Titanium Alloys for High Temperature Applications - A Symposium Dedicated to the Memory of Dr. Martin Blackburn
Edited by M.W. Peretti, D. Eylon, U. Habel, and G.C. Keijzers
TMS (The Minerals, Metals & Materials Society), 2006
THE ROLE OF JOMINY TESTS IN UNDERSTANDING
TRANSFORMATIONS IN TiAl-BASED ALLOYS
David Hu, M H Loretto, Aijun Huang and Xinhua Wu
IRC in Materials, The University of Birmingham, Edgbaston B15 2TT, UK
Keywords. TiAl alloys, End-quenching, Transformations.
Abstract
The influence of composition and of grain size on the transformation of the high temperature
alpha phase during cooling TiAl-based alloys has been studied using end-quench Jominy tests.
Longitudinal sections of these quenched samples provide information, over a wide range of
cooling rates, on the influence of composition and of the alpha grain size on the solid state
transformations. The transformations occurring during cooling lead to two-phase structures
(lamellar, Widmanstätten, feathery) and to single-phase structures (massive gamma or to retained
alpha) and the factors which influence the balance between these two types of transformation
have been investigated. The observations show that the phases formed on cooling can be strongly
influenced by diffusion occurring during cooling and by the availability of nucleation sites. The
factors favouring the formation of fully massively transformed samples have been clarified and
some limited tensile data obtained are presented which have been obtained from fully massive
samples which have been HIPped (aged) after quenching.
Introduction
There have been many papers published which deal with the influence of cooling rate and
composition on solid-state transformations in TiAl-based alloys [1-6]. The response of this class
of alloys to imposed cooling rates is important both to improve our fundamental understanding
and because significant microstructural changes, which lead to changes in mechanical properties,
can be obtained by changing the thermal history of samples. Examples of these changes are
given in this paper. Some attempts at generating CCT and TTT curves have been published, but
the generation of such curves is a very time-consuming process.
An alternative approach, which allows the generation of much of the data required for CCT
curves, is to end-quench samples and to examine changes in microstructure along the length of
the end quenched sample, which can then be correlated with (measured) changes in cooling rate
along the length of the sample. This technique of Jominy quenching was developed for steels [7]
and has been used occasionally for assessing the response of Ti-based alloys to different cooling
rates [8, 9]. Recently a considerable amount of work has been carried out on TiAl-based alloys
driven both by the desire to understand the significance of different elements on these kinetics,
but also by the fact that appropriate heat treatment of massively transformed samples can lead to
significant improvements in the properties of cast samples [10, 11]. Thus if a fully massive
structure can be retained by rapidly cooling a cast component (whilst avoiding cracking)
subsequent ageing (carried out as the HIPping cycle used for structural castings) can lead to
improved tensile properties. It is of interest therefore to define compositions which lead to an
increased tendency for the massive transformation to occur and conversely to identify what
alloying elements suppress the massive transformation. Hence the main focus in this paper is to
25
identify those factors which strongly influence the massive transformation and this has been
done mainly using end quenching in a Jominy facility although other quenching methods have
also been used where appropriate.
Experimental.
A number of alloys based on Ti-46Al-8Nb has been used in this work and their compositions,
including oxygen-content and details for the methods used to produce the various alloys are
summarised in table 1.
