Abstract
The reproducibility of thermally induced reversible martensitic transformation, occurring in a Cu–15Zn–6Al (mass-) shape memory alloy during heating, was studied on three different fragments of martensitic alloy subjected to complex thermal cycling comprising four series of five heating–cooling cycles, applied on a differential scanning calorimetry (DSC) device, between room temperature (RT) and three maximum temperatures: 453, 463 and 473 K respectively. After each series of heating–cooling cycles the fragments were naturally aged at RT for 37 days (∼3·2 Ms). Thermograms of DSC revealed a gradual diminution in the magnitude of martensite reversion to parent phase, accompanied by decreasing tendencies of both critical temperatures and transformation rates, which reflect a gradual loss of thermal memory, associated with ‘amnesia’ occurrence. Scanning electron microscopy observations coupled with energy dispersive X-ray spectroscopy performed at the end of complex thermal cycling revealed that ‘amnesia’ occurrence was associated with changes in the morphology of martensite plates, from parallel plate-like to interblocking needle-like, and with an increase of chemical fluctuations, as an effect of the intensification of atomic migration.
Keywords
Introduction
Polycrystalline Cu–Zn–Al based shape memory alloys (SMAs) gained commercial applicability due to their thermal memory behaviour, characterised by recovery strains and stresses as large as 4 and 400 MPa respectively.1 As compared to NiTi alloys, which are by far the most successful SMAs to date, Cu–Zn–Al alloys have lower cost, better electrical and thermal conductivities,2 and good workability in processing yet inferior shape memory properties.3 However, several drawbacks, caused by metallurgical processing,4 training5 or cycling,6 limited the development of the applications of Cu–Zn–Al based SMAs to single triggering devices such as couplings7 or high work output thermal actuators8 for fire protection9 or for temperature monitoring in freezing chambers.10 The main disadvantage in exploiting the applications of Cu–Zn–Al based SMAs, as functional materials, is diffusion controlled martensite stabilisation, which occurs on successive heating even in a low temperature range.11 This stabilisation phenomenon causes the shift of the critical temperatures for reverse martensitic transformation (As and Af) to higher temperatures.12 In a recent study, on a Cu–Zn–Al SMA subjected to thermal cycling with free air cooling, thermal memory degradation was associated with an alteration of martensite reversion to parent phase which experienced three variation tendencies, during repetitive heating:
to shift to lower temperatures becoming less intense, up to total extinction
to slow down more than 12 times in the first part and almost four times at the end of the process
to form lath martensite disposed as interlocking needles.13
The present paper aims to further analyse the behaviour, during complex thermal cycling, of a Cu–Zn–Al SMA which was developed as a potential candidate for electric actuators, under the form of either bending lamellas14 or long stroke lamellar helical springs.15 Considering that, due to diffusion controlled phenomena either intermediate or equilibrium phases, such as bainite16 or α-phase with specific orientation17 and morphology,18 were formed during martensite aging, the present experiments will comprise thermal cycles intercalated with prolonged natural aging at room temperature (RT).
Experimental
Lamellas of a Cu–15Zn–6Al (mass-, as all compositions will be expressed hereinafter) SMA, with typical thickness of 10−3 m, were obtained by induction melting, casting, hot rolling, instant water quenching, machining and homogenisation (1070 K/18 ks/water), according to a previously detailed procedure.19 After this thermomechanical processing, the structure of the alloy under study was fully martensitic and contained no noticeable precipitates, neither by optical13 or transmission electron microscopy nor by X-ray diffraction.16 From such a martensitic lamella, after careful mechanical removal under water cooling of any marks of superficial corrosion, three fragments weighing less than 50×10−6 kg each were cut, with appropriate caution in order not to alter their thermomechanical history. Each of the three fragments was subjected to several series of thermal cycles performed on a Netzsch calorimeter type DSC 200 F3 Maia with temperature accuracy of 0·1 K. The device was calibrated with Bi, In, Sn and Zn standards. The measurements were performed under Ar protective atmosphere using corresponding correction curves, for each thermal series. The thermal cycling series comprised:
heating from RT to a maximum temperature, with a rate of 1·67×10−1 K s−1
isothermal maintaining for 180 s
cooling to 300 K at a cooling rate of 8·33×10−2 K s−1, then the heating maintaining cooling sequence was repeated four times, in such a way that, in all, five thermal cycles were applied.
