Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

Acta Materialia 58 (2010) 6411–6420

Role of severe plastic deformation on the cyclic reversibilityof a Ti50.3Ni33.7Pd16 high temperature shape memory alloy

B. Kockar a,c, K.C. Atli a, J. Ma a, M. Haouaoui a, I. Karaman a,⇑,M. Nagasako b, R. Kainuma b

a Department of Mechanical Engineering, Texas A&M University, MS 3123, College Station, TX 77843, USAb Department of Materials Science, Tohoku University, Sendai 980-8579, Japan

c Department of Mechanical Engineering, Hacettepe University, Beytepe, 06800 Ankara, Turkey

Received 7 June 2010; received in revised form 3 August 2010; accepted 3 August 2010

Abstract

The present work focuses on the effect of microstructural refinement on the thermo-mechanical cyclic stability of a Ti50.3Ni33.7Pd16

high temperature shape memory alloy (HTSMA) which was severe plastically deformed using equal channel angular extrusion (ECAE).The grain/subgrain size of the high temperature austenite phase was refined down to about 100 nm, the lowest reported to date in HTS-MAs. The increase in strength differential between the onset of transformation and the macroscopic plastic yielding after ECAE led to anotable enhancement in the cyclic stability during isobaric cooling–heating experiments. The reduction in irrecoverable strain levels wasattributed to the increase in critical stress for dislocation slip due to the microstructural refinement during the ECAE process.� 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: TiNiPd; High temperature shape memory alloys; Severe plastic deformation; Equal channel angular extrusion; Equal channel angular pressing

1. Introduction

NiTi shape memory alloys (SMAs) have revolutionizedthe development and use of active materials in the last40 years by providing large reversible shape changes, highactuation forces, and large elastic strains as a result of ther-moelastic martensitic transformations. The development ofSMA actuators has recently been a priority in space, auto-motive and power generation applications, where highoperating temperatures are needed. However, the low oper-ation temperatures (<100 �C) and relatively large transfor-mation hysteresis of binary NiTi limit their utility in hightemperature applications [1]. Luckily, the transformationtemperatures of NiTi can be controlled by the addition ofalloying elements. For instance, Pd addition at more than

1359-6454/$36.00 � 2010 Acta Materialia Inc. Published by Elsevier Ltd. All

doi:10.1016/j.actamat.2010.08.003

⇑ Corresponding author.E-mail address: ikaraman@tamu.edu (I. Karaman).

10 at.% increases the transformation temperatures to above100 �C [2,3].

NiTi-based ternary alloys with the addition of one ofPd, Pt, Zr, Au, or Hf have been the most studied high tem-perature shape memory alloys (HTSMAs) to date [2–5].The addition of Zr and Hf could be more favorable as athird alloying element due to their low cost, however, NiTialloys with Zr and Hf exhibit unstable shape memory prop-erties and their thermal and stress hystereses are large [4,5].The addition of Pt to a NiTi binary alloy would massivelyincrease the cost. Therefore, Pd could be a better choice asa third alloying element considering the increase in trans-formation temperatures and the decrease in thermal hyster-esis [2,6].

Dislocation plasticity that may accompany martensitictransformation is an important anticipated problem inNiTiPd HTSMAs, affecting the stability of the shapememory effect and superelasticity. A low critical stress fordislocation slip at high temperatures can easily cause the

rights reserved.

6412 B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420

accommodation of transformation shear and volumechange by dislocation slip, and cause dissipation of storedelastic energy during transformation. Eventually, this leadsto an apparent permanent deformation, dimensional insta-bility, and a lack of superelasticity [7,8]. Well-developeddislocation substructures, desired crystallographic texture,refined grain sizes, and coherent precipitates, all of whichimprove the critical stress for slip (CSS), have beenreported as necessary conditions for cyclic and dimensionalstability and for superelasticity in HTSMAs [9,10]. Fourstrategies have been proposed to achieve an increase inCSS: (1) addition of a quaternary element, (2) age harden-ing of the parent phase, (3) thermo-mechanical treatmentand (4) severe plastic deformation (SPD) [10].

A limited number of studies have reported on themechanical and shape memory properties of NiTiPd alloys.Cai et al. [7] showed that the shape memory behavior ofTiNiPd alloys was fairly good when deformed at roomtemperature, however, became relatively poor with increas-ing deformation temperature in martensite due to adecrease in CSS. Lindquist and Wayman reported 6%unconstrained shape recovery [6]. Khachin et al. [11]showed 4% complete strain recovery, which was introducedby applying 200 MPa torsional stress in Ni13Ti50Pd37.Otsuka and co-workers studied the shape memory effectin Ti50Pd50 [8]. They reported poor shape memory behav-ior, which could again be attributed to a low CSS possiblyleading to a high density of slip upon deformation, in addi-tion to the twinning in martensite [8]. NiTi alloys with 40–50 at.% Pd show only 0.5% shape recovery when loaded intension [1].