Table 1 showing the compositions and process routes used to produce the samples in this work
Alloy composition (at%) Process details
O (wtppm)
Grain size (Pm)
Ti-46Al-8Nb
Plasma ingot
520
850
Forged ingot
520
250
Ti-46Al-8Nb-0.25B
Plasma ingot, cast as 30mm bar
780
320
Ti-46Al-8Nb-0.5B
Plasma ingot cast as 30mm bar
790
220
Ti-46Al-8Nb-0.5B
Forged ingot
600
90
Ti-46Al-8Nb-1B
Plasma ingot cast to 30mm bar
620
200
Ti-46Al-8Nb
Melted in CaO crucible
1500
1000
Standard Jominy end quenching specimens with a diameter of 25mm and a net length of 97mm,
were machined from ingots, from forged billets and from the cast samples. The cooling rate
profile along the axis in TiAl Jominy-end quenching specimen was established in earlier work,
as being between 4-350°Cs-1, varying continuously with the distance from the quenching end
[12]. The cooling rate measured in a Jominy specimen is termed the ‘nominal cooling rate’
because the cooling rate clearly varies with the sample temperature and it was determined from
the cooling rate at 900°C. These nominal cooling rates were obtained from samples quenched
only from 1150°C in order to limit the reaction between TiAl and the embedded thermocouples
and samples quenched from the alpha phase field (typically about 1300°C) would be expected to
cool through 900°C slightly more slowly. One set of non-standard samples (Ti-46Al-8Nb with
520 and 1500ppm oxygen) which were used for end quenching were only 20mm in diameter and
80mm long because of the size of available materials, but calibration experiments showed that
the microstructures obtained were similar to those in standard Jominy samples. End quenched
specimens were sectioned along the axis after quenching and the microstructures along the axis
correlated with distance from the quenching end and thus with the nominal cooling rate. Cooling
rates beyond the range available from Jominy tests were achieved with furnace cooling, oil
quenching, water quenching and iced brine quenching of 10x10x10mm samples and in some
specific cases salt bath quenching were used.
Microstructural assessment of the samples has been carried out using optical microscopy,
analytical scanning and transmission electron microscopy. Tensile tests have been carried out on
some samples after quenching and ageing in order to assess the change in properties that could
be obtained by heat treatments alone rather than by thermomechanical processing.
Results and Discussion.
The Influence of Grain Size.
Two Ti-46Al-8Nb (at%) samples with the same oxygen content (520ppm) and with different
grain sizes were used for these experiments (see table 1). The coarse-grained sample was taken
directly from a plasma double-melted 25kg ingot. The finer-grained sample was made by forging
26
sections from this ingot at 1150°C to 70% compression at a strain rate of 5x10-3s-1 followed by a
two-step heat treatment with a short incursion into the alpha phase field before quenching. The
average grain sizes were 850Pm (standard deviation 400Pm) and 250Pm (standard deviation
125Pm).
Figure 1 shows the influence of grain size at a cooling rate of 180°Cs-1 and it is clear that the
coarse grained sample is fully massively transformed whereas the fine grained sample contains
lamellar regions at the pre-existing alpha grain boundaries and massive gamma in grain centres.
(a)
(b)
Figure 1 Optical micrographs showing microstructures obtained from (a) fine grained and (b)
coarse grained samples of B-free Ti-46Al-8Nb cooled at 180°Cs-1.
Figure 2 summarises the observations made on the different grain-size samples and several
conclusions are apparent. The cooling rate, for which a fully massive microstructure is formed,
extends over a far wider range in the coarse-grained sample than in the fine grained sample. The
cooling rates which give rise to a lamellar microstructure extend to higher cooling rates in the
finer grained sample. The onset of the massive transformation at higher cooling rates (about
500°Cs-1) is not influenced by grain size. Finally, only a very small amount of the Widmanstätten
microstructure is seen in the fine grained sample.
Massive J
Widmanstätten L
Volume Fraction
Retained D2
Feathery L
Lamellar
1
468,
850μm
0
1
468,
250μm
0
1
10
100
1000
Nominal Cooling Rate (°Cs-1)
Figure 2 Microstructure-cooling rate relationship from Jominy tests on coarse-grained and fine
grained samples of B-free Ti-46Al-8Nb.
