Académique Documents
Professionnel Documents
Culture Documents
Manuscript Draft
Manuscript Number:
Title: Precipitation strengthening of Nb-stabilized TP347 austenitic steel by a dispersion of secondary
Nb(C,N) formed upon a short-term hardening heat treatment
Article Type: Research Paper
Keywords: Steel, precipitation, ageing, transmission electron microscopy, mechanical properties
Corresponding Author: Dr. Ralph Spolenak,
Corresponding Author's Institution:
First Author: Christian Solenthaler
Order of Authors: Christian Solenthaler; Mageshwaran Ramesh; Peter J Uggowitzer; Ralph Spolenak
Abstract: A dense intra-granular dispersion of small secondary Nb(C, N) precipitates in commercial
TP347 Nb-stabilized austenitic steel was formed during a short-term hardening heat treatment in the
temperature window T = 1050C - 950C, starting with isothermal annealing at 1050C and followed
by continuous cooling from 1050C to 950C. The precipitates are semi-coherent in cube-on-cube
orientation relationship with the austenitic matrix, of characteristic faceted octahedral shape with
{111} facets and typically d = 15 4.6 nm in size. The experimentally determined volume fraction f =
1.4 0.2 10-3 correlates well with thermodynamic calculations. The precipitation gives a significant
strengthening of the steel from y initial ~ 529 42 MPa to y hardened ~ 703 39 MPa,
corresponding to a strength increment y ~ 174 MPa (~ 33%) in good agreement with the prediction
for Orowan strengthening. It is suggested that maximizing f and minimizing d within realistic limits
should further improve the strengthening effect towards an estimated maximum y hardened max ~
859 MPa. The present work is relevant to the exploration of the application potential of short term
hardening heat treatments for precipitation strengthening austenitic steels.
Cover Letter
Editor
Materials Science and Engineering A
The central aspect of the reported research is that upon specific, controlled shor t term hardening heat
treatments of commercial TP347 steel dense intragranular dispersions of very fine secondary Nb(C,N)
carbonitrides can be formed with a strong increase of strength as the consequence. Experimental
observations include careful TEM work and are combined with a detailed and coherent discussion based on
thermodynamic calculations and hardening models to present a fundamental assessment of the pr ecipitate
strengthening potential of this technically very important material. It is concluded th at short term
hardening heat treatments may provide an alternative to long term ageing treatments. This work
contributes substantially to the basic understanding of the relationship between heat treatment,
microstructure, strength and application potential of Nb-stabilized austenitic steels. It is relevant (i) on
the metallurgical side to the optimal control of thermal treatments and to the future development of
improved, precipitate-strengthening austenitic steels, and (ii) on the engineering side to the s ave
evaluation and extrapolation of engineering data for the current application of TP347 steel.
We are hoping for a positive decision and remain sincerely yours
Ralph Spolenak
Manuscript
Click here to view linked References
Paul Scherrer Institute, Laboratory for Nuclear Materials, CH-5232 Villigen-PSI, Switzerland.
Laboratory of Metal Physics and Technology, Department of Materials, ETH Zurich, Vladimir-Plelog-Weg 4,
CH 8093 Zurich, Switzerland
Abstract
A dense intra-granular dispersion of small secondary Nb(C, N) precipitates in
commercial TP347 Nb-stabilized austenitic steel was formed during a short-term hardening
heat treatment in the temperature window T = 1050C 950C, starting with isothermal
annealing at 1050C and followed by continuous cooling from 1050C to 950C. The
precipitates are semi-coherent in cube-on-cube orientation relationship with the austenitic
matrix, of characteristic faceted octahedral shape with {111} facets and typically d = 15 4.6
nm in size. The experimentally determined volume fraction f = 1.4 0.2 10-3 correlates well
with thermodynamic calculations. The precipitation gives a significant strengthening of the
steel from y initial ~ 529 42 MPa to y hardened ~ 703 39 MPa, corresponding to a strength
increment y ~ 174 MPa (~ 33%) in good agreement with the prediction for Orowan
strengthening. It is suggested that maximizing f and minimizing d within realistic limits
should further improve the strengthening effect towards an estimated maximum y hardened max
~ 859 MPa. The present work is relevant to the exploration of the application potential of
short term hardening heat treatments for precipitation strengthening austenitic steels.
1.
Introduction
The Nb-stabilized austenitic steel AISI TP347 is widely used in the power generating
industry for piping and tubing systems in power-plants. There are two main application fields
to be distinguished: (i) at intermediate service temperatures of typically 340C in nuclear
power-plants (e.g. cooling circuits of light water reactors) where an optimal combination of
good mechanical properties and high resistance against inter-granular corrosion is the
primary aim, and (ii) at high service temperatures of typically 700C and above in fossil-fired
steam power-plants (e.g. superheater tubes) where high creep strength is an additional
concern [1, 2].
