Comparison of Methods for Quantification

of Topologically Close-Packed Phases in Ni-Based

Superalloys
A.S. WILSON, K.A. CHRISTOFIDOU, A. EVANS, M.C. HARDY, and H.J. STONE

The ability to quantify accurately the formation of topologically close-packed phases in

nickel-based superalloys is key to assessing their thermal stability and ensuring that their
performances will not deteriorate during long-term exposure at high temperatures. To
investigate the effectiveness of synchrotron XRD for the detection of such minority phases in
Ni-based superalloys, the commercial polycrystalline alloy RR1000 was analyzed following
exposures of varying times at 800 C. Data were collected from both solid samples and extracted
residues, and additional laboratory X-ray diffraction was performed on the residues. The minor
phases were successfully detected in solid samples using synchrotron radiation, and a
comparison of the results from these quantification methods shows that the extraction method
gives results of the right order of magnitude to reflect the phase quantities that are present in the
alloy. However, the results indicate that the synchrotron route is not a suitable method for the
quantification of phases present in quantities less than approximately 0.3 wt pct.

https://doi.org/10.1007/s11661-019-05442-3
 The Author(s) 2019

I. INTRODUCTION fractions. As such, these alloys are becoming increas-

ingly prone to detrimental TCP phase formation. This
POLYCRYSTALLINE nickel-based superalloys are has motivated several recent studies to investigate TCP
used for hot section components in turbine engines, and phase formation in polycrystalline alloys.[3,5,6] Such
therefore experience long periods of time at elevated precipitation must generally be avoided, which necessi-
temperatures during service. These conditions promote tates balancing the requirement of improved mechanical
the formation of brittle, intermetallic topologically performance and environmental resistance with the need
close-packed (TCP) phases that can be detrimental to to avoid compromising the microstructural stability.[7]
alloy performance.[1] Mechanisms by which TCP phases Key to the assessment of thermal stability during
have been reported to compromise mechanical perfor- alloy development is the ability to accurately quantify
mance include interfacial decohesion at precipitate the amount of TCP precipitation that occurs, allowing
interfaces,[2] brittle fracture of TCP precipitates acceler- direct comparison of performance between alloys. A
ating crack propagation,[3] and stress concentration at number of different possible methods for performing
precipitates causing microcrack generation.[4] Histori- this quantification can be found in the literature, the first
cally, TCP phase formation has been a problem for of which is areal analysis. This involves calculating the
single crystal nickel-based superalloys due to their high average area fraction of a particular phase across a large
refractory contents and c¢ volume fractions. However, number of images that are representative of the alloy’s
new generations of engines present more demanding microstructure, and has previously been used to quan-
operational environments for polycrystalline alloys, tify the precipitation of TCP phases.[8] However, there
requiring the development of novel alloys that contain are various problems with this technique, the most
higher refractory element contents and c¢ volume significant of which is the requirement to be able to
consistently distinguish the phase of interest from other
phases present in the sample. This is not trivial for TCP
phases as one type of TCP phase can be similar in
A. S. WILSON, K. A. CHRISTOFIDOU and H. J. STONE are
with the Department of Materials Science and Metallurgy, University morphology and composition to other TCP phases, as
of Cambridge, Cambridge CB3 0FS, UK. Contact e-mail: well as certain carbides and borides, which makes it
[email protected] A. EVANS is with the Bundesanstalt für difficult to distinguish them using scanning electron
Materialforschung und -prüfung (BAM), 8.5 - Micro-NDT, Unter microscopy (SEM). Selected area diffraction in the
den Eichen 87, 12205 Berlin, Germany. M. C. HARDY is with the
Rolls-Royce plc, PO Box 31, Derby DE24 8BJ, UK.
transmission electron microscope (TEM) allows for
Manuscript submitted May 24, 2019. definitive identification of these phases, but the limited
Article published online October 15, 2019

