Comparative study of deformation-induced martensite stabilisationvia martensite reorientation and stress-induced martensitic

transformation in NiTi

Geraldine Tan, Yinong Liu*

School of Mechanical Engineering, The University of Western Australia, Nedlands, WA 6907, Australia

Received 1 June 2003; accepted 25 November 2003

Abstract

This paper presents a comparative study of the effects of deformation via stress-induced martensitic transformation and viamartensite reorientation on the transformation behaviour of a polycrystalline NiTi alloy. It was found that the two deformationmodes had very similar effects on the thermomechanical behaviour of the alloy after the deformation, including the recovery of

deformation upon subsequent heating, the development of two-way memory effect, and the stabilisation of the martensite. It wasestablished in this study that the transformation deformation, either stress-induced martensitic transformation or the martensitereorientation, is incomplete at the end of the stress plateau and continues in the homogeneous deformation afterwards. This study

also provided a direct comparison of the effects of the two deformation modes on the transformation heat. It was found in bothcases the heat effect of the reverse transformation of the oriented martensite, stress-induced or reoriented, decreased with increasingpre-deformation, in agreement with previous study on deformation via stress-induced martensitic transformation but in contra-diction to a previous study on deformation via martensite reorientation.

# 2004 Elsevier Ltd. All rights reserved.

Keywords: B. Martensitic transformations; B. Shape-memory effects; F. Calorimetry; F. Mechanical testing

1. Introduction

It has been reported in recent years that deformationcauses stabilisation to thermoelastic martensites in anumber of shape memory alloys [1–5]. The stabilisationeffect is manifested as an increase of the temperature forthe reverse transformation from deformed martensite toaustenite on heating. This stabilisation effect, however,is observed only for the first reverse transformationafter deformation. In subsequent thermal transforma-tion cycles, the transformation temperature is found torestore to the original value before the deformation.Associated with the stabilisation effect, other changes tothe transformation behaviour have also been observed,including:

1. decrease of temperature interval for the first

reverse transformation and increase of the inter-val for transformations in subsequent cycles [3,5];

2. changes of latent heat for the first reverse trans-

formation, increasing in one case of sheardeformation via martensite reorientation (MR)[3] and decreasing in other cases of tensiledeformation via stress-induced martensitictransformation (SIMT) [4] and rolling [6]; and

3. development of two-way memory effect in

subsequent transformation cycles [7]

The stabilisation effect has been observed in a numberof shape memory alloys, including TiNi, TiNiNb andCuAlNi [5,8], for various deformation modes, includingcold-rolling, tension, shear and compression [1–3,5]and via stress-induced martensitic transformation ormartensite reorientation process [3,4].Various hypotheses have been postulated to explain

the deformation-induced stabilisation effect. Lin et al.

0966-9795/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.intermet.2003.11.008

Intermetallics 12 (2004) 373–381

www.elsevier.com/locate/intermet

* Corresponding author. Tel.: +61-8-9380-3132; fax: +61-8-9380-

1024.

E-mail address: liu@mach.uwa.edu.au (Y. Liu).

attributed the martensitic stabilisation observed aftercold working to structural defects, such as dislocationsand vacancies, generated during deformation. Thesedefects increased the frictional resistance to the auste-nite/martensite phase boundary movement, thus inhi-biting the reverse transformation. They suggested that,upon heating, the annihilation of these defects and theformation of a different, defect-free, martensitic struc-ture resulted in the disappearance of the stabilisationeffect in subsequent thermal cycles. However, even at40% deformation, which yields a stabilisation effect of�120 K, heating did not exceed 493 K. It is unlikelythat, at this temperature, there would be significantalteration to dislocation structures leading to the dis-appearance of the stabilisation effect. Furthermore, sta-bilisation effect has been observed for deformationlevels as low as 2% in shear and for a magnitude ofseveral degrees.Piao et al. studied the stabilisation effect in CuZnAl

single crystals and NiTiNb and NiTi polycrystals. Theyproposed that the relaxation of the elastic strain energystored in the multi-variant martensite was responsiblefor the stabilisation effect. For single crystals therelaxation occurs during deformation by martensitevariant reorientation without plastic deformation. Forpolycrystalline specimens the relaxation of internalelastic energy is achieved during deformation with someplastic deformation, due to the constraint of grainboundaries. However, quantitative analysis of calori-metric measurements [3,4,6] suggests that the relaxationof internal elastic energy alone is insufficient to accountfor the degree of stabilisation observed, both in terms ofshift of transformation temperature and of the magni-tude of changes of transformation heat.Picornell et al. reported on the stabilisation of reor-

