Available online at www.sciencedirect.com
Acta Materialia 60 (2012) 5049–5055
www.elsevier.com/locate/actamat
A quantitative atom probe study of the Nb excess at prior
austenite grain boundaries in a Nb microalloyed strip-cast steel
Peter J. Felfer a,⇑, Chris R. Killmore b, Jim G. Williams b, Kristin R. Carpenter b,
Simon P. Ringer a, Julie M. Cairney a
a
Australian Centre for Microscopy and Microanalysis, The University of Sydney, Madsen Building F09, NSW 2006, Camperdown, Australia
b
Metallurgical Technology, BlueScope Steel, Five Islands Rd, NSW 2500, Port Kembla, Australia
Received 6 March 2012; received in revised form 5 June 2012; accepted 6 June 2012
Available online 22 July 2012
Abstract
Most modern HSLA steels rely on the effect of Nb in steels to achieve the properties desired for a specific application. While the role
of Nb in forming precipitates has been well characterized, its role in a solid solution is less well understood due to the difficulty of obtaining quantitative experimental data. In the current work, site-specific atom probe tomography was used to quantify the amount of Nb
Ò
present at prior austenite grain boundaries in a commercial strip-cast steel, produced via the Castrip process. This was compared to
the amount of Nb found at ferrite–ferrite grain boundaries that had formed during the transformation from austenite to ferrite. With
the interfacial excess Nb measured, thermodynamic calculations were carried out and compared to the change in transformation
temperature obtained by dilatometry, with reference to a comparable Nb free, strip-cast steel.
Ó 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Atom probe tomography; Bainitic steel; Grain boundary segregation; Nucleation of phase transformations; Strip-casting
1. Introduction
In many microalloyed steels, Nb is used to improve the
mechanical properties through grain refinement, precipitation hardening and the production of microstructures
containing acicular ferrite, bainite and martensite. This
happens through conditioning of the austenite grain
structure and/or shifting of the austenite (c)–ferrite (a)
transformation to lower temperatures. The role of Nb in
controlling the austenite microstructure through precipitation of Nb(C,N) is firmly established and understood on
the micrometer length scale, as reviewed by DeArdo [1].
The microstructural evolution of these precipitates has
largely been characterized using transmission electron
microscopy. It has been established that Nb(C,N) precipitation in the austenite regime takes place almost exclusively
at lattice defects such as grain boundaries and dislocations,
⇑ Corresponding author. Tel.: +61 2 9351 7679.
E-mail address: peter.felfer@sydney.edu.au (P.J. Felfer).
owing to the large lattice mismatch between Nb(C,N) and
both a and c Fe [2]. These lattice defects are usually introduced by hot deformation, such as rolling and forging, and
precipitation typically takes >100 s to occur.
In steels where the processing time/temperature/deformation scheme is too rapid for precipitation to occur, Nb
will act in solid solution through segregation to lattice
defects, mainly grain boundaries. Nb in solid solution
lowers the c!a transformation temperature, leading to
non-polygonal ferrite microstructures and increased yield
strength, with a pronounced dependency on cooling rates.
This effect has generally been attributed to solute drag of
the prior austenite grain boundaries. Suehiro et al. [3]
found that, in a 0.045 at.% Nb steel, solute drag significantly lowered the transformation temperature if the cooling rate was >10 K s1.
The effect of Nb in solution is more difficult to
characterize than particle precipitation, and is therefore
less well understood. Its impact on transformation and
microstructural evolution in the austenite regime has been
1359-6454/$36.00 Ó 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.actamat.2012.06.013
5050
P.J. Felfer et al. / Acta Materialia 60 (2012) 5049–5055
experimentally established, as mentioned above, but experimental data, both quantitative and qualitative, on the
distribution of Nb within the microstructure is rare. This
is due to the small amount of Nb involved and the
atomic-scale nature of the segregation processes. The atom
probe has both the chemical sensitivity and the necessary
resolution to study the Nb distribution within the microstructure. Some work has previously been undertaken in
this area using atom probe field ion microscopy (APFIM)
[4] and atom probe tomography (APT) [5]. APFIM was
used by Palmiere et al. [4] to determine the Nb bulk concentration within various steels in order to determine the
solubility product. Their analysis reported a bulk solubility
that was significantly lower than values determined by
other methods, but the method by which they derived their
Nb concentrations was not documented in the paper.
