TWO-WAY SHAPE MEMORY EFFECT AND ITS DEGRADATION DURING THERMAL CYCLES IN Ni-Ti
ALLOYS
DVOSMERNI SPOMINSKI EFEKT V Ni-Ti ZLITINAH IN NJEGOVA DEGRADACIJA MED TOPLOTNIMI CIKLI
HEINRICH SCHERNGELL, A. C. KNEISSL
Institute of Physical Metallurgy and Materials Testing, University of Leoben, A-8700 Leoben, Austria
Prejem rokopisa - received: 1997-10-01; sprejem za objavo - accepted for publication: 1997-10-21
This work presents a study of the degradation of the Two-Way Shape Memory Effect (TWSME) due to working cycles. An intrinsic TWSME was induced in wire specimens of a near equiatomic Ni-Ti al!oy by thermomechanical training. The deveiopment of the two-way strain vvas analysed and discussed with respect to the different training parameters and the preliminary heat treatment of the samples. Subsequent to the training, the stability of the obtained TVVSME was examined by thermal cycling under stress-free conditions. Possible reasons and mechanisms for the degradation of the effect during thermal cycling are discussed.
Key words: shape memory, two-way-effect, Ni-Ti alloy, stability, martensitic transformation, fatigue
V delu se predstavlja raziskava degradacije dvosmernega spominskega efekta (TWSME) zaradi delovnih ciklov. Intrinsičen TWSME je bil induciran v žice iz equiatomske Ni-Ti zlitine s termomehanično obdelavo. Razvoj dvosmerne deformacije je bil raziskan in ocenjen glede na obremenilne in predhodno toplotno obdelavo vzorcev. Po obremenitvah je bila stabilnost dosežene TWSME preiskana s toplotnim cikliranjem brez napetosti. Razpravlja se o možnih vzrokih in mehanizmih degradacije efekta med termičnim cikliranjem.
Ključne besede: oblikovni spomin, dvosmerni efekt, zlitina Ni-Ti, stabilnost, martenzitna premena, utrujenost
1 INTRODUCTION
The term Shape Memory refers to the ability of certain alloys, to recover large strains perfectly either right after unloading (pseudoelasticity) or after heating (pseudoplasticity). Several alloys can show this phe-nomenon as a consequence of martensitic transformation1. In the čase of the intermetallic phase Ni-Ti, sueh a transformation occurs between the high temperature modification (austenite) vvith a CsCl-type B2 superlattice and the lovv temperature phase (martensite) vvith a mono-clinic BI9' strueture2. Furthermore, a TWSME may be obtained in these alloys after a suitable thermomechanical treatment, vvhich is often termed training. This spe-cial behaviour refers to the ability to produce spontane-ous shape changes on heating as vvell as on cooling, both vvithout the application of external forces. Consequently, trained elements can directly be used as temperature-sen-sitive actuators. The actuators are usually electrically ac-tuated and cooled by natural convection. One important eriterion for the aetual use of Shape Memory compo-nents in sueh multiple cycle applications is the stability of the funetional parameters throughout application. A reliable component should exhibit constant transforma-tion-temperatures and a two-way strain independent of number of cycles. In this vvork, the stability of an intrinsic TWSME during repeated thermal cycling has been investigated.
2 EXPERIMENTAL
The investigations have been carried out on a binary Ni-Ti alloy vvith a nominal composition of Ni-50,3 at% Ti. The material vvas melted from pure metals in an arc-furnace, hot extruded and cold-dravvn vvith intermediate annealings. The as-received vvire material vvas cold vvorked by 13,5% vvith a diameter of 3 mm. Wire specimens of 20 cm in length vvere prepared and some of them annealed at 550°C / 20 min. The transformation temperatures vvere deduced from resistivity measurements and are listed in Table 1. Ali transformation temperatures are above room temperature, vvith Ms, Mf indi-cating the start and finish temperature for the forvvard transformation, and As, Af deseribing the reverse transformation. The mechanical properties of the alloy vvere investigated by tensile tests carried out at room temperature. The obtained stress-strain-curves of the martensitic phase are illustrated in Figure 1. The training procedure vvas simulated on a tensile testing machine. The specimens vvere fixed at room temperature, loaded vvith the constant training stress a,rain and repeatedly thermally cy-cled betvveen Mf and a temperature above the highest temperature, at vvhich martensite can be stress-induced (M<j). Heating vvas done by Joules-Effect, cooling by natural convection in air. According to the applied training StreSS CTirain and number of training cycles Ntrain, different magnitudes of the TWSME vvere obtained. Some of the values are given in Table 2. Subsequently to the training, the magnitude of the TWSME vvas measured
strain [%]
Figure 1: Stress strain curves of the martensitic modification
and its degradation during repeated stress-free thermal cycling studied.