Nb in TP347 steel forms Nb-carbonitrides Nb(C,N) (in the following referred to as
NbX) due to the strong affinity between Nb and C and between Nb and N. NbX has a fcc
crystal structure, type Fm3m (like the austenitic matrix), with a lattice parameter a0
NbX
between 0.447 nm (NbC) and 0.437 nm (NbN) [3]. NbX forms as primary NbXP during
solidification of the steel and as secondary NbXS due to the decreasing Nb-solubility in the
austenitic matrix with decreasing temperature. NbX formation provides two beneficial
effects: (i) It stabilizes the steel against inter-granular corrosion because it prevents the
precipitation of Cr23C6 at grain boundaries and concomitant local Cr-depletion. (ii) Intragranular dispersions of fine secondary NbXS may result in substantial precipitation hardening
and improvement of creep strength. Volume fraction and particle size of hardening NbXS
depend on the heat treatment history of the material, including processing and long-term
service exposure at high temperature. Nb:C ratios of the steel composition of 8:1 10:1
(wt.%) are normally assumed as optimal to precipitate as much NbXS as possible while still
maintaining a low coarsening rate and sufficient stabilization against inter-granular corrosion
[4 7].
Experimental
2.1.
steel supplied as seamless pipes with 219 mm outer diameter and 18 mm wall thickness. The
pipes were fabricated by the Pilgering process, solution annealed at 1100C for 1 h and water
quenched before delivery. The major chemical composition is given in Table 1. TEM
investigations of the as-delivered material revealed that (i) the grain interior was free from
any secondary phases and (ii) dislocations were heterogeneously distributed in planar arrays
on selected {111} slip planes in a moderate density of ~ 1.75 0.25 109 cm-2. For the
present study tubular test specimens (8 mm outer diameter, 5.5 mm inner diameter, 20 mm
3
long) were annealed at 1050C for 1 h in vacuum, followed by continuous cooling to 950C
at a rate of 10 Kmin-1, and subsequent fast cooling in argon atmosphere to room temperature
at 85 Kmin-1. This heat treatment causes a significant hardening of the steel due to an intragranular dispersion of small secondary NbX precipitates and is therefore referred to as
hardening heat treatment, HHT.
2.2.
Methods
Transmission electron microscopy (TEM), including selected area electron diffraction
(SAD) and energy dispersive x-ray analysis (EDX), was applied to characterize the
morphology of the precipitate dispersion and to identify the precipitates unambiguously as
NbX. TEM was performed with a Philips CM 200 at 160 kV and with a FEI Technai F30ST
at 300 kV. TEM samples were thin foil disc samples, electro-polished in a solution of 90%
acetic acid and 10% perchloric acid at 40 V / 10C, and mounted on a low background
double tilt holder. Particle size d was measure manually from kinematical bright field images
(i.e. from images without strain field contrast) and from dark field images formed with NbX
{200} reflections, always from images with the beam direction z = <100>. Since the particles
are of {111} faceted octahedral shape, d was defined as the <110> edge length of the
octahedron. The detailed procedure is given in section 3.2.2. Sample thickness t for the
determination of the NbX volume fraction was derived from the number of thickness fringes
at grain boundaries observed in dynamical {200} and {220} bright field images, assuming
extinction distances for fcc Fe of g {200} = 47 nm and g {220} = 70 nm [17]. For EDX
spectral imaging a beam diameter of 0.5 nm and a step width of 2 nm were chosen. The effect
of the precipitate dispersion on the strength of the steel was evaluated by means of Vickers
micro-hardness measurements on material with and without HHT (series of 20 indents at a
load of 1 N, indents separated by 3 indent diameters). The obtained hardness values, HV,
were converted to approximate yield strength values, y, using the relation y = 3.03 HV
recommended for austenitic stainless steels [18].
3.
Results
3.1.
Microstructure and strength increase y exp after the hardening heat treatment
After the HHT a dense intra-granular dispersion of small NbX precipitates is observed
(Fig. 1; phase identification see section 3.2.1). The particle distribution appears to be rather
homogeneous in images with the beam direction z = <100> (Figs. 1a, 1c and 1d). However,
images with the beam direction z = <110> (Fig. 1b) show clearly that the particles form
preferentially on selected {111} planes. These densely populated planes are separated from
each other by an approximate spacing of ~ 300 500 nm, over which the particle density is
significantly lower. The particles give under near dynamical TEM bright-field imaging
conditions a strong strain field contrast, which is much wider than the true particle size.
Hardness measurements reveal a significant strengthening of the steel (Fig. 2) from y
initial
Apart from the strong matrix reflections (big spots, strong lines), each pattern contains a set
of weak reflections from the precipitates (small spots, dashed lines). Though weak, the
diffracted intensities from the precipitates are clearly visible and the patterns are well defined.