iented and detwinned g0 martensite in Cu–Al–Ni singlecrystals during compression via both SIMT and MR [5].They argued that once detwinned g0 martensite hadformed, original interfaces between the parent b phaseand the twinned g0 phase no longer existed and, there-fore, the return path is effectively destroyed. This loss ofboundary coherency makes it difficult for the deformedmartensite to return to the parent phase, resulting instabilisation and the observed burst-like nature of thereverse transformation. It is also noted from theirresults that the transformation behaviour after defor-mation, observed using DSC, was similar regardless ofwhether SIMT or MR deformation mode was applied.Whereas it is reasonable that boundary coherency is lostduring stress-induced transformation from b to det-winned g0, the applicability of the hypothesis to the caseof MR, where no parent-martensite phase boundariesexist, is uncertain.Liu and Favier studied deformation by MR in shear

and proposed that deformation not only relaxes theinternal elastic energy that is stored in thermal marten-

site, but also creates an internal elastic energy in thereoriented martensite that opposes its reverse transfor-mation [3]. Furthermore, they proposed that in poly-crystalline materials, an irreversible energy componentis unavoidable during deformation due to the necessityfor plastic deformation as an accommodation mechan-ism between neighbouring grains of mis-matched var-iant orientations. This hypothesis of internal plasticdeformation is supported by observation of a perma-nent two-way memory effect developed after the defor-mation [7]. Liu and Favier also found that indeformation via MR, the heat effect associated with thereverse transformation of the stabilised martensite, asmeasured by differential scanning calorimetry (DSC),increased with increasing deformation strain. Thisincrease was associated with the increase in elastic andirreversible energy and the actual amount of materialparticipating in the transformation. This is, however, incontrast to the findings in the case of deformation viastress-induced martensitic transformation [4]. Theendothermic heats for the reverse transformations wereobserved to decrease in SIMTs but an explanation forthis was not provided. In this work, it was also notedthat two reverse transformation peaks were observed onheating, calling into question the microscopic uni-formity of deformation. This phenomenon was notobserved in the case of shear [3] suggesting a more uni-form reorientation process. It should also be noted thatin tension, Luders type deformation is commonlyobserved, however, it was not observed in shear defor-mation. While, it is believed that the existence ofLuders-like deformation is dependent on sample geo-metry, its presence indicates a macroscopically localiseddeformation process. Microscopically, the deformationmechanism is unclear, however, it is expected that itwould differ for deformation via reorientation anddeformation via transformation.Almost all applications of shape memory alloys

involve transformation-related deformation via eitherMR or SIMT and shape recovery of deformed marten-site. The critical temperature for the shape recovery onthe reverse transformation, either induced by heating,as in the case of shape memory effect, or by unloading,as in the case of pseudoelasticity, and the magnitude ofhysteresis (stress window for pseudoelasticity, forexample [9]) are critical design parameters for manyapplications of shape memory alloys. For example, Xuet al. recognised the need for higher transformationtemperatures in the fabrication of NiTi shape memoryalloys/carbon fibre reinforced plastics (SMA/CFRP)smart composites due to the need for curing at elevatedtemperatures without inducing the reverse transforma-tion prior to application [10]. Furthermore, the deter-mination of the temperatures and the hysteresis is offundamental importance in the determination of ther-modynamic parameters. Therefore, an understanding of

374 G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381

this phenomenon is of essential technical importance aswell as scientific significance. This paper presents acomparative study of the effects of deformation viaSIMT and MR on the transformation behaviour andshape memory effect of a commercial polycrystallineNiTi alloy.

