Scripta Materialia 52 (2005) 13411346

www.actamat-journals.com

Grain size eect on multiple-stage transformations of a

cold-rolled and annealed equiatomic TiNi alloy
S.H. Chang a, S.K. Wu a,b,*
, G.H. Chang b

a
Department of Materials Science and Engineering, National Taiwan University, 1, Roosevelt Rd., Sec. 4, Taipei 106, Taiwan, ROC
b
Department of Mechanical Engineering, National Taiwan University, Taipei 106, Taiwan, ROC

Received 8 November 2004; received in revised form 30 January 2005; accepted 7 February 2005
Available online 2 March 2005

Abstract

Ti50Ni50 shape memory alloy cold-rolled and annealed at specic conditions can exhibit a four-stage martensitic transformation.
There are two small and two large peaks which are associated with the B2 ! R ! B19 0 transformations of large grains in the spec-
imen center and that of small grains near the rolling surface, respectively.

2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Shape memory alloy; Martensitic transformation; Cold working; Annealing; Dierential scanning calorimetry

1. Introduction the stress elds around coherently precipitated Ti3Ni4

particles. Khalil-Alla et al. [5] found two distinct steps
TiNi alloys are known as the most important shape of B2 ! R and R ! B19 0 , and an additional step with
memory alloys (SMAs), with good shape memory eect B2 ! B19 0 in the dierential scanning calorimetry
(SME), superelasticity (PE) and damping capacity (DC) (DSC) cooling curves of aged Ni-rich TiNi alloys. They
[1]. It is well known that the martensitic transformation proposed that the MST comes from the inhomogeneous
of an equiatomic TiNi alloy comprises a single-stage distribution of Ti3Ni4 precipitates.
martensitic transformation from the parent B2 phase Su and Wu [6] have obtained a complex four-stage
to the B19 0 martensite phase. After conducting a ther- MST in a 20% cold-rolled and annealed Ti49Ni51 alloy,
momechanical treatment or adding a third element, and they suggested that the MST is the result of the
the transformation sequence can change into a two- combined eect of severe cold-working and long-time
stage transformation, from B2 to premartensite R-phase annealing. Kim et al. [7] reported a low temperature
and then to B19 0 upon cooling [2]. Recently, the mar- aging-induced two-stage R-phase transformation in
tensitic transformation of Ni-rich TiNi SMAs was found Ti49.1Ni50.9 alloy, and they identied that the rst R-
to have more than two distinct steps, and is known as phase transformation was associated with precipitates
the multi-stage martensitic transformation (MST). while the second R-phase transformation could be
Bataillard et al. [3,4] attributed the cause of MST to attributed to matrix regions far from precipitates.
According to the aforementioned studies, all the com-
*
plex MSTs appear in Ni-rich TiNi alloys and have to
Corresponding author. Address: Department of Materials Science do with heterogeneities caused by Ti3Ni4 precipitates.
and Engineering, National Taiwan University, 1, Roosevelt Rd., Sec.
4, Taipei 106, Taiwan, ROC. Tel.: +886 2 2363 7846; fax: +886 2 2363
In this study, we nd a four-stage MST in a cold-rolled
4562. and annealed Ti50Ni50 alloy. In the Ti50Ni50 alloy, nei-
E-mail address: [email protected] (S.K. Wu). ther Ti3Ni4 precipitation nor non-uniform precipitation

1359-6462/$ - see front matter
2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.scriptamat.2005.02.006
1342 S.H. Chang et al. / Scripta Materialia 52 (2005) 13411346

distribution is observed in the specimen. The complex method [8]. Ten parallel straight lines along or normal
MST behavior discussed in this study was identied by to the rolling direction were used to count the number
DSC and dynamic mechanical analyzer (DMA) tests; of grains intercepted and then the average was taken
microstructure evidence is provided to elucidate this to quantify the average grain size.
uncommon four-stage MST.