Of particular interest is the distribution of Nb at prior
austenite grain boundaries (PAGBs), which poses an even
greater experimental challenge due to their distribution
within the microstructure. A study, using focused ion beam
(FIB)-based preparation of specimens containing ferrite–
ferrite grain boundaries (FFGBs) in Nb- and Mo-containing model alloys for APT was carried out by Maruyama
et al. [5]. After equilibrating at 800 °C, followed by quenching, they measured an enrichment of 0.97 atoms nm2 at
the grain boundaries. Maruyama et al. used a voltagepulsed energy-compensated atom probe with an acceptance
angle of 8° at 75–85 K and 15% pulse fraction (PF), the
ratio of pulsing voltage to DC bias voltage. Their alloy
contained 0.087 at.% Nb and 0.0028 at.% C. Clustering
analysis was also carried out, but no significant clustering
of Nb at the boundary was found. This might be due to
the low amount of C or due to atomic displacements
caused by the FIB preparation [6].
In the current paper, APT was utilized in conjunction
with FIB preparation to measure the interfacial excess of
Nb at prior austenite grain boundaries in a rapidly cooled
strip-cast steel produced via the CastripÒ1 process. Since
the measurement of the interfacial excess cannot be carried
out in situ during the transformation, we extracted several
prior austenite grain boundaries after cooling and compared
the interfacial excess to other ferrite–ferrite grain boundaries that were formed during the c!a transformation.
suppress all Nb precipitation [7]. During processing, the
steel could be cooled rapidly (40–60 °C s1) through the
austenite to ferrite transformation temperature region, to
around 600 °C, followed by run-out-table cooling at a
minimum 4–6 °C s1, and then coiled. The strip was measured to reach 500 °C after about 40 s. No hot deformation
was carried out on these particular specimens, hence no
deformation-induced precipitation would be expected in
the austenite region.
Fig. 1 shows the microstructure of this steel compared to
the microstructure of a steel without Nb, but with otherwise identical composition and processing conditions.
The Nb-free steel showed mainly polygonal ferrite and
some acicular ferrite. After the addition of Nb, the microstructure shifted towards low-temperature reaction products from the austenite decomposition, mainly acicular
ferrite and bainite. This resulted in an increase in the yield
strength from 355 to 460 MPa [7].
Dilatometry work was previously carried out by Killmore et al. [9] for the 0.08 wt.% Nb (0.48 at.% Nb) steel
and the continuous cooling transformation (CCT) diagram
is reproduced here—see Fig. 2. Fig. 2 showed that the effect
of Nb (i.e. more low-temperature reaction products) was
most prominent at higher cooling rates, including those
achieved in the CastripÒ process, as indicated by the cooling curve plotted as a solid black line. Polygonal ferrite was
not observed at cooling rates larger than 10 K s1. This is
consistent with the threshold value found by Suehiro [3],
suggesting that solute drag on the austenite grain boundaries caused by Nb was the dominant effect. The phases
indicated on the CCT diagram (Fig. 2) were based on
metallographic determination after the ISIJ nomenclature
for low-carbon steels [10].
Microstructural studies carried out by Xie [7,8] using
TEM and APT have shown that no significant Nb(C,N)
precipitation was observed in this steel in the initial state
after cooling within the bulk of the material, even when
hot rolling was carried out. It is therefore concluded that
the change in microstructure is predominantly caused by
Nb in solution influencing the nucleation of ferrite at prior
austenite grain boundaries.