2. Experimental procedure

A Ti–50.5at.%Ni alloy in wire form of f 3 mm wasused in this study. The as-received material wasannealed at 1073 K for 1.8 ks followed by quenching inwater. Transformation temperatures of heat-treatedspecimens were determined by DSC to be Ms=292 K,Mf=279 K, As=307 K and Af=322 K.Tensile testing was carried out using an Instron 4301

testing machine. Samples were deformed both in mar-tensitic state via a martensite reorientation process andin austenitic state via a stress-induced martensitictransformation process. Specimens for SIMT were tes-ted in a hot air box at 307 K, �15 K above the Ms

temperature. The testing temperature was approachedby cooling from above 328 K to ensure an austeniticstructure. For martensitic reorientation transformation,samples were initially dipped in liquid nitrogen to attaina martensitic structure prior to deformation at 293 K,�15 K below the As temperature. Local deformationwas measured using an Instron extensometer with agauge length of 50 mm. Deformation was performed ata strain rate of 2�10�4 s�1, which was sufficiently low toavoid the adverse heating effect of the stress-inducedphase transformation on the deformation behaviour[11]. After deformation, the samples were kept at roomtemperature such that no transformation occurred priorto subsequent measurements.

Thermal dilatation measurement was carried outusing a Netzsch DIL 402EP dilatometer. Heating wascontrolled at a rate of 10 K/min. No accurate coolingrate control was applied. DSC measurements werecarried out using a Perkin-Elmer DSC4 calorimeter inargon atmosphere at a heating/cooling rate of 10 K/min. Prior to measurement deformed samples wereimmersed in liquid nitrogen so that any residual auste-nite would transform to form martensite thermally.Specimens for DSC and thermal dilatation measure-ments were cut from the tensile specimens using alow speed diamond-cutting wheel to avoid additionaldeformation.

3. Results

Fig. 1 shows the stress–strain behaviour and the sub-sequent thermal transformation behaviour after thedeformation of an annealed sample. The sample wasdeformed in martensitic state via MR. Similar curveswere obtained for samples deformed in austenitic statevia SIMT. The deformation behaviour exhibited 4 dis-tinct stages. Stage I is the elastic deformation of thestarting structure. The elastic deformation was followedby Luders-like deformation behaviour, identified by astress plateau (Stage II). The deformation mechanism ofStage II can be either MR or SIMT, depending on thestarting structure. Further deformation proceeded via auniform non-linear deformation in stages III beforereaching an apparent yield by plastic deformation instage IV. Samples were loaded to a certain pre-strainin various stages and then unloaded. The strain recov-ered upon unloading consisted of an elastic strain, "el,and a transformation-related strain, termed the mechan-ical recovery, "mech. Subsequent heating produced a

Fig. 1. Thermomechanical behaviour of annealed Ti–50.5at.%Ni wire deformed to 20% strain via SIMT.

G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381 375

thermal recovery strain, "th. The sum the three repre-sents the total strain recovery, shown in Fig. 1 as "re,whereas the sum of "th and "mech constitutes the totalstrain recovered as a result of the crystallographic rever-sibility of the martensitic transformation. Subsequentthermal dilatation measurements, involving a heatingand cooling cycle, revealed a two-way memory strain,"tw. The occurrence of the first reverse transformationafter deformation at a higher temperature compared tothe reverse transformation in the subsequent cycle indi-cates a stabilisation effect of the deformed martensite.

3.1. Deformation via stress-induced martensitictransformation

Fig. 2 shows stress–strain curves of specimensdeformed to various strains via SIMT. The specimendeformed to within the stress plateau region displayed adistinct demarcation on its oxidised surfaces betweenthe macroscopically undeformed and deformed sections,demonstrating the localisation of the deformation [12].One specimen was deformed to a global strain of 4%,from which samples from both within and outside themacroscopically deformed sections were taken for fur-ther analysis. After deformation to �6.7% strain, loca-lised deformation bands had completely propagatedthroughout the entire gauge length, hence all of thespecimens deformed to beyond this point had uniformdeformation.Fig. 3 shows thermal dilatation measurements of

samples pre-strained via SIMT. For these specimens,the strain recovery upon heating, "th, was associatedwith the reverse transformation of the stress-inducedmartensite. The curves have been set to zero strain atthe austenitic state after the first reversion for easy pre-sentation. It is evident that the first reverse transforma-

tion of the stress-induced martensite occurred at highertemperatures compared to the reverse transformation insubsequent cycles. The sample marked as 4%(a) wastaken from within the macroscopically deformed regionof the specimen. It is seen that the transformationbehaviour as well as strain recovery of this sample werepractically the same as the sample deformed to the endof the stress plateau at 6.4%. The sample marked as4%(b) was taken from outside of the macroscopicallydeformed region of the specimen. It is seen that thissample exhibited a very small strain recovery and vir-tually no two-way memory by the dilatation measure-ment, indicating that the deformation status of thissample was equivalent to the deformation at the onsetpoint of the stress-plateau.Fig. 4 shows DSC measurements of the transforma-

