Phase Change Behavior of Nitinol Shape Memory Alloys REVIEWS

REVIEWS 437 Joachim Strittmatter Prof. P. Gümpel, J. Strittmatter, Fachhochschule Konstanz, Brauneggerstr. 55, D-78462 Konstanz (Germany) E-mail: [email protected] Prof. J. M. Gallardo Fuentes, EscuelaUniversidad Superior de de Sevilla, Ingenieros, Camino de losIsla Desenbrimientos de s/n la Cartuja, E-41012 Sevilla (Spain) Crystallography of SMAs: Thermoelastic martensites are ,* ential Scanning Calorimetry (DSC)Testing and are Electrical also Resistance usedtures. to determine transformation tempera- characterised by theirwhich low can be energy driven byThe and small glissade needle temperature or interfaces, structure stress changes. of martensite essentially consists of ± [*] Fig. 1. (a) Betaand phase D, crystal. after (b) cooling Self-accommodatingnant and when twin-related transformation stress variants, to is A,recovers applied. martensite. B, its original Upon C, (c) shape heating, Variant [2]. the A material becomes reverts domi- to the beta phase and [1] . Paul Gümpel WILEY-VCH VerlagGmbH, D-69469 Weinheim, 2002 1438-1656/02/0707-0437 $ 17.50+.50/0 , Ó -curve, as shown in Figure 2. e in Figure 2) varies with the alloy system. Differ- 1 T

JosØ María Gallardo Fuentes

In trained SMAs two way shape memory effect (TWME) Shape Memory Alloys (SMAs) are materials that have the ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 1. General Characteristics Among other specialpredetermined characteristics shape Shape when Memory heated.influenced Alloys by In (SMAs) thermal fact, have and thevarious thermo-mechanical the materials, phase treatments. ability which change Up can to to of generallybased now, be return an alloy classified SMEs systems to existing into have and noble-metal a element been non-metallica based, SMAs. discovered can Cu-based, detailed In in Fe-based, strongly review this Ni-Ti- be over papertreatments a the upon general Ni-Ti-Cu the overview phase system of change will the behavior. be Ni-Ti system given and with special regard to the influence of heat By Influence of Heat and Thermomechanical Treatments Shape Memory Alloys Phase Change Behavior of Nitinol Most of the transformationtemperature occurs range, over although atransformation the relatively beginning during narrow and heatingover end or a of cooling much the also larger actually exhibits temperature hysteresis extends in range. thatand the The on transformations cooling transformation on do heating not(shown overlap. This as transformation hysteresis can be observed in theis sense needed that to no mechanical completecling. deformation the The strain phase cycle,temperature but transformation but only does thermal over not cy- awith occur range each at of alloy a system. temperaturesing A single which transformation method temperatures varies of is to determiningmen thermally and under cycle load, nam- a producing speci- a T- ability to returnBeing cold to or a belowtheir their predetermined martensitic transformation forms shape temperatures (Fig. orstrength when 1b), and in can they heated. be have quite easily a(Fig. deformed very 1c). into any However, low new when yield shape oneformation temperature, alloy it is undergoes heated a change aboveture, in its crystal reverts trans- struc- to(Fig. austenite 1a). If and the SMAmation, recovers faces it any can its resistance generate during extremely previous thisknown large transfor- as forces. shape one This way process shape memory is effect (OWME) Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys REVIEWS

Fig. 2. Typical transformation versus temperature curve for a specimen under constant Fig. 3. Typical stress-strain curves at different temperatures relative to the transforma-

load (stress) being cooled and heated. T1: transformation hysteresis; Ms: martensite tion, showing (a) austenite, (b) martensite, and (c) pseudoelastic behavior [2]. start; Mf: martensite finish; As: austenite start, Af: austenite finish [2].

twin-related, self-accommodating variants (Figure 1). The line CD, and shape recovery occurs, not upon the applica- shape change among the variants tends to cause them to elim- tion of heat (even though Md > T3 > Af is maintained) but inate each other. As a result, little macroscopic strain is gener- upon a reduction of stress. ated. It is also possible to obtain a stress-induced martensite

below Md-temperature (Md > T > Af). In that case, or when stressing a self-accommodating structure, the variant that can 2. The Binary Ni-Ti System transform and provide the greatest shape change in the direction of the applied stress is stabilised and becomes dominant 2.1 Description of Ni-Ti-based Alloys in the configuration (Figure 1c). This process creates a macro- Up to now, SMEs have been discovered in various materi- scopic strain, which is recoverable as the crystal structure als, which can generally be classified into noble-metal based, [2] reverts to austenite during reverse transformation. Cu-based, Fe-based, Ni-Ti-based alloy systems and non-me- The mechanical properties of shape memory alloys vary tallic shape memory materials. In this paper a detailed review greatly over the temperature range spanning their transfor- over the Ni-Ti- and the Ni-Ti-Cu system will be given, with mation. This is seen in Figure 3, where simple stress-strain special attention devoted to the influence of heat treatments curves are shown for a nickel titanium alloy that was tested upon the phase change behavior.[4] in tension below, slightly above, and well above its trans- SMAs based on nickel and titanium have to date provided formation temperature range. The martensite is easily de- the best combination of material properties for most commer- formed to several percent of strain at quite a low stress, cial applications. Buehler at NOL, the Naval Ordnance Labo- whereas the austenite (high temperature phase) has much ratory, first discovered the SME in the Ni-Ti system in 1962 higher yield and flow stresses. The dashed line on the mar- and thus the system is named NiTi-NOL or Nitinol, which is tensite curve in Figure 3 indicates that upon heating after commonly used when referring to Ni-Ti-based alloys. This removing the stress, the sample remembered its unstrained discovery led to a rapid growth of interest in the shape mem- shape and reverted to it as the material transformed to aus- ory phenomenon. A series of extensive reports documented tenite. No such shape recovery is found in the austenite the properties of Ni-Ti-based alloys, mainly in wire form, and phase upon straining and heating, because no phase change provided much useful information for the designer. However occurs. In Figure 3c, the material is tested slightly above its since then there has been significant progress in the under- transformation temperature. At this temperature, the so- standing of the alloy, particularly with regard to the effects of called superelastic effect occurs. This effect is also called processing and heat treatment on the mechanical and shape many times pseudoelasticity, pseudoelastic effect or supere- memory behavior.[5] lasticity.[3] It is caused by the stress-induced formation of some martensite above its normal temperature. Because it 2.1.1 Metallurgy of Binary Ni-Ti-based Alloys has been formed above its normal temperature, the martensite reverts immediately to undeformed austenite as soon Ni-Ti-based alloys are ordered intermetallic compounds as the stress is removed. This process provides a very based on the equiatomic composition, of 50 at.-% or 55 wt.-% springy, ªrubber likeº elasticity in these alloys. It then im- Ni. This compound usually exists as a metastable phase mediately strains and exhibits the increasing strain at con- down to room temperature. Consequently, in contrast to cop- stant stress behavior, seen in AB. Upon unloading, though, per based alloys, no betatising and quenching are necessary the material reverts to austenite at a lower stress, as seen in to prevent the decomposition into other phases at intermedi-

438 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys REVIEWS ate temperatures. However, at low temperatures the homoge- 450 nous range of the NiTi compound is very narrow and so the 400 Ni-Ti-based alloys often contain precipitates of a second 350 intermetallic phase distributed in the matrix, either Ni3Ti on 300 the Ni-rich side or Ti2Ni on the Ti-rich side. The maximum solution capacity is 57 at.-% Ni at the 1118 C Eutectic point, 250 but on the Ti-rich side of the alloy system, the capacity is only 200 50 at.-% Ti. The concentration range of technical interest is Stress (MPa) 150 [6] between 49 and 52 at.-% Ni. 100