3. Experimental results
2. Experimental procedures
3.1. DSC and DMA measurements
An equiatomic Ti50Ni50 alloy was prepared by conven-
tional vacuum arc remelting. The as-melted ingots were Fig. 1 shows the cold-rolled Ti50Ni50 alloy annealed at
hot-rolled at 850 C into 2 mm thick plates, and then 500 C for dierent time intervals. When the annealing
the plates were solution-treated at 850 C for 2 h followed time is short, as shown in Fig. 1(a) and (b), two peaks
by quenching in water. The oxide layer of the solution- characterize the forward and reverse transformations.
treated plates was removed and the specimen thickness Upon increasing the annealing time to 1 and 3 h, the
was reduced to 1.9 mm. Thereafter, the plates were cold- transformation occurs in four stages on cooling, as
rolled along the hot-rolling direction at a rolling speed shown in Fig. 1(c) and (d). After annealing for 24 h, these
of 10 m/min. No annealing process was conducted during four peaks combine and become two large peaks; and the
the cold-rolling to avoid the occurrence of recrystalliza- two peaks in the reverse transformation after 1 and 3 h
tion. The thickness of the plate decreased from 1.9 mm merge to one single large peak. The transformation
to 1.24 mm and reached a nal 35% thickness reduction. behavior of Fig. 1 agrees with what has been reported
Then the cold-rolled plates were cut into test specimens, on the transformation behavior of cold-rolled and an-
sealed in an evacuated quartz tube and annealed at nealed near equiatomic TiNi SMAs [911] except that
500 C and 650 C in salt baths for dierent time intervals. the four-stage MST shown in Fig. 1(c) and (d) has never
Transformation temperatures of cold-rolled and an- been reported before. As the annealing temperature is in-
nealed specimens were determined by DSC tests using creased to 650 C (not shown), the tendency of the trans-
the TA Q10 DSC equipment at a constant cooling rate formation sequence is similar to that shown in Fig. 1. A
(10 C/min). Specimens for DMA experiments were cut four-stage MST also appears when the annealing is per-
in the direction along the rolling direction (RD). Tan d formed at 650 C for 90 s and for longer annealing time
was measured by TA 2980 DMA equipment using heat- these four peaks gradually merge to two peaks.
ing and cooling rates of 3 C/min. The dimension of the Fig. 2 shows DSC and DMA results of the cold-rolled
specimens used in the DMA tests was 40 5 1.24 mm3, specimen annealed at 500 C for 3 h. Fig. 2(a) shows the
and the amplitude and frequency used were 2 lm and DSC curve of Fig. 1(d) and Fig. 2(b) indicates the corre-
2 Hz, respectively. The cross-sectional microstructures sponding DMA results for tan d. As shown in Fig. 2(b),
of the specimens were prepared by the normal metallo- four transformation peaks which match the transforma-
graphic procedures and then observed by a Nikon tion peaks in DSC curve are observed; there is a small
Microphoto-FXA optical microscope (OM). From the shift between the results of the two methods which is
OM images, the average grain size of cold-rolled and an- attributed to dierent cooling/heating rates and speci-
nealed specimens was estimated by the linear intercept men sizes used in DSC and DMA testing. In addition

Fig. 1. DSC curves for cold-rolled Ti50Ni50 alloy annealed at 500 C for dierent time intervals.
 S.H. Chang et al. / Scripta Materialia 52 (2005) 13411346 1343