Table 1: Transformation temperatures obtained by resistivity measurements
HT MS(°C)		Mf(°C) AS(°C)	Ar(°C)
none 52		38 82	141
550°C 55		42 84	127
Table 2: TWSME achieved vvith different numbers of training cycles			
and heat treatments			
HT	^train	(MPa) Ntrain	£2W (%)
550°C	50	20	1,65
none	100	20	1,18
550°C	100	20	1,80
550°C	100	10	2,15
3 RESULTS AND DISCUSSION
It is very useful to study the strain temperature relationship in order to understand the shape memory behaviour. Figure 2 illustrates the strain-temperature-curves as they have been recorded by microstrain measurement during the training cycles. On cooling the loaded wire specimen from a temperature above Md, martensite is stress induced on reaching the Ms-temperature. Nor-mally, the Ni-Ti martensite consists of 24 variants, that are crystallographically equivalent3. But the applied stress causes a preferred nucleation and growth of those variants, that have the highest compatibility vvith the strain-field. This orientation of the martensite results in a macroscopic strain in direction of the applied stress, as can be observed in Figure 2. On heating, the alloy trans-forms back to the B2 structure of the austenitic modification, forced to restore the associated original shape. A careful look on Figure 2 shovvs, that the restoration does not succeed completely, but there is a small amount of irreversible strain. This irreversible strain consists as vvell of locally stabilised martensite as of true plastic strain4. During the grovvth of the martensite plates, stress
concentrations may arise on the B2/B19' interfaces, lo-cally exceeding the true plastic yield strength. Actually, this generation of dislocations during the ther-momechanical training procedure is a precondition for the development of the two-way effect. The generated dislocation pattern is a characteristic feature of the de-formed lovv-temperature shape, vvhich is repeatedly formed during training. On subsequent stress-free thermal cycling, these dislocations are effective in guiding the martensite formation, causing the martensite to form at least partly oriented. The resulting two-way strain e2w has the same direction as the reversible strain during the training, but it is smaller in magnitude, as the efficiency of dislocations in the formation of a preferential martensite morphology is vveaker compared to the stress field caused by an external force.
Furthermore, it can be observed in Figure 2 that the temperature interval (Af - Mf) increases throughout training. This is thought to be likevvise due to the increase of the dislocation density. The increase of the internal stress that is connected vvith it reduces the mobility of the B2/B19' interfaces. As the martensitic transformation in near equiatomic NiTi occurs by the way that the martensite plates grovv continuously on cooling and shrink on heating, the rise of frictional stress leads to an increase of Ar and a decrease of Mf.
The development of the TWSME can be observed best by studying the course of the different strains during training. Figure 3 shovvs the labelling of the different strains describing a training cycle. If one follovvs the course of these strains throughout training, it becomes evident, that a high temperature and a lovv temperature shape is being established. The pseudoplastic strain epp increases very fast during the first fevv cycles as a conse-quence of the stress-biased martensite formation. A cer-tain airain is able to orient a certain amount of martensite variants. Consequently, epp saturates after having reached a certain magnitude. To increase a,rain beyond a certain limit has no sense in the čase of a thermomechanical training procedure as one obtains excessive irreversible strain, leading to a deterioration of the high temperature
'i5
vi
temperature [°C]
Figure 2: Illustration of some training cycles
annealed specimen	- first heating
cM|n = 100 MPa	- 1. training cycle
............ 2. training cycle
................------- 8. training cycle
\	-----------20. training cycle
temperature -►
Figure 3: Schematic illustration of a training cycle indicating the labelling of the different strains: et: total strain, £pp: pseudoplastic strain, ep: irreversible strain
Number of training cycles
Figure 4: Development of the different strains during training, annealed specimen, 50 MPa
shape. It has been shovvn by other authors that generally those training parameters (combination of training stress and number of training cycles) give the highest two-way memory, vvhich result in the least permanent strain, vvhile producing a full one-way strain5. The maximum transfor-mation strain for a certain heat-treatment condition may be estimated from the extension of the martensite plateau in the stress-strain curve in Figure 1. As can be observed from Figure 4, 50 MPa is not enough to take advantage of the vvhole orientation capacity. The epp in this plot is considerable smaller than the extension of the martensite plateau in Figure 1, vvhich reaches about 4%. It is evi-dent from Figure 5 that an increase of atrain to 100 MPa leads to a epp of the aimed magnitude. It is interesting to note in this plot that the pseudoplastic strain decreases after having passed a maximum value after about 4 training cycles. Apparently, the irreversible strain rises too fast in the čase of 100 MPa. This is quite important as it has been concluded elsevvhere5, that there exists a near proportionality betvveen the pseudoplastic strain and the obtained two-way strain. This suggests that in order to
Number of training cycles
Figure 5: Development of the different strains during training, annealed specimen, 100 MPa
Number of training cycles
Figure 6: Development of the different strains during training, cold vvorked specimen, 100 MPa
maximise e2w, fevver than 20 training cycles should be applied. This is confirmed by the values listed in Table 2. In the čase of atrain =100 MPa, the two-way strain obtained after 10 training cycles is larger than after 20 cy-cles. Figure 6 shovvs the course of the strains for the vvork hardened specimen. Due to the high internal stresses, the o[rain is not as efficient in biasing the martensite formation as in the annealed specimens. A pseudoplastic strain of only 2,5% is reached and it has neither saturated nor passed a maximum after 20 training cycles. It may therefore be concluded that epp would increase further, if training vvas continued.