The presence of oxygen is often ignored when considering 50 the microstructure of Ni-Ti-based alloys. Titanium is very 0 reactive, particularly in the molten state, and some oxygen is 02468 invariably present in the alloy. Oxygen decreases the stoichio- Strain (%) Transverse metric range of the NiTi compound and can unexpectedly Longitudinal result in compositions within a three-phase field. Thus Ni Ti 3 Fig. 4. Martensitic stress strain curves of specimens taken longitudinally and transver- can also be present, for example in a Ti-rich alloy. Further- sely from rolled sheet [7]. more, the oxide Ti4Ni2O is isostructural with the intermetallic [7] Ti2Ni, which can make unique phase identification difficult. none exists and the stress levels are much higher. Some Concerning transformation temperatures related to SME the further effects of thermo-mechanical treatments on the SME binary Ni-Ti system is very sensitive to delicate variations of behavior are described in section 2.2. the stoichiometric composition. A deviation of 0.1 at.-% effects The length of the martensite plateau also determines the a change of transformation temperature of 10 C.[5] Thus, strain at which the transition to the third region of the stress depending on the desired Ms, precise composition control of strain curve occurs, and can therefore affect the amount of up to one tenth and one hundredth of a percent is necessary. memory strain. Thus although it is often stated that up to 8 % strain is heat recoverable, the actual amount depends on the alloy, its thermo-mechanical processing, testing direction and 2.1.2 Mechanical and Shape Memory Properties of Ni-Ti-based deformation mode. In many cases, a conservative design Alloys would use less than 8 % strain. Like most SMAs, Ni-Ti-based alloys show marked differ- With regard to transformation temperatures, nowadays ences in mechanical behavior depending on whether they are temperatures from ±100 C to 100 C, depending on composi- tested in the austenitic or the martensitic phases. The stress- tion and thermo-mechanical treatment, are common for com- strain curve of the martensite can be divided into three well- mercial applications of Ni-Ti-based alloys, having a hysteresis defined regions. An initial low plateau results from the stress width of typically 30±50 K. The transformation temperatures induced growth of one martensite orientation. This process oc- usually increase with an increasing load in a linear fashion. curs by untwinning unfavourably oriented neighbours. At This slope, as well as being of thermodynamic significance, is higher stresses there is a second region that is usually linear, also important when choosing an alloy for a particular appli- although not purely elastic. It is believed that the deformation cation. Concerning applied load/transformation temperature mechanism in this stage is a mixture of elastic deformation of relationship, the stress rate of Ni-Ti-based alloys can also the untwinned martensite, together with the formation of new change significantly from alloy to alloy, covering a range orientations of martensite which intersect those already pre- from 2.5 MPa/C to over 15 MPa/C.[7] sent, and which provide additional heat recoverable strain.[7] The transition to the third region is a result of the onset of irreversible plastic deformation, as in the case of the yielding of all conventional metals. Thus the maximum amount of heat recoverable or memory strain is obtained by initially deforming to the end of stage two. If larger deformation Mar tens itic strains are used, then the reversible martensitic deformation Phas e processes and the dislocations resulting from plastic flow ) Change interact and the memory strain decreases, see the following. R-Phas e

Depending on the precise nature of the alloy and particu- Strain (% larly its prior thermo-mechanical history, the martensitic plateau can vary from a continuous curve with an inflection point to a clear horizontal plateau with a sharp yield point and upturn. Figure 4 shows curves from rolled sheet, where it can be seen that testing in the longitudinal direction results Temperature (°C) in a well-defined plateau, whereas in the transverse direction Fig. 5. Premartensitic (R-phase) transformation in a strain-temperature diagram [5].

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 439 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

2.1.3 R-Phase Transition of the twin boundaries, while annealing rearranges these dislocations. Ni-Ti-based alloys show the phenomena of a premartensi- Figure 6 shows the effect on austenitic yield strength of tic transition, which is named R-phase in most accounts. This annealing temperature after 40 % cold work in the martensite can be confirmed either by Differential Scanning Calorimetry of a 50.6 at.-% Ni alloy. It can be seen that a rapid decrease is (DSC) or electrical resistance testing, whereby a strong anom- REVIEWS observed in the range 350 to 450 C, followed by a more grad- aly in the curve appears due to R-phase transition. Figure 5 ual decrease up to 850 C. At the same time, M increases, and shows that R-phase transition is also coupled to a strain s again the increase is rapid between 350 and 450 C. Compar- recovery, which takes place with or without small hysteresis ing the two graphs in Figure 9 it is apparent that for a given of 1±2 K. Although the R-phase effect is much inferior to the alloy, although processing can increase the yield strength, it normal SME (max. about 1 %), it is important for many appli- is done at the expense of a lower M . cations such as thermal actuators, effecting a high number of s One of the principal improvements upon the properties of cycles without injury.[8] Ni-Ti-based alloys obtained by processing is cyclic behavior. When the same binary alloy annealed at 850 C, with a strain corresponding to an initial stress of 150 MPa being applied, is subjected to one cycle of deformation and reheating, a large 2.2 Influence of Heat Treatments and Thermo-mechanical amnesia of around 1.5 % is noticed. In other words a single Processing on the Shape Memory Effect behavior of memory cycle with this stress leads to an irreversible defor- Ni-Ti-based Alloys mation of 1.5 %. Conducting the same test after cold working Only a few years after the discovery of the SME in Ni-Ti- 40 % and annealing at 400 C results in amnesia of around based alloys, a process patent was filed, increasing the yield 0.5 %. strength of these alloys by cold working. Since then, the com- The strain obtained on cooling and the amnesia on heating bination of cold work in the martensite together with a subse- are plotted in Figure 7 as a function of stress after cold work- quent anneal has been extensively explored as a way of ing and annealing at 350 C and 450 C respectively. The improving the SMA characteristics. 350 C anneal gives significant amnesia at all useful working Cold work alone, i.e. without the annealing step, destroys stress levels. For low temperature anneals where the amnesia the martensitic plateau on the stress strain curve. Thus a is low, the data obtained from the first cycle, as shown here, material cold worked 20 % in the martensite has a very high approximates the cyclic behavior after many cycles. However yield strength, but its shape memory properties are poor in where the amnesia on the first cycle is high, the cyclic stabil- that only very low recoverable strains are possible.[9] Anneal- ity is poor.[7] ing will restore the memory effect, but decrease the yield When choosing a Ni-Ti-based alloy for an actuator type of strength. The choice of amount of cold work and the actual application, it must be remembered that for a given composi-

annealing temperature establish the combination of those two tion, a high Ms-temperature is obtained by annealing out any properties. It is concluded that cold work introduces a high effects of prior cold work. The cyclic stability is therefore density of ªrandomº dislocations, which hinder the mobility poor. Devices triggering at or above 100 C are unlikely to have mechanical stability over tens of thousands of cycles. However, it 1200 90 should be remembered that this is a 1100 mechanical instability, where the 50,6at%Ni,CW40% 80 strain per cycle varies at high work- 1000 ing loads. There is no metallurgical in- 900 70 stability leading to shifts in the re- 50,6at%Ni,CW40% 800 sponse temperature as a result of prolonged exposure to around 700 60 100 C, in contrast to Cu-based SMAs. 600 Ms at 150 MPa If superelastic behavior at ambient 50 500 temperature is required, then Ms

Austenitic Yield Stress (MPa) must be lowered, and the alloy shall 400 40 be nickel-riched. Another way of low- 300 ering Ms is a solution treatment fol- 200 30 lowed by rapid cooling and cold 300 400 500 600 700 800 900 300 400 500 600 700 800 900 working. But it has to be taken into Annealing Temperature (°C) Annealing Temperature (°C) account, that this will also provide Fig. 6. Austenitic yield stress of 50.6 at.-% Ni alloy, cold worked 40 % and then annealed for 30 mins. at the tempera- additional hardening from the pre- tures indicated (left). Transformation temperature Ms (measured at 150 Mpa) as a function of annealing temperature of the same alloy (right) [7]. cipitation of nickel-rich phases.