For the three fragments maximum heating temperature was 453, 463 and 473 K respectively. After a series of five thermal cycles, the fragments were removed from differential scanning calorimeter and stored at RT (natural aging) for 37 days, a period of time which was approximated as 3·2 Ms, as it will appear hereinafter. In all, each specimen fragment underwent four series of five thermal cycles intercalated with three periods of 3·2 Ms natural aging at RT. The differential scanning calorimetry (DSC) data, recorded during each heating of the thermal cycles were evaluated with Proteus software, provided by Netzsch, which enabled the use of tangent method for critical temperature determination.
At the end of DSC scans, each of the three fragments was encapsulated into Mécaprex KM-U cold mounting resign, before being ground up to 2400 mesh and automatically polished for 1·8 ks, on a Metkon Forcipol 1V machine, with 0·04μ alumine suspension. The three metallographically prepared specimens were analysed by scanning electron microscopy on a SEM VEGA II LSH (TESCAN) microscope, coupled with an EDX QUANTAX QX2 (ROENTEC) detector.
Results and discussion
The 20 DSC thermograms, corresponding to the heating stages within the four series of five thermal cycles applied up to 453 K, to the first specimen fragment, are illustrated in Fig. 1. It is noticeable that, except for the first series, from Fig. 1a, where the transformation tends to shift to higher temperatures, the thermograms from the second to fourth series (Fig. 1b–d) are quite similar. The variation tendencies of martensite reversion to parent phase, within these three series, are to shift to lower temperatures and to gradually decrease in intensity, in good accordance with the first of the three above mentioned variation tendencies reported in the previous study.13 Owing to the intercalation of the stages of 3·2 Ms RT natural aging, between each series of five thermal cycles, this tendency became more and more obvious, with increasing the number of aging periods. In other words, the tendency of reverse martensitic transformation to shift to lower temperatures and to decrease in intensity during thermal cycling, became more intense with increasing the total time of natural aging, in the order of the second, third to fourth series (Fig. 1b–d).
Heat flow variations during heating stages within four series of five consecutive thermal cycles up to 453 K, intercalated with periods of 3·2 Ms of natural aging
The second previously reported variation tendency, regarding the slowdown of reverse martensitic transformation during thermal cycling, has been verified by means of Fig. 2 which shows the derivatives of heat flow variation designated as Derivative Differential Scanning Calorimetry (DDSC) with temperature, corresponding to the 20 DSC charts from Fig. 1. In this case, as well, the decreasing tendency of martensite reversion rate became more intense with increasing the number of natural aging stages, in the order of the second, third to fourth series (Fig. 2b–d). Therefore, it seems that, with increasing the duration of natural aging, martensite reversion to parent phase, which actually represents the so called thermal memory of SMAs, tends to shift to lower temperatures to become less intense and to slowdown during thermal cycling performed up to 453 K.
Derivatives of heat flow variations during heating stages within four series of five consecutive thermal cycles up to 453 K, intercalated with periods of 3·2 Ms of natural aging
Up a certain degree, the evolution tendencies of DSC thermograms corresponding to the behaviour of the other two specimen fragments subjected to thermal cycling up to 463 and 473 K respectively, were similar to those discussed in Fig. 1. However, the change in the aspect of the DSC thermograms, corresponding to the series of thermal cycles performed up to 463 K, was much more severe, when comparing the first to the fourth series. Moreover, in the fourth series, the last three cycles experienced solid state transitions of very low intensity, which suggest that only a very small amount of thermally induced martensite reverted to parent phase during the heating of the third to fifth cycles. It has been obvious that reverse martensite transformation diminished much faster, with increasing maximum heating temperature. In the case of 473K cycling, reverse martensitic transformation disappeared completely in the third cycle of the third series and the degradation of the martensite reversion has been associated with ‘amnesia’ in contrast to thermal memory which was associated with thermally induced reversion of martensite to parent phase, during heating.20 The DSC thermograms corresponding to four series of consecutive thermal cycles up to 463 K and 473 K are displayed in Supplementary Material 1 and Supplementary Material 2 respectively: http://dx.doi.org/10.1179/1743284711Y.0000000099.S1 andhttp://dx.doi.org/10.1179/1743284711Y.0000000099.S2. Therefore, even if the thermal cycles proceeded, no evidence of solid state transition was noticeable on the DSC thermograms of the fourth series. Table 1 summarises all calorimetric data recorded during the four series of five thermal cycles, intercalated with periods of 3·2 Ms natural aging, performed up to three maximum temperatures, 453, 463 and 473 K respectively.