Several attempts at enhancing the shape memoryresponse of NiTiPd alloys have been reported. Yang andMikkola [12] examined the effect of boron addition onthe shape memory characteristics of a Ni22.3Ti50.7Pd27 alloyand found 90% shape recovery for 2–3% applied strainunder compression. Boron had no real effect on the shapememory characteristics. However, they determined anincrease in ductility at room temperature and attributedthis to grain refinement due to boron addition [12].Another method to improve the shape memory character-istics of the NiTiPd alloys is to use slightly off-stoichiome-tric compositions and generate homogeneously distributedTi2Ni precipitates in a Ni19.4Ti50.6Pd30 alloy [13]. 90%shape recovery was observed when a Ni19.4Ti50.6Pd30 alloywas deformed to a total strain of 6% at 473 K. This wasabout 10% higher than that of Ti50Ni30Pd20, in which pre-cipitation was not observed. Such an increase was attrib-uted to the hardening effect of the homogeneouslydistributed Ti2Ni precipitates [13]. Moreover, certainthermo-mechanical treatments have been applied to modifythe microstructure and strengthen the material via grainrefinement and dislocation hardening. Golberg and co-workers reported some improvement in the shape memoryproperties of Ti50Ni30Pd20 after cold rolling and subse-quent heat treatment [14,15]. They claimed superelasticityfor the first time in Ti50Ni30Pd20 after annealing the alloy

at 673 K and testing at 535 K [14,15]. Cai et al. [7] studiedthe effect of thermal cycling on the shape memory proper-ties of Ni19.4Ti50.6Pd30 and found that Ms temperature,total transformation strain, and irrecoverable strainincreased with the number of cycles. The change in theseparameters occurred quickly in the first 40 cycles and thentended to stabilize [7].

As an alternative to cold working, precipitation harden-ing and thermo-mechanical training, microstructural refine-ment using severe plastic deformation (SPD) should alsobring about good cyclic reversibility and shape recovery inNiTiPd HTSMAs. SPD via high pressure torsion (HPT)leads to the formation of nanosized grains and amorphiza-tion in bulk binary NiTi samples, however, HPT samplesare too small to investigate the shape memory propertiesof the alloys [16–18]. On the other hand, SPD via equal chan-nel angular extrusion (ECAE) has several advantages overHPT, such as producing larger samples, introducing uni-form deformation, and allowing some control over grainmorphology and crystallographic texture. Our previousstudies on ECAE of Ni-rich NiTi, equiatomic NiTi and TiN-iHf HTSMAs have shown that SPD via ECAE enhances thecyclic stability and shape recovery behavior under relativelyhigh stress levels by reducing the grain size and inducing finedislocation substructures [10,19–22].

Most of the aforementioned studies on TiNiPd alloysdemonstrated shape recovery of these alloys under stress-free conditions, and presented the isothermal mechanicalproperties and microstructure. Most applications, how-ever, require shape recovery under applied stress and/orthermal cyclic stability under constant loads. Only oneresearch group has recently reported the shape memoryresponse of Ni19.5Ti50.5Pd30 HTSMAs and correspondingwork outputs during few thermal cycles under variousloads, who observed irrecoverable strains at relatively highstress levels [23,24].

Thus, this work presents the first report on the SPD of aTi50.3Ni33.7Pd16 alloy produced by ECAE. We introducethe influence of ECAE on the evolution of microstructure,and cyclic reversibility and shape recovery during thermalcycling under stress. The material was ECAE processedat 425 �C following route E for four passes. The resultsshow that the cyclic stability was dramatically enhancedafter the ECAE process.

2. Experimental procedure

The Ti50.3Ni33.7Pd16 alloy was prepared by vacuuminduction melting. A cylindrical billet was then hot rolledat 925 �C down to 1.2 cm diameter. 10 or 12.5 cm long barswere canned in 304 stainless steel (SS) before ECAE. BothTi (commercial purity grade 2) and 316 SS were explored ascanning materials prior to using 304 SS. Ti was too soft atthe processing temperature and caused the enclosed SMAbillets to undergo severe shear localization during multi-pass ECAE. The high strain hardening of 316 SS at theextrusion temperature substantially increased the press

B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420 6413

load after the second or third ECAE pass. 304 SS experi-enced lower strain hardening and thus proved to be theright choice, given the limited temperature window forextrusion imposed by these hard to work SMAs and theload limit of the ECAE die at high extrusion temperatures.

The canned billets were kept for 30 min in the ECAE diebefore the first pass and for 15 min prior to successivepasses to allow sufficient heating and to minimize therecovery/recrystallization as much as possible. Before thefirst pass and between the successive passes the die and bil-let were lubricated to reduce the friction between them andbetween the different moving parts of the die. The wholeset-up was enclosed in a furnace which was preheated to425 �C. The extrusions were conducted in a nitrogen atmo-sphere to minimize potential oxidation. ECAE was con-ducted following route E, having a billet rotationsequence of +180�, 90�, 180� at 425 �C. Route E wasselected because it results in the largest volume of uni-formly processed material (i.e. the largest product yield)and the greatest number of equiaxed grains [25]. 425 �Cwas the lowest temperature at which samples withoutmicro/macro shear localization could be produced. FourECAE passes were performed for each billet, since at leastfour or a higher number of passes is required to obtain asignificant volume fraction of high angle grain boundariesin bcc structures during ECAE [26,27]. The billet was waterquenched after each ECAE pass to minimize recovery dur-ing slow cooling. Fig. 1 shows one of the processed billetsthat had a 10 cm processed length.