440 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys REVIEWS 6 8

7 5 450 °C 350 °C 6

4 5

3 4 Strain (%) Strain (%) 3 2 2

1 1

0 0 0 100 200 300 400 0 100 200 300 400 Stress (MPa) Stress (MPa) Fig. 7. The strain obtained on cooling (upper curve) and the amnesia (lower curve) as a function of stress for a Ni-Ti based alloy cold worked 40 % and annealed at 350 C (left) and the data after annealing at 450 C (right). [7]

2.2.1 The Influence of Thermomechanical Treatment on R-Phase Both Af and ep increased dramatically for all three materials Transition with the Standard Superelastic faring the worst and the Cr- doped again faring the best. The heights of the loading and An increase in cold deformation of Ti-Ni alloy results in unloading plateaus decreased uniformly for all three alloys the increase in both tensile strength and R-phase transition. with an increase in time at 400 C. Overall, exposure times at As R-phase becomes stable, the SME will be improved and 400 C should be kept at or below 2 minutes if at all possible, the decay of memory effect will also be controlled at fatigue although if PTFE coating is required the decrease in the pla- test. The increase in annealing temperature results in a de- teau heights can be offset by selecting an alloy with higher crease of R-phase transition range. In that case, the stability of starting plateaus such as the Cr-doped alloy.[9] R-phase and the SME become poor. When annealing at 600 C, the recrystallisation occurs and no R-phase is found. Many of the previous facts were confirmed during an 2.2.2 Influences of Several Training Methods investigation by Shape Memory Applications Inc., Santa Especially in order to obtain the possibility of two-way Clara, California, USA, conducted into improvements of Ni- shape recovery, certain training methods have to be applied to Ti-based alloys for medical applications.[9] The effect of heat common SMAs. The training of SMAs is described as a ther- treatment of three experimental alloys as listed in Table 1 momechanical treatment with the intent to make it possible for were tested with regard to various coating operations, includ- the alloy to remember both the high and the low temperature ing polyurethane and PTFE coating. shapes, repeatedly. Under normal circumstances, a SMA Polyurethane coating typically exposes the material to remembers its high temperature shape, upon heating recover- temperatures of about 200 C for a few minutes, while PTFE ing its austenitic shape, but immediately forgets the low coating processes are carried out at temperatures in the temperature (deformed) shape. However, it can be ªtrainedº neighbourhood of 400 C. To study these heat treatment to remember the low temperature shape as well. This is accom- effects in general, heat treatment trials were conducted at plished essentially by leaving some ªremindersº of the de- 200 C, 300 C, and 400 C to study the effects on critical NiTi formed low temperature condition in the high temperature properties. At 200 C, all three materials are relatively stable, phase. In general, these procedures are based on repetition of a even for times as long as 60 minutes. While the effect even of thermomechanical cycle through the transformation region long 300 C heat treatments on the stress plateaus was insig- nificant, some measurable effects on Af and ep (mentioned as amnesia above) were noted. At times as low as 15 minutes, Table 1. both the Af and the ep of the three alloys began to increase noticeably. In general, the effect appeared to be most dramatic for the high nickel material and least dramatic for the Standard High Chromium Superelastic Nickel Doped Cr-doped alloy. Overall, it is recommended to keep exposure times at temperatures of approximately 300 C to a maximum Nickel Content 55.8 wt% 56.0 wt% 55.7 wt% of about 5 to 10 minutes. Heat treatments at 400 C had sub- Ternary Addition ± ± 0.2 wt% Cr stantial effects on all three alloys at times as low as 2 minutes.

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 441 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

and therefore are referred to as training. The widely accepted heating. Although effective, this method is a little more com- mechanism of the two-way memory effect (TWME) is based plicated than the others that have been described. on the observation that cycling the material through the (5) TWME training by constrained temperature cycling of transformation region generates complex dislocation arrange- deformed martensite: this is a variation on the previous meth- ments. Based on this observation, the TWME has been attribut- od, which is perhaps somewhat easier to carry out in terms of

REVIEWS ed to the microscopic stress fields accompanying these disloca- temperature control, and is probably the most commonly tion arrangements. The residual stress fields favour the centre used training method at present. The sample is deformed

and the beginning of the growth of preferential variants. Con- below Mf, thus producing a stress-corrupted martensitic sequently, during further cooling, preferential variants grow microstructure. The sample is then constrained in the

without any internal or external assistance, thus causing deformed condition and heated to above Af. The sample is [10] TWME. There are several common ways to train TWME: typically cycled from below Mf to above Af a number of (1) TWME Training by overdeformation while in the times, with the sample all the while constrained in the origi-

martensitic condition: the alloy is cooled below Mf, and while nal deformed shape, to complete the training routine. This in the martensitic state, is severely bent to well beyond the training method proves to be particularly effective and is rel- usual strain limit for completely recoverable shape memory. atively straightforward to carry out. When reheated to the austenite range, the alloy will not completely recover the original shape, due to the excessive deformation of the martensite. As the shape memory strain 2.3 Influence of Alloy Elements in the Ni-Ti System limit is exceeded, a partial loss of memory results (amnesia). However, if cooled again to the martensite range, the alloy The binary standard Ni-Ti-based alloys deliver quite con- will spontaneously move part of the way back towards the vinient properties in the transformation temperatures range overdeformed shape. between ±30 and +80 C. They exhibit superelasticity and/or (2) Training by shape memory cycling: this procedure con- shape memory behavior, combined with an excellent biocom- [11] sists simply of repeatedly carrying out shape memory cycles patibility (in the case of proper treated surfaces, but also until the two-way behavior begins to be demonstrated. One good enough in normal conditions, even if sometimes dis- [12] shape memory training cycle would consist of the component cussed controversely ) and are therefore successfully applied for orthodontic devices and other medical applica- being cooled to below Mf, deformed to a level below the heat- [3,12] recoverable shape memory strain limit, then heated to recover tions. Besides this, other applications require transforma- the original high temperature shape. After perhaps 5 to 10 of tion temperatures beyond those of the binary system (e.g. these shape memory cycles have been carried out, (the direc- sensors), or a wider or narrower hysteresis (e.g. actuators). tion of the strain must be the same during each training This can be attained by the addition of third and fourth ele- cycle), the component will begin to spontaneously change ments to the binary alloy system. shape on cooling, moving in the direction in which it has been The addition of third or fourth elements to Ni-Ti-based consistently deformed during the training cycles. The amount alloys provides a powerful tool for controlling the properties of spontaneous shape change on cooling will be significantly and can be used to: less than that which was being induced in the shape memory ± control transformation temperatures deformation step. Typically the spontaneous shape change ± increase the stability of Ms with respect to thermal history will perhaps be 1/5 to 1/4 of the training strain; for example, ± control the hysteresis width if the strain induced during training was 6 %, the sponta- ± increase austenitic strength neous TWME strain is likely to be no more than 1 or 2 %. ± reduce or increase martensitic strength (3) Training by pseudoelastic cycling (PE): it consists of ± increase two-way effect ability repeatedly stress-inducing martensite by loading and unload- High temperature transformation was the basis of many investigations in recent times.[13±20] Alloy systems such as Ni- ing the austenite above the Af-temperature, but below Md where pseudoelastic (or superelastic) behavior is expected. Ti-Pd, Ni-Ti-Pt, Ni-Ti-Hf, Ni-Ti-Zr, Ni-Ti-Au and Ni-Ti-Al As for the training method above, the number of training have been studied to explore their properties, including engi- cycles required is typically in the order of 5 to 10, and the sub- neering considerations like manufacturing, phase stability, sequent spontaneous shape change on cooling and heating is mechanical stability, and cost. a fraction of the training strain. (4) TWME training by combined SME/PE cycling: a partic- 2.3.1 Nickel-Titanium-Palladium System and Nickel-Titanium- ularly effective training routine for TWME has been found Platinum System that combines some of the features of methods 2 and 3. The component is first deformed in the austenite condition, then These two alloy systems can be counted among the high