Compared with the patterns from the matrix, always the same common properties are
observed: (i) the same symmetry and orientation, correlating with the fcc crystal lattice, type
Fm3m, and (ii) the smaller reflection spacing RNbX = 0.82 Rmatrix, indicating a larger lattice
parameter a0
NbX
= 1.22 a0
matrix.
matrix
0.3597 nm [19], the lattice parameter of the precipitates is obtained as a0 NbX = a0 matrix Rhkl
matrix
/ Rhkl
NbX
experimental value fits well with the established value for NbC, a0 NbC = 0.447 nm [3]. The
slight deviation (1.8%) may be partially due to the restricted precision which is inherent to
the evaluation of such electron diffraction patterns, however it is consistent with the tendency
a0 Nb(C, N) < a0 NbC. The secondary NbX particles grow according to Fig. 3 in cube-on-cube
orientation relationship with the matrix, i.e. the cubic unit cells of matrix and NbX are
parallel. Since these diffraction experiments were performed under dynamical multiple beam
conditions, multiple diffraction may occur between matrix and NbX (marked with crosses in
Fig. 3a and Fig. 3c). The dominating presence of Nb in the precipitates was qualitatively
verified with EDX element distribution maps.
3.2.2. Shape and size dexp of the secondary NbX precipitates
The characterization of shape and size of the particles requires TEM images formed
under well-controlled imaging conditions. At first the strong strain-field contrast from the
particles has to be suppressed. This can be achieved with kinematical bright-field images and
with dark-field images using diffracted intensity from the precipitates. In the present study
both imaging modes were applied. Additionally, low-index beam directions z must be chosen
(actually z = <100> and z = <110>), because otherwise the true particle shape cannot be
recognized.
Fig. 4 shows two kinematical dark field images formed with NbX {200} reflection of
the same group of precipitates in two different sample orientations. Beam directions are z =
[001] (Fig. 4a) and z = [011] (Fig. 4b). Selected particles are labeled A E for reference. The
sample tilt of 45 around the axis [100] reveals clearly the typical {111} faceted octahedral
particle shape with edges along <110>. While some particles form almost regular
octahedrons (though with slightly blunted or truncated edges), others are elongated along one
<110> edge. The particles appear in the [001]-projection of image Fig. 4a with their square to
rectangular (001) cross-section bounded along the two perpendicular <110> directions [110]
and [110], in the [011]-projection of image Fig. 4b with their rhombic (011) cross-section
bounded by the two planes (111) and (111) at a 70.5 angle. The dashed lines in Fig. 4b
mark the traces of these two {111} planes which are parallel to the beam direction [011] and
perpendicular to the plane of the image. While in Fig. 4a all particles are visible over a
projected width of approximately 50 100 nm as separate objects (though partially
overlapping), all particles (excepting particle A) appear essentially along one line in Fig. 4b,
i.e. all are arranged on the same plane (111). The characteristic thickness fringes from the
particles display position and inclination of the {111} facets. These fringes are sensitive to
shape modulations and trace local deviations from the ideal octahedral shape (blunted or
truncated edges). As shown in Fig. 5 many particles possess essentially the octahedral shape
but with truncated tips, involving {100} facets. Since the particles are in cube-on-cube
orientation relationship with the matrix, simultaneous diffraction from particles and matrix
may cause Mor fringes such as those visible from particles A and B in Fig. 5b. The Moir
fringes appear perpendicular to the diffraction vector ghkl (here {200}) with a fringe distance
dMoir = dp dm / (dp dm) [20], where dp and dm are the interplanar spacing of precipitate and
7
matrix, respectively. With the lattice parameters a0 NbX = 0.439 nm and a0 matrix = 0.3597 nm
(section 3.2.1) the predicted value is dMoir = 0.996 nm, in agreement with the observed
distance.
The particle size d was determined from kinematical bright-field images and from
dark-field images formed with a NbX {200} reflection, always from images with the beam
direction z = <100>. For the purpose, the particle volume V* was calculated for each particle
from the measured perpendicular <110> edges (d1 and d2, d2 d1), assuming perfect {111}
facets, and V* was then normalized to the edge length d of a regular octahedron according to
! ! 2
! ! ! !
3
=
+
=
3
2
2
! !
(1)
Fig. 6 shows the distribution of the particle size d (<110> octahedron edge length) for
a data base of 1568 particles. The average particle size dexp corresponds to dexp = 15 4.6 nm.