tion behaviour after deformation. The measurementswere conducted by heating from room temperature. Allsamples had been cooled in liquid nitrogen prior to theDSC measurements. Similar to the thermal dilatationmeasurement, sample 4%(b) exhibited similar transfor-mation behaviour to the undeformed specimen, practi-cally no stabilisation effect was observed, whereassample 4%(a) exhibited similar transformation beha-viour to the specimen deformed to 6.4%. For the speci-mens deformed to 6.4% and 8% [also including sample4%(a)], a small endothermic peak was measured priorto the major reverse transformation on the first heating,as indicated by the arrows. This small peak, occurringat a similar temperature to the reverse transformation ofthe thermally formed martensite in the as-annealedsample, is attributed to the thermally formed martensiteupon cooling in liquid nitrogen from the residual auste-nite surviving the pre-deformation. It is evident that theintensity of this small peak decreased with increasingpre-strain, indicating a progressive disappearance of

Fig. 3. Thermal dilatation measurements of samples after deformation

at 307 K via SIMT. The open circles indicate the starting points of the

measurements.

Fig. 2. Stress–strain curves of annealed Ti–50.5at.%Ni wire deformed

at 307 K via SIMT.

376 G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381

residual austenite after the stress plateau. This observa-tion suggests that the stress-induced martensitic trans-formation was incomplete at the end of the stressplateau and continues in Stage III.

3.2. Deformation via martensite reorientation

Fig. 5 shows the tensile stress–strain curves of samplesdeformed to various strains via MR. A comparison ofstress–strain behaviour between MR and SIMT isshown in the inset. In comparison with SIMT, the initialloading of MR in stage I exhibited a marked non-lineardeformation of �1.7% prior to the onset of the Luders-like deformation, whereas the Luders-type deformationstarted at �0.3% initial strain in the case of SIMT. Thetotal strains at the end of Stage II were �5.0% for MRand �6.7% for SIMT. The stresses of the deformationplateau are not directly comparable between the twodeformation modes, due to the fact that the stress forSIMT is temperature-dependent. The stresses in StagesIII and IV were practically the same for the two cases.

Thermal dilatation measurements of the deformedspecimens are shown in Fig. 6. The thermomechanicaltransformation behaviour of these specimens was simi-lar to those deformed via SIMT. The stabilisation effectis also evident. It is observed that the sample deformedto 5.3%, corresponding to the end of the stress plateau,showed very little two-way memory strain, unlike thespecimen deformed to 6.4% strain via SIMT, whichshowed a two-way memory strain of �1%.Fig. 7 shows DSC measurements of the transforma-

tion behaviour of the deformed specimens. Similar tothe case of SIMT, a small endothermic peak was alsoobserved prior to the major reverse transformation oninitial heating of the samples pre-strained to 5.3% and8%. In this case, there was no residual austenite in thedeformed specimens, because the specimens had been

Fig. 4. DSC measurements of samples after deformation at 307 K.

The samples had been immersed into liquid nitrogen prior to the

measurements. The open circles indicate the starting points of

the measurements.

Fig. 5. Stress–strain curves of annealed Ti–50.5at.%Ni wire deformed

at 293 K via MR. The samples had been immersed into liquid nitrogen

prior to the deformation.

Fig. 6. Thermal dilatation measurements of samples after deformation

at 293 K via MR. The open circles indicate the starting points of the

measurements.

G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381 377

cooled in liquid nitrogen prior to the deformation atbelow As. This peak is attributed to the original thermalmartensite that had not been reoriented or detwinned bythe deformation, suggesting that the martensite reor-ientation and detwinning deformation was micro-scopically inhomogeneous.

