Effect of Heat Treatment on Microstructure and Thermal Properties of Cu-based Shape Memory Ribbons

The aim in this work was to investigate the change in microstructure, phase transformation temperatures, and thermal proper- ties due to the quenching of the investigated Cu-Al-Mn and Cu-Al-Mn-Ti alloys in ribbon form. This paper presents the results of microstructure analysis and thermal properties of Cu-Al-Mn and Cu-Al-Mn-Ti shape memory alloys produced in ribbon form by melt spinning technique. The microstructural analysis was carried out before and after quenching. After casting of the investigated alloys, annealing at 900 °C for 30 min was performed, followed by water quenching. The microstructural analysis was carried out by optical and scanning electron microscopy equipped with an energy dispersive spectrometer and by X-ray diffractometer. Thermodynamic calculation of a ternary Cu-Al-Mn system in equilibrium condition was performed using Ther- mo-Calc 5 software. Phase transformation temperatures were determined by differential scanning calorimetry and electrical resistance measuring. The results of microstructural analysis show the presence of martensite microstructures before and after quenching in the Cu-Al-Mn alloy, while in the Cu-Al-Mn-Ti alloy martensite microstructure exists only after quenching. Phase transformation temperatures decreased after quenching and titanium addition.


Introduction
Commercially attractive shape memory alloys (SMAs), such as copper-based shape memory alloys, have practical applications because of their characteristic properties, i.e., superelasticity (SE), shape memory effect (SME), two-way memory effect (TWME), and damping properties. 1 In addition, alloys like Ni-Ti with shape memory effect can be used in different ways to control civil structures. 2 Usually these alloys can be used in the form of bars, wires or plates. Mali et al., 2 in their review on shape memory alloys, concluded that SMAs have many opportunities and represent the construction materials of the future. Quader et al. 3 stated that there are different techniques available to shape memory alloys production: (i) casting -vacuum arc remelting, vacuum induction melting or electron beam melting, (ii) powder metallurgy -conventional processes or additive manufacturing. Compared to other shape memory alloys, copper-based alloys have the following main advantages: (i) lower cost than nickel-based alloys, and (ii) better SME and SE properties than iron-based alloys. Copper-based shape memory alloys are easier to produce than nickel-based alloys, and are less expensive. However, a great disadvantage of some copper-based shape memory alloys, like Cu-Al-Ni or Cu-Zn-Al, is brittleness and insufficient cold-working. Improvement of ductility has been achieved with the addition of manganese, so Cu-Al-Mn with low aluminium content shows good ductility. Practical applications of Cu-Al-Mn shape memory alloys can be applied as SE, SME, and TWME materials, as high damping material or as material with high thermal and electrical conductivities, e.g., for medical and electrical devices. 1 The reason for this application is microstructure and texture control such as grain size. Hasnat el al. 4 gave a review of utilising shape memory alloy in structural safety. In their work, they concluded that the production of SMAs and devices based on SMAs was still comparatively expensive, which could be main reason for restriction of application of this alloy in structural engineering. Civil engineering structures are often very large and require a lot of material. According to literature, 5 Cu-Al-Mn SMAs are appropriate for application in civil engineering because of their high damping capacity. Kumar et al. 6 in investigations of changes in properties of Cu-Al-Mn shape memory alloys mentioned that quaternary addition of different elements can have a significant effect on alloy properties. Adding some alloying elements, such as Mn, Ti, Fe, Zr, and B can significantly improve mechanical properties, especially ductility. [7][8] In addition, the addition of manganese has been proven to enhance thermoelastic and pseudoelastic behaviours. Magnetic properties can be a function of the addition of a fourth element (e.g., Ag, Be, Co, Ga, Sb or Zr). 9 The behaviour of shape memory alloys is significant under various thermal conditions due to the development of martensite that occurs in these alloys. The Ni-Ti alloys were often exposed to heat treatment (annealing) at 550 °C in order to create conditions with a recovered microstructure in comparison to cold-working conditions. 10 On the other hand, the Cu-based shape memory alloys are usually annealed in β-region at 800 or 900 °C to achieve preconditions for the formation of shape memory effect. 10,11 Canbay et al. 12 mentioned that copper-based shape memory alloys Effect of Heat Treatment on Microstructure and Thermal Properties of Cu-based Shape Memory Ribbons show transformation to martensite on cooling, and transformation to austenite on heating, and during the cooling process close packed structures are characterised by order, such as 6R, 18R, and 2H type. It is known that, in Cu-Al-Mn shape memory alloys, there exists the disordered β-phase which is stable at high temperatures. This phase transforms into martensite with quenching. Transformation is started by reactions: β(A2) → β 2 (B2) → β 1 (L2 1 ). 13 At the end of transformation, three different types of martensite can be formed: α 1 '(3R), β 1 '(18R) and γ 1 '(2H). Ordered structures, like B2 − , DO 3 − or L2 1 , transform into close-packet martensite structures (6R, 18R and 2H) with cooling. The type of martensite depends on alloy composition. Type β 1 ' martensite is predominant at lower aluminium content, while type γ 1 ' is mainly formed at higher content of aluminium. In addition, β 1 ' and γ 1 ' martensites co-exist in the intermediate range. Canbay et al. 14 characterised ternary 82.20 Cu -12.67 Al -3.13 Mn (wt.%), and obtained presence of both types of 18R and 2H martensite phases. In addition, Canbay et al. 15 mentioned that aging process has significant effect on transformation temperatures. The alloys show SME because the martensite re-orientation under external force causes a thermoelastic martensite transformation. Mallik et al. 16 mentioned that Zn and Ni, as quaternary additions, increase the martensite transformation temperatures, while Fe, Cr, Si, Mg, and Ti decrease them.