2. Materials and methods
Polished steel specimens were etched with a mixture of
2 ml 40% HF in 100 ml 30% H2O2 and rinsed in 30%
H2O2 for optical microscopy. The atom probe samples
were fabricated using a FIB-based, site-specific methodology presented in Ref. [6] that involves positioning the grain
boundary after locating it in the transmission electron
microsope (FEI Quanta 200 3D and a JEOL 2100 microscope were used). Fig. 3 shows a region from which a number of grain boundary specimens have been prepared (the
specific regions from which the specimens have been fabricated are labeled A1, A2 and F in the inset image at the
bottom right of the figure). The specimens were first annular milled with a 30 kV beam and then the grain boundaries
2.1. The material and its thermal history
The steel investigated was a low-carbon microalloyed
Ò
steel, produced by the Castrip process, with a chemical
composition of 0.14% C, 0.84% Mn, 0.4% Si and 0.048%
Nb (all compositions are in at.%). In this steel, the effect
of Nb in solid solution is of particular interest, since the
cooling during processing could be made rapid enough to
1
CASTRIP is a registered trademark of Castrip LLC.
2.2. Sample preparation and atom probe experiment
5051
P.J. Felfer et al. / Acta Materialia 60 (2012) 5049–5055
Fig. 1. Comparison of the microstructure of two strip-cast steels. The only difference in the composition is the Nb content and identical processing
conditions were used. The steel in (a) has no Nb, (b) has 0.048 at.% Nb.
(a)
(c)
FFGB
120 nm
F
PAGB
A2 A1 F
50 µm
Fig. 2. CCT of the investigated steel determined by dilatometry and
metallographic analysis of the resulting microstructure (adopted from Ref.
[9]). The cooling curve of the actual process is indicated with a solid black
line. Note that the temperature of onset of the transformation increases by
150 °C if the cooling rate is lowered from 20 to 10 °C s2.
were positioned close to the tip using a 5 kV beam. Since an
increase in specimen survival rates was observed if the
annular milling is carried out at lower accelerating voltages, later samples (denoted A3 and A4) were milled with
a 10 kV beam and finished with a 5 kV beam. The atom
probe runs were carried out using an Imago (now Cameca)
LEAP 3000 Si, operating at 40 K with a PF of 20%. The
low-kV milled samples were probed at 20 K, which changed the charge state ratio of Nb3+/Nb2+ towards the higher
charge state but did not otherwise affect the analysis.
Fig. 3a shows an overview of the microstructure (FIB
secondary electron (SE) image). This steel has a very
complex microstructure, consisting of acicular ferrite and
bainite, often making the identification of prior austenite
grain boundaries challenging. In this location, the prior
austenite grain boundary is apparent, even though during
transformation some ferrite grains have apparently grown
across the boundary, for example in the region marked X
in Fig. 3b. In Fig. 3b (which is also a FIB-SE image) the
detail of the location of extraction is shown and the sites
of the samples are marked. The difference in contrast
between Fig. 3a and b is caused by the presence of a thin
(b)
20 µm
Fig. 3. The area extracted for atom probe. (a) is an overview of a polished
and etched sample, with the prior austenite grain boundary marked by the
two dashed lines. The detailed view (b) shows the actual extraction site
with the origin of the datasets labeled F, A1 and A2 marked. A region
where a ferrite grain boundary has grown across the austenite grain
boundary is marked with a cross. Inset (c) shows a TEM micrograph of
the sample used to acquire the dataset F, with the prior austenite grain
boundary marked PAGB and the ferrite–ferrite grain boundary marked
FFGB. After taking this micrograph, additional sample length was
removed, so that data from the upper grain boundary, which is located
around 120 nm away from the prior austenite grain boundary, was
acquired in the atom probe.
oxide layer, which was sputtered away before the image
in Fig. 3b was acquired. The orientation contrast in
Fig. 3b makes the position of the prior austenite boundary
more obvious. As shown in the image of the atom probe tip
(Fig. 3c, inset in the top right of the figure), the ferrite–ferrite boundary (marked FFGB in Fig. 3c) measured in sample F was located 120 nm away from the prior austenite
boundary (marked PAGB in Fig. 3c). This boundary has
formed during the c!a transformation and can therefore
be used to compare the segregation levels at a boundary
that existed before the transformation to the segregation
that occurred during the subsequent cooling. The close
proximity of the boundaries in this case allows for a comparison without ambiguities about variations caused by
differences in the thermal history.