As-received and ECAE processed specimens were ther-mally cycled five times between 20 �C and 200 �C using aPerkin–Elmer Pyris I differential scanning calorimeter.The heating–cooling rate during these experiments wasset to 10 �C min�1.

Transmission electron microscopy (TEM) studies wereconducted using a JEOL JEM-2000 EX-II microscopeoperated at an accelerating voltage of 200 kV. The TEMsamples were mechanically ground down to a thicknessof 100 lm and twin jet electro-polished with a 8 vol.% per-chloric acid, 72 vol.% acetic acid, 12 vol.% ethanol and8 vol.% ethylene glycol solution at room temperature.Bright field images and diffraction patterns of the as-received and ECAE processed samples were recorded atroom temperature to investigate the martensite twin struc-

Fig. 1. The digital image of the ECAE processed Ti50.3Ni33.7Pd16 high temperaprocessed for up to four passes at 425 �C using route E.

ture. In order to examine the austenite phase, the sampleswere heated to above the austenite finish (Af) temperatureusing an in situ TEM heating stage.

The isothermal monotonic loading experiments undercompression and thermo-mechanical characterizationexperiments under tension were conducted using an MTS810 servo-hydraulic test frame. Cooling and heating ofthe samples was achieved by conduction through the grips,which were heated using heating bands and cooled usingliquid nitrogen flowing through copper tubing wrappedaround the grips. The heating–cooling rate was10 �C min�1. A high-temperature extensometer with agauge length of 12.54 mm was used to record the axialstrains by attaching its ceramic extension rods directly tothe gage section of the tensile samples. Compression spec-imens with the dimensions of 4 � 4 � 8 mm and small dog-bone-shaped tension samples with a gauge section of1.5 � 3 � 8 mm were cut using wire electro-dischargemachining (EDM) from the uniformly deformed regionof the billets. For the compression experiments a miniatureextensometer with a 3 mm gage length was used to measurethe strain directly on the sample. The temperature wasmeasured using a thermocouple attached to the samples.

Three types of experiments were conducted on the as-received and ECAE processed samples:

1. Isothermal monotonic loading experiments under com-pression in austenite to determine the effect of ECAEon the strength levels. The test temperature was 20 �Cabove the martensite start temperature (Ms) of eachsample measured under a 50 MPa load (Mr¼50 MPa

s ),which was selected to ensure thermodynamically equiv-alent conditions for the two different microstructures.The strain rate was 5 � 10�4 s�1.

2. Isobaric cooling–heating experiments under increasingtensile stress levels with 50 MPa increments to determinethe shape memory characteristics, such as transforma-tion temperatures, recovered transformation and irre-coverable strain levels as a function of stress, and todetermine the threshold stress level for the onset of irre-coverable strain.

3. Isobaric cooling–heating experiments up to 10 cyclesunder 150 MPa load to evaluate the cyclic stability ofthe as-received and ECAE processed samples.

ture shape memory alloy billet. The billet was 10 cm long and was ECAE

6414 B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420

3. Experimental results

3.1. Microstructural evolution

The grain size of the as-received sample was impossibleto determine using optical microscopy at room temperaturesince the microstructure consisted of martensite variantsand twins, as shown in Fig. 2. The martensitic structuredid not allow the detection of the prior austenite grainboundaries. Additionally, revealing the austenite grainboundaries was impossible in the as-received sample usingin situ TEM after heating the sample above the Af due tothe large grain size. Figs. 2 and 3 show the bright fieldTEM images and corresponding selected area diffraction(SAD) patterns of the as-received and ECAE processedsamples at room temperature in the martensitic state. Bothsamples had a B19 orthorhombic structure, as expected for

500nm

(a)Fig. 2. Room temperature (a) bright field TEM image and (b) correspondingsample. The martensitic structure was internally twinned B19 orthorhombic w

(a)500nm

Fig. 3. Room temperature (a) bright field TEM image and (b) correspondingfour passes following route E. The twin thicknesses are much smaller than th

the martensite in these ternary SMAs [2,6]. The ECAE pro-cessed sample had a notably smaller martensite twin thick-nesses than the as-received sample. Such refinement afterECAE was also observed in binary NiTi [19]. Fig. 4 showsthe austenite structure in the ECAE processed samples at200 �C with an average grain size on the order of 100 nm.

3.2. Differential scanning calorimetry

Transformation temperatures under stress-free condi-tions for both cases were determined using differentialscanning calorimetry (DSC) measurements shown inFig. 5. The transformation temperatures of the first andfifth cycles and the thermal hysteresis of the as-receivedand ECAE processed samples are summarized in Table 1.The Ms temperature of the as-received sample was93.2 �C and that of the ECAE processed sample was about20 �C lower. Thermal hysteresis was around 20 �C in both

Incident beam//110B19

(b)selected area diffraction (SAD) pattern of the as-received Ti50.3Ni33.7Pd16

ith {1 1 1} type 1 twins.