cooled to below Mf while holding the induced strain in the temperature SMAs because replacing Ni either by Pd or Pt sample (including the elastic strain), then heated up to re- has similar effects on the shape memory properties. It was cover the original shape. If this routine is repeated a number observed that small additions of these elements cause a slight

of times, TWME will be obtained on subsequent cooling and decrease of Ms-temperature, and even shift it below 0 C, but

442 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

higher additions strongly elevate the transformation tempera- Table 2. DSC measured transformation temperatures of several Ni-Ti-Pd alloys [22]. REVIEWS [21] tures compared to other elements. The Ms-temperature ranges from ±26 C (for 10 at.-% Pd) to 563 C (for 50 at.-% Sample Compound Ms/C Mf/C As/C Af/C Pd) and from ±10 C (10 at.-% Pt) to 1040 C (for 50 at.-% 1TiPd 459 427 497 533 Pt),[21] see Figure 8. The Table 2 demonstrates that rising 50 50 replacement of Ni by Pd results in rising transformation tem- 2 TiNi10Pd40 321 587 327 366 [22] peratures. 3 TiNi20Pd30 108 80 109 153

For all the observed alloy systems, a OWME effect has 4 TiNi22Pd28 Ð Ð 132 180 been achieved and also the training of TWME has been possi- 5 TiNi27Pd23 50 25 45 85 ble e.g. for the Ti-30 at.-% Ni-20 at.-%Pd alloy. Some alloys recovered nearly 100 % from deformation and reached up to 4 % of shape memory for one±way; others recovered only a fraction of applied strain or broke due to micro-cracks from a certain cycle, the dislocations become saturated and the hot rolling. two-way memory remains unchanged with a further increase Effect of training strain: TWME-training of Ni-Ti-Pd-based in the number of cycles. For the same reason, the increase in alloys was studied in bending tests.[23] In this bending test the the training strain also induces more dislocation arrays and specimen was (1) deformed against a cylindrical rod to a finally produces a larger TWME for the same number of given constant strain at room temperature, (2) unloaded, (3) cycles.[23] heated to Af +30C and finally, (4) quenched in water. The angles after unloading, heating, and quenching, were measured during each cycle, along with TWME, the transformation 2.3.2 Nickel-Titanium-Hafnium System reverse strain and the permanent strain. The variations of recovery angles with the number of training cycles and the Of particular interest is the Ni-Ti-Hf system because of its training strain are shown below in Figure 9a. The angle low cost and high transformation temperature potential. increases rapidly in the first few training cycles, then changes Transformation temperatures up to Ap=622 K (=349 C) were very slowly as the number of trainings increase, finally reach- recorded, exhibiting good shape memory properties.[13] ing a saturation value. Figure 9b shows the effect of the train- Stress-Strain behavior: the stress-strain behavior of binary ing strain on the TWME. It is apparent that the saturation val- Ni-Ti-based alloys normally exhibits well-defined yielding, a ue increases as the training strain increases. Figure 9 stress plateau with little work hardening and a strain of about [14] demonstrates that the training strain and the number of 8%. In contrast, for an investigated Ni49Ti36Hf15 alloy, the cycles directly determine the TWME. stress plateau completely disappears and high work harden- As stated before, the mechanism of the TWME originates ing is constantly observed instead. The high work hardening from the complex dislocation arrangements, generated by will tend to result prematurely in dislocation slip during the cycling the material through the transformation region. As martensite variant reorientation or the stress-induced marten- the number of cycles increases, the dislocation increases and sitic transformation and thereby worsen the shape memory its arrangement becomes more and more stable. Therefore, properties of the investigated alloy. the TWME increases as the number of cycles increased. After Tensile properties: Figure 10 shows the variation of the r0.2

stress with tensile temperature. Clearly, the r0.2 stress versus temperature curve is approximately S-shaped, a general fea- 1100 ture observed in SMAs. The negative temperature dependence

1000 of the r0.2 stress is associated with the martensite variant reori- 900 entation in the low temperature range below about 510 K 800 (237 C) and with dislocation slip in the high temperature range above about 590 K (317 C), respectively, while the posi- 700 Ti50 (Ni(50-x) Ptx) 600 tive temperature dependence of the r0.2 stress between 510± 500 590 K (237 C±317 C) is attributed to the stress-induced mar- 400 tensitic transformation. The dr0/dT of the experimental alloy 300 is found to be smaller than that for Ni-Ti-based alloys. Com- Ms Temperature (°C) pared with other SMAs, the present experimental alloy exhib- 200 Ti50 (Ni(50-x) Pdx) its a relatively high critical stress for the martensite variant 100 reorientation. This is considered to be disadvantageous to the 0 shape memory properties of the Ni Ti Hf alloy.[24] -100 49 36 15 01020304050Figure 10 also shows the variation in the elongation with tensile temperature for the Ni Ti Hf alloy. Clearly, the Composition (x) 49 36 15 elongation is strongly dependent on the tensile test tempera- Fig. 8. Plot of Ms temperatures as a function of composition for Ti-Pd-Ni and Ti-Pt-Ni alloys [21]. ture. Around room temperature, the elongation is about 10 %.

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 443 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

14 14 a b 3.8% 12 12 Saturation Angle Value

10 10

REVIEWS 2.3%

8 8

6 6 1.9% TWME(Degree) TWME (degree)

4 4 1.7% 1.3% 2 2

0 0 0 5 10 15 20 25 11,522,533,54 Number of cycles Training Strain % Fig. 9. The two-way memory effect of Ti-30 at.-%Ni-20 at.-%Pd alloy as a function of the number of cycles (a), and the variation of the saturation value due to training strain (b) [23].