In the same way as depicted in Fig. 3a, two more prior
austenite grain boundaries have been extracted (A3 and
5052
P.J. Felfer et al. / Acta Materialia 60 (2012) 5049–5055
(a)
(b)
Cumulative diagram
Σ of Nb atoms
1000
ana
dire lysis
ctio
n
800
600
400
200
0
excess = 170
0
P
5
10 x105
Σ of all atoms
Nb
Fig. 4. Atom map of the dataset A2, with P and Nb displayed (a). P is
displayed since the grain boundary is not obviously discernible if only Nb
is shown. In the cumulative diagram (b) along the analysis direction
marked in (a) the interfacial excess of Nb becomes apparent and can easily
be quantified. The dashed box in (a) marks the outline of the cylindrical
region used for the analysis in (b).
A4). In order to be able to compare the levels of Nb at
these boundaries to random grain boundaries in the bulk,
electropolished samples were also prepared using standard
methods [11], imaged in the transmission electron microscope and the closest grain boundary positioned for APT
analysis using a 5 kV beam in the FIB.
2.3. Measuring the interfacial Nb excess
The analysis of the data was carried out using Cameca
IVAS software for reconstruction, custom MATLAB
(Mathworks Inc.) scripts and the 3Depict point cloud analysis software [12] for the analysis of the segregation. An
atom map of the reconstruction obtained from the dataset
A2 is displayed in Fig. 4a. In this figure, detected elements
within the mass ranges of both Nb and P are displayed. P is
shown because it can clearly be seen at the grain boundary
in the reconstructed data. The weaker Nb segregation is
not so easy to visualize.
The interfacial excess was quantified using cumulative
diagrams (commonly known as the “ladder method”, see
e.g. [13]). An example of a cumulative diagram for Nb in
the boundary A2 is given in Fig. 4b. The interfacial excess
was calculated in the plane of highest solute concentration,
which can be determined with sufficient statistics in datasets from modern wide-angle atom probes. This method
is quite insensitive to a constant background caused, for
example, by a peak overlap, which makes it ideal for the
determination of the Nb excess in steels, where the Nb3+
peak is within the tail of the Fe2+/Si+ peaks owing to the
mass resolution of the instrument used. It is also insensitive
to errors in the positioning of the atoms such as preferential retention [24] as long as they are small compared to
the size of the dataset. Hence, the largest source of error
with respect to the interfacial excess is the uncertainty in
area of the boundary that has been captured, originating
from errors in the reconstruction, an estimation of which
is derived below. The derivation is based on the reconstruction algorithm after Bas et al. [14], although improved
algorithms for wide-angle atom probes by Geiser et al.
[15] have been used for this reconstruction. The Bas
algorithm has a much simpler analytical formulation and
the differences between the protocols are small compared
to the errors considered here.
The reconstruction of the atom probe data relies on the
accurate determination of three parameters, the detection
efficiency, , the field factor, kf, and the image compression
factor, . When the detection efficiency of the instrument has
been properly determined (g = 0.57 in the instrument
used), the field factor kf and the image compression factor
n remain as the most significant sources of error. If a pole
structure is visible in the field desorption pattern (FDP), n
can be calibrated [16], and R as the remaining parameter
can be determined using known lattice spacings or postmeasurement TEM imaging, significantly reducing these
sources of error in the reconstruction. For arbitrarily oriented interfaces, the error in area determination will be a
combination of the error in the x,y coordinate determination and z coordinate error, which are influenced differently
by errors in n and kf. Even for an ideal cylindrical specimen, if the evaporation field changes during the measurement of a grain boundary due to a change in crystal
orientation, additional reconstruction errors can arise if
the radius is derived from the voltage necessary to maintain
a constant evaporation rate [17]. Furthermore, kf and n
generally are not constant during an APT experiment.