40 lm. The grains near the rolling surfaces are ber-

structured and their grain boundaries are not well
dened. This is because the two surfaces are in direct
contact with the rollers and suer more plastic deforma-
tion than the central region.
Fig. 3(b) shows the cross-sectional microstructure of
the cold-rolled specimen annealed at 500 C for 3 h.
Fig. 3(c) and (d) show the enlarged views of the regions
at the center and near the rolling surface of Fig. 3(b),
respectively. The average grain sizes calculated from
Fig. 3(c) and (d) are about 70 lm and 7 lm in the RD
and, 15 lm and 7 lm in its normal direction, respec-
tively. Fig. 3(e) shows the microstructure of the cold-
rolled specimen annealed at 500 C for 24 h. From
Fig. 2. (a) DSC and (b) DMA curves for cold-rolled Ti50Ni50 alloy
annealed at 500 C for 3 h. Fig. 3(e), the average grain sizes in the specimen center
and near the rolling surface are calculated to be about
72 lm and 53 lm in RD and 17 lm and 12 lm in its nor-
to the four transformation peaks which appear in the
mal direction, respectively. Here, the grain sizes in the
DSC tests, there is also one extra peak obtained at about
center of the sheet and that near the rolling surface are
75 C in Fig. 2(b). This peak was not observed in the
approaching each other, i.e. the grain size distribution
DSC curve and is known as a relaxation peak [2].
in the whole specimen becomes more uniform.
3.2. Microstructure observations
4. Discussion
Fig. 3 shows OM images of the cross-sectional micro-
structure of the cold-rolled Ti50Ni50 alloy with and with- 4.1. Identication of four-stage MST in a cold-rolled and
out annealing at 500 C for 3 h. Fig. 3(a) illustrates the annealed Ti50Ni50 alloy
as cold-rolled Ti50Ni50 alloy and it can readily be no-
ticed that the microstructure in the specimen center is 4.1.1. DSC results
quite dierent from that near the specimens rolling sur- Fig. 4 shows several partial DSC transformation cy-
faces. The grains in the specimen center are elongated cles of the cold-rolled specimen annealed at 500 C for
along the RD and the average grain size is about 3 h. There are four transformation peaks on cooling,

Fig. 3. OM micrographs of the cross-sections of the microstructure of cold-rolled Ti50Ni50 alloy with and without annealing: (a) as cold-rolled
specimen; (b) annealed at 500 C for 3 h; (c) enlarged central region of (b); (d) enlarged central region near the rolling surface of (b); and (e) annealed
at 500 C for 24 h.
1344 S.H. Chang et al. / Scripta Materialia 52 (2005) 13411346