With increasing tensile test temperature, the elongation grad- 2.3.3 Nickel-Titanium-Zirconium System ually increases and reaches a peak of about 30 % at about Since the high temperature SMAs of the Ni-Ti-Hf system 510 K (237 C), slightly above A -temperature, and then de- f were produced, the development of economic high tempera- creases rapidly to a minimum at about 560 K (287 C). The ture SMA has attracted great interest. Studies on Ni-Ti-Zr- maximum elongation is possibly attributed to transforma- based alloys have shown that the phase-transformation tem- tion-induced plasticity.[24] perature could be enhanced dramatically by adding Zr. It has Sufficient ductility is one of the necessary conditions for been reported that the phase transformation temperatures SMAs to exhibit excellent shape memory characteristics. As is (M , A , A ) increase up to 240 C, as with Hf addition, with known, the memorised strain or the recoverable strain is s s p Zr contents up to 30 at.-%. The microstructure, thermomecha- essentially determined by the martensite shape strain. If an nical behavior and shape memory properties of Ni-Ti-Zr- alloy is brittle, this kind of potential strain cannot be based alloys are very similar to Ni-Ti-Hf-based alloys.[19] exploited. Ni-Ti-based alloys are known to be exceptionally ductile. However, they tend to be brittle when alloyed with Zr, Hf, Au, Pt and Pd. Consequently, when designing high 2.3.4 Nickel-Titanium-Iron System temperature shape memory alloys in a Ni-Ti-X system a com- Contrary to the addition of Niobium, the addition of Iron promise has to be made: to obtain a Ms-temperature as high to binary Ni-Ti system, results in a dramatic decrease in as possible and meanwhile to keep the alloy sufficient ductile. transformation temperature, and a considerable decrease of From this point of view, Ni-Ti-Hf-based alloys appear to be hysteresis width. Furthermore, the addition of Fe provokes much more attractive. suppression of the martensitic phase to the favour of R-phase Shape memory properties: The shape recovery ratio R for transition and increasing ductility results in good cycle stabil- different tensile strain is listed below. The tensile deformation ity. This alloy composition thus has excellent properties for temperature is 80 C. It can be seen that the magnitude of the actuator applications. It was shown that a decrease in trans- strain ep greatly influences the shape recovery ratio R. A com- formation temperature is achieved with rising substitution of plete shape recovery is observed until about 3 % strain. With Ni by Fe.[22] further increase of the strain, the shape recovery ratio gradually decreases (see Table 3). In order to test whether the 2.3.5 Nickel-Titanium-Niobium System experimental alloy exhibits pseudoelasticity the specimen

was strained slightly above Af. No pseudoelasticity was ob- The Ni-Ti-Nb-based alloys are mainly advantageous for served, when the specimen was cycled. pipe-joint couplings, due to its wide hysteresis. It was found to have sufficient mechanical and fatigue strengths to be used as a structural element at elevated temperatures in water. Table 3. The shape recovery ratio R for different strains ep: A detailed study of a Ni-Ti-Nb-based alloy with a chemical composition of 51 wt.-% Ni, 38 wt.-% Ti, and 11 wt.-% Nb e % 2.01 2.93 3.51 4.03 4.97 6.01 p, was conducted at the Mechanical Engineering Research Lab. R,% 100 100 97.8 95.5 87.2 83.1 in Hitachi, Japan.[25] The following principle properties were determined:

444 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys REVIEWS 35 700

30 Martensite 600 Austensite 25 500 ) 20 400

15 300 Stress (MPa Elongation (%) 10 200 Elongation 5 100 Stress

0 0 274 372 449 474 494 520 554 594 Fig. 10. Temperature dependence of the r0.2 stress and the Temperature (K) elongation for the Ni49Ti36Hf15 alloy [24].

The shape-recovery finishing temperature at a pre-strain N = 100 cycles to appear the R-phase transformation. Ti52- of 12 % was around 77 C. Ni47Al alloy exhibits good shape recovery, which can reach

Young's modulus increased with the temperature, up to 50 % at the Af-temperature and gradually increase to 80 % at 400 C. 300 C ageing temperature. His shape recovery is much better

The fatigue strength at 288 C in air and in water was than that of Ti51Ni49 alloy below 130 C, but exhibits the same slightly higher (about 1.2 times) than at 20 C in air. behavior above 200 C. The shape recovery of this alloy in- The martensitic and austenitic transformation tempera- creases during the ageing: it can reach 90 % for the specimens tures (Ms, Mf, As, Af) differ, as usual for SMAs, between aged at more than 60 h at 400 C. before and after shape recovery. Those after shape recovery are listed as:

Ms Mf As Af 3 Ni-Ti-Cu System ±111 C ±158 C ±76 C ±33 C It is possible to substitute copper for nickel in an equiatomic Ni-Ti-based alloy. This ternary alloy system is of partic- The tensile strength and r proof stress increased with 0.2 ular interest. As copper is a neighbour of Ni in the periodic the temperature from ±60 to 20 C, they were almost constant table, it readily changes the properties of a Ni-Ti-Cu-based between 20 and 288 C, and they decreased with the increas- alloy up to a concentration of 30 at.-% Cu. In contrast to addi- ing temperature at above 288 C. Therefore three temperature tions like Co, Fe or Cr, Cu maintains the same austenitic regions can be characterised for the mechanical strengths of phase temperature, instead of depressing it to lower tempera- this Ni-Ti-Nb-based alloy: below 20 C, between 20 and tures. Beneficial modifications of Cu are in particular a nar- 288 C, and above 288 C. rower hysteresis and lower martensitic yield strength. Figure 11 shows the effect of overdeformation of a Ni-Ti- Nb-based alloy. It can be mentioned that a total strain exceed- ing 8 % of total strain, results in a considerable loss of recovered strain. Furthermore, the recovery temperature rises 3.1 Crystal Structure of Ni-Ti-Cu-based Alloys nearly constantly with deformation strain up to 14 %.[26] The NiTi phase has a limited solubility for either Ni or Ti. The Cu-Ti binary system has a phase diagram very similar to that of the Ni-Ti system. However the austenitic phase of 2.3.6 Nickel-Titanium-Aluminium System NiTi has a B2 cubic structure, whereas the CuTi compound is [27] An investigated Ti52Ni47Al1 alloy has lower transforma- tetragonal. As a consequence, more than 30 at.-% Cu will tion temperatures compared to those of the binary Ti51Ni49 result in the presence of both tetragonal and cubic phases. alloy, but higher than those of equiatomic or Ni-rich Ti-Ni When Cu is substituted for Ni in Ni-Ti-based alloys, the over- alloys. They decreased further with an increased ageing time all Ti concentration must remain close to 50 at.-%, in order to at 400 C. The R-phase transformation seemed to be more eas- yield a single-phase material desirable for shape memory ily promoted by both cold rolling and thermal cycling in this applications. alloy. R-phase transformation appeared within the ten first In this alloy system the formation of the premartensitic cycles, compared with Ti51Ni49 alloy, which needs more than rhombic R-phase does not occur. The R-phase effect, taking

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 445 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

8 80

7,5 70

7 60

REVIEWS 6,5 50

6 40

5,5 30 Recovery Strain (%)

5 Recovery Temperature20 (°C)

4,5 10

4 0 46810121416 4 6 8 10121416 Total Strain (%) Deformation Strain (%)