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 50A, DECEMBER 2019—5925

size of examinable area precludes the use of this method attempt at quantitative evaluation of the accuracy of the
for the determination of bulk volume fraction. results produced using this method.
An alternative method for quantifying constituent The high energy and brilliance of synchrotron X-ray
phases in a material is to obtain an X-ray diffraction radiation allow for the sampling of much larger volumes
(XRD) pattern and then perform Rietveld refinement. of material, and the detection of lower volume fractions
However, due to the exceedingly low volume fractions of of phases, than is possible with laboratory XRD. In this
TCP phases generally present in commercial Ni-based study, diffraction patterns from thermally exposed solid
superalloys, combined with the low symmetry of their samples of the commercial polycrystalline Ni-based
crystal structures, it is not possible to reliably detect and superalloy, RR1000, were collected using synchrotron
characterize these phases using laboratory XRD of a radiation. The aim was twofold: to assess the minimum
solid sample. As a result, the commonly used quantifi- phase fraction that can be detected and quantified in a
cation method for TCP phases is the electrolytic solid sample using this method, and to compare these
extraction of the minor phases via preferential dissolu- results with the quantities obtained using the standard
tion of the c and c¢ phases[9,10] to produce a residue extraction method.
containing a much higher volume fraction of the phases
of interest, on which laboratory XRD can be carried
out. Combining the results of Rietveld refinement of the II. EXPERIMENTAL PROCEDURE
obtained diffraction pattern with the mass of the sample
before and after extraction, and of the extracted residue, The composition of the RR1000 used in this study is
allows the weight fraction of each minor phase in the given in Table I. The alloy was produced via powder
original material to be calculated. processing by ATI powder metals and supplied by
The extraction method has been used in various Rolls-Royce plc. The processing route consisted of the
studies of TCP formation in Ni-based superalloys.[11–14] production of 270 mesh powder followed by Hot
However, a number of potential issues have been Isostatic Pressing at 1107 C and 100 MPa for 4 hours
identified that cast doubt on the reliability of quantita- to form a billet. Cylindrical samples of ~ 5 mm diameter
tive information obtained using this route. The most were obtained from the billet for this work. The c¢ solvus
significant of these is the potential for loss of TCP temperature of RR1000 has previously been reported to
precipitates, and other phases, at multiple stages during be 1147 C[16] and, therefore, a super-solvus solution
the extraction and filtration processes. Examples of heat treatment of 1 hour at 1170 C was performed in
routes for such loss include the inability to capture air. The samples were cooled at a cooling rate of 0.5 to
particles smaller than the pore size of the filter paper 2 C per second from the solution temperature to
during the filtration process, the potential for unrepre- 900 C; below 900 C, the samples were air-cooled.
sentative sampling of the phases in the residue during Long-term isothermal exposures were performed on
transfer to a Si single crystal wafer before XRD, and the RR1000 samples at 800 C for durations varying
difficulties in ensuring the complete removal of residue from 15 to 5000 hours, with the material encapsulated in
from all of the equipment used for extraction and glass ampoules under an Ar atmosphere to minimize
filtration. In addition, multiple weighing steps must be oxidation. SEM samples were produced from the
carried out which all have associated errors. Attempts material using standard metallographic preparation
have been made in previous studies to give an approx- techniques of mounting in conductive resin, grinding
imate value for the size of the random error associated and polishing, culminating with a 20 to 30 minute polish
with such results by repetition of extractions on multiple with colloidal silica. The microstructures of these
samples of the same material, with reported values of ± samples were characterized with back scattered electron
0.3 wt pct[15] and ± 0.2 wt pct.[12] These are significant (BSE) imaging in an FEI Nova NanoSEM and compo-
errors given that reported values for TCP phase frac- sitional data from the phases present were obtained
tions using this method are generally lower than 2 using energy dispersive X-ray spectroscopy (EDX) with
wt pct, and that phase fractions that are lower than the a Bruker detector.
error values have been reported in both of these studies. Two types of samples were prepared from the aged
The use of a consistent experimental methodology material for XRD analysis. Solid cuboidal samples with
following the ASTM experimental standard E963-95[9] a cross-sectional width of ~ 1 mm were produced by
can minimize some of the aforementioned random electrodischarge machining (EDM) and the surface layer
errors. However, the capacity for operator and proce- was then manually ground off to remove any re-cast
dural variation, even while adhering to this standard, layer caused by the machining process. The second type
makes reliable quantitative comparison between results of sample was powder consisting of a concentrated
obtained in different studies difficult. In addition, the residue of minor phases produced from the aged
inability to obtain phase fraction data directly from the material by electrolytic extraction. Electrolytic phase
solid sample means that it is not known how closely extractions were performed using an electrolyte of 1 pct
results obtained through the extraction method repre- tartaric acid in 10 pct HCl-methanol and a current
sent the phase quantities present in the solid material. density of ~ 1 mA mm 2 for 6 to 8 hours, following
Thus far in the literature, some of these potential ASTM standard Practice E963-95.[9] The electrolyte was
problems with the electrolytic extraction method have filtered through a polytetrafluoroethylene-type filter
been identified, but to date, there has not been any with a pore size of 0.45 lm using a water aspirator
pump to obtain the required residue. The residue was

5926—VOLUME 50A, DECEMBER 2019 METALLURGICAL AND MATERIALS TRANSACTIONS A

 Table I. Composition of RR1000 in Atomic Percent

Ni Cr Co Mo Al Ti Ta C B Hf
RR1000 50.5 16.5 18.4 3 6.4 4.3 0.6 0.13 0.08 0.16
All elements measured by XRF except for B by ICP-OES and C using the combustion infrared detection method.