(b)

Incident beam//101 B19

SAD pattern of the Ti50.3Ni33.7Pd16 sample ECAE processed at 425 �C forose in the as-received sample.

(a) (b) 200nm Incident beam//011B2

200

011

Fig. 4. Bright field TEM image of the B2 austenite structure at 200 �C taken from the Ti50.3Ni33.7Pd16 sample ECAE processed at 425 �C to four passesusing route E showing the refined grain size of the order of 100 nm.

Fig. 5. DSC responses of the (a) as-received and (b) ECAE processedTi50.3Ni33.7Pd16 alloy and the transformation temperatures in the firstcycle are shown in the graphs.

B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420 6415

materials and was much lower than that for binary NiTi[19]. There was almost no change in the transformationtemperatures of the as-received sample with number ofcycles. Similar observations were made in the ECAE pro-cessed case, with the exception of the first heating cycle.The first heating cycle showed martensite stabilizationdue to the heavily deformed structure, which caused addi-tional resistance to reverse transformation upon first heat-ing [6]. The austenite start (As) temperature of the ECAE

processed sample in the fifth cycle was approximately4 �C lower than that in the first cycle.

3.3. Isothermal monotonic uniaxial response under

compression

Fig. 6 displays the monotonic compression response ofthe as-received and ECAE processed Ti50.3Ni33.7Pd16 sam-ples. The experiments were conducted at 20 �C above theMs temperature of each sample, which was measured under50 MPa load (Mr¼50 Mpa

s ). The samples were first heatedabove Af to ensure that they were fully austenitic, and thencooled down to the test temperature. This temperature wasselected to clearly distinguish the regions of stress-inducedmartensitic transformation and macroscopic plastic defor-mation of martensite in the stress–strain response, in addi-tion to the samples with different microstructures to beunder thermodynamically similar conditions. The compres-sion test results, rather than tension, are presented herebecause both materials failed under tension before theyreached the second plateau in the stress–strain response,which indicates macroscopic plastic deformation of mar-tensite. The early failure in tension was likely to be dueto the relatively high fraction of carbides in the present ter-nary alloy.

Fig. 6 clearly shows that the macroscopic yield strengthof martensite, rM

y , increased after ECAE, even though thecritical stress required to induce martensite, rSIM, did notnotably change. This dissimilar trend gave rise to a clearincrease in the strength differential, Dr, between thesetwo stress levels upon ECAE. Such an increase should bea consequence of an increase in the CSS due to work hard-ening and grain refinement. Note that rM

y is not discerniblein Fig. 6 for the case of the ECAE processed sample, eventhough the sample was still intact at the high stress levelsduring the experiment. This was because of the load limitof the test frame, which did not allow the application ofhigher stresses for the present compression sample geome-try. It is likely that rM

y for the ECAE processed sample is

Table 1Martensitic transformation temperatures and thermal hysteresis of the as-received and ECAE processed Ti50.3Ni33.7Pd16 high temperature shape memoryalloy samples at the first and fifth thermal cycles, extracted from the DSC data. A�1st and M�1st are the transformation peak temperatures during the firstheating and cooling cycles, respectively.

Material As1st (�C) Ms

1st (�C) As5th (�C) Ms

5th (�C) As5th–As

1st (�C) Ms5th–Ms

1st (�C) A1st�–M1st

� (�C)

As-received 99 93.2 99 92.5 0 �0.7 19.3ECAE 4E at 425 �C 61.5 72.6 57.6 71.4 �3.9 �1.2 21.6

2000

1500

1000

500

0

Stre

ss (M

Pa)

Strain (%)

As-Received ECAE 4E at 425ºC

2%

Ti50.3Ni33.7Pd16

SIMMy σσσ −=Δ

Fig. 6. Monotonic compression response of the as-received and ECAEprocessed Ti50.3Ni33.7Pd16 alloys under compression at (Mr¼50 Mpa

s +20 �C). This temperature was selected so that the samples were subjectedto similar thermodynamic conditions.

(a)

(b)

Stra

in (%

)

20016012080400Temperature (°C)

50 MPa

150 MPa

250 MPa

100 MPa

200 MPa

300 MPa Ti50.3Ni33.7Pd16As-Received5%

Stra

in (%

)

20016012080400Temperature (°C)

Ti50.3Ni33.7Pd16ECAE 4E at 425°C

50 MPa

150 MPa

250 MPa

350 MPa

100 MPa

200 MPa

300 MPa5%

Fig. 7. Strain vs. temperature responses of the Ti50.3Ni33.7Pd16 alloy undervarious constant stress levels during isobaric cooling–heating experiments:(a) as-received and (b) ECAE 4E at 425 �C samples.

6416 B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420

significantly higher than 1500 MPa. Regardless of this, theresistance to plasticity is obviously higher in the ECAEprocessed sample, and this should point to an enhancedtransformation reversibility and cyclic stability.