However, the change in evaporation field at the boundary
has been found to be small in grain boundaries and the
change in n and kf can be minimized by reconstructing only
the region around the grain boundary. Hence, these effects
will not be taken into account. This will be different for
phase boundaries and other general interfaces. Even
though n and kf have different physical origins, in the
reconstruction, only the term n /kf = ct 2 appears and so
only an error in ct will change the measured area. Investigations by Gault et al. [18] on various materials, including
the steel investigated in this paper, have shown a standard
deviation of 0.02 for ct if the reconstruction is calibrated
from the FDP and the relative radius change of the specimen in the reconstructed volume is less than 2. If ct cannot
be calibrated, average values calibrated on other samples
have to be used. The standard deviation r for the data published by Gault [18] for ct was 0.06, with an average ct of
0.43, and hence if a Gaussian distribution of the values is
assumed with a 95% confidence interval, ct = 0.43 ± 0.12.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
A
cosh
2
ð1Þ
¼ ct ðct sinhÞ þ
A0
c2t
The change in area caused by the deviation of ct from
unity is given in Eq. (1), where A is the area of the grain
boundary in the reconstruction and A0 is the area of the
2
ct = total compression, the deviation of the projection/evaporation
field ratio from that of a small sphere emitting atoms onto a surrounding
concentric sphere.
5053
P.J. Felfer et al. / Acta Materialia 60 (2012) 5049–5055
2
0°
0
-2
0.4
15°
0.5
0.6
0.7
30°
-4
45°
θ
-6
60°
-8
-10
90°
θ
Fig. 5. Dependence of the determined relative area error in an atom probe
reconstruction on ct = n/kf and interface tilt h with respect to the measurement axis. The relative error in area is obtained by multiplying the error
estimate for ct with the value on the ordinate corresponding to ct (see text).
grain boundary in a reconstruction with ct = 1, and h is the
angle between the boundary plane and measurement axis
(Fig. 5).
@A
Fig. 5 shows the derivative @cA0t plotted for different angles
of the interface with respect to the measurement axis. This
value can be used for a first-order approximation of the
error of a flat interface, using Eq. (2). The relative error
@A
in area is obtained by multiplying the @cA0t value for a certain
ct (the average value of 0.43 for uncalibrated reconstructions) and boundary inclination from Fig. 5 with the potential error in ct. Fig. 5 also reveals the large influence of the
interface orientation on the potential error. Interfaces oriented approximately 80° from the axis are least sensitive
to changes in ct for the average value of ct of 0.43.
A
D
@A
A
¼ Dct 0
A0
@ct
S was not included, as its peak overlaps with O. For all
grain boundaries measured, Si or Mn accounted for most
of the interfacial excess. This is not surprising, since these
elements are also the main alloying elements (0.84% Mn,
0.4% Si) in this steel. In the datasets where the grain
boundary was captured close to perpendicular to the
measurement direction, all the segregating elements
appeared to have a distribution around the grain boundary
with a full width at half maximum (measured above base
concentration) of 2–4 nm. This is shown in Fig. 6. This is
somewhat broader than two atomic layers, which are often
quoted as the grain boundary thickness, but consistent with
observations in multicomponent systems made by other
authors using atom probe [19–23]. Since some of the
species exhibit preferential retention of solute atoms on
the atom probe tip (the concentration profiles have been
corrected after a protocol presented in Ref. [24]) these values are only approximate.
For these elements, a comparison of the calculated interfacial excess values revealed no clear difference in the total
amount of solute found at the different types of grain
boundaries. Just one FFGB (F(B2)) was found to have
interfacial excess values that were lower than at the other
boundaries. This may indicate that this boundary has
formed at relatively low temperatures, limiting the amount
of solute that was able to diffuse to this boundary, but this
was not reflected in the Nb excess. Overall, the cumulative
interfacial excess of the substitutional elements ranges
from <10% of a monolayer to 65% of a monolayer. Interestingly, the relative contribution of the individual elements
that are contained in the steel varies strongly from boundary to boundary. This variation may be attributed to the
ð2Þ
C
3. Results and discussion
Mn
Fig. 6 shows representative concentration profiles for
alloying elements that were observed to segregate to the
boundaries (apart from Nb), including C, Si, Mn and P.