3.3. Comparisons between SIMT and MR deformation

Fig. 8 shows the recovered strains determined fromthe mechanical and thermomechanical measurementsfor both deformation modes. Recoverable strains areshown as open shapes for the case of SIMT deformationand as solid shapes for MR. The behaviour of bothappeared very similar. The dashed sections of the curvesindicate the expected global recoveries for deformationswithin the range of Luders-type stress plateau. It is seenthat "tr continued to increase with increasing pre-strainin Stage III after the stress plateau, reaching a max-imum of �8% after a deformation to 10%. Uponentering into Stage IV, whereas there was a significantreduction in "th, "tr decreased only slightly, compen-sated by a large increase of "mech. The differencebetween "re and the reference line at 45� is the unrecov-

ered plastic strain, "pl. It is evident that plastic strainwas present even at low levels of deformation. Itsmagnitude increased with increasing pre-strain. The factthat "re remained nearly constant in Stage IV indicatesthat deformation in this stage was nearly purely plastic,i.e., the process of SIMT or MR had stoppedcompletely.Fig. 9 shows the transformation heat, as measured

from the DSC curves, for both deformation modes.QthM is the heat of the thermal martensite (the smallpeaks shown in Figs. 4 and 7) detected in the firstreverse transformation after deformation. Q 1h is thetotal heat of the reverse transformation on the firstheating (including the small peak) and Q2h is the heat ofthe reverse transformation on the second heating. It isevident that the effects of the two deformation modeswere practically identical. QthM decreased with pre-strain in Stage III and disappeared in Stage IV. BothQ1h and Q2h decreased with pre-strain continuously in

Fig. 7. DSC measurements of samples after tensile deformation at 293

K. The samples had been immersed into liquid nitrogen prior to the

measurements. The open circles indicate the starting points of

the measurements.

Fig. 8. Recovery of deformation and two-way memory strain induced

by deformation.

Fig. 9. Measurements of transformation enthalpies after deformation.

378 G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381

Stages III and IV, with Q 2h being consistently lessthan Q 1h. After 20% deformation Q2h had decreased to12 J/g, �50% of its original value prior to deformation.Fig. 10 shows the characteristic transformation tem-

peratures measured at the maxima of heat flux from theDSC curves, with the open and solid symbols repre-senting the cases of SIMT and MR, respectively. TM-A

1

is the temperature of the first reverse transformationafter deformation, TA-M

2 is the temperature of the for-ward transformation in the second transformation cycleafter the first reversion, and TA-M

2 is the temperature ofthe reverse transformation in the second cycle. It is evi-dent that the two deformation modes exhibited almostidentical effects on the transformation temperatures.TM-A1 increased continuously with pre-strain well aboveTA-M2 , demonstrating the deformation-induced stabili-

sation of the oriented martensite. TA-M2 decreased only

slightly whereas TA-M2 showed more marked decrease,

leading to a slight increase in transformation hysteresis.

4. Discussion

4.1. Deformation mechanisms in various deformationstages

A clear difference in the stress–strain behaviour inStage I between SIMT and MR is evident. In the case ofSIMT, the linear stress–strain behaviour seems to indi-cate that this section is almost purely elastic. DSC andTMA measurements seem to confirm this. Sample4%(b), which represents the deformation strain at theend of Stage I deformation (or onset of Stage II defor-mation), exhibited behaviour similar to the undeformedsample, i.e., barely discernible stabilisation effect andtwo-way memory strain. This is in contrast to the caseof MR deformation, where a marked non-linear defor-

mation preceded Stage II, suggesting that homogeneousreorientation had occurred prior to the localisedLuders-like deformation. Sample 1%MR exhibited anoticeably larger thermal recovery strain as well as amore obvious stabilisation effect, as compared to Sample4%(b). It is logical to expect such differences betweenthe two deformation modes. In the case of SIMT, theparent and product phases, austenite and oriented mar-tensite respectively, correspond to two distinct freeenergy levels. Deformation via transformation relies onstress (resolved shear stress on a transformation shearsystem) levels achieving a critical value dependent onfree-energy differences. In the case of MR, multiplevariants of various orientations exist. Due to the differ-ences in orientation, it is natural to expect that some of thevariants may be very easy to reorient under the influenceof an external stress whereas some others may requirehigher local stress levels, leading to a wide stress windowfor the start of the reorientation deformation process. Italso appears that the magnitude of the strain precedingthe Luders-type stress plateau increases with increasingcross-section area of specimens in tension [13].The deformation mechanism in stage II, where