Fig. 11. The effects of overdeformation on the recovery strain and the recovery temperature (As) of Ni-Ti-Nb.[26]

place nearly without a hysteresis, is of interest where accu- concentrations of Cu does not change the Ms-temperature racy and cycle stability is recommended. But the suppression significantly.[4] The addition of copper results in various

of R-phase in Ni-Ti-Cu system is considered to be advanta- alloys, all with approximately the same Ms-temperature or a geous for electrically actuated SMA wires. While heated un- delicate increase, and reduced A-temperatures, so that der constant load, hysteresis of Ohm-resistance of an actuator transformation hysteresis is much smaller: about 25 C at wire is noticeably lowered by the absence of the R-phase tran- 10 at.-%Cu facing 48 C at 3 at.-% Cu[30] see Figure 12. For sition.[5] actuator applications, amounts of about 10 wt.-%Cu substi- Ternary alloys with copper amounts about 7 at.-% trans- tuting Ni are used. The range of possible transformation form in a one-step and alloys containing more than 8 at.-% temperatures is placed between ±100 C and +100 C, having Cu transform in a two-step fashion: a hysteresis of 20±25 C.[5] below 7 at.-%Cu: B2 (cubic structure) ® B19¢ (monoclinic) above 8 at.-%Cu: B2 (cubic structure) ® B19 (orthorhom- bic) ® B19¢ (monoclinic) 3.3 Influence of Cu on Phase Transition Properties The martensitic B2 ® B19 phase appears with a strong transformation peak, which is favourable for actuator appli- Yield strength: Another property influenced by copper cations.[5] additions is the yield strength of the martensite, i.e. the stress level at which the twins will re-orient. The yield strengths of binary Ni-Ti-based alloys and a Ni-Ti-10 %Cu alloy were tested in both the martensitic and austenitic phases (see 3.2 Influence of Cu on Transformation Temperatures Table 4).[29] and Hysteresis Although processed to provide similar austenite yield As mentioned in chapter 2, the binary Ni-Ti system is very strengths, the martensitic strength of the Cu containing alloy sensitive to delicate variations in the stoichiometric composi- is almost half that of the binary. This can be important for tion. A deviation of 0.1 at.-% may effect a change of transfor- cyclic applications. One way of redeforming a device when it mation temperature of 10 C. Thus, depending on the desired cools back to the martensitic phase is to provide a reset

Ms-temperature, the precise composition control of up to one spring. On heating, the SMA can do useful external work, but tenth and one hundredth of a percent is necessary.[4] How- some of the available energy must be stored in this reset ever the sensitivity of the transformation temperatures is no- spring. Subsequent cooling of the SMA will cause it to soften ticeably reduced in Ni-Ti-Cu system. Compared to the binary when transforming to martensite, and if designed correctly systems, the alloy composition is controlled less closely. It al- the reset spring will then apply sufficient force to return it to lows for easier production of commercial quantities of materi- its original low temperature position. The softer martensite of [28,29] al having a controlled Ms±temperature for actuator use. the ternary alloy requires a lower force, so less work is dissi- The addition of even small concentrations of many third pated per cycle. The larger the strength differential between elements to Ni-Ti-based alloys results in a large change in the austenite and martensite phases, the larger the work that

the Ms-temperature. In contrast, substitution of even large can be done by the SMA during recovery.

446 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys REVIEWS 60 80

50 75

40 70

(°C) NiTi f 30 65 NiTiCu Ms (°C) -M f A 20 60

10 55

0 50 1234567891011 010203040 Cycle No. At.-% Cu Fig. 13. The Ms-temperature is less sensitive to transformation cycling for the ternary Fig. 12. Influence of Copper on hysteresis width (Af -Mf) in Ni-Ti alloys. [30] (10 %Cu) alloy than for the binary alloy.[31]

Cycling stability and fatigue: The stability of the properties shape memory strain depends on plastic pre-strain. The of SMAs during cycling is a major concern. A reliable compo- TWME (etw) in three types of alloys is plotted as a function of nent should exhibit a constant Ms independent of cycling. plastic strain in Figure 15. In Ti-50.0 at.-%Ni, for example, the

Secondly, the amount of actually recovered shape change (or TWME increases with increasing ep until reaching 12 % plas- strain) should be independent of the number of cycles. Any tic strain and decreases with further increasing ep. The strain inability of a shape memory component to retain these con- etw shows a maximum of 6 % at 10 % plastic strain in Ti- stant properties can be defined as ªFatigueº. 45 at.-%Ni-5 at.-%Cu and a maximum of 5 % at 7 % plastic

The cyclic dependency of the Ms-temperature of annealed strain in Ti-40 at.-%Ni-9 at.-%Cu. binary and ternary (10 % Cu) alloys are shown in Figure 13. Besides the fact that TWME of Ni-Ti-Cu-based alloys de-

As indicated, the Ms-temperature of the binary increases by pends on the amount of pre-strain, it has been shown that around 20 C during the first 10 thermal cycles. A further TWME also depends on alloy composition. Figure 16 shows cycling does not significantly change Ms-temperature. The the maximum TWME observed in Ti-Ni-Cu-base alloys alloys ternary alloy, however, exhibits an Ms-temperature, which is with Cu-contents of 0 % to 10 %. The strain etw of Ti-Ni-Cu- relatively constant to cycling.[31] based alloys in this investigation is about twice as large as A comparison of the cyclic dependence of the recoverable that of the Ti-Ni binary alloy. This data could mean that Cu strains of the binary and ternary alloys obtained during the addition is effective for increasing the two-way shape memo- transformation due to an applied load is shown in Figure 14. ry strain. Seen from a critical point of view, no cycle stability Data is shown for two stress levels, with the strains of the has been mentioned in this investigation, which is important binary alloy decreasing rapidly with the number of cycles. for any application. Conversely, Stöckel[5] indicates that the However, the values for the Ni-Ti-Cu-based alloy remained relatively constant. The reasons for these differences in cyclic behavior are not clear. One possibility is associated 5 with an incomplete cycle. Heating was terminated at 100 C, NiTi which was below Af±temperature for the binary. Thus some NiTiCu martensite would be retained in this case, but not in the 172,4 MPa ternary. On the other hand, M ±temperature for the Ni-Ti-Cu- f 4 275,8 MPa based alloy is well below Ms±temperature and cooling to room temperature does not complete the martensite transformation for the ternary alloy. This is one reason why the initial cycle strain is less.[29] 3 Transformation Strain (%) 3.4 Effect of Cu Addition on Two-Way Shape Memory Effect in Ni-Ti-based Alloys 2 The TWME is also affected by adding Cu as a third ele- 1 10 100 1000 10000 ment in Ni-Ti-based alloys. An investigation was conducted Cycle No. at the Institute of Materials Science, University of Tsukuba, Fig. 14. The recoverable strain exhibits less fatigue due to transformation cycling for Japan, to clarify this effect.[32] It is clear that the two-way the ternary (10 %Cu) alloy than for the binary alloy [31].

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 447 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

Table 4. 7

6 Ni-Ti-based alloy Ni-Ti-10%Cu

Martensite (tested at 25C) 208 MPa 106 MPa (%) 5 tw

REVIEWS Austenite (tested at 200C) 1053 MPa 1177 MPa 4

2 TiNiCu with Cu<7,5at%

maximum reachable TWME is reduced by the addition of Max. of Elongation TiNiCu with Cu>7,5at% copper. In the Ni-Ti-10 %Cu-alloy a TWME of less than 1 % 1 after 10 cycles has been measured. The substitution of Ni for Cu in the Ni-Ti-based alloys results in minor structural 0 changes and numerous modifications of the shape memory 024681012 properties. However, the basic SME is observed in alloys con- Cu-content (at%) taining up to around 30 % Cu. Some advantages of adding Cu Fig. 16. Effect of Cu content on the maximum of the two-way shape memory strain [32]. are a narrower transformation hysteresis and a more constant