then cleaned using methanol and finally dried in air, and difference between the r and M23C6 precipitates is
kept covered to prevent contamination. illustrated in Figure 2 and quantified in Table II for the
Synchrotron XRD was performed on the ID22 alloy following exposure for 5000 hours at 800 C. It
high-resolution powder diffraction beamline at the can be seen that the r precipitate is primarily enriched in
European Synchrotron Radiation Facility (ESRF). Mo as well as Cr compared with the surrounding matrix
During data collection, the solid samples were contained phase, whereas the M23C6 precipitate is enriched pri-
in Kapton capillaries, and the extracted powder samples marily in Cr with a lower level of Mo. The given M23C6
were contained in 1 mm diameter borosilicate glass composition does not include C as C cannot be
capillaries. The capillaries were spun at 787 rpm to measured accurately using EDX but if the stoichiomet-
reduce any preferred orientation effects on the diffrac- rically expected C content is included then the compo-
tion data acquired. A multianalyzer stage containing 9 sitional trends observed with respect to the r phase
Si analyzer crystals was used with a beam energy of 50 remain the same. An additional difference between the
keV and a beam size of 1 mm by 0.9 mm. Data were two phases is that the M23C6 phase is depleted in Co,
collected over a 2h range of 1 to 34 deg at a scan speed whereas the r phase contains Co at similar levels to
of 0.5 deg/min for the powder samples and 0.75 deg/min those found in the matrix.
for the solid samples. Multiple scans were performed for
each sample and combined to produce the pattern for
B. XRD Data Analysis
analysis.
Laboratory XRD was additionally carried out on the Initial Rietveld refinements were performed on the
extracted residues using a Bruker D8 DAVINCI diffrac- synchrotron XRD data for the extracted material. The
tometer with a Cu Ka source and a position-sensitive structural and site occupancy information obtained was
detector. The extracted residue was deposited on a subsequently used as the starting point for the refine-
single-crystal Si wafer, and the scans were carried out ment of the synchrotron data from the solid samples and
over a range of 10 to 120 deg 2h with a step size of the laboratory XRD data from the extracted material.
0.02 deg and dwell time of 2 seconds. Rietveld analysis The phases included in the refinements were the c and c¢
of the acquired diffraction data was performed using phases as well as the minor phases, M23C6, r and TiC, as
Topas-Academic software.[17] The equilibrium compo- identified using SEM. In addition, HfO2 was identified
sitions of the phases were predicted by thermodynamic in the XRD spectra, consistent with previous reports of
modeling using ThermoCalc[18] software with the HfO2 formation in RR1000.[12]
database TTNi8.[19] The compositions used in the Rietveld refinement for
the c and c¢ phases were those predicted for this alloy at
800 C using ThermoCalc software with the database
III. RESULTS TTNi8 (see Table III). For the c¢ phase, this was
combined with information from Stoloff[20] concerning
A. Microstructural Analysis which elements occupy which type of site within the
ordered structure. Ni and Co were allocated solely to the
SEM BSE images of the microstructure of RR1000
Ni site; Ti, Ta, and Al only to the Al site; and Cr and
after thermal exposures of varying durations at 800 C
Mo, which can occupy both sites, were distributed
are shown in Figure 1. The phases present in these
between them to ensure all site occupancies were equal
microstructures were identified using their appearance
to 1. In both the c and c¢ phases, the site occupancies
and phase contrast in BSE imaging combined with
were not allowed to refine, as this made the refinement
compositional information from EDX, and their pres-
unstable.
ence was confirmed by analysis of the XRD data. These
In addition, the compositions of the M23C6 and r
phases are identified by arrows in the inset in
phases as obtained by EDX on samples following
Figure 1(d). It can be seen that after 15 hours at
exposure for 5000 hours at 800 C were used in the
800 C the predominant minority phase is MC-type
refinements (see Table II). In both phases, the levels of
carbides, which appear bright white in the BSE images
Al, Ti and Ta detected were very low (below 3.5 at. pct)
and are located primarily at prior particle boundaries.
and were considered likely to be the result of X-rays
After 1000 hours of exposure at 800 C, small amounts
originating in the surrounding c¢ phase. Therefore, it
of both dark M23C6 phase and bright r phase are
was decided not to include these in the analysis, and the
present on the grain boundaries, alongside a small
relative amounts of the remaining elements (Ni, Co, Cr,
quantity of MC carbides. Following both 2525 and 5000
and Mo) were re-scaled. This decision is supported by
hours of exposure at 800 C, a higher fraction of larger
compositional data collected using atom probe tomog-
discrete r precipitates can be seen on the grain bound-
raphy found in the literature, which show very limited
aries, alongside the other phases. The compositional

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 50A, DECEMBER 2019—5927

Fig. 1—BSE SEM images of the microstructures of RR1000 after exposure at 800 C for (a) 15 h, (b) 1000 h, (c) 2525 h, and (d) 5000 h. The
MC, M23C6, and r phases are identified in the inset in (d).

