room temperature (see
Fig. 4 and Table 2 ). The complete austenite transformation into lower
bainite would take approximately 4 hours at 230°C [2] . This would pro-
vide an enhanced dimensional stability to meet bearing requirements in
the aerospace industry, but at the expense of longer production times
than the QT microstructure.
The QP process can accelerate the lower bainite formation from
austenite by introducing a prior quenching step below M
S
, with ~20%
time reduction for full bainitic transformation [ 52 , 53 ]. In this work, the
Koistinen-Marburger expression predicts an austenite fraction of 𝑓
𝛾~70
vol.% and a fraction of athermal martensite of 𝑓
𝛼~30 vol.%, whereas
the austenite fraction present at room temperature decreases with in-
creasing partitioning temperature from 220°C to 260°C, but remains in
all cases below 20 vol.%. Consequently, austenite transformation into
bainitic ferrite takes place during the partitioning step in QP microstruc-
tures and leads to carbon diusion from the newly formed bainitic ferrite
into austenite ahead of the moving interface during the transformation.
However, the austenite carbon content in QP microstructures is higher
than in B microstructures for the same austempering/partitioning tem-
perature and time. This dierence points to the additional carbon ingress
into austenite from the neighboring martensite, conrmed by the lower
carbon content in solution in QP martensite with respect to QT marten-
site. Carbon partitioning into austenite after martensite transformation
was predicted in Fe-C alloys based on the constrained carbon equilib-
rium criterion [54] and observed experimentally in QT microstructures
of SAE 52100 steel during thermal aging above 180°C using in-situ SXRD
[14] . Additionally, the lower austenite fraction in the B-220C relative
to the QP-220C microstructure could be ascribed to the lower austenite
thermal stability in B-220C as the result of its lower carbon content and
potentially lower stabilization eect from the surrounding lower bai-
nite, causing a fraction to transform into martensite during cooling to
room temperature.