27
Influence of Boron Content on Transformations During Continuous Cooling of Ti-46Al-8Nb
Double-melted plasma ingots of 25kg were produced containing 0, 0.5 and 1at% B and sections
from the 0.5B and 1B ingots were remelted in a cold wall induction furnace and cast into 30mm
diameter bars. Bars containing nominally 0.25at%B were prepared by remelting equal amounts
of the 0.5B and 0B alloys in the cold wall induction furnace and casting into 30mm diameter
bars. A section of the Ti-46Al-8Nb-0.5B ingot was forged at 1150°C at a strain rate of 5x10-3s-1
to 70% reduction in height. The oxygen content for all of these samples was between 600 and
790wtppm and the grain sizes for the cast bars of the 0.25B, the 0.5B and the 1B were
respectively 320Pm, 220Pm and 200Pm. The forged billet 0.5B had an average grain size of
90Pm after solution treatment. These details are summarised in table 1.
Figure 3 shows some examples of the different microstructures observed along the lengths of the
end-quenched samples containing different levels of boron. It should be noted that the grain sizes
in the cast samples are similar except for the B-free sample from the double-melted ingot. For a
cooling rate of 180°Cs-1 the B-free alloy is fully massive but the B-containing alloys sample
contain some lamellar patches (figures 3(a) and (c) ). At 17°Cs-1 the B-containing samples are
mainly lamellar with some massive gamma areas in large grains but the coarse-grained B-free
sample is mainly transformed to massive gamma (figures 3(b), (d)).
(a)
(b)
(c)
(d)
Figure 3 Typical microstructures of Ti-46Al-8Nb-xB alloys cooled at different nominal cooling
rates. (a) 0B, 180°Cs-1, (b) 0B, 17°Cs-1, (c) 1B, 180°Cs-1, and (d) 1B, 17°Cs-1.
The results obtained from these samples are summarised in figure 4, which also contains data
from more slowly cooled samples (air-cooled) and more rapidly cooled samples (waterquenched) beyond the dotted lines in this figure.
28
Retained D2
Feathery L
Lamellar
Massive J
Widmanstätten L
1
+0B,
850μm
0
Volume Fraction
1
+0B,
250μm
0
1
+0.25B,
320μm
0
1
+0.5B,
220μm
0
1
+1B,
200μm
0
1
10
1000
100
-1
Nominal Cooling Rate (°Cs )
Figure 4
Microstructure-cooling rate relationships for Ti-46Al-8Nb samples with different
boron levels. All grain sizes are similar as indicated except for the coarse grained B-free sample.
The following conclusions can be drawn from this figure. The cooling rate required for the onset
of the massive transformation is virtually constant for these different samples and corresponds to
about 500°Cs-1. The formation of Widmanstätten and feathery microstructures which occur at
intermediate cooling rates in the coarse-grained B-free alloy are suppressed in the B-containing
sample and in the fine grained B-free sample. The cooling rates over which only massive gamma
is formed is far wider in the coarse-grained B-free sample than in all other samples shown here.
Finally and importantly, the cooling rates over which the lamellar structure is formed increases
to high cooling rates in the B-containing and in the fine-grained B-free samples.
Similar observations have been carried out on the two different grain-size (220 and 90Pm)
samples of Ti-46Al-8Nb-0.5B and these also show that the quenching rates over which massive
gamma is formed extend to slower rates in the coarser grained sample. The factor which limits
the formation of massive gamma in the fine grained sample is the formation of the lamellar
structure at higher cooling rates than is found in the coarser grained sample. This is illustrated in
figure 5.
Influence of Oxygen on Transformations During Continuous Cooling
Samples of Ti-46Al-8Nb containing 520 and 1500ppm oxygen, both of which had grain sizes of
about 1000Pm have been end quenched and as suggested in earlier work it has been found that
oxygen strongly influences the transformations that occur during cooling from the alpha phase.
In both samples the microstructure observed at the highest cooling rate samples is mostly
retained alpha, with massive gamma mostly at the pre-existing alpha grain boundaries (figures
6(a) and (b)). At slower cooling rates the low oxygen sample transforms to a fully massive
29
structure. In contrast in the higher oxygen sample increasingly large volume fractions of fine
lamellae with some massive gamma in the grain centres is formed (figures 6(c) and (d)).