Fig. 2—BSE SEM image of a grain boundary in RR1000 following exposure for 5000 h at 800 C along with the associated SEM-EDX
elemental distribution maps. The M23C6 and r phases are identified in the BSE SEM image.

solubility for Al and Ta in the r phase[5] and no positions (see Table IV), so the atoms were distributed
solubility for Al and Ti in the M23C6 phase.[21] The evenly between these in proportion with the number of
M23C6 structure contains 4 non-equivalent metal atom each type of site in a unit cell and the overall phase

5928—VOLUME 50A, DECEMBER 2019 METALLURGICAL AND MATERIALS TRANSACTIONS A

 Table II. r and M23C6 Compositions in Atomic Percent Obtained by EDX from a Polished Specimen Following Thermal
Exposure for 5000 h at 800 °C (The Reported Error is the Standard Deviation of 10 Measurements Made Across a Selection of
Precipitates of the Relevant Phase)

Ni Cr Co Mo Al Ti Ta
M23C6 Bal. 54.2 ±6.0 8.7 ±1.2 6.1 ±0.5 2.3 ±0.6 3.4 ±0.6 0.3 ±0.1
r Bal. 36.1 ±0.4 22.7 ±0.2 10.7 ±0.1 1.6 ±0.0 2.4 ±0.0 0.1 ±0.0

Table III. c and c¢ Phase Compositions at 800 °C Predicted allocated primarily to the 12 coordinated sites and then
Using ThermoCalc with Database TTNi8 (Atomic Percent) to fill the other sites as required. The site occupancies
were then allowed to refine for all elements on each site,
Ni Cr Co Mo Al Ti Ta under the constraints of each site occupancy being equal
c 42.6 26.2 25.2 3.6 2.0 0.3 0 to 1 and the total composition remaining equal to the
c¢ 65.7 1.8 9 0.3 12.2 9.6 1.4 measured composition. The atomic positions used to
start the refinement were taken from Yakel[24] for an
Fe-, Cr-, Ni-, Mo-, and Mn-containing r phase (see
Table V). Seven atomic position parameters were
Table IV. Crystal Structural Information for M23C6 Taken allowed to refine as detailed by Yakel.[24] The compo-
from Medvedeva et al.[22] with Atomic Positions Taken from sitions and atomic positions for TiC and HfO2 were kept
CIF File ICSD-154719 fixed.
Isotropic temperature factors were refined for each
Wyckoff Coordination phase as refining individual temperature factors for each
Site Position x y z Number atom caused very large errors in the values returned. A
M1 4a 0 0 0 12 stable refinement was obtained while allowing lattice
M2 8c 0.25 0.25 0.25 16 parameters, peak shapes, and temperature factors to
M3 32f 0.38 0.38 0.38 13 refine for all phases and also refining site occupancies
M4 48h 0 0.171 0.171 14 for M23C6 and r and atomic positions for the r phase
C 24e 0.272 0 0 8 for the synchrotron data acquired from the 1000-hour,
2525-hour, and 5000-hour-extracted residues.
The synchrotron data obtained from the 5000 hours
at 800 C extracted residue are displayed in Figure 3(a),
Table V. Crystal Structural Information for r Taken from
along with the calculated fit and the residual. The
Joubert[25] with Atomic Positions Taken from Yakel[24]
residual is the difference between the experimental data
Wyckoff Coordination and the calculated fit, and therefore its relatively small
Site Position x y z Number size shows that a good fit to the experimental data has
been obtained. For the primary region of interest,
A 2a 0 0 0 12 shown in Figure 3(b), the peaks are labeled according to
B 4f 0.39800 0.39800 0 15
which phase they originate from. All major peaks have
C 8i1 0.46371 0.13057 0 14
D 8i2 0.73868 0.06572 0 12 been identified and fitted, although it was not possible to
E 8j 0.18302 0.18302 0.2512 14 identify the origin of several minor peaks. The M23C6, r,
and c/c¢ peaks are clearly identifiable and can be fitted
with a large degree of confidence, but the MC peaks are
both broad and difficult to separate from peaks of other
composition. These site occupancies were allowed to phases, causing larger errors to be obtained in the fits of
refine while the total site occupancy of each site was this phase. However, the predominant minor phase in
constrained to be equal to 1 and the overall composition the sample exposed for 15 hours is MC carbide (see
was constrained to be equal to the measured Figure 1(a)) and, therefore, in data from the extract of
composition. this sample, the MC peaks are more intense and well
The crystal structure of the r phase has the P42/mnm defined. As a result, MC phase information obtained
space group and contains 5 non-equivalent positions from fitting the data for the 15-hour sample was used to
with coordination numbers varying from 12 to 15 (see inform the fits for the data from the longer duration
Table V). It is known that Group Vb or VIb atoms that samples. It should be noted that, while only one peak
are deficient in electrons and usually have larger radii has been labeled as resulting from the MC phase in the
preferentially occupy the sites with a coordination displayed region of the spectrum in Figure 3(b), overall
number of 15, while the electron-rich, smaller atoms more than ten peaks were fitted from this phase.
from Group VIIb or VIII prefer to occupy sites with a Table VI gives the lattice parameters and atomic coor-
coordination number of 12.[23] As a result of these dinates for the r phase and the M23C6 phase, which were
ordering preferences, the initial site occupancies were obtained from Rietveld refinement of the synchrotron
calculated by allocating all of the Mo and Cr to the 14 X-ray data from the extracted residue of the 5000-hour
and 15 coordinated sites, whereas the Ni and Co were