3.2.3. Volume fraction fNbX exp of the secondary NbX precipitates
The experimental determination of fNbX
exp
particle density is not uniform: Local measurements give local values as a reasonable
estimate but are not fully representative for the true integral value. (ii) Measuring the sample
thickness t with the applied thickness fringe method (section 2.2) involves an estimated
uncertainty in the order of at least extinction distance g, i.e. at least 10 20 nm. This
may be critical for very thin sample areas, but well acceptable for the investigated thicker
areas with t ~ 250 - 300 nm. Table 2 summarizes the results of five measurements (five
different sample areas). Listed are the lateral dimensions x y of the analyzed sample area,
the range of sample thickness trange over x y, the mean sample thickness tmean over x y, the
analyzed sample volume VS = x y tmean, the volume VNbX of all precipitates within VS, the
number n of particles within VS and the volume fraction fNbX = VNbX / VS. The result
8
according to Tab. 2 will be used for the further argumentation as an approximate range fNbXexp
= 1.4 0.2 10-3.
4.
Discussion
4.1.
understood as a sequence of three separate stages. Stage 1: Solution annealing at 1100C for
1h, followed by water quenching to room temperature. Stage 2: Isothermal annealing at
1050C for 1h. Stage 3: Continuous cooling from 1050C to 950C at a moderate cooling rate
of 10 Kmin-1, followed by rapid cooling to room temperature at 85 Kmin-1. While stage 1 was
part of the fabrication process of the tube material, stages 2 and 3 correspond to the actual
HHT. The observed intra-granular dispersion of small NbX precipitates must have been
formed upon stages 2 and 3 because it was absent in the as-delivered pipe material. Fig. 7
provides the relevant thermodynamic facts for the precipitation behavior, calculated for the
temperature range 1200C 800C using the database PanFe2012 [21]. Shown are a) the
element contents of Nb, C and N in the matrix solid solution together with b) the volume
fraction f of precipitated NbX. Numbers 1, 2 and 3 on the curves refer to the above
mentioned stages of the heat treatment history of the material, arrows on the curves pointing
from 1 to 2 and from 2 to 3 mark the decremental change of the Nb-solubility CNb (T) in the
matrix and the incremental change of the volume fraction fNbX (T) of precipitated NbX during
the HHT. Assuming equilibrium, the following NbX-populations are predicted:
Stage 1: Primary NbXP formed upon solidification of the steel and secondary NbXS1 formed
in the temperature range T 1100C during the fabrication process. The Nb-solubility in the
matrix approaches at 1100C CNb (1100C) = 0.160 wt.%. The major amount (72%) of the Nbcontent of the steel is already precipitated as coarse primary NbXP and coarse secondary
9
NbXS1, corresponding to a volume fraction fNbX (1100C) = 4.64 10-3. After water quenching to
room temperature, the remaining Nb is present in the matrix supersaturated solid solution. It
is reasonable to assume that the NbXP and NbXS1 populations will not be affected by the
following HHT stages 2 and 3.
Stage 2: Secondary NbXS2 formed due to the decomposition of the supersaturated solid
solution during isothermal annealing at 1050C for 1h. The Nb-solubility in the matrix
approaches the equilibrium content CNb (1050C) = 0.111 wt.% and the NbX volume fraction the
equilibrium amount fNbX (1050C) = 5.17 10-3. These NbXS2 particles may tend to overage due
to the comparatively high temperature and long time of stage 2.
Stage 3: Secondary NbXS3 formed due to the continuous decrease of the Nb-solubility in the
matrix upon continuous cooling from 1050C to 950C. The Nb-solubility in the matrix
decreases to CNb (950C) = 0.047 wt.% and the NbX volume fraction increases to fNbX (950C) =
5.96 10-3. Fast cooling from 950C to room temperature conserves the NbXS2 and NbXS3
populations.
Accordingly, the increase of fNbX upon the applied HHT, i.e. the relevant equilibrium
value fNbX th for the observed hardening NbX-dispersion, is predicted as fNbX th = fNbXS2 +
fNbXS3 = 1.32 10-3, with fNbXS2 = fNbX (1050C) fNbX (1100C) = 0.53 10-3 and fNbXS3 = fNbX (950C)
fNbX (1050C) = 0.79 10-3. After the completed sequence of heat treatment stages 1, 2 and 3
the precipitation potential of the sample material is nearly exhausted since almost all (92%)
of the total Nb-content of the steel is precipitated as NbXP + NbXS1 + NbXS2 + NbXS3,
summing up to the total volume fraction fNbX (950C) = (4.64 + 0.53 + 0.79) 10-3 = 5.96 10-3.
As discussed later in section 4.2.2, aims to maximize strength may require the
maximization of fNbX th . The only option is then to extend the temperature window T of the
heat treatment, i.e. to make use of a higher difference CNb(T) > CNb(1100C / 950C) of the Nb-
10
solubility in the matrix. Since at the lower limit T = 950C of the present T the precipitation
potential of the steel is already almost exhausted, T must be extended beyond the upper
limit to temperatures well above 1100C. This means that the present sample material must
first be solution annealed at Tsol > 1100C and quenched to room temperature before the
HHT. Basically, a high Tsol provides a high Nb-supersaturation after quenching and
correspondingly much NbXS may be precipitated upon the subsequent HHT. However, a high
Tsol may involve detrimental grain growth and may also promote susceptibility to intergranular corrosion. For the further argumentation in section 4.2.2 it is therefore assumed that
the upper limit of an optional T does not exceed 1150C. The two extensions of T that will
be considered later are T = 1100C 900C and T = 1150C 900C. The correlated
approximate equilibrium values for fNbX th (T) are fNbX th (1100C 900C) = 1.6 10-3 and fNbX th
(1150C 900C)
4.2.