Summary of thermodynamic parameters determined on DSC charts recorded during four series of five thermal cycles, intercalated with 3·2 Ms natural aging
(1) Critical temperature for the start of martensite reversion to parent phase;
(2) Critical temperature for 50 martensite reversion to parent phase;
(3) Critical temperature for the end of martensite reversion to parent phase;
(4) Absorbed heat per unit mass;
(5) Variation rate, in time, of heat flow per unit mass during the first part (beginning) of martensite reversion to parent phase;
(6) Variation rate, in time, of heat flow per unit mass during the second part (end) of martensite reversion to parent phase.
It should be noted that the values of critical transformation temperatures of martensite reversion to parent phase, previously determined for the same alloy by measuring the variations of electric resistance versus temperature, were As = 381 K and Af = 407 K.17 All critical temperature values listed in Table 1 range within these limits. A closer look at the values of critical temperatures and transformation rates from Table 1 reveals the following variations tendencies:
all temperatures tend to increase after the first cycle and to decrease during subsequent cycles
during 453 K cycling, temperature decrease was so slow that, even after 20 cycles, the values of critical temperatures were still higher than those from the first cycle
during 463 K cycling, the critical temperatures decreased much faster so that after 20 cycles they became lower than those in the first cycle
during 473 K cycling, the critical temperatures decreased so intensely that, after 12 cycles, when ‘amnesia’ occurred, they were lower than initially
in any situation, the transformation rate decreased during cycling.
In order to better illustrate the effects of natural aging caused on thermal cycling behaviour of the three specimen fragments, Fig. 3 compares the DSC thermograms of the fifth cycle of each of the four series (designated as 1 to 4). Figure 3a shows that, in the case of 453 K cycling, martensite reversion to parent phase tends to shift to lower temperatures and to become less intense in a rather regular and continuous manner, in such a way that the distances between the four curves are almost equal. This suggests that each intercalation of 3·2 Ms natural aging altered thermal memory behaviour in a similar way. By comparing the evolution of the values of specific energy absorption in the fifth cycle, listed in Table 1, from the first to the fourth series, they varied from 4·327 to 4·331 kJ kg−1 (second series), then to 2·963 kJ kg−1 (third series) and finally to 2·512 kJ kg−1 (fourth series). By neglecting the small increase from the second series, it seems that, as compared to the first series, the intensity of martensite reversion decreased with ∼32 in the third series and with almost 42 in the fourth series. In contrast to Fig. 3a, Fig. 3b and c show a rapid degradation of thermally induced martensite reversion to parent phase. Thus, in the case of 463 K cycling, Fig. 3b illustrates marked decreases of the intensity of martensite reversion after the intervention of each additional 3·2 Ms natural aging stage. The corresponding values of specific energy absorption in the fifth cycle, from Table 1, were: 2·223 kJ kg−1 (first series), 1·103 kJ kg−1 (second series), 0·284 kJ kg−1 (third series) and 0·165 kJ kg−1 (fourth series), which suggest that, as compared to the first cycling series, martensite reversion intensity decreased with ∼50 in the second series, 87 in the third series and almost 93 in the fourth. This marked alteration of thermal memory behaviour confirms the completion of ‘amnesia’ in the fourth series of 463 K cycling. Moreover, in the case of 473 K cycling, ‘amnesia’ clearly occurred in the third series, after applying twelve thermal cycles, when martensite reversion was no longer noticeable.
Comparisons of heat flow evolutions during fifth heating cycle in each of the four series intercalated with periods of 3·2 Ms natural aging (1: first, 2: second, 3: third, 4: fourth) applied
In order to make a connection between ‘amnesia’ occurrence and the changes induced by complex cycling in the microstructure of the Cu–Zn–Al SMA under study, the three specimens fragments were analysed by SEM and the typical metallographic aspects are described in Figs. 4–6.
Images (SEM) of specimen subjected to four series of five thermal cycles up to 453 K, intercalated with periods of 3·2 Ms natural aging
Images (SEM) of specimen subjected to four series of five thermal cycles up to 463 K, intercalated with periods of 3·2 Ms natural aging
Images (SEM) of specimen subjected to four series of five thermal cycles up to 473 K, intercalated with periods of 3·2 Ms natural aging
Figure 4 illustrates the typical SEM micrographs of the fragment which underwent 20 thermal cycles to 453 K and a total period of 9·2 Ms natural aging. Figure 4a shows a diamond type structure commonly observed in quenched Cu based SMAs, comprising plate-like or spear-like parallel martensite plates.21 This means that, in spite of the calculated intensity decrease of martensite reversion, which reached almost 42, complex cycling to 453 K did not cause sensible changes in the structure of Cu–15Zn–6Al SMAs. Even in the detail of Fig. 4b, no evidence of martensite plate interlocking13 can be observed at the level of secondary plates which are contained in the primary ones.