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 50A, DECEMBER 2019—5929

 could be stray scatter from the sample passing between
the crystals in the analyzer and reaching the detector.
This is more of a problem at high beam energies, such as
used in this experiment, as the analyzer crystals are at a
low angle with respect to the diffracted beams. Also, it is
possible that the presence of short-range order in these
phases may contribute to these background features. In
addition, the peaks from the minor phases of interest
were of sufficiently low intensities that individual peaks
had to be fitted separately and then the parameters for
that phase fixed in order to obtain a stable refinement.
The synchrotron X-ray data obtained from the 5000
hours at 800 C solid sample is shown in Figure 4. As
expected from the relative fractions of the phases in the
sample, the intensity of the peaks corresponding to the
minor phases is very weak compared to the intensity of
the c/c¢ peaks. Closer examination of the region of the
spectrum displayed in Figure 4(b) reveals that the
majority of minor peaks were correctly fitted but it
must be noted that the fit obtained to both the peaks
and the background is inferior to that obtained for the
data from the extracted residue. Figure 5 combines the
spectra obtained from the solid specimens into one plot
where they are stacked and slightly offset for ease of
visualization. It can clearly be seen that the peaks
associated with the MC phase are present in the 15-hour
specimen, but then gradually reduce in intensity after
longer durations, whereas the M23C6 and r peaks are
not visible after 15 hours but appear in the 1000-hour
specimen and then increase in intensity for the
2525-hour and 5000-hour specimens. In the spectrum
for the 1000-hour specimen, some of these minor phase
peaks can be distinguished by eye, but the majority
cannot be fitted, indicating that the phase fractions are
too low to be reliably quantified in the 1000-hour solid
sample using this technique.
Quantitative results for the weight fractions of a
selection of the minor phases obtained by Rietveld
refinement from the diffraction data are displayed in
Figure 6 and detailed in Table VII. According to the
Fig. 3—X-ray diffraction (XRD) data for the extracted residue after results obtained from the solid samples, the quantities of
thermal exposure for 5000 h at 800 C, displayed with the calculated the r and M23C6 phases increase with the increasing
fit and the residual. (a) Data collected with synchrotron XRD, (b) a exposure time, and the quantity of MC carbides
magnified view of a selected angular range of the synchrotron XRD
data with phase identification markers, and (c) data collected with
decreases (as they are only detectible in the 15-hour
laboratory XRD. specimen). Comparing the results of the two measure-
ment methods on the extracted residues shows very
similar results for the r and M23C6 phases although they
diverge somewhat for the MC phase, particularly for
sample. The laboratory X-ray data obtained from the low exposure durations. Comparing the results from the
extracted residue of the same specimen is shown in solid samples and the extracted residues reveals mostly
Figure 3(c). This data is comparable to the synchrotron the same trends for all of the phases. While the trends
X-ray data but is not of quite such a high quality. It are reproduced, some differences in the magnitude of the
should be noted that in order to obtain a good fit to the values obtained via the different methods can be seen. In
laboratory X-ray data it was necessary to use the addition, an unexpected decrease in r precipitation
parameters obtained from refining the higher quality between 2525 hours and 5000 hours of exposure is
synchrotron data as the starting point for the observed in the results from the extracted residue. To
refinement. investigate this discrepancy, image analysis of BSE
Fitting the data obtained from the solid samples was images of the microstructures of the 2525 hour- and
more challenging for several reasons. The background 5000-hour-exposed specimens was performed to deter-
contained a large number of humps of various sizes, mine the area fraction of r. An area of 0.92 ± 0.06 pct
which had to be modeled using individual pseudo-voigt was found for the 2525-hour specimen and 0.69 ±
peaks. It is suggested that the source of these humps 0.04 pct for the 5000-hour specimen, where the reported

5930—VOLUME 50A, DECEMBER 2019 METALLURGICAL AND MATERIALS TRANSACTIONS A

 Table VI. Lattice Parameters and Atomic Coordinates for
the r and M23C6 Phases Obtained from Refinement of the
Synchrotron X-ray Diffraction Data for the Extracted Residue
of the Sample Thermally Exposed for 5000 h at 800 °C

Atomic Coordinates

Phase Lattice Parameters (Å) x y z

r a = 8.91 0 0 0
c = 4.61 0.396 0.396 0
0.464 0.128 0
0.74 0.066 0
0.183 0.183 0.248
M23C6 a = 10.72 0 0 0
0.25 0.25 0.25
0.38 0.38 0.38 Fig. 5—Synchrotron XRD patterns for solid samples following
0 0.171 0.171 various thermal exposure durations at 800 C. The patterns are
0.272 0 0 offset vertically and horizontally for ease of visualization such that
the numbers on the axes apply only to the 15-h spectra.

Fig. 6—Weight Percent of minor phases in RR1000 at 800 C

calculated from data collected by different experimental methods
(solid line = solid sample measured using synchrotron XRD, dashed
line = extracted sample measured using synchrotron XRD, dotted
line = extracted sample measured using laboratory XRD).