2
4
concentration (at%)
1
3
0
2
F
1
0
0
3.1. Distribution of alloying elements other than Nb
A1
P x 10
concentration (at%)
The main focus in this work was the measurement of the
amount of Nb at the grain boundaries, but the total amount
of segregated solute and the distribution of the individual
elements around the boundary was also of interest. While
determining the total amount of solute (the interfacial
excess) is relatively easily achievable, measuring the distribution of the solute atoms about the interface is complicated by the preferential retention of certain solutes
during the field evaporation of the samples [24]. Furthermore, due to the mass resolution of the atom probe used
to acquire the data, the Nb3+ peak (in which most of the
Nb can be found), is within the tail of the Fe and Si peaks.
This does not influence the quantification of the interfacial
excess, but renders accurate spatial mapping of Nb difficult.
3
Si
5
10
15
20
25
30
35
40
distance (nm)
Fig. 6. Concentration profiles of the main alloying elements in the steel
around a PAGB (A1) and a FFGB (F), corrected after a protocol
published in Ref. [24]. The P concentration profile has been magnified by a
factor of 10.
5054
P.J. Felfer et al. / Acta Materialia 60 (2012) 5049–5055
different amounts of equilibrium segregation expected,
depending on the crystallographic nature of the boundary.
of Nb was not caused by non-equilibrium segregation,
but rather by equilibrium segregation.
3.2. Findings from the chemical composition of the datasets
in the vicinity of the boundaries
3.3. Nb excess at prior austenite vs. ferrite–ferrite grain
boundaries and the implications for the fcc–bcc
transformation
When a material is cooled rapidly from its melting temperature, as occurs in these specimens using the CastripÒ
process, non-equilibrium segregation is a common
phenomenon. In contrast to equilibrium segregation,
non-equilibrium segregation does not just influence the
composition in the grain boundary plane and immediately
adjacent, but in a region that can stretch up to several
microns, with a pronounced dependency on cooling rate
and start temperature. This is well established for B in both
Fe- and Ni-based alloys [25], where the distribution can be
mapped using autoradiography, but has not, to the knowledge of the authors, been investigated for Nb. If this is the
case, different amounts of Nb should be found close to a
PAGB, compared to a FFGB. In Fig. 7, the measured
Nb concentrations of the datasets that contain a PAGB
(omitting the enriched region around the boundary) and
datasets that have been collected in the immediate proximity (<50 nm)3 of PAGBs (A1–A5) are compared to those
collected from specimens in random locations within the
material (F, F(B1), F(B2)).
Lower concentrations of Nb were indeed found in the
vicinity of the PAGBs, but the dataset acquired around
120 nm away from the PAGB (F) showed a level of Nb
comparable to the datasets acquired in the bulk. For all elements other than Nb, no difference in composition between
the bulk and PAGB regions was found. The above distance, 120 nm, matches the approximate integrated diffusion distance (di) in ferrite for cooling from 650 °C to
room temperature (calculated using values from Ref. [1]),
where 650 °C is selected as it is the transformation start
temperature (see Fig. 2). This indicates that this depletion
Although the segregation levels for most species vary
widely between individual boundaries, this is not the case
for Nb (Fig. 8). Consistent values were found for both
the PAGBs and the FFGBs. A possible explanation for this
observation is the high segregation energy involved,
leading to saturation largely independent of the grain
boundaries’ actual free volume, as higher-energy sites
would also be occupied by Nb atoms. The average interfacial excess at the FFGBs was found to be significantly
lower than at the PAGBs, at 0.10 ± 0.035 atoms nm2.
The consistency of the values for the FFGBs is remarkable
given that the samples were taken from very different areas,
one being close to PAGBs and others from random positions that can be assumed to be located in the bulk.
At the PAGBs, the average value is 0.25 ± 0.02 atoms
nm2, more than double the Nb found at the FFBGs. Even
though the error for the average interfacial excess (provided by Fig. 5) was fairly low, there were some values that
deviated from that value by more than the margin of error.