= 2.3 10-3.
th
= 1.32 10-3. No
prediction is made for the special morphology of the precipitation. Assuming equilibrium, the
theoretical value fNbX
exp
/ fNbX
th
th
exp
exp
differ by fNbX
measurement error for the sample thickness t. The heterogeneous distribution of the particles
may substantially contribute to the error as well: While fNbX th refers to an integral property,
fNbX exp corresponds to a local result which gives only an approximate estimate.
11
th
decrease CNb(T) in the matrix upon HHT, the availability of nucleation sites and the kinetics
of ageing will control the spatial distribution, the size and the shape of the particles. It is well
known that dislocations cause a significant acceleration of nucleation and growth of
precipitates [22-26]. The investigated sample material in the as-delivered state contains
dislocations in a moderate (i.e. sufficient but not abundant) density of ~ 1.75 0.25 109
cm-2 (equal to ~ 15 20 dislocation lines per m2). However, these dislocations are not
homogeneously distributed but concentrated in planar arrays on selected {111} planes.
Accordingly, nucleation and evolution of the NbXS population will result in a heterogeneous
particle distribution with areas of high particle density on selected {111} planes, separated by
areas of low particle density, as observed (Fig. 1b): Each grain is in fact divided into
alternating stronger and weaker domains. As a consequence, it has to be expected that the
present precipitate dispersion cannot provide the full advantage of strengthening and
homogeneous slip distribution.
The observed characteristic {111} faceted octahedral particle shape has been reported
in the literature as an intermediate stage of shape evolution during ageing of NbC precipitates
at 925C in an austenitic model alloy Fe-30Ni-0.1Nb-0.1C [27] and during ageing of VC
precipitates at 900C in an austenitic steel Fe-12Mn-0.3V-0.8C at 900C [28, 29]. Particle
shape was found to evolve with increasing particle size (i.e. with prolonged ageing) from
elliptical or spherical to {111} faceted octahedral and later to {100}-truncated octahedral
approaching the shape of 14-sided tetra-kai-decahedrons. This shape evolution has been
explained in terms of the changing relative contributions of elastic strain energy and
interfacial energy to the total energy of a matrix / particle pair with increasing particle size:
The equilibrium shape for a given precipitate volume is the particular shape for which the
sum of elastic strain energy and interface energy approaches a minimum [30]. However, not
12
yet fully understood is why in the early stage of faceting obviously the {111} facets are
favored over the {100} facets, since for fcc metals the {100} facets should have a somewhat
lower energy than the {111} facets [28, 29, 31]. The present work confirms that the
previously reported observations of particle shape of NbC and VC in the intermediate stage
of ageing are also typical for NbXS in commercial TP347 steel. A new aspect may be the
observation of anisotropic particle growth along {111} / <110>, i.e. particles can grow
asymmetrically while still maintaining a pure {111} faceted (elongated) octahedral shape.
4.2.2. Dispersion morphology and strength increase y exp
This section aims to evaluate the compatibility between the measured morphological
properties of the observed NbXS precipitate dispersion (particle size dexp and volume fraction
fNbX exp) and the measured strength increase y exp.
According to the diffraction results (section 3.2.1), the observed 15 4.6 nm large
NbXS particles display a cube-on-cube orientation relationship with the matrix, i.e. they are
semi-coherent. This means, basically, that dislocations may either cut the particles or must
bypass the particles through the Orowan mechanism. In order to evaluate the strengthening
mechanism and the strengthening effect both possibilities have to be discussed. For the
purpose, the coherency stress coherency (Eq. 2) and the Orowan stress orowan (Eq. 3) are
compared [32]:
!"!!"!#$% = 4
!"#$%& =
! !
! !"# ! !
=4
2
! !
! !
(2)
2
2
~
~
(3)
2
3
2
where f is the NbXS precipitate volume fraction, d the NbXS particle size (<110> octahedron
edge length), G the shear modulus (76 GPa), b the Burgers vector (0.254 nm), a0
NbX
the
13
lattice parameter of NbX (0.439 nm), a0 m the lattice parameter of the matrix (0.3597 nm),
the misfit parameter, k a constant (0.84) and L the average distance between the particles.