Fig. 4—Synchrotron X-ray diffraction data for the solid sample after
between 2525 and 5000 hours, while the intensity of
thermal exposure for 5000 h at 800 C, displayed with the calculated the c/c¢ peaks is comparable across the two spectra.
fit and the residual. (a) The complete angular range over which
refinement was performed and (b) a magnified view of a selected
angular range with phase identification markers.
IV. DISCUSSION
A. Assessment of Results Obtained by Different Methods
error is the standard error obtained from image analysis
of 30 randomly selected regions for each sample. These The majority of the quantitative results obtained from
results appear to support the unexpected trend seen in diffraction analysis show an increase in r and M23C6
the results obtained from the extracted residues. How- precipitation (Figure 6), which agrees with the observa-
ever, consideration of the relative intensity of the r tions made using SEM analysis (Figure 1). A notable ex-
peaks in the solid spectra for the 2525- and 5000-hour ception to this trend is the decrease in the quantity of the
samples reveals a clear increase in their intensity r phase fraction obtained between 2525 and 5000 hours
of exposure in the data presented from the extracted
residues. This is unexpected and disagrees with the

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 50A, DECEMBER 2019—5931

Table VII. Quantitative Data for Weight Percent of Phases data obtained from the solid samples and thus better
Obtained from Rietveld Refinement of the Diffraction Data for quantification of the accuracy of the results obtained
All Three Acquisition Methods from the extraction method.
Comparison of the results from synchrotron XRD of
Exposure Duration (h) XRD Method r M23C6 MC the solid samples with those reported from a previous
5000 synchrotron (s) 0.87 0.46 0.00 investigation into the stability of RR1000 by Hunt[12]
synchrotron (e) 0.58 0.34 0.12 reveals a considerable discrepancy in the phase fractions
laboratory (e) 0.52 0.33 0.19 obtained. Hunt reported significantly higher r phase
2525 synchrotron (s) 0.47 0.28 0.00 fractions following the same exposure conditions using
synchrotron (e) 0.72 0.28 0.15 the method of electrolytic extraction and laboratory
laboratory (e) 0.75 0.31 0.20 XRD. For example, a r fraction of ~ 4.69 larger than
1000 synchrotron (e) 0.29 0.17 0.25
that found in this study (2.15 wt pct) after ~ 2500 hours
15 synchrotron (s) 0.00 0.00 0.61
synchrotron (e) 0.00 0.00 0.34 at 800 C was reported. Comparison of the material and
laboratory (e) 0.00 0.00 0.74 experimental conditions reported by Hunt to those used
in the current study reveals one difference, which is the
Data collected from solid sample (s) or extracted residue (e). prior processing of the material and resultant grain size.
The current study was performed on hot isostatically
pressed material after 15 hours at 800 C, whereas Hunt
used forged material after 16 hours at 800 C. Assuming
results obtained from the solid samples, which show the equiaxed grains, this results in a 1.449 larger theoretical
continuation of the expected trend of increasing r grain boundary density in Hunt’s material, calculated
precipitation with time. A possible explanation of this from grain size measurements according to ASTM
behavior would be the metastability of the r phase, standard method E112-13.[26] Grain boundaries provide
however, no evidence to support this has been found. preferential sites for r precipitation so a higher grain
Examination of the relative r peak intensities in the boundary density could result in more r precipitation
spectra from the 2525- and 5000-hour solid samples explaining part of the observed discrepancy.
confirms that there is an increase in the quantity of the r
phase in these samples. While the area analysis results
suggest a decrease in phase fraction in the SEM samples, B. Site Occupancies in the r and M23C6 Phases
the possible inaccuracy of this measurement method Information about preferential elemental segregation
when the phase fractions in both samples are so similar to different sites within the r and M23C6 phases can be
makes it a less compelling piece of evidence. Therefore, obtained following the refinements. While the exact
the apparent decrease in r precipitation shown by values obtained vary with repeated refinements of the
analysis of the extracted residues could be an example of data, some significant preferential elemental segregation
the possible problems and inaccuracies with using the can be observed consistently for both phases. The final
extraction method for quantitation of very small quan- site occupancies obtained for the M23C6 phase differ
tities of these phases in commercial alloys. Alternatively, considerably from the originally assigned random dis-
it is possible that this behavior is associated with tribution of the elements in proportion to the overall
sample-to-sample variations, as errors of ± 0.3 wt pct[15] phase composition. A significant difference is the segre-
and ± 0.2 wt pct[12] for phase fractions have been gation of the majority of the Mo content to site M2,
reported for this in the literature. However, it was not which has the highest coordination number of 16. This
possible to investigate this further in the current study may be explained by the relative atomic sizes of the
due to limitations on the availability of synchrotron elements involved; Ni, Co and Cr are all of similar sizes,
beam time. but Mo is significantly larger, hence its preference for
The values of the phase fractions obtained by these the higher coordinated site. The other incidence of
different measurement methods are encouragingly sim- severe segregation of an element to a site was found to
ilar in magnitude (see Figure 6). This is particularly be Cr segregating to site M3 but a mechanistic origin for
apparent for the M23C6 phase and suggests that the this has not yet been identified.
extraction route can provide quantitative results, which For the r phase, the most significant observation is
are representative of the phase fractions present in the the preferential distribution of Mo to site B. Once again,
solid material. This additionally implies that there is not this can be explained by the relative atomic sizes of the
a significant amount of precipitate loss during the elements as Mo is the largest element and, therefore,
various stages of the extraction and filtration processes. preferentially occupies the B site, which has the largest
However, the size of the variation between the results coordination number of 15. This effect has previously
obtained by the different methods for the MC and r been reported by Yakel.[24]
phases is significant given the very low phase fractions
we are considering in this material. Therefore, the
repetition of this experiment using a less stable alloy, C. Factors Impacting the Quality of Data Fitting
which precipitates much larger quantities of r after long A good fit was obtained to the synchrotron data
exposure times, may be warranted for a more conclusive acquired from all of the samples except the 1000-hour
assessment. The presence of larger quantities of the exposure, particularly for the extracted residues. This is
minor phases would allow more accurate fitting of the shown by the small size of the residual produced by