3.4. Isobaric cooling–heating experiments

The dimensional and cyclic stability and recovered trans-formation strain levels of Ti50.3Ni33.7Pd16 are important fac-tors for practical applications when the alloy is acting againsta force. Therefore, isobaric cooling–heating experimentswere conducted, during which the test specimen was heldunder constant stress while it underwent continuous coolingand heating at a rate of 10 �C min�1. After each additionalcooling and heating cycle the stress was increased by50 MPa increments while in the austenitic state. Fig. 7 showsthe strain vs. temperature response of both the as-receivedand ECAE processed samples. Fig. 8 schematically illus-trates how the recovered transformation and irrecoverablestrain (eirr) values and transformation temperatures weredetermined from the isobaric cooling–heating experiments.The recovered transformation strain was measured at tem-perature (Af + Ms)/2 between the extrapolated baselinestrain levels of the austenite and martensite phases. Trans-formation hysteresis was determined at the half of the recov-ered transformation strain level, as shown in Fig. 8. Asummary of the recovered transformation and irrecoverablestrain levels, and thermal hysteresis is presented in Fig. 9aand b, respectively. The maximum recovered transformationstrain level reached 3.3% at a stress level of 250 MPa in theas-received sample, however, this was accompanied by an

irrecoverable strain of 0.57%. In fact, irrecoverable strainbegan to appear in the as-received sample even under100 MPa stress. The eirr levels of the ECAE processed sampleshown in Fig. 9a demonstrate dramatic improvements overthe as-received case. No eirr was observed at 100 MPa andnone appeared until 250 MPa, while the recovered transfor-mation strain was lower by 0.4% at 250 MPa compared withthat of the as-received sample.

Another intriguing observation from the isobaric cool-ing–heating experiments was the lower thermal hysteresisvalues in the ECAE processed samples above 200 MPaload. In addition, the hysteresis was quite stable and almostindependent of the applied stress in the processed samples.On the other hand, the thermal hysteresis of the as-receivedsample increased significantly with stress, especially above50 MPa. Thermal hysteresis is an indication of how muchenergy dissipation occurs during forward and reversetransformation. Apparently, ECAE significantly stabilizestransformation, such that energy dissipation and, thus,

Thermal Hysteresis

Recovered Transformation Strain

Ms

Mf

IrrecoverableStrain

Stra

in (%

)

Temperature (°C)

Fig. 8. Schematic illustration of the recovered transformation andirrecoverable strain, Ms and Mf temperatures, and thermal hysteresis ina typical isobaric cooling–heating experiment.

(a)

(b)

5

4

3

2

1

0

Stra

in (%

)

35030025020015010050Stress (MPa)

As-Received ECAE 4E at 425°C

Ti50.3Ni33.7Pd16Recovered Transformation Strain

Irrecoverable Strain As-Received ECAE 4E at 425°C

35

30

25

20

15

10

Ther

mal

Hys

tere

sis

(ºC)

350300250200150100500Stress (MPa)

Ti50.3Ni33.7Pd16 As-Received ECAE 4E at 425°C

Fig. 9. (a) Recovered transformation and irrecoverable strain levels and(b) thermal hysteresis of the as-received and ECAE processedTi50.3Ni33.7Pd16 alloy as a function of constant tensile stress levelsdetermined from the isobaric cooling–heating experiments.

350

300

250

200

150

100

50

0

Stre

ss (M

Pa)

15014013012011010090807060Temperature (ºC)

Ti50.3Ni33.7Pd16 ECAE 4E at 425ºC As-Received

8.4 MPa °C-1

7.3 MPa °C-1

Fig. 10. Effect of temperature on the critical stress for the martensitictransformation in the ECAE processed and as-received Ti50.3Ni33.7Pd16

alloy.

(a)

(b)

Stra

in (%

)

20016012080400Temperature (°C)

Ti50.3Ni33.7Pd16As-Received

150MPa1%

Stra

in (%

)

20016012080400Temperature (°C)

Ti50.3Ni33.7Pd16ECAE 4E 425°C

1%

150MPa

Fig. 11. Cyclic strain vs. temperature response of Ti50.3Ni33.7Pd16 under150 MPa load for 10 cycles: (a) as-received and (b) ECAE 4E at 425 �Csamples.

B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420 6417

thermal hysteresis, did not change with increasing stress.The only exception occurred above 350 MPa, at whichthe thermal hysteresis of the ECAE processed sample nota-bly increased.

To reveal the effect of ECAE on the onset of martensitictransformation, a critical stress vs. temperature diagramwas constructed, as shown in Fig. 10, using the results ofthe isobaric cooling–heating experiments in Fig. 7. rSIM

strongly depends on temperature, with a positive slopeaccording to the modified Claussius–Clapeyron (CC) equa-tion. The CC relation is defined as [28]:

dr=dT ¼ �ðDH=T 0etrÞ ð1Þwhere DH is the transformation enthalpy, T0 is the chemi-cal equilibrium temperature for the transformation, andetr is the transformation strain. dr/dT values were deter-mined to be 8.4 and 7.3 MPa �C�1 for the ECAE processedand as-received samples, respectively.