Ms-temperature, less dependent on concentration variations and cyclic fatigue. Furthermore, the Cu-containing ternary alloy also has a martensite phase with lower yield strength, 4.1 Influence of the Applied Load upon the Af-Temperature thereby requiring a lower reset force for some cyclic applications. Therefore, alloys with about 10 % Cu seem to establish In order to examine the influence of the loading on the as a favourable standard for actuator applications. phase change behavior while heating, wires with a length of 290 mm were cycled in a test facility. The wires were fixed with six screws at each end in two special grippers and had a free length of 250 mm. One of the grippers was connected on 4 NiTiCu Wires its other side with a steel wire. Over two rolls the steel wire Several experiences have been made by the authors with was led from a basket into a bath filled with an 80 % glycer- NiTiCu-wires from the AMT-company (Herk-de-Stad, Bel- ine-water-mixture, in which the NiTiCu-wire was submerged gium) with the following characteristics:[33] Round wires with in entirely. The other gripper was fixed in this bath. So it was 0.43 mm diameter, ªtrained with the one-way-effect with possible to charge the wire with different loadings by filling the weights in the basket. Then the glycerine-water-liquid effect stability (incomplete two-way-effect)º with an Af at approx. 75 C in the unloaded condition and a maximum was heated up very slowly to 160 C and the deformation was deformation of 3.4 %. measured by means of an inductive way sensor at the chang- ing height of the basket. Af-temperature was measured ac-

cording to Figure 2. The dependence of this Af-temperatures on preload has been plotted in Figure 17. A mathematical ex- 6 pression can be derived considering a linear relationship: Ti-45Ni-5Cu 5 Ti-40Ni-9Cu Af = 61.9 C + F 1.71 C/N (1) Ti-50Ni 4

(%) where F is the applied preload in N. tw 3 Taking into account the wire cross section, the inverse slope can be expressed as 4 MPa/K, which is similar to other [7]

Elongation reported values for binary Ni-Ti-based alloys. 2 It has also been possible to produce a three dimensional graph considering all the measured data e, T, L (Figure 18). 1 With such a diagram it is possible to determine the maximum percentage contraction of this wire in dependence of certain 0 values of the parameters load and temperature. Values of 0102030 maximum recoverable strain are also dependent on preload. Plastic strain (%) In the range of tested preloads (112 to 461 MPa) maximum Fig. 15. Effect of plastic strain on the maximum of two-way shape memory strain in Ti-Ni-Cu-based alloys [32]. recoverable strain of between 3.5 to 4.5 % can be obtained.

448 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

After these experiences several heat treatments were car- REVIEWS Load Dependency of the Af Temperature ried out on the NiTiCu-wires in order to examine the influ- 180 ence on the phase change temperatures and to explain the 160 poor results obtained with the as-supplied material. In the 140 first group of tests wires without any heat treatment were 120 examined at a constant temperature of 75 C at a constant 100 80 stress of 112 MPa. Then the wires were objected to the follow- 60 ing different additional treatments in a tempering furnace: Tem perature (°C ) f 40 ± Heat treatment of 150 C/10 s A 20 ± Heat treatments of 150 C/10 s and 325 C/10 s 0 ± Heat treatments of 150 C/10 s and 325 C/10 s and 350 C/ 0 1020304050607010 s. Load (N) The results showed that the heat treatments at 150 C/10 s and 325 C/10 s lead to the best behavior concerning the con- Fig. 17. Load dependence of the Af-Temperature. traction. The brief overheating of the wire considerably be-

yond the Af-temperature leads to 100 % austenitic phase. After cooling the specimen, the phase change temperatures 4.2 The Influence of Heat Treatments on the Phase Change are in the areas that are determined by the composition of the Behavior alloy. (No remaining effect of this heat treatment could be observed. The heat treatment with the highest temperatures did In order to examine the influence of various heat treat- not give as good results concerning the contraction. It may be ments on the phase change behavior on the above mentioned suggested, that damage to the lattice was already in prog- NiTiCu-wire many tests were carried out at our own labora- ress). tories. In the second group of tests the wires were treated with First of all, the as-supplied wires were tested in order to various heat treatments and then thermally cycled in the assess the relationships between contraction, preload and water-bath with a constant stress of 112 MPa. So it was possi- temperature. Loaded with different weights the pre-cooled ble to plot the hysteresis graph with all the temperature specimens without any prior heat-treatment were submerged points of the phase changes and a comparison to the wire in the glycerine-water-liquid at different constant tempera- without any heat treatment could be made. After every test tures. It could be seen that at constant temperature a higher the specimen was taken out of the test plant and heat-treated load leads to smaller contractions and that at constant load a with the specification of the next step of the heat treatment. higher temperature leads to bigger contractions, see Fig- The examined steps were: ure 19. It also can be noticed, that the contraction (emax = 0.4 %) of the tested wires were far away from the expected ± Without any heat treatment values (emax/unload = 3.4 %, as certified by the supplier). ± Heat treatment of 150 C/10 s

Force-/Temperature-Dependency of the Contraction

4,5

3,5

2,5

1,5 Contraction [%] 4-4,5 1 3,5-4 3-3,5 0,5 2,5-3 0 2-2,5 154 1,5-2 144 134 124 1-1,5 114 104 0,5-1

F=16.26[N] 94 84 0-0,5

F=25.30[N] 74 64 Temperature [˚C] 54 Load [N] F=34.36[N] 44 F=43.39[N]

F=56.91[N] Fig. 18. Force-temperature-dependency of the contraction

F=65.94[N] translate.

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 449 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

Contraction over the Load at Different Temperatures (Untreated) This means that e.g. by holding a piece of metal during a defined period of time at a certain temperature the same 0,45 results can be obtained as by holding the same piece of metal 0,4 T=65°C during a longer period of time at a lower temperature. This 0,35 T=70°C substitution is possible in both directions. T=80°C REVIEWS 0,3 This effect is also found in the NiTiCu-wire examined in

0,25 our tests: the relatively high temperature during the short heat treatments in the hardening furnace was replaced by the 0,2

Contraction (%) low temperature in the bath combined with the long holding 0,15 time, caused by the slow heating. It was therefore not very 0,1 surprising, that all the graphs plotted at constant temperature 0,05 showed exactly the same contraction value, regardless of the

0 applied heat treatment. In other words: already the first step 010203040506070of heat-treatment is good enough to produce a recovery of Load (N) the SME in the expected range. Fig. 19. Contraction versus load at different temperatures (as-supplied).

± Heat treatment of 200 C/10 s 5 Concluding Remarks ± Heat treatment of 250 C/10 s The effect of thermomechanical treatments on phase ± Heat treatment of 300 C/10 s change behavior of Nitinol alloys were reviewed. Some con- ± Heat treatment of 350 C/10 s cluding remarks are as follows: I) Depending on the details of the alloy and its prior ther- It has to be taken into account that in each test the same momechanical history the martensitic plateau resulting in a wire was used. So the wire was not only subjected to the heat stress-strain test can vary from a continuous curve to a clear treatment of the actual step, but also to all of the previous horizontal plateau. Also the length of the plateau, what af- heat treatments. Figure 20 shows the contraction versus the fects the heat recoverable strain, is affected. load in dependence of the temperature. The results show Increasing the cold work reduces the heat recoverable clearly that a heat treatment leads to a considerably improved strain, but increases the yield strength. maximum contraction, in this case of 4.75 %. To explain this On the other hand, a prior annealing treatment reverts the

improvement it can be supposed, that the supplied wires, above mentioned phenomena, increasing Ms temperature but probably stored for a long time without any thermal cycling, reducing the mechanical stability below tens of thousands have experienced an evolution in their microstructure in such cycles. a way, that martensitic phase was estabilized. This is a well Concerning training methods of the TWME five thermo- known phenomena in material science, e.g. in the quenching mechanical training treatments are used. of steels. In material science it is also known that in these II) An extensive review of different additions to Ti-Ni base fields as the treatment of ageing the factor ªtemperatureº can compositions was presented, stressing the main differences be replaced by the factor ªtimeº multiplied by certain factors. observed in SME properties.