Retained D2
Volume Fraction
Lamellar
Massive J
1
0.5B,
220μm
0
1
0.5B,
90μm
0
1
10
1000
100
-1
Nominal Cooling Rate (°Cs )
Figure 5 Microstructure-cooling rate relationship obtained from Jominy tests on coarse-grained
and fine grained samples of Ti-46Al-8Nb-0.5B. The data between the dotted lines represent the
data obtained from the end-quenched samples. Data at slower and faster cooling rates was
obtained from air cooled and water quenched samples.
(a)
(b)
(c)
(d)
Figure 6 Optical micrographs showing the influence of cooling rate on Ti-46Al-8Nb containing
(a) and (c) 520ppm O and (b) and (d)1500ppmO. High cooling rates (a) and (b).Slow cooling
rates (c) and (d).
30
Influence of Alloy Composition on Transformations During Continuous Cooling
A number of different alloys have been quenched and there is a clear trend in the behaviour
where the addition of significant amounts of elements such as Nb increases the range of cooling
rates over which massive gamma is formed because diffusion is slowed down and during even
relatively slow cooling rates the lamellar transformation cannot take place and massive gamma
(or in extreme cases retained alpha) is formed [10].
Tensile Properties of Samples Aged After Quenching to Produce a Fully Massive Structure
The properties of cast samples can be improved using the data obtained from end-quenching tests
to select the suitable alloys. Thus if a fully massive sample is aged in the two phase region
during HIPping the properties exceed those of HIPped castings and are comparable with
thermomechanically processed samples (see table 11). These observations have driven the
research summarised here since when casting technology of TiAl-based alloys is improved it will
be useful to be able to choose alloys which respond to this sort of heat treatment for use as cast
structural components. On the basis of the end-quenching work saltbath quenching is being used
in order to avoid cracking whilst still cooling quickly enough through the massive start
temperature to produce fully massively transformed samples.
Table 11. Comparison of the tensile properties of variously processed TiAl-based alloys
Alloy
Process route
UTS
0.2% proof
Ductility
MPa
stress MPa
Ti48Al2Cr2Nb
Forged/HT
347
312
0.5
Forged/OQ/HIP*
622
425
1.3
Ti47Al2Nb1W1Mn0.2Si
Cast/HIP
469
402
0.5
Cast/OQ/HIP
527
428
0.9
Ti47Al2Nb1W1Mn0.2Si1B
Forged/HT
477
377
0.8
Ti46Al5Nb1W
Cast/HIP
494
473
0.25
Extruded./SBQ/HIP*
581
528
0.7
Cast/HIP
537
0.1
Ti46Al8Nb
Cast/SBQ/HIP
567
525
0.5
* These thermomechanically processed samples which have been subsequently quenched and aged
represent the upper bound for the properties of cast samples which are HIPped after quenching. They
have been used to illustrate the potential of quenching and ageing because currently the properties of
TiAl-based castings are commonly dominated by casting defects rather than by microstructure.
General Discussion
The observations reported here have shown that the influence on solid state transformations in
TiAl-based alloys of grain size, boron additions and oxygen content must be considered in
addition to the influence of major alloying additions. The role of grain size is perhaps unexpected
inasmuch as the massive transformation can be nucleated at grain boundaries and finer grained
alpha would on that basis be expected to transform to massive gamma more rapidly than coarser
grained samples. As is clear from the Jominy data the massive transformation requires the same
(fast) cooling rate to initiate it but the formation of lamellar begins at much higher cooling rates
in the finer grained samples. Hence the window for formation of massive gamma is reduced in
the finer grained alloy. Since it is known that the massive gamma requires lamellae to nucleate at
grain boundaries [13] it has been argued that when the grain size is small enough the diffusional
growth of these lamellae can go to completion, since long range diffusion is not involved, before
31
any massive gamma can form [14]. It is clear from the results presented here that the influence of
boron on the massive transformations is felt through its influence on grain size.