3.5. Thermal cycling experiments under constant stress

In thermal cycling experiments, the samples were held at150 MPa for 10 thermal cycles with a heating–cooling rate

(a)

(b)

6

5

4

3

2

1

0

Stra

in (%

)

10987654321Number of Cycles

Cumulative Irrecoverable Strain Strain - As-Received ECAE 4E at 425°C

Ti50.3Ni33.7Pd16

Recovered Transformation Strain As-Received ECAE 4E at 425°C

130

120

110

100

90

80

70

Ms T

empe

ratu

re (º

C)

10987654321Number of Cycles

Ti50.3Ni33.7Pd16Martensite Start Temperature

As-Received ECAE 4E at 425ºC

Fig. 12. (a) Recovered transformation and cumulative irrecoverable strainand (b) martensite start temperature (Ms) evolution of the as-received andECAE processed Ti50.3Ni33.7Pd16 alloy as a function of number of cyclesunder 150 MPa load.

6418 B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420

of 10 �C min�1. 150 MPa was chosen as the stress level forthe thermal cycling experiments because none of the samplesdemonstrated eirr at and below 100 MPa. However, under150 MPa load, the as-received samples started to show eirr,while the ECAE processed sample did not exhibit a notableone. The thermal cycling behavior of the as-received andECAE processed samples are shown in Fig. 11a and b upto 10 cycles. Fig. 12a demonstrates the recovered transfor-mation and irrecoverable strain levels as a function of num-ber of cycles, determined from the strain vs. temperaturecurves in Fig. 11. Under 150 MPa load, no eirr appeared inany cycle and the recovered transformation strain did notchange with the number of cycles for the ECAE processedsample, while for the as-received case, eirr obviouslyincreased and the recovered transformation strain levelevolved only slightly with the number of cycles.

The evolution of Ms temperature as a function of thenumber of cycles was also constructed and is shown inFig. 12b. ECAE stabilized the Ms temperature, which is apositive indication for the implementation of ECAE pro-cessed TiNiPd alloys in high force actuation.

4. Discussion of the results

In this work the main goal was to reveal the effect ofsevere plastic deformation via ECAE and associated micro-

structural refinement on the thermo-mechanical cyclic sta-bility of a ternary TiNiPd alloy. The main observations canbe summarized as follows:

1. The average grain size of the samples ECAE processedfor four passes at 425 �C was of the order of 100 nm.To the best of the authors’ knowledge, this is the lowestgrain size reported for a NiTiPd HTSMA. Similarmicrostructures could possibly be obtained using con-ventional processing techniques, such as wire drawing,however, there are no reports on the microstructuresof these difficult-to-work alloys processed using conven-tional techniques. In addition, an added benefit ofECAE processing is conservation of the workpiececross-section, which makes it possible to fabricate bulkactuators, such as torque tubes, from NiTiPd HTSMAs.

2. The twin mode in martensite was found to be {1 1 1}type I, and this did not change after ECAE.

3. The transformation temperatures decreased after ECAEprocessing. Strengthening due to microstructural refine-ment stabilized the thermal hysteresis, which no longerchanged with increasing constant stress levels up to350 MPa during thermal cycling.

4. ECAE led to a significant improvement in the thermalcycling stability and a reduction in the irrecoverablestrain levels under constant stress.

The rationale and mechanisms responsible for theseobservations are discussed in detail below.

4.1. Effect of ECAE on microstructural evolution

The TEM images and the diffraction patterns in Figs. 2and 3 showed that ECAE has no effect on the martensitetwinning mode in the Ti50.3Ni33.7Pd16 alloy. However, thetwin thickness after the processing was much smallerthan that of the as-received sample. In addition, therewas a notable refinement of the grain size after ECAE.Grain refinement in binary equiatomic NiTi alloys usingECAE triggers the formation of (0 0 1)m compound twinswith a few nanometers thickness in martensite, in addi-tion to the common f11�1g type I twins [19], while thegrain refinement in the Ti50.3Ni33.7Pd16 alloy did notaffect the twinning mode. The reason for such a differ-ence may be related to the different martensite structuresof NiTi and TiNiPd alloys, which are B190 and B19,respectively. (0 0 1)m compound twinning is a latticeinvariant solution for the two-stage transformationsequence B2 ? R ? B190 in binary NiTi [29]. It is notknown whether the same is true for the B2 ? B19transformation.

The mechanism of grain refinement during the ECAEprocess should be severe plastic deformation together withcontinuous dynamic recovery/recrystallization. Severe plas-tic deformation at elevated temperatures causes thedevelopment of new grains as a result of a gradual increase inmisorientations between the subgrains. Further deformation

B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420 6419

reduces the grain size continuously and the volume fractionof newly generated ultrafine grains increases [19].

4.2. Effect of ECAE on transformation temperatures

It is known that the transformation temperaturesincrease and thermal hysteresis decreases upon Pd additionto binary NiTi alloys [6,30]. However, the effect of Pd addi-tion on the cyclic stability of NiTi as a function of thermalcycles has not been studied in detail. It is interesting toobserve excellent thermal cycling stability of the transfor-mation temperatures in DSC experiments of the as-receivedTi50.3Ni33.7Pd16 samples without application of anythermo-mechanical treatment. Previous studies on equi-atomic NiTi and TiNiHf alloys [10,19] have shown thatthe thermal stability of transformation temperatures is onlyimproved after thermo-mechanical treatment in thesealloys and their thermal hysteresis was much higher thanthat of the present alloy.