The argumentation will be restricted to a qualitative approach for several reasons: (i)
The application of Eq. 2 makes only sense for precipitates which are sufficiently large to
form discrete NbX crystals. It has been reported [27] that for NbC in an austenitic matrix the
critical size for the transition between pre-precipitation clusters (which do not yet possess the
three-dimensional NbC crystal structure) and well developed NbC crystals seems to be in the
range of ~ 2 3 nm. (ii) Eq. 3 is strictly valid only for spherical particles in random (i.e.
homogeneous) distribution. For the present heterogeneous particle distribution Eq. 3 may
give an overestimate of the effective yield strength increment since dislocations can locally
move through grain volume where the effective particle (obstacle) density is significantly
lower than suggested by the integral (average) value. (iii) Dislocations are assumed to move
exclusively on {111} slip planes where they always meet the <110> octahedron edge length
as the relevant obstacle size. (iv) y exp is transformed to a shear stress exp ~ 1/3 y exp ~
58 14 MPa.
Fig. 8 shows the comparison of cutting stress coherency and bowing stress orowan,
calculated according to Eq. 2 and Eq. 3, respectively, for the lower and the upper limit of the
experimentally determined NbXS volume fraction fNbX exp = 1.4 0.2 10-3 . The two curves
intersect at a very small particle size d < 1 nm. As already mentioned above, this picture does
obviously not reflect reality throughout. However, its only but important purpose is to
demonstrate the strong effect of the large misfit ( ~ 22 %) between NbX and matrix, as soon
as discrete NbX crystals are formed: The coherency stresses may very rapidly reach levels
which are prohibitively high for the dislocation cutting process. Though the exact range of
the particle size at the hardness maximum cannot be derived from Fig. 8., it is reasonable to
14
conclude that dislocations cannot cut the observed 15 4.6 nm large particles, even if the
coherency stresses might be relieved to some extent by misfit dislocations in the precipitate /
matrix interface. This conclusion is supported by the suggestion in the literature [23] that
dislocations should interact with carbide particles larger than ~ 4 nm in diameter through the
Orowan bypass mechanism.
All available experimental data (NbXS particle size dexp = 15 4.6 nm, NbXS volume
fraction fNbX exp = 1.4 0.2 10-3 and strength increase exp ~ y exp / 3 ~ 58 14 MPa are
summarized in Fig. 9 and shown together with the curves for the predicted strength increase
orowan, calculated according to Eq. 3 for fNbX
exp
marks as area of interest the overlap of the scatter ranges of dexp and exp. The orowan
curves approach an almost diagonal intersection through this area, suggesting a surprisingly
good qualitative agreement between dexp, fNbX exp , exp and orowan.
It remains to qualify the present precipitate dispersion and its strengthening effect
compared to a maximum NbXS precipitation strengthening potential that may be expected for
commercial TP347 steel within a realistic range of d and fNbX. For the purpose, the strength
increment y is calculated and compared for a variation of d and fNbX. To stay within
realistic limits two further assumptions are made: (i) The variation of the temperature
window T of the heat treatment is chosen according to section 4.1 as T1 = 1100C
950C, T2 = 1100C 900C and T3 = 1150C 900C. i.e. considered as equilibrium
NbXS volume fraction are f 1NbX = 1.3 10-3, f 2NbX = 1.6 10-3 and f 3NbX = 2.3 10-3. (ii) The
variation of d is chosen to cover the range of the experimentally observed value dexp = 15
4.6 nm, i.e. considered as mean particle size are d = 10.4 nm, d = 15 nm and d = 19.6 nm. All
variables and results are summarized in Tab. 3, where T is the temperature window, fNbX
the increase of fNbX within T, orowan and y ~ 3 orowan the predicted strength increment,
15
y initial = 529 MPa the strength of the steel before the HHT, and y hardened = y initial + y.
Within this approach the maximum achievable strength increment corresponds to y max ~
330 MPa and the maximum strength of the steel after the HHT to y hardened max ~ 859 MPa.
The strengthening effect of the present precipitate dispersion approaches y exp ~ 0.52 y
max
and y hardened ~ 0.82 y hardened max. Increasing fNbX from 1.3 10-3 to 2.3 10-3 at constant d
max
and y
max
hardened
and y
hardened
~ 0.90 y
~ 0.88 y
hardened max
hardened max
(constant fNbX ),
respectively. A closer approach to y max and y hardened max requires both increasing fNbX and
decreasing d. The advantage of the high precipitate volume fraction fNbX = 2.3 10-3 gets lost
when the average particle size approaches d ~ 20
5.
Conclusions
The present work deals with the precipitation hardening in a commercial TP347 Nb-
stabilized austenitic steel due to a dense intra-granular dispersion of small secondary NbXS
precipitates. The precipitates were formed upon a short-term hardening treatment in the
temperature window T = 1050C 950C, starting with isothermal annealing at 1050C and
followed by continuous cooling from 1050C to 950C. The morphology of the dispersion, its
thermodynamic background and its strengthening effect were investigated and discussed. The
main findings and conclusions are as follows.