Luders-type deformation occurs, has been regarded asmartensite reorientation or stress-induced martensitictransformation depending on the starting structure. Themagnitude of the strain is determined by the crystal-lography of the transformation. However, it is of inter-est to note that the average lengths of the stress plateauare different for the two cases, 4% for MR and 6.4% forSIMT. Despite this difference, on the other hand, thetotal maximum recoverable strain, achieved at a pre-strain of �10% in Stage II for both cases, is practicallythe same (Fig. 8). This implies that the stress plateau ismerely a manifestation of the localisation of deforma-tion, instead of an exclusive process of transformationor martensite reorientation.This is consistent with the observation of the continued

increase of the recoverable strain with pre-deformationin stage III, which implies that the deformation mechan-ism of SIMT or MR, depending on the individual case,continues in stage III. This is consistent with previousobservations [7,14,15]. However, an explanation to thedistinctive change in stress–strain characteristics fromthe Luders-like deformation in Stage II to the homo-geneous deformation in Stage III, despite the continuationof the deformation mechanisms, is yet to be established.Given that a two-way memory effect started to developwith deformation to beyond Stage II and that theoccurrence of a two-way memory effect is an indicationof the establishment of a directional internal residualstress field, it is possible that the transition from Stage IIto Stage III is caused by the need to induce massiveinternal plastic deformation with further deformation.The small endothermic peaks measured by DSC at

similar temperatures to that of the reverse transformation

Fig. 10. Measurements of transformation temperatures after deformation.

G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381 379

of undeformed sample are attributed to thermalmartensite, which is either the part that had not beenreoriented in the case of MR or is formed from theresidual austenite surviving the pre-deformation inthe case of SIMT. The heat effect of the small peaks forthe specimens deformed to the end of the stress plateau,in both cases is �3 J/g, accounting for �12% of thetotal latent heat of the transformation in the unde-formed specimen, 23 J/g. This suggests that �12% ofresidual austenite or of non-oriented martensite wasretained at the end of the stress plateau. However,recoverable transformation strain, "tr, increased from4.2% at the end of the stress plateau to a maximum of5.7%, an increase by 26% in Stage III. This impliesthat, in addition to the further transformation of theresidual austenite or the reorientation of the retainedthermal martensite, further detwinning of stress-inducedmartensite or reoriented martensite produced in Stage IIalso took place in Stage III. This is consistent with thecontinued increase of TM-A

1 in Stage III.Summarising the above, it is evident that Stage III is a

mixed stage of further SIMT or MR, further detwinningof stress-induced martensite or reoriented martensite,elastic deformation and plastic deformation. Thus, thetransition between Stage III and Stage IV should bedefined at the end of the production of transformationstrain, instead of an ‘‘apparent’’ yield.Stage IV is a stage of pure plastic deformation

(neglecting the small increase in elastic strain). It isgenerally regarded that severe plastic deformationimpairs shape memory recovery. However, it is ofinterest to note that the total transformation strain, "tr,after reaching a maximum of 5.7% at 10% pre-strain,remained nearly constant for deformations up to 20%,despite a significant decrease of thermal recovery "th. Itis evident in Fig. 8 that this is due to the increase of"mech, demonstrating an exchange from thermal recov-ery upon heating to simultaneous mechanical recoveryupon unloading. This is consistent with the observationof ‘‘linear’’ pseudoelasticity after severe cold rolling inpolycrystalline NiTi [16].

4.2. Stabilisation effect

It is evident that both deformation via MR anddeformation via SIMT cause martensite stabilisation.There appears to be little difference between the twodeformation modes in their effects on the transforma-tion behaviour, including strain recoveries and two-waymemory effect (Fig. 8), transformation latent heat(Fig. 9) and transformation temperatures (Fig. 10). Thissuggests that, from a micromechanical point of view,stress-induced martensite and reoriented martensitehave similar structures. The transformation heat mea-surements show that the heat of the reverse transfor-mation on both the first heating and the second heating

decreased with increasing pre-strain for both deforma-tion modes. The decrease of Q1h for the case of SIMT isconsistent with a previous measurement of NiTi wiredeformed via SIMT in tension [4]. The decrease of Q1h

for the case of MR, however, is in contrast to a previousstudy of NiTi plate deformed via MR in shear [3], whereQ1h increased with increasing pre-strain. The reason forthis disagreement is unclear. The decrease of Q2h

appears to be consistent with the effect of cold rolling[6,17]. However, the dominant mechanism of deforma-tion during cold rolling is severe plastic deformationand it is unlikely that the causes of the changes totransformation heat in these two cases are the same.The transformation temperature measurements are

consistent with all previous studies: TM-A1 increases

continuously with increasing pre-strain in Stages III andIV whereas TA-M

2 and TA-M2 decrease moderately with

increasing pre-strain. It is seen that TM-A1 increased by

�15 K after deformation to the end of the stress pla-teau for both deformation modes whereas afterdeforming to 20% the total increase is �36 K, i.e., themajority of the stabilisation effect, as quantified bythe increase of TM-A