5932—VOLUME 50A, DECEMBER 2019 METALLURGICAL AND MATERIALS TRANSACTIONS A

Rietveld refinement of the data (see Figure 3). However, limit is also affected by whether the peaks are isolated or
when assessing this quantification technique it is impor- in a region containing many other phase peaks as well as
tant to consider any factors that may have impacted the the broadness of the peaks. These last two factors
quality of the obtained fits and thus the accuracy of the explain why the MC carbides can only be quantified in
reported phase fractions. These factors include the phase the 15-hour solid sample and not in the solid material
compositions used and the presence of a small number following longer durations. The MC peaks are very
of unidentified peaks and will be discussed below. broad compared to those of the other phases and some
The r and M23C6 compositions used for analysis of of them coincide with peaks of the r and M23C6 phases,
the data for all of the specimens were obtained by which obscure these peaks and makes them difficult to
SEM-EDX on a polished sample after thermal exposure fit accurately. The 15-hour solid sample contains a
for 5000 hours at 800 C (Table II). Compositions were higher fraction of the MC phase and does not contain
obtained from the alloy in this condition because it the r and M23C6 phases, allowing the MC phase to be
contained the largest precipitates of these phases and fitted with a high degree of confidence. Overall, we can
thus minimized the complications that may arise as a see that, given the right phase and combination of other
result of the interaction volume from which X-rays were phases present, phase fractions as low as ~ 0.3 wt pct
detected including surrounding phases such as c and c¢. can be quantified in solid material for some phases using
However, the thickness of the precipitates is not known, this method, showing the power of this technique for
so there is always the potential for compositional detecting small quantities of phases.
contamination from phases below the precipitate. To
determine whether the compositions of these phases
vary significantly depending on the exposure time, they V. CONCLUSIONS
were also measured in the material exposed for 2525
hours using the same method. The compositions The measurement route used here involving syn-
obtained were very similar for both phases, with the chrotron diffraction of solid specimens allows for the
only significant difference being an apparent change in detection of very low phase fractions (~ 0.3 wt pct) of
the Ni:Co ratio. Using the compositions obtained from minor phases in a nickel-based superalloy. The exact
the 2525-hour-exposed sample as alternative constraints detection limit depends on the phase under considera-
on the site occupancies for the 2525-hour-extracted tion as it is affected by various factors including the
sample data produced, no significant improvement in positions of the peaks relative to the other peaks in the
the fit compared to that obtained using the 5000-hour data and the peak breadths.
compositions. The high quality of the fit obtained to the The phase fractions obtained for the solid and
r and M23C6 peaks in all of the data sets using the extracted samples were of a similar order of magnitude
compositions obtained by EDX from the 5000-hour providing encouraging evidence that the results pro-
sample meant that these data were judged to be duced by the extraction method are generally represen-
sufficiently representative of the true composition for tative of the phases present in the material. The low
the required purpose. phase fractions of r present in all of the RR1000
A small number of low intensity peaks in the data for samples prevented a more detailed assessment of the
the 5000-hour extract have not been accounted for by accuracy of the method from being made. However, it
the identified phases (see Figure 3b) and similar occur- can be concluded that synchrotron diffraction analysis
rences can be seen in all of the data sets, although the of solid samples is not a suitable route for the quantifi-
peak positions are not consistent across all of the cation of phases present in quantities below ~ 0.3 wt pct
spectra. It has not been possible to identify the phases in commercial alloys, as these result in peaks with
which these peaks are originating from so this will have extremely low intensities that can be difficult to fit with
a small effect on the calculated quantities of the confidence.
identified phases. However, the relatively small size of
the peaks involved suggests that this effect is likely to be
minimal.
ACKNOWLEDGMENTS
D. Synchrotron XRD of solid samples as a route for TCP
This work was supported by the Rolls-Royce
quantitation
EPSRC Strategic Partnership under EP/M005607/1,
Some useful conclusions can be drawn from these EP/H022309/1 and EP/H500375/1. The underlying
results as to the lower limits on the quantity of minor research data can be found at https://doi.org/10.17863/
phases that can be detected using this method, as well as CAM.44528.
the factors that affect these limits. Considering the
values obtained from refinement in Table VII, we can
see that the phase fraction required for detection in the
OPEN ACCESS
solid sample depends on the phase involved. For
example, r levels that are expected from the extraction This article is distributed under the terms of the
results to be ~ 0.3 wt pct cannot be quantified in the Creative Commons Attribution 4.0 International
solid 1000-hour sample but M23C6 levels of ~ 0.3 wt pct License (http://creativecommons.org/licenses/by/4.0/),
can in the solid 2525-hour sample. This quantification