The role of oxygen is complex but the results clearly show that its influence becomes more
significant at grain boundaries if slower quenching is used. This has been interpreted in terms of
the increased segregation of oxygen to grain boundaries during slow quenching [15], which
results in the retention of alpha which can form fine lamellae during subsequent slow cooling.
The presence of these fine lamellae removes grain boundaries as nucleation sites for massive
gamma and sites within grain centres are then able to come into operation as is evident from
figure 6(d). These observations can be used as the basis for a CCT curve for TiAl-based alloys as
shown in figure 7
Ls
Lf
Temperature
Decrease of
grain size
Increase of oxygen or
of slow diffusers.
Massive
AC
SBQ
IBQ
Low temperature Fine Ls
Fine Lf
WQ
Time
Figure 7. Schematic CCT curves for TiAl-based alloys. AC is air cool. SBQ is saltbath quench,
WQ is water quench and IBQ is iced brine quench. The arrows indicate the sense of movement
of boundaries. The shape of the massive region indicates that increased amounts of massive
gamma are formed at lower cooling rates within this temperature range. Low temperature fine
lamellae are formed if retained alpha is cooled slowly through the low temperature range as
indicated.
Conclusions
1
Jominy end-quenching is a very useful method of obtaining data concerning the influence
of composition on the kinetics of transformations which occur on cooling TiAl-based alloys.
2
If the production of a fully massively transformed sample is required the oxygen-content
should be as low as possible (say about 500wtppm), the grain size should be reasonably large
(say about 1000Pm) and slow diffusers such as Nb should be added.
3
Generic CCT curves have been constructed which attempt to capture the data obtained
from the Jominy tests.
32
Acknowledgements
The work reported in this paper has been supported by EPSRC, Nb products Ltd, The Royal
Society through its joint grant scheme with China, and by the EC through support for IMPRESS.
Other colleagues’ contributions, especially those by D Novovic, are gratefully acknowledged
References
1
A Denquin and S Naka. Acta Mat. 1996 44, p353-365
2
G Ramanath and V K Vasudevan. Mat Res Soc Symposium Proc 1993 228, p223-228
3
S. Jones and M J Kaufman. Acta Mat 1999, 41, p378-398
4
D Veeraraghavan and V K Vasudevan Acta Materialia 1999 47, p 3313-3330
5
Y Yamabe, M Takeyama and M Kikuchi In Gamma Ti aluminides (eds YW Kim, R
Wagner and M Yamaguchi) TMS San Diego CA 1995 p111
6
W Lefebrve, A Loiseau and A Menand Met Trans A 2003, 34A, p2067-2075
7
ISO Standard, ISO 642:1999, Steel-hardenability test by end quenching (Jominy test)
8
J Moulin, R Molinier, R Syre, in RI Jaffee, NE Promisel editors, The Science
Technology and Applications of Titanium, Pergamon Press, Oxford 1970, pp779-782
9
T Ahmed, HJ Rack. Materials Science and Engineering A, 1998, A243, 206-211
10
Hu D, In: Lutjering G, Albrecht J, editors. Ti-2003 Science and Technology. Weinheim:
Wiley-VCH; 2004, pp2369-2376
11
X. Wu, D. Hu, Scripta Materialia. 2005, 52 (8), pp731-734
12
D Hu, A J Huang, A Gregoire, X Y Li, Xinhua Wu and M H Loretto. In: Nie JF, Barnett
M (Eds), Proceedings of ICAMP3, The Institute of Materials Engineering Australasia.
2004, pp 172-175.
13
X D Zhang, S Godfrey, M Weaver, M Strangwood, P Threadgill, M J Kaufman and M H
Loretto. Acta Materialia 1996, 44, p3723-3734
14
Y Q Sun Met Trans 1998, 29A, p2679-85.
15
A Huang, M H Loretto, D Hu and X Wu (submitted for publication)
33