1 , occurred during deformation inStages III and IV. Considering that it is expected thatthe stored internal elastic energy in the self-accom-modating thermal martensite is released largely by theend of the stress plateau, this experimental evidenceappears to support the hypothesis that internal plasticdeformation is a significant contribution to the stabili-sation effect. It is interesting to note that despite thedecrease in transformation heat, TA-M

2 and TA-M2

decreased only slightly in the same range of pre-strain.It is known that both transformation temperatures andtransformation heat, as measured by DSC, are related tothe elastic and irreversible energies of the transformation.

5. Conclusion

The main findings and conclusions derived from theexperimental evidences of this study include the following:

1. Stress-induced martensitic transformation or

martensite reorientation is incomplete at the endof the stress plateau, implying that the stressplateau is only a manifestation of localisation ofdeformation. Both SIMT and MR are micro-scopically inhomogeneous, with localised pocketssurviving the deformation over the stress plateau.

2. Deformation in Stage III consists of (i) further

SIMT of the residual austenite or further reor-ientation of the retained martensite, dependingon the starting structure, (ii) further detwinningof the stress-induced martensite or reorientedmartensite produced in Stage II, as well as plasticand elastic deformations.

380 G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381

3. The two deformation modes, stress-induced mar-

tensitic transformation and martensite reorienta-tion, have very similar effects on the transformationbehaviour, implying structural similarities betweenmartensites produced by the deformation.

4. The heat of the reverse transformation of

deformed martensite decreases with pre-strain forboth deformation modes. This is consistent withprevious studies for SIMT but contradictory toprevious studies for MR. Explanations to boththe decrease and the disagreement with previousstudies (in the case of MR) are unclear.

5. Transformation heat in subsequent transforma-

tion cycles is also decreased as a result of thepre-deformation, whereas transformation tem-peratures decreased only slightly. Both transfor-mation temperatures and transformation heatare related to the elastic and irreversible energiesof the transformation. However, trends ofchanges of the two parameters in this study donot appear to provide a clear indication to thevariation of the free energy components causedby the deformation.

References

[1] Lin HC, Wu SK, Chou TS, Kao HP. Acta Metall Mater 1991;

39:2069.

[2] Piao M, Otsuka K, Miyazaki S, Horikawa H. Materials Trans-

actions, JIM 1993;34(10):919–29.

[3] Liu Y, Favier D. Acta Mater 2000;48:3489.

[4] Liu Y, Tan GS. Intermetallics 2000;8:67.

[5] Picornell C, Pons J, Cesari E. Acta Mater 2001;49:4221.

[6] Lin HC, Wu SK. Met Trans A 1993;24A:293.

[7] Liu Y, Liu Y, Humbeeck JV. Acta Mater 1998;47:199.

[8] Piao M, Miyazaki S, Otsuka K. Materials Transactions, JIM

1992;33:344.

[9] Liu Y, Galvin SP. Acta Mater 1997;45:4431.

[10] Xu Y, Otsuka K, Toyama N, Yoshida H, Nagai H, Kishi T.

Scripta Mater 2003;48:803.

[11] McCormick PG, Liu Y, Miyazaki S. Mater Sci Eng A 1993;

A167:51.

[12] Shaw JA, Kyriakides S. Acta Mater 1997;45:683.

[13] Sun QP, Li ZQ, Tse KK. Solid Mechanics and Its Applications

2001;89:113.

[14] Miyazaki S, Otsuka K, Suzuki Y. Scripta Metall 1981;15:287.

[15] Miyazaki S, Kimura S, Takei F, Miura T, Otsuka K, Suzuki Y.

Scripta Metall 1983;17:1057.

[16] Duerig TW. Engineering aspects of shape memory alloys.

London: Butterworth-Heinemann Ltd; 1990.

[17] Morawiec H, Stroz D, Chrobak D. J Phys IV 1995;5:C2/205.

G. Tan, Y. Liu / Intermetallics 12 (2004) 373–381 381