The situation is a little different in the case of the fragment which was subjected to 20 thermal cycles to 463 K, as shown in Fig. 5. Figure 5a reveals the occurrence of long martensite plates overlapping on primary plates. The detail shown in Fig. 5b contains primary martensite plates and differently inclined secondary martensite needles, blocked by primary plates, which could be regarded as an effect of thermal cycling.21 Similar formations were encountered in the case of a quenched Cu–26·1Zn–4·8Al SMA, aged twice to 423 K. In this case, as an effect of aging, new martensite formed at an inclined angle to the existing martensite.22
As shown in Fig. 6, the above mentioned aspects are even more obvious in the case of the microstructure of the specimen fragment subjected to 20 thermal cycles to 473 K. Thus, martensite interlocking between differently oriented needle-like martensite formations is generally illustrated in Fig. 6a. The detail from Fig. 6b shows that most of the martensite plates were replaced by needles which have all almost the same width, being characteristic to lath martensite. This complete interlocking tendency, previously reported in the case of differently oriented lath martensite needles, can be associated with the absence of both thermoelastic accommodation of martensite and coherency between martensite and parent phase,13 in such a way that the presence of thermal memory has been replaced by ‘amnesia’.
One of the possible reasons for this alteration of martensitic structure, as an effect of complex cycling, could be atom redistribution, which is considered as the most important time dependent process occurring in Cu–Zn based SMAs in a lower temperature range.23 In order to compare the evolutions of atom distribution in the three specimen fragments, energy dispersive X-ray spectroscopy (EDX) profiles were determined. A typical result is given in Fig. 7, which shows the fluctuations of Cu, Zn and Al amounts on a distance above 30 μm. It is rather obvious that, with increasing maximum temperature for thermal cycling from 453 to 463 K and further to 473 K, an apparent intensification in atomic migration occurred. Considering that similar exchanges of Cu, Zn and Al atoms were also reported in the case of another Cu–Zn–Al SMA when heated to 423 K,24 it is expectable that the Cu–15Zn–6Al SMA under study should experience a similar behaviour.
Evolution of EDX profiles of compositional variations of Cu, Zn and Al in structure of specimens subjected to four series of five thermal cycles intercalated with periods of 3·2 Ms natural aging
Based on these observations, it was possible to corroborate ‘amnesia’ occurrence with an intensification of atomic migration which is the opposite of diffusionless martensitic transformation. Therefore, with the increase of both the total period of natural aging and the maximum temperature of thermal cycling, reverse martensite transformation to parent phase is no longer noticeable and ‘amnesia’ takes the place of thermal memory.
Conclusions
Three fragments, cut from the same lamella of martensitic Cu–15Zn–6Al SMA, were subjected to complex cycling performed on a DSC device and were further analysed by SEM–EDX. Each of the three specimen fragments underwent four series of five thermal cycles up to 453, 463 and 473 K respectively, intercalated with periods of 3·2 Ms natural aging (in all, there were 9·6 Ms of natural aging, for each specimen). During heating, endothermic solid state transitions occurred which were ascribed to martensite reversion to parent phase,25 which represents the mechanism of thermal memory behaviour in SMAs. As compared to the critical temperatures for the end of martensite reversion, Af = 407 K, determined by electrical resistivity variations,17 the above three maximum heating temperatures were higher with 46, 56 and 66 K respectively. With increasing the number of thermal cycles, reverse martensite transformation experienced the tendencies to shift to lower temperatures, to slow down and decrease in intensity, which was associated with a gradual diminution of thermal memory, reaching almost 42 in the case of 453 K cycling and almost 93 for 463 K cycling. These tendencies were accentuated by the increase of maximum temperature of thermal cycling. This severe degradation of thermal memory was associated with ‘amnesia’, which was accompanied by sensible intensification of atomic migration and the formation of interlocking lath martensite needles, as summarized in Graphical abstract and Highlights displayed in Supplementary Materials 3 http://dx.doi.org/10.1179/1743284711Y.0000000099.S3 and 4http://dx.doi.org/10.1179/1743284711Y.0000000099.S4 respectively.
Footnotes
Acknowledgements
B. Pricop acknowledges financial support from EURODOC project.