1. The precipitates form semi-coherent in cube-on-cube orientation relationship with the
matrix. They are of characteristic faceted octahedral shape with {111} facets and
typically dexp = 15 4.6 nm in size. The experimentally determined volume fraction
fNbX exp = 1.4 0.2 10-3 correlates well with the thermodynamic prediction for the
equilibrium value fNbX th = 1.32 10-3. The particles nucleate and grow preferentially
16
on selected {111} planes of the matrix due to the heterogeneous distribution of the
available dislocations as nucleation sites.
2. The precipitation results in a significant strengthening of the steel by y ~ 174 MPa
from y
initial
~ 529 42 MPa to y
hardened
strength increase of ~ 33% in good agreement with the prediction for Orowan
strengthening.
3. The strengthening effect may be maximized by increasing the precipitate volume
fraction fNbXS and decreasing the precipitate size d. A significant increase of fNbXS can
only be obtained over a higher Nb-supersaturation in the austenite: The present
sample material has to be solution annealed at T > 1100C and quenched to room
temperature before the HHT. For an optional extended temperature window T =
1150C 900C it is fNbXS ~ 2.3 10-3. Assuming d ~ 10.4 nm, the strengthening
effect may be maximized to an estimated y
hardened max
17
the (wrong) assumption that the microstructure is free from any crucial secondary
phases. Such uncertainty about microstructure may cause ambiguity in the evaluation
and extrapolation of engineering data and consequently serious impact on the safety of
structural integrity assessments.
References
[1]
[2]
D. Dulieu, The role of niobium in austenitic and duplex stainless steels, In: Niobium:
Science and Technology, Niobium 2001 International Symposium, Orlando, 2001,
975-999.
[3]
W.B. Pearson, A handbook of lattice spacings and structures of metals and alloys, vol.
1, Pergamon Press, New York 1958.
[4]
F.B. Pickering, S. Keown, Niobium in stainless steels. In: H. Stuart, editor, Niobium,
Proc. Intern. Symposium. Warrendale, PA: Met. Soc. AIME; 1981, 11131142.
[5]
[6]
[7]
S.R. Keown, F.B. Pickering, Creep strength in steel and high temperature alloys, The
Metals Society, London,1974.
[8]
W.O. Binder, Columbium in Iron and Steel, in: Boron, calcium, columbium and
zirconium in iron and steel, Wiley, New York / London,1957, p. 103-414.
[9]
[10]
[11]
[12]
[13]
[14]
J. Erneman, M. Schwind, P. Liu, J.O. Nilsson, H.O. Andren, J. Agren, Acta Mater. 52
(2004) 4337-4350.
[15]
[16]
[17]
[18]
J.T. Busby, M.C. Hash, G.S. Was, J. Nucl. Mater 336 (2005) 267-278.
[19]
D. Kalkhof, M. Grosse, M. Niffenegger, H.J. Leber, Fatigue Fract Eng. Mater. Struct.
27 (2004) 595607.
[20]
[21]
[22]
[23]
H.S. Zurob, C.R. Hutchinson, Y. Brechet, G.R. Purdy, Acta Mater. 50 (2002) 30753092.
[24]
H.S. Zurob, C.R. Hutchinson, Y. Brechet, G.R. Purdy, Mater. Sci. Eng. A 382 (2004)
64-81.
[25]
[26]
C.M. Sellars, E.J. Palmiere, Mater. Sci. Forum. 500-501 (2005) 3-14.
[27]
[28]
[29]
[30]
[31]
[32]
19
Figure captions
Fig. 1.
TP347, HHT. TEM bright-field images (a and b) and dark-field images formed with a NbX
{200} reflection (c and d) showing a dense intra-granular dispersion of small secondary NbX
precipitates. The beam directions are z = <001> in a, c and d, and z = <011> in b. The
precipitates are preferentially arranged on selected {111} planes (traced by dashed lines in b)
and give under near dynamical bright-field imaging conditions a strong strain-field contrast
which is much wider than the true particle size.
Fig. 2.
Comparison of approximate yield strength y of TP347 before and after the hardening heat
treatment. y values are derived from Vickers micro hardness values, HV, converted as y =
3.03 HV [18]. The observed strength increase corresponds to y exp ~ 174 MPa.
Fig. 3.
TP347, HHT. SAD electron diffraction patterns recorded from 0.5 m wide areas containing
only small precipitates, providing unambiguous evidence that the precipitate dispersion
consists of fcc NbX formed in cube-on-cube orientation relationship with the austenite
matrix. The beam directions are a) z = <110>, b) z = <100> and c) z = <111>. Schematic
drawings give the indexing and mark the pattern unit cells for matrix and NbX.
Fig. 4.