These deviations are most likely caused by the different
crystallography of the individual boundaries and are therefore not random. An influence from a different thermal history or compositional variations of the individual
boundaries is also possible, but unlikely since the CastripÒ
process leads to very uniform solidification and cooling.
Nb in solution is considered to inhibit the formation of
polygonal ferrite by segregating to the austenite grain
boundaries, rendering them less potent nucleation sites
for ferrite and effectively lowering the transformation temperature. As a consequence, in the Nb-alloyed steel, only a
0.35
FFGB
PAGBs
FFGB
0.3
0.25
0.25±0.02
1200
atoms/nm2
Nb concentration (ppm)
1600
PAGBs
800
400
0.2
0.10±0.035
0.15
0.1
0.05
0
A1
A2
A3
A4
A5
A6
F
F(B1) F(B2)
0
Fig. 7. Nb concentration in datasets in the vicinity of (but not containing)
PAGBs (A1–A6) and at FFGBs both 120 nm away from a PAGB (F)
and in the bulk (F(B1), F(B2)), excluding the boundary itself. Due to the
segregation, the Nb level is lower at the PAGBs. The measured Nb
concentrations are higher than the nominal one (480 ppm) due to the
convolution of the Nb3+ peak with the Fe2+ peaks.
A1
A2
A3
A4
F
F(B1)
F(B2)
sample ID
Fig. 8. Interfacial excess of Nb in the various datasets. A1–A4 are datasets
acquired from prior austenite grain boundaries; F, F(B1) and F(B2) are
datasets acquired from regular ferrite–ferrite boundaries.
P.J. Felfer et al. / Acta Materialia 60 (2012) 5049–5055
5055
small amount of polygonal ferrite was observed and the
majority of the microstructure was a mixture of acicular
ferrite and bainite, lower transformation products. The
value of 0.25 atoms nm2 that we have detected is about
a quarter of the value of 0.97 atoms nm2 found by Maruyama et al. [5] for the case of FFBGs for an alloy with
almost twice the Nb content after equilibrating at 800 °C
for 104 s. This suggests that a relatively small amount of
Nb is required to inhibit the transformation of polygonal
ferrite. Due to the fact that polygonal ferrite grows across
PAGB, the interfacial excesses in these areas of the PAGBs
cannot be measured by our technique. This is unfortunate,
as it would be interesting to determine if regions where
polygonal ferrite was favoured had lower levels of Nb
segregation and hence a lower barrier to overcome.
Maruyama et al. [5] also calculated the segregation
enthalpy of Nb in ferrite from their experiment, obtaining
a value of 38 ± 2 kJ mol1. This value, in conjunction
with the interfacial excess measured in this work, can be
used for a quantitative assessment of the suppression of
the phase transformation. Simply put, the amount of
energy needed to move the segregated atoms from a grain
boundary position to a bulk position as a ferrite grain is
growing across the boundary must be compensated by an
increase in Gibb’s energy. This means higher supercooling
is necessary for the ferrite nuclei to grow, suppressing the
transformation. Using a linear approximation for the
increase in Gibb’s free enthalpy for the transformation
from austenite to ferrite in pure Fe (Ref. [26] and references
therein), the change in transformation temperature can be
calculated for the interfacial excess of 0.25 atoms nm2.
This yields a change of 97 ± 8 °C, which is in good agreement with the experimental value of 150 °C (Fig. 2).
the lowering of transformation temperature of 150 °C
due to the addition of 0.08 wt.% Nb to the base steel, as
was determined by dilatometry.
4. Conclusions
[12]
The role of Nb in solid solution in a commercial stripcast steel, produced via the CastripÒ process, was investigated by means of APT. Prior austenite grain boundaries
were extracted using a FIB-based lift-out technique and
compared to grain boundaries in the ferrite microstructure
that formed during the transformation from fcc to bcc. The
microchemistry around the grain boundaries showed no
distinction between prior austenite and ferrite–ferrite
boundaries apart from Nb, which showed a clear difference. The chemical composition of the regions around
the grain boundaries gave no indication that non-equilibrium segregation was involved in this steel. At ferrite–ferrite grain boundaries, an average of 0.1 ± 0.035
atoms nm2 was measured, whereas prior austenite grain
boundaries showed an average coverage of 0.25 ± 0.02
atoms nm2. This is significant compared to the equilibrium value in ferrite, obtained at 800 °C (0.087% Nb) by
Maruyama et al. of 0.97 atoms nm2. Thermodynamic
data has been used to calculate the supercooling that is necessary to compensate for the energy of segregation. This
amounts to 97 ± 8 °C, which is in good agreement with
[13]
[14]
Acknowledgements
The authors would like to thank Kelvin Xie for providing the micrographs in Fig. 1, and Mr. Joffrey Longour for
programming the interfacial excess quantification program.