All the transformations associated with the peaks

shown in Fig. 4 are highlighted. From the gure, the
transformation behavior of the cold-rolled and
500 C 3 h annealed Ti50Ni50 alloy is delineated as fol-
lows. When the specimen is cooled to about 31 C, a rel-
atively small volume transforms from B2 to R1 rst.
After being cooled to 20 C, the untransformed B2 re-
gion of the specimen starts to transform from B2 to
R2. Both R1 and R2 are the premartensite R-phase,
but they are transformed from dierent regions of the
specimen and R2 is the dominant one. Upon the contin-
uation of cooling down to 3 C, R1 transforms to B1901 .
After the temperature is cooled to 16 C, the dominant
R2 transforms to B1902 . Both B1901 and B1902 are the
Fig. 4. Overlapped partial DSC cycles for cold-rolled Ti50Ni50 alloy monoclinic martensite, but they are transformed from
annealed at 500 C for 3 h. R1 and R2, respectively. During the reverse transforma-
tion on heating, B1902 and B1901 transform back to B2 at
37 C and 50 C, respectively. The above discussion
and we interrupt the cooling process at each transforma- indicates that two pairs of B2 ! R ! B19 0 occur on
tion peak and immediately heat the specimen up to ob- cooling and two pairs of B19 0 ! B2 appear on heating.
tain the corresponding DSC heating curve. As shown in
Fig. 4, the solid line represents the full DSC cycle, and 4.1.2. DMA results
the dotted line represents the DSC cycle which was inter- From Fig. 2(b), four tan d peaks at 18.1 C, 4.5 C,
rupted at the rst transformation peak (about 31.3 C), 7.2 C and 23.4 C correspond with four DSC trans-
and a new peak appears at 37.1 C on heating which is formation peaks in Fig. 2(a). It was reported that the
not found in the full cycle. These two peaks are typical internal friction, Q1/tan d, of R M B19 0 transformation
B2 M R1 due to their small hysteresis transformation is much larger than that of the B2 M R transformation
temperature (about 6 C) and small transformation at the same strain amplitude and frequency [13]. From
enthalpies (about 1.4 J/g). Since the transformation Section 4.1.1, peaks at 18.1 C and 4.5 C in Fig. 2(b)
enthalpies are lower than the normal values, 5 J/g [12], are associated with B2R and peaks at 7.2 C and
it is concluded that the transformation only takes place 23.4 C are attributed to R ! B19 0 . Peak at
in a relatively small volume of the specimen. When the 23.4 C is much larger than that at 4.5 C and peak
specimen is cooled to the second transformation peak, at 7.2 C is also much larger than that at 18.1 C.
a peak appears at 28.2 C on heating. Subsequently a Therefore, the DMA result of Fig. 2(b) also provides
R1 ! B2 transformation peak occurs at 37.5 C. Obvi- more evidence to support the four-stage MST which is
ously, the new peak developed at 28.2 C upon heating comprised of two pairs of B2R ! B19 0 on cooling, as
corresponds to the second transformation peak on cool- identied in Section 4.1.1 by DSC result.
ing, and therefore, together with the second peak on
cooling it describes B2 M R2 transformation characteri- 4.2. Reasons causing the four-stage MST in a cold-rolled
zed by a small hysteresis width and small transformation and annealed Ti50Ni50 alloy
enthalpies.
When the specimen is continually cooled to the third The reasons for the MST behavior in Ni-rich TiNi al-
transformation peak at about 3.8 C, there are also two loys have been widely discussed, and a common conclu-
peaks observed on subsequently heating. The R2 ! B2 sion of this complex MST is related to the
peak is still obtained at 28.3 C upon heating, but the inhomogeneous formation of aging-induced Ti3Ni4
R1 ! B2 transformation peak at 37 C disappears, and precipitates. However, in this study, no Ti3Ni4 particles
another new peak appears at 49.8 C. Therefore, the precipitate in the equiatomic Ti50Ni50 specimen. Conse-
third peak on cooling can be characterized as a R1 ! quently, the inuence of Ti3Ni4 precipitates and their
B1901 transformation peak because the R1 ! B2 peak heterogeneous precipitation on the MST behavior
disappears at 37 C and a new B1901 ! B2 peak appears should be excluded.
at 49.8 C on heating. When the specimen is cooled to the From the DSC and DMA results in this study, the
fourth peak, the transformation peak of R2 ! B2 at four-stage MST can only be observed in some specic
28 C gradually decreases and a new peak is created at annealing conditions, for example annealing at
about 37 C. Consequently, the fourth peak on cooling 500 C 1 h, 500 C 3 h and 650 C 90 s. As men-
and the new 37 C peak on heating represent the trans- tioned in Section 4.1, the four-stage MST on cooling
formations R2 ! B1902 and B1902 ! B2, respectively. comprises two pairs of B2 ! R ! B19 0 transforma-
 S.H. Chang et al. / Scripta Materialia 52 (2005) 13411346 1345