TP347, HHT. Kinematical TEM dark-field images formed with a NbX {200} reflection. The
same group of particles is shown in two different sample orientations. The beam directions
are a) z = [001] and b) z = [011]. The sample tilt of 45 around the [100] axis reveals clearly
the {111} faceted octahedral shape of the particles. The characteristics of the thickness
fringes display position and inclination of the {111} planes. Selected particles are labeled A
E for reference. The dashed lines in b) mark the traces of the planes (111) and (111)
perpendicular to the plane of the image. All particles, excepting particle A, are arranged on
the same plane (111).
20
Fig. 5.
TP347, HHT. Kinematical TEM bright-field image, beam direction z = <011>. a) Group of
precipitates showing the characteristic {111} faceted octahedral shape, partially with
truncated tips, involving {100} facets. b) Outset of a): The distance between Moir fringes
visible in particles A and B correlates with the difference of the {200} interplanar spacing
between matrix and NbX. The dashed lines mark the traces of the {111} planes perpendicular
to the plane of the image.
Fig. 6.
Histogram of the measured particle size d (<110> octahedron edge length) determined
according to the procedure given in section 3.2.2. The average particle size corresponds to
dexp = 15 4.6 nm. The database includes 1568 particles.
Fig. 7.
Precipitation behavior of TP347 in the temperature window 1200C 800C, calculated
using the Pandat database PanFe2012 [21]. Shown are a) the content of Nb, C and N in the
matrix and b) the volume fraction of precipitated NbX. Numbers 1, 2 and 3 on the curves
refer to the heat-treatment history of the sample material, namely: solution annealing at
1100C followed by quenching to room temperature (stage 1), solution annealing at 1050C
(stage 2) followed by continuous cooling to 950C (stage 3) before fast cooling to room
temperature. While stage 1 was part of the fabrication process of the tube material, stages 2
and 3 correspond to the actual hardening heat treatment.
Fig. 8.
Evaluation of the strengthening mechanism: Comparison of cutting stress coherency and
bowing stress orowan, as predicted according to Eq. 2 and Eq. 3 for the upper and lower limit
of the actual volume fraction fNbX
exp
unrealistically small particle size d < 1 nm, the picture demonstrates that the cutting stress
approaches rapidly levels which are prohibitively high as soon as discrete NbX crystals form.
Dislocations should therefore interact with the observed 15 4.6 nm large particles through
the Orowan bypass mechanism.
21
Fig. 9.
Evaluation of the compatibility of experimental data (dexp, fexp, y
exp)
compared to the
predicted strength increase orowan according to Eq. 3. The orowan curves approach an
almost diagonal intersection through the area bounded by the scatter-ranges of dexp and exp
(marked as a dashed rectangle), suggesting good qualitative agreement between the observed
morphology of the NbXS precipitate dispersion and the measured strengthening effect.
22
Tables
Si
Mn
Cr
Ni
Nb
0.058
0.20
1.72
0.025
0.006
17.4
10.4
0.571
0.045
sample
area x y
(nm2)
sample
thickness
trange (nm)
sample
thickness
tmean (nm)
sample
volume VS
108 (nm3)
NbX volume
VNbX 103
(nm3)
number of
particles n
NbX volume
fraction fNbX
10-3
980 x 980
240 - 270
255
2.45
345.49
760
1.447
980 x 980
300 - 330
315
3.03
481.04
385
1.590
720 x 720
240 - 270
255
1.32
200.72
158
1.518
720 x 720
270 - 300
285
1.48
231.98
208
1.571
720 x 720
60 - 90
75
0.39
46.85
84
1.205
Tab. 2. Experimental determination of the volume fraction fNbX of the secondary NbX
dispersion.
23
T C
fNbX
d
(nm)
orowan
(MPa)
y
(MPa)
y / y initial
(%)
y hardened
(MPa)
y hardened /
y initial
1100-950
0.0013
10.4
82
246
47
775
1.47
1100-900
0.0016
10.4
91
273
52
802
1.52
1150-900
0.0023
10.4
110
330
62
859
1.62
1100-950
0.0013
15
57
171
32
700
1.32
1100-900
0.0016
15
63
189
36
716
1.35
1150-900
0.0023
15
77
231
44
760
1.44
1100-950
0.0013
19.6
44
132
25
661
1.25
1100-900
0.0016
19.6
50
150
28
679
1.28
1150-900
0.0023
19.6
60
180
34
709
1.34
Tab. 3. Optional variation of the heat treatment window T and effect of particle size d and
volume fraction fNbX on the strength increment y and on strength y hardened.
24
Figure 1
Click here to download high resolution image
Figure 2
Click here to download high resolution image
Figure 3
Click here to download high resolution image
Figure 4
Click here to download high resolution image
Figure 5
Click here to download high resolution image
Figure 6
Click here to download high resolution image
Figure 7a
Click here to download high resolution image
Figure 7b
Click here to download high resolution image
Figure 8
Click here to download high resolution image
Figure 9
Click here to download high resolution image