The authors gratefully acknowledge the technical and
scientific input of the AMMRF node at the University of
Sydney, as well as the support through the travel and access program of the AMMRF. This work was funded by
the Australian Research Council (ARC).
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
DeArdo AJ. Int Mater Rev 2003;48:371.
Jack DH, Jack KH. Mater Sci Eng 1973;11:1.
Suehiro M, Liu ZK, Agren J. Acta Mater. 1996;44:4241.
Palmiere EJ, Garcia CI, Deardo AJ. Metall Mater Trans A-Phys
Metall Mater Sci 1994;25:277.
Maruyama N, Smith GDW, Cerezo A. Mater Sci Eng A-Struct Mater
Prop Microstruct Process 2003;353:126.
Felfer P, Alam T, Ringer SP, Cairney JM. Microsc Res Technol
2012;75:484.
Xie K, Yao L, Zhu C, Cairney J, Killmore C, Barbaro F, et al. Metall
Mater Trans A 2011:1.
Xie Y, Zheng TX, Cairney JM, Kaul H, Williams JG, Killmore CC,
et al. Scr Mater. doi:10.1016/j.scriptamat.2012.01.029.
Killmore C, Carpenter K, Kaul H, Williams J, Edelman D, Campbell
P, et al. In: AIST, editor. MS&T 2010 proceedings. Houston, TX;
2010.
Bainitic committee of ISIJ. Atlas for bainitic microstructures, vol. 1;
1992.
Miller MK, Cerezo A, Hetherington MG, Smith GDW. Atom probe
field ion microscopy. New York: Oxford University Press; 1996.
Haley D, Ceguerra A. 3Depict point cloud analysis program; 2011.
threedepict.sourcefourge.net.
Krakauer BW, Seidman DN. Phys Rev B 1993;48:6724.
Bas P, Bostel A, Deconihout B, Blavette D. Appl Surf Sci 1995;87–
8:298.
Gault B, Haley D, de Geuser F, Moody MP, Marquis EA, Larson DJ
et al. 2011;111:448.
Gault B, de Geuser F, Stephenson LT, Moody MP, Muddle BC,
Ringer SP. Microsc Microanal 2008;14:296.
Chen YC, Seidman DN. Surf Sci 1971;27:231.
Gault B, Loi ST, Araullo-Peters VJ, Stephenson LT, Moody MP,
Shrestha SL, et al. Ultramicroscopy 2011;111:1619.
Baik S-I, Olszta MJ, Bruemmer SM, Seidman DN. Scr Mater
2012;66:809–12.
Sha G, Ringer SP, Duan ZC, Langdon TG. Int J Mater Res
2009;100:1674–8.
Seto K, Larson DJ, Warren PJ, Smith GDW. Scr Mater
1999;40:1029–34.
Felfer P, Ringer SP, Cairney JM. Ultramicroscopy 2011;111:435.
Alam T, Felfer P, Chaturvedi M, Stephenson L, Kilburn MR,
Cairney JM. Metall Mater Trans A 2012. http://dx.doi.org/10.1007/
s11661-012-1085-9.
Felfer P, Gault B, Sha G, Stephenson LT, Ringer SP, Cairney JM.
Microsc Microanal 2012;18:1.
Harries DR, Marwick AD. Philos Trans R Soc Lond Ser A-Math
Phys Eng Sci 1980;295:197.
Chen Q, Sundman B. J Phase Equilib 2001;22:631.