tions. OM observations in Fig. 3 show that the grain size From the viewpoint of the cold-rolling eect, the degree
distribution along the specimens thickness direction is of plastic deformation at the specimen center should be
inhomogeneous, in which the grain size at the central re- much less than the 35% thickness reduction of the whole
gion is apparently larger than that near the rolling sur- specimen. As a result, the recrystallized grains in the cen-
face due to the cold-rolling eect. This implies that the tral region should be larger than those near the rolling
four-stage MST, i.e. the two pairs of B2 ! R ! B19 0 surface under the same annealing condition. At the same
transformations, comes from the B2 ! R ! B19 0 trans- time, defects and dislocations induced by the severe cold-
formations of grains in the central region and those of rolling of grains near the specimens surface should be
grains near the rolling surface. much more than those in the center. It is well known that
In order to clarify the eect of grain size on the MST, these cold-rolling induced defects/dislocations can sup-
the specimen shown in Fig. 3(b) is ground from the bot- press premartensitic and martensitic transformations to
tom rolling surface to remove the smaller grains in one lower temperatures [9]. Therefore, B2 ! R and R !
rolling side. At the same time, the larger grains in the B19 0 transformations corresponding to small grains on
central region are also ground. The thickness of the cooling take place at lower temperatures than those cor-
specimen is now reduced from 1.2 mm to 0.4 mm. responding to large grains, as shown in Fig. 5.
DSC test has been conducted on this ground specimen,
and the result is shown in Fig. 5. Fig. 5(a) is the same
specimen as Fig. 1(d), and Fig. 5(b) is the same specimen 5. Conclusions
as Fig. 5(a), but now has been ground to 0.4 mm thick-
ness. From Fig. 5(b), one can nd that two small trans- The transformation behavior of Ti50Ni50 shape mem-
formation peaks disappear, providing the evidence that ory alloy after 35% cold-rolling and subsequent recrys-
the four-stage MST is associated with the small and tallization annealing is investigated in this study. The
large grains induced by the severely cold-rolled and an- four-stage MST appears at specic annealing conditions
nealed Ti50Ni50 alloy. In other words, in Fig. 5(a), two having two large and two small peaks in the cooling
small peaks on cooling are associated with B2 ! curves of DSC and DMA tan d tests. Microstructural
R ! B19 0 caused by large grains in the central region observations reveal that the grain size distribution along
of the specimen, but two large peaks on cooling, the specimens thickness direction is inhomogeneous due
although they are also associated with B2 ! R ! B19 0 , to the cold rolling eect, in which the grain size at the
are caused from small grains near the rolling surface of central region is apparently larger than that near the
the specimen. rolling surface. Two small peaks in four-stage MST
The reason why the transformation peaks corre- are related to the B2 ! R ! B1901 transformations of
sponding to larger grains are smaller than those corre- large grains and two large peaks are also associated with
sponding to smaller grains comes from the fact that the the B2 ! R ! B19 0 transformations, but now they cor-
volume of large grains in the central region is less than respond to small grains near the rolling surface. The vol-
that of small grains near the rolling surface. This is be- ume of large grains in the specimen center is much less
cause the thickness of large grains in the specimen center than that of small grains near the rolling surface. This
is only about 50100 lm, which is comparatively less feature accounts for the appearance of two small and
than the whole specimen thickness of about 1200 lm. two large peaks on cooling.

Acknowledgement

The authors gratefully acknowledge the nancial sup-

port for this research provided by the National Science
Council (NSC), Taiwan, Republic of China, under
Grants NSC92-2216-E002-008.

References

[1] Wayman CM, During TW. In: During TW, Melton KN, Stockel
D, Wayman CM, editors. Engineering aspects of shape memory
alloys. London: Butterworth-Heinemann; 1990. p. 320.
[2] Wu SK, Lin HC, Chou TS. Acta Metall Mater 1990;38:95.
Fig. 5. DSC curves for cold-rolled Ti50Ni50 alloy annealed at 500 C [3] Bataillard L, Gotthardt R. J Phys III 1995;5:C8C647.
for 3 h: (a) original specimen with 1.2 mm thickness (the same as Fig. [4] Bataillard L, Bidaux J-E, Gotthardt R. Philos Mag 1998;78:327.
1(d)); (b) ground specimen to 0.4 mm thickness. [5] Khalil-Alla J, Dlouhy A, Eggeler G. Acta Mater 2002;50:4255.
1346 S.H. Chang et al. / Scripta Materialia 52 (2005) 13411346

[6] Su PC, Wu SK. Acta Mater 2004;52:1117. [10] Thoma PE, Kao MY, Fariabi S, AbuJudom DN. In: Proceedings
[7] Kim JI, Liu Y, Miyazaki S. Acta Mater 2004;52:487. of ICOMAT 92. p. 917.
[8] ASTM StandardsStandards test method for determining aver- [11] Treppmann D, Hornbogen E, Wurzel D. In: Proceedings of
age grain size. Annual book of ASTM standards, vol. 03.01. 2003. ICOMAT 95. p. C8-569.
p. 256. [12] Lin HC, Wu SK. Scripta Metall Mater 1991;25:1295.
[9] Khelfaoui F, Guenin G. Mater Sci Eng A 2003;355:292. [13] Lin HC, Wu SK, Yeh MT. Metall Trans A 1993;24:2189.