Recently, Delville et al. [31] showed that a TiNiPd alloywith 11 at.% Pd possessed very low thermal hysteresis, dueto the enhanced compatibility between the single variantmartensite and austenite. Better compatibility allows twin-less martensite formation and minimizes the overall energyof the interfaces. Therefore, the improved stability of trans-formation temperatures and decrease in thermal hysteresisin the present alloy, compared with binary alloys, can beattributed to the enhanced lattice compatibility betweenthe austenite and martensite phases [31,32].

The transformation temperatures were decreased afterECAE processing, as summarized in Table 1. This dropis due to severe plastic deformation of the alloy, whichleads to grain refinement and formation of dislocation sub-structures. In such a microstructure multiple phase frontsshould nucleate and propagate, which would require moreenergy (i.e. a reduction in the transformation tempera-tures). Grain boundaries and high dislocation densitiesact as barriers to the propagation of phase interfacesand, thus, necessitate the nucleation of new phase fronts.A decrease in transformation temperatures due to grainrefinement has also been reported in equiatomic NiTi afterECAE processing [19].

In summary, after ECAE the transformation tempera-tures decreased and thermal hysteresis increased slightly.Note, however, that the above is only valid understress-free conditions. Under applied stress the situationis slightly different during thermal cycling. From the iso-baric cooling–heating experiments it can be seen that thetransformation hysteresis increased with external stress inthe as-received material, while it was quite stable in theECAE processed sample. This was a consequence ofstrengthening the material (as clearly shown in Fig. 6)due to microstructural refinement, such that even underhigh stress levels the dislocation plasticity and/or addi-tional defect generation was negligible. Thus, energy dis-sipation does not notably change with increasing stress inthe ECAE case.

4.3. Effect of ECAE on recoverable transformation and

irrecoverable strains

In this section the mechanism responsible for the reduc-tion in recovered transformation and irrecoverable strains,and the enhancement in cyclic stability after ECAE, duringthe isobaric cooling–heating and thermal cycling experi-ments will be discussed.

The recovered transformation strain levels of theECAE processed samples were lower at each stress levelin the isobaric cooling–heating experiments. A possiblereason is the very fine grain structure in the ECAE pro-cessed samples, in which full transformation from B2 toB19 martensite may not be attained. This argument stemsfrom the work by Waitz et al. [33], who investigated theeffect of ultrafine grain sizes on martensitic transforma-tion in binary NiTi. They proposed that grains smallerthan about 60 nm did not martensitically transform, evenupon large undercooling [33,34]. In addition, in order tosatisfy the compatibility across grain boundaries, theultrafine grains in the ECAE processed samples maynot undergo full reorientation/de-twinning upon cooling.Incomplete reorientation/de-twinning leads to a highervolume fraction of self-accommodated, twinned martens-ite, limiting the achievable recoverable strain levels.Another reason for the lower recoverable transformationstrain levels in the ECAE processed samples could becrystallographic texture. However, since the ECAE tem-peratures were high, significant texture evolution wouldnot be expected. Moreover, texture formation after fourpasses of ECAE in several bcc and fcc materials has beenshown to be not as strong as the texture usually observedafter conventional cold working. Also, the slope of thestress–temperature curve in Fig. 10 changed only slightlyafter ECAE, indicating weak or negligible texture evolu-tion. Therefore, crystallographic texture cannot have afirst order effect on the recovered transformation strainlevels in the present case.

From Figs. 9a and 12a it can be observed that therewas a notable decrease in the irrecoverable strain levelsafter ECAE. This was expected because, as shown inFig. 6, the CSS after ECAE increased as a result ofmicrostructural refinement. On the other hand, the as-received samples demonstrated an increase in theirrecoverable strain levels with number of cycles in thethermal cycling experiments, similar to the increase withconstant applied stress in the isobaric cooling–heatingexperiments.

An intriguing finding from the thermal cycling experi-ments is the enhanced cyclic stability in terms of transfor-mation temperatures after ECAE processing. Thetransformation temperatures of the ECAE processedsamples did not change with number of cycles, as pre-sented in Fig. 12. This can also be attributed to theincrease in CSS and, thus, the suppression of additionaldefect generation and a deterioration in the shape mem-ory characteristics.

6420 B. Kockar et al. / Acta Materialia 58 (2010) 6411–6420

5. Summary and conclusions

The present work has investigated the effect of severeplastic deformation by ECAE on the thermal cyclic stabil-ity of a Ti50.3Ni33.7Pd16 alloy. The effects of microstructuralrefinement on the transformation temperatures understress-free conditions, recovered transformation and irre-coverable strain levels under isobaric conditions, and theircyclic stability under a constant stress have been revealed.In the light of the experimental results the following con-clusions can be drawn from the present study:

1. Severe plastic deformation by ECAE led to a refinementof the microstructure, including grain and twin size. Thegrain/subgrain sizes achieved were the lowest reportedto date in NiTiPd HTSMAs.

2. Near perfect stability of the transformation tempera-tures during thermal cycling under stress-free conditionswas attained in the as-received NiTiPd alloy. The stabil-ity of the transformation temperatures under constantstress levels was enhanced after ECAE. The ECAE pro-cessed samples exhibited good thermal cyclic stability inthe constant stress thermal cycling and isobaric cooling–heating experiments. This improvement could be attrib-uted to the increase in CSS due to microstructuralrefinement.