In order to repeatedly achieve the same lovv temperature shape during stress-free thermal cycling it is essen-tial, that the transformation follovvs the same transforma-tion paths over and over again. The thermoelastic martensitic transformation occurs sequentially. Those martensite variants that have formed first, are the last to revert. A thermodynamic balance exists at ali times on the B2/B19' interfaces betvveen chemical and not chemical energy terms. The latter one is essentially the elastic
2,2 n
2,0-
annealed specimen
Training: onji = 100 MPa / NMin= 20
free thermal cycles
1,4-
300 600 900 1200 Number of thermal cycles
1500
1,6-
ST 1.4-
i CN
CJJ
1,2" 1,0-
cold worked specimen
Training: a,rain = 100 MPa / Ntrai„ = 20
free thermal cycles
300 600 900 1200 Number of thermal cycles
1500
Figure 7: Degradation of the TWSME during thermal cycling, Figure 8: Degradation of the TWSME during thermal cycling, cold annealed specimen	worked specimen
strain energy developed as the microstructural units of new phase nucleate and grow in the original phase6. By this energy term, the substructure can exercise some influence on the selection of variants, that are assisted in their nucleation and grovvth. In an asymmetric substructure, as it has been produced during training, this leads to the formation of a microstructure during cooling, that vvould normally correspond to a strained condition, ex-hibiting the associated strain e2w.
Figure 7 illustrates the change of the two-way strain due to thermal cycling. There is a rather strong decay during the first 50 to 100 cycles, vvhich gradually de-creases by further cycling. After about 500 cycles, the degradation rate becomes fairly small. Considering the importance of the asymmetric microstructure for the ap-pearance of the TWSME, it appears very likely, that the decrease of e2w is the consequence of changes in the substructure. Investigations by Miyazaki et al.7 shovved, that thermal cycling causes a rearrangement and annihilation of existing dislocations as vvell as the introduction of ad-ditional ones. Both processes change considerably the dislocation pattern that has been created during the training cycles and vvill therefore lead to an increasing amount of unfavourably oriented martensite. Some hints about the actually controlling mechanism may be obtained by comparing the degradation of the annealed specimen in Figure 7 to a cold vvorked one in Figure 8. While the initial decay cannot be avoided, the decrease in the latter stage shovvs a sensitivity from the initial dislocation density. In the annealed specimen, the two-way strain keeps degrading throughout the investigated range of 1500 thermal cycles vvhereas in the cold vvorked sample it stays almost stable. It may be suggested from this observation, that the degradation in this stage is primar-ily due to the introduction of dislocations. In the cold vvorked sample it is more difficult to generate additional dislocations as the dislocation density is already very high. This results in a more stable TWSME in the latter stage.
On the other hand, the sharp decay at the beginning of thermal cycling appears to be not affected by a higher dislocation density. Moreover, it seems to be a transient phenomenon vvhich is limited to the first 50 to 100 cy-cles, reaching a saturation at the end of the initial period. It appears very likely that the decrease in this range cor-responds to a continual change in the dislocation substructure - by rearrangement and annihilation - until a stable configuration, representative of the saturated state, is reached.
It should be further noted, that even on variation of the training parameters, the characteristic features of the degradation curve remain more or less the same. It becomes evident, that for the applied training procedure, the initial microstructure of the alloy influences the sta-bility of the intrinsic TWSME far more than the training parameters.
4 CONCLUSIONS
With the chosen training procedure, a maximum two-way strain of 2,1% could be achieved. The induced intrinsic TWSME shovvs a sharp decay in the early stage of cyclic application. This is ascribed to the rearrangement and annihilation of dislocations during the first fevv thermal cycles until a new saturated state is reached. In the latter stage, the degradation rate is smaller and it shovvs a dependence from the preliminary heat-treatment. It is suggested that the controlling mechanism in this stage is the introduction of additional dislocations. For the prac-tical point of vievv it may be concluded from this vvork, that trained shape memory elements should be cycled several times before application until the mechanism, re-sponsible for the sharp decay in the initial stage, has been exhausted. The stability in the latter stage, espe-cially in cold vvorked specimens, should be suitable for technical applications. The preliminary heat-treatment appears to be the crucial parameter to optimise the TWSME regarding magnitude and stability.
ACKNOWLEDGEMENT
The authors gratefully acknowledge financial support of this work by the "Jubilaumsfond der Čsterreichischen Nationalbank".
3 S. Miyazaki, K. Otsuka, ISIJ Int., 29 (1989) 353 4R. Stalmans, J. van Humbeeck, L. Delaey. Acta metali, mater., 40 (1992) 501
5 Y. Liu, P. G. McCormick, Acta metali, mater., 38 (1990) 1321
6J. Perkins, Mat. Sci. Eng., 51 (1981) 181
7S. Miyazaki, Y. Igo, K. Otsuka, Acta metali, 34 (1986) 2045
5 REFERENCES
1	E. Hombogen, Pract. Met., 26 (1989) 279
2	J. Perkins, Metali. Trans., 4 (1973) 2709