Contraction over the Load at Different Temperatures (Heat Treated)

5 4,5 4 3,5 3 2,5 2

Contraction (%) 1,5 1 0,5 0 0 20406080100120140 Load (N) Fig. 20. Contraction versus load at different temperatures T=65°C T=70°C T=75°C T=80°C T=85°C (heat-treated).

450 ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 Gümpel et al./Phase Change Behavior of Nitinol Shape Memory Alloys

III) For Ni-Ti-Cu-based own experimental work showed [16] K. H. Wu, Z. Pu, H. K. Tseng, F. S. Biancaniello, in REVIEWS an increase in the applied stress of 4 MPa results in Af critical SMST-94: Proceedings of the First International Conference temperature increase of 1K. on Shape Memory and Superelastic Technologies, Pacific Short annealing treatments at intermediate temperatures Grove, CA, USA; 1994, pp. 61±66. (from 150 C/10 s to 350 C/10 s) lead to the best value con- [17] K. H. Wu, L. Sun, in SMST-94: Proceedings of the First In- cerning contraction. Nevertheless these very short treatments ternational Conference on Shape Memory and Superelastic showed no signifficant differences in other properties. Technologies, Pacific Grove, CA, USA; 1994, pp. 67±72. [18] Y. Q. Wang, W. Cai, W. H. Zhu L. C. Zhao, in SMST-97: Received: July 06, 2001 Proceedings of the Second International Conference on Shape Final version: February 23, 2002 Memory and Superelastic Technologies, Pacific Grove, CA, ± USA; 1997; pp. 77±82. [1] Shape Memory Applications, Inc.; Webpage: www.- [19] Y. Gao, Z. J. Pu, K. H. Wu, in SMST-97: Proceedings of sma-inc.com, 2380 Owen Street, Santa Clara, CA 95054, the Second International Conference on Shape Memory and USA. Superelastic Technologies, Pacific Grove, CA, USA; 1997; [2] D. E. Hodgson; Shape Memory Applications, Inc.; Ming pp. 83±88. H. Wu, Memry Technologies; and Robert J. Biermann, [20] A. Shelyakov, A. Gulyaev, P.Potapov, E. Svistunova, Harrison Alloys, Inc.; ªShape Memory Alloysº; Web- D. Hodgson, N. Matveeva, J. Cederstrom, in SMST-97: page: www.sma-inc.com. Proceedings of the Second International Conference on Shape [3] J. Van Humbeeck; Adv. Eng. Mater. 2001, 3, 837. Memory and Superelastic Technologies, Pacific Grove, CA, [4] F. Fischer, Thesis, Escuela Superior de Ingenieros Uni- USA; 1997; pp. 89±94. versidad de Sevilla, 2000. [21] P. G. Lindquist, C. M. Wayman, in Engineering Aspects [5] D. Stöckel, in Legierungen mit Formgedächtnis (Eds: of Shape Memory Alloys (Eds: T. W. Duerig, K. N. Mel- D. Stöckel, W.J. Bartz), Kontakt und Studium, Band 259, ton, D. Stöckel, C. M. Waymann), Butterworth-Heine- Expert Verlag, Ehningen (Germany) 1988, pp. 42-50. mann, London 1990, pp. 58±68. [6] M. Mertmann, in VDI-Forschungsberichte Reihe 5, Ni-Ti- [22] J. J. Moore, H. C. Yi, in Shape-Memory Materials and Phe- Formgedächtnislegierungen für Aktoren der Greifertechnik, nomena ± Fundamental Aspects and Applications (Eds: 1996, pp. 9±13. C. T. Liu, H. Kunsmann, K. Otsuka, M. Wuttig), Materi- [7] K. N. Melton; in Engineering Aspects of Shape Memory al Research Society (MRS); Pittsburgh; USA; 1992; Alloys (Eds: T. W. Duerig, K. N. Melton, D. Stöckel, pp. 331±336. C. M. Waymann), Butterworth-Heinemann, London [23] K. H. Wu, L. Sun, in SMST-94: Proceedings of the First In- 1990, pp. 21±34. ternational Conference on Shape Memory and Superelastic [8] E. Hornbogen, in Legierungen mit Formgedächtnis (Eds: Technologies, Pacific Grove, CA, USA; 1994, pp. 67±72. D. Stöckel, W.J. Bartz), Kontakt und Studium, Band 259, [24] Y. Q. Wang, Y. F. Zheng, W. Cai, L. C. Zhao; Skripta Expert Verlag, Ehningen (Germany) 1988, pp. 14±16. Mater. 1999, 40, 1327. [9] S. M. Russel, D. E. Hodgson, F. Basin, in SMST-97: Pro- [25] O. Oyamada, K. Amano, K. Enomoto, Int. J. Solid Me- ceedings of the Second International Conference on Shape chanics Mater. Eng. 1999, 42, 243. Memory and Superelastic Technologies, Pacific Grove, CA, [26] T. W. Duerig, K. N. Melton, J. L.Proft, in Engineering As- USA; 1997; pp. 429±436. pects of Shape Memory Alloys (Eds: T. W. Duerig, K. N. [10] J. Perkins, D. Hodgson, in Engineering Aspects of Shape Melton, D. Stöckel, C. M. Waymann), Butterworth-Hei- Memory Alloys (Eds: T. W. Duerig, K. N. Melton, nemann, London 1990, pp. 130±136. D. Stöckel, C. M. Waymann), Butterworth-Heinemann, [27] S. F. Hsieh, S. K. Wu, J. Mater. Sci. 1999, 34, 1659. London 1990, pp. 195±206. [28] O. Mercier, K. N. Melton, Met. Trans. 1979 10A, 387. [11] S. A. Shabalovskaya, Proc. ICOMAT '95, J. Phys. IV 1995, [29] W. J. Moberly, K. N. Melton, in Engineering Aspects of C8, 1199. Shape Memory Alloys (Eds: T. W. Duerig, K. N. Melton, [12] D. Mantovani; JOM 2000 (10), 36. D. Stöckel, C. M. Waymann), Butterworth-Heinemann, [13] S. M. Russel, F. Sczerzenie, in SMST-94: Proceedings of London 1990, pp. 46±57. the First International Conference on Shape Memory and [30] H. Funakubo, Shape Memory Alloys, Gordon and Breach Superelastic Technologies, Pacific Grove, CA, USA; 1994, Sci. Pub. 1987. pp. 43±48. [31] J. L. Proft, K. N. Melton, T. W. Duerig: Proc. Of the Inter. [14] S. M. Tuominen, in SMST-94: Proceedings of the First In- MRS Conf., Tokyo, Japan 1987. ternational Conference on Shape Memory and Superelastic [32] S. Miyazaki, S. Chujo, in SMST-94: Proceedings of the Technologies, Pacific Grove, CA, USA; 1994, pp. 49±54. First International Conference on Shape Memory and Super- [15] J. H. Mulder, J. Beyer, P. Donner, J. Peterseim, in SMST- elastic Technologies, Pacific Grove, CA, USA, 1994, 94: Proceedings of the First International Conference on pp. 73±77. Shape Memory and Superelastic Technologies, Pacific [33] S. Glaeser, Diploma Thesis, Fachhochschule Konstanz, Grove, CA, USA; 1994, pp. 55±60. 1996.

ADVANCED ENGINEERING MATERIALS 2002, 4, No. 7 451