3. Microstructural refinement also led to a notabledecrease in the irrecoverable strain levels in isobariccooling–heating experiments and thermal cycling undera constant stress.

4. The recovered transformation strain levels of theTi50.3Ni33.7Pd16 alloy decreased after ECAE due tothe reduced transformation volume and/or grainboundary constraints that limit martensite reorienta-tion/de-twinning.

Acknowledgements

This work was supported by the US Office of Naval Re-search under contract no. N00014-03-M-0332 and partiallyby the National Aeronautics and Space Administration un-der grant no. NNX07AB56A.

References

[1] Ma J, Karaman I, Noebe RD. Int Mater Rev 2010;55:257.[2] Donkersloot HC, Van Vucht JH. J Less-Common Met 1970;20:83.[3] Lo YC, Wu SK, Wayman CM. Scripta Metall Mater 1990;24:1571.

[4] Olier P, Brachet JC, Bechade JL, Foucher C, Guenin G. J Phys IV1995;8:741.

[5] Thoma PE, Boehm JJ. Mater Sci Eng 1999;273–275:385.[6] Lindquist PG, Wayman CM. In: Duerig TW, Melton KN, Stockel D,

Wayman CM, editors. Engineering aspects of shape memoryalloys. Boston, MA: Butterworth; 1990.

[7] Cai W, Tanaka S, Otsuka K. Mater Sci Forum 2000;327:279.[8] Otsuka K, Oda K, Ueno Y, Piao M, Ueki T, Horikawa H. Scripta

Metall Mater 1993;29:1355.[9] Hornbogen E. J Mater Sci 1999;34:599.

[10] Kockar B, Karaman I, Kim JI, Chumlyakov Y. Scripta Mater2006;54:2208.

[11] Khachin VN, Matveeva NM, Sivokha VP, Chernov DB, Yu K.Doklady Akad Nauk SSSR 1981;257:167.

[12] Yang WS, Mikkola DE. Scripta Metall Mater 1993;28:161.[13] Shimizu S, Xu Y, Okunishi E, Tanaka S, Otsuka K, Mitose K. Mater

Lett 1998;34:23.[14] Golberg D, Xu Y, Murakami Y, Morito S, Otsuka K, Ueki T, et al.

Scripta Metall Mater 1994;30:1349.[15] Golberg D, Xu Y, Murakami Y, Morito S, Otsuka K. Intermetallics

1995;3:35–46.[16] Waitz T, Antretter T, Fischer FD, Simha NK, Karnthaler HP. J

Mech Phys Solids 2007;55:419–44.[17] Waitz T, Kazykhanov V, Karnthaler HP. Acta Mater

2004;52:137–47.[18] Waitz T. Acta Mater 2005;53:2273–83.[19] Kockar B, Karaman I, Kim JI, Chumlyakov YI, Sharp J, Yu C. J

Acta Mater 2008;56:3630–46.[20] Kockar B, Karaman I, Kulkarni A, Chumlyakov Y, Kireeva IV. J

Nucl Mater 2007;361:298.[21] Karaman I, Karaca HE, Maier HJ, Luo ZP. Metall Mater Trans A

2003;34:2527.[22] Karaman I, Kulkarni AV, Luo ZP. Philos Mag 2005;85:1729.[23] Noebe R, Padula II S, Bigelow G, Rios G, Garg A, Lerch B. In:

Smart structures and materials 2006: active materials; behavior andmechanics, San Diego, CA, Bellingham WA: SPIE; 2006. p. 279–91.

[24] Padula A, Bigelow G, Noebe R, Gaydosh D, Garg A. In: MitchellMR, Berg B, editors. SMST 2006: Proceedings of the internationalconference on shape memory and superelastic technologies. MetalsPark, OH: ASM International; 2006.

[25] Barber RE, Dudo T, Yasskin PB, Hartwig KT. Scripta Mater2004;51:373.

[26] Niendorf T, Canadinc D, Maier HJ, Karaman I. Metall Mater TransA 2007;38:1946.

[27] Niendorf T, Canadinc D, Maier HJ, Karaman I. Int J Fatigue2008;30:426.

[28] Liu Y, McCormick PG. Acta Metall Mater 1994;42:2401.[29] Sehitoglu H, Jun J, Zhang X, Karaman I, Chumlyakov Y, Maier HJ,

et al. Acta Mater 2001;49:3609.[30] Liu Y, Kohl M, Okutsu K, Miyazaki S. Mater Sci Eng A

2004;378:205.[31] Delville R, Schryvers D, Zhang Z, James D. Scripta Mater

2009;60:293.[32] Cui J, Chu YS, Famodu OO, Furuya Y, Hattrick-Simpers J, James

RD, et al. Nature 2006;5:286.[33] Waitz T, Antretter T, Fischer FD, Simha NK, Karnthaler HP. J

Mech Phys Solids 2007;55:419.[34] Waitz T, Kazykhanov V, Karnthaler HP. Acta Mater 2004;52:137.