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 50A, DECEMBER 2019—5933

which permits unrestricted use, distribution, and 11. H.S. Lee, D.S. Kim, K.B. Yoo, and K.S. Song: Met. Mater. Int.,
reproduction in any medium, provided you give 2012, vol. 18, pp. 287–93.
12. D.W. Hunt: Stability & mechanical properties of a nickel-base
appropriate credit to the original author(s) and the turbine disc alloy, University of Cambridge, Cambridge, 2001.
source, provide a link to the Creative Commons 13. S.E. Kim, M.P. Jackson, R.C. Reed, C. Small, A. James, and
license, and indicate if changes were made. N.K. Park: Mater. Sci. Eng. A, 1998, vol. 245A, pp. 225–32.
14. R.J. Mitchell, C.M.F. Rae, and S. Tin: Mater. Sci. Technol., 2005,
vol. 21, pp. 125–32.
15. R.A. Mitchell: Development of a New Powder Processed Ni-Base
REFERENCES Superalloy for Rotor Disc Application, University of Cambridge,
Cambridge, 2004.
1. R. Reed: The Superalloys Fundamentals and Applications, Cam- 16. A.J. Manning: Development of a Polycrystalline Ni Base Superalloy
bridge University Press, Cambridge, 2006, pp. 51–53. for Gas Turbine Disc Applications, University of Cambridge,
2. Q. Shi, X. Ding, J. Chen, X. Zhang, Y. Zheng, and Q. Feng: Cambridge, 1999.
Metall. Mater. Trans. A, 2014, vol. 45A, pp. 1665–69. 17. A.A. Coelho: TOPAS-Academic. Version 5. 2012, http://topas-ac
3. Y. Liu, Y. Wu, M. Kang, M. Wang, M. Li, H. Gao, J. Wang, ademic.net/.
B. Sun, and Y. Ning: Mater. Sci. Eng. A, 2018, vol. 736A, 18. J.O. Andersson, T. Helander, L. Hoglund, P.F. Shi, and
pp. 438–52. B. Sundman: Calphad, 2002, vol. 26, pp. 273–312.
4. Z. Zhang and Z. Yue: J. Alloys Compd., 2018, vol. 746, pp. 84–92. 19. Thermo-Calc Software, Thermotech Ni-based Superalloys Data-
5. S. Antonov, J. Huo, Q. Feng, D. Isheim, D.N. Seidman, base version 8.0.
R.C. Helmink, E. Sun, and S. Tin: Mater. Sci. Eng. A, 2017, 20. N.S. Stoloff: Int. Met. Rev., 1989, vol. 34 (4), pp. 153–83.
vol. 687A, pp. 232–40. 21. D. Tytko, P. Choi, J. Klower, A. Kostka, G. Inden, and D. Raabe:
6. R. Krakow, D.N. Johnstone, A.S. Eggemen, D. Hunert, Acta Mater., 2012, vol. 60, pp. 1731–40.
M.C. Hardy, C.M.F. Rae, and P.A. Midgley: Acta Mater., 2017, 22. N.I. Medvedeva, D.C. Van Aken, and J.E. Medvedeva: Comput.
vol. 130, pp. 271–80. Mater. Sci., 2015, vol. 96, pp. 159–64.
7. A.S. Wilson: Mater. Sci. Technol., 2016, vol. 33, pp. 1108–18. 23. H.L. Yakel: Acta Crystallogr., 1983, vol. B39, pp. 20–28.
8. J.X. Yang, Q. Zheng, X.F. Sun, H.R. Guan, and Z.Q. Hu: Scr. 24. H.L. Yakel: Acta Crystallogr., 1983, vol. B39, pp. 28–33.
Mater., 2006, vol. 55, pp. 331–34. 25. J.M. Joubert: Prog. Mater. Sci., 2008, vol. 53, pp. 528–83.
9. E04 Committee: Standard Practice for Electrolytic Extraction of 26. E04 Committee: Test Methods for Determining Average Grain
Phases from Ni and Ni-Fe Base Superalloys Using a Hydrochlo- Size, ASTM International, 2014.
ric-Methanol Electrolyte, ASTM International, 2010.
10. M.J. Donachie, O.H. Kriege: Phase Extraction and Analysis in
Superalloys-Summary of Investigations by ASTM Committee E4 Publisher’s Note Springer Nature remains neutral with regard to
Task Group 1, 1972. jurisdictional claims in published maps and institutional affiliations.

5934—VOLUME 50A, DECEMBER 2019 METALLURGICAL AND MATERIALS TRANSACTIONS A