Materials
The Cu-Al-Mn and Cu-Al-Mn-Ti shape memory ribbons used in this investigation were cast by the melt spinning process. Production of the investigated alloys was performed using technically pure components: Cu (purity 99.9 %), Al (purity 99.5 %), Mn (purity 99.8 %), and Ti (purity 99.8 %). Certain weights of the pure components were re-melted by induction melting in a graphite crucible. The produced Cu-Al-Mn and Cu-Al-Mn-Ti ingots were used as input material for ribbon productions. The dimensions of solidified ingots were φ45×55 mm and about 680 g in weight. The ingots were smelted and injected through a rectangular nozzle onto a cold surface of the melt spinner. The melting temperature of the ingots was around 1450 °C. Under argon overpressure in the graphite crucible, the molten alloy was sprayed through the nozzle on a rotating disc made of copper alloy. Because of very fast heat transfer, ribbons with width 3.5-5.0 mm and thickness 0.12-0.30 mm were produced. The chemical composition of the produced alloys is presented in Table 1.

Methods
After casting, the ribbons were solution annealed at 900 °C in the electro-resistance chamber furnace for 30 min, and then quenched in room-temperature water. The investigated alloys were heated at 900 °C in order to carry out the quenching in water from high-temperature β phase region, and thus create the preconditions for the formation of martensite. Microstructural characterisation of samples was carried out by optical microscopy (OM) using an Olympus GX 71s, scanning electron microscopy (SEM) using TES-CAN VEGA 5136MM equipped with a device for energy dispersive spectroscopy (EDS) (measurement conditions: SEI, HV = 20.00 kV, WD = 20.00 mm), and X-ray diffraction (XRD) analysis by Shimadzu XRD 6000 diffractometer (measurement conditions: 40.0 kV, 30.0 mA, scan range 2θ: 20-100°). The samples for metallographic analysis were prepared by mechanical grinding (papers 120-1200) and polishing (water suspension of 0.3 μm Al 2 O 3 ), and were then etched using a solution of 2.5 g FeCl 3 , 48 ml methanol, and 10 ml HCl. XRD analysis was done with CuKα radiation on a Philips PW 1830 to identify the phases present in the samples.
Thermodynamic calculation for Cu-Al-Mn was carried out by Thermo-Calc software based on the CALPHAD method. Thermal properties were analysed on a Mettler Tolledo 822e differential scanning calorimeter (Mettler Toledo, Greifensee, Switzerland). The samples were tested in a temperature range from -100 to 250 °C in two heating/cooling cycles at a rate of 10 °C min −1 , under N 2 atmosphere, (40 ml min −1 ). The heating/cooling rate of 10 °C min −1 was applied for DSC measurements because it is a standard heating/cooling rate for metal samples analysis and determination of transformation temperatures. For the measurement of electrical resistance, the D.C. method was used. By this method, the ribbons were coiled on a 50 mm long ceramic tube. During the heating period, the electrical resistance was measured in the tube furnace, and temperature change was observed by the thermocouple Pt-Pt 10% Rh. The heating from room temperature up to 200 °C was controlled by the Eurotherm control system. The electrical resistance was simultaneously calculated at known electric current through the sample and a measured voltage drop.

Microstructural characterisation
Optical micrographs of Cu-Al-Mn alloy before and after quenching show a certain increase in grain size in quenching state (Figs. 1 and 2). Optical micrographs of the Cu-Al-Mn-Ti alloy before quenching show a significantly different microstructure than the Cu-Al-Mn alloy. The Cu-Al-Mn alloy before and after quenching shows the presence of a martensitic microstructure. In contrast, Cu-Al-Mn-Ti alloy optical micrographs cannot detect the formation of martensite before quenching ( Fig. 1). By annealing and quenching (900 °C/30 min/water), a martensitic microstructure was achieved (Fig. 2).
More detailed microstructural characterisation was performed by scanning electron microscopy and energy dispersive spectroscopy. The SEM micrographs of the Cu-Al-Mn and Cu-Al-Mn-Ti alloys before and after quenching are shown in Figs The Cu-Al-Mn-Ti alloy before quenching shows no presence of martensite, and its microstructure consists of a two-phase region of α + β 1 . Grain structure with α + β phase, before quenching, clearly indicates that this alloy has potential to exhibit the martensite formation and shape memory behaviour after quenching. By annealing at 900 °C, a single-phase β 1 structure of Cu-Al-Mn-Ti alloy was produced, and by quenching, it transformed into martensite. Quenching at 900 °C was carried out expecting formation of martensitic microstructures. This formation was observed in both quenched samples. Moreover, in the microstructure, the presence of residual untransformed β 1 phase can be observed. If copper-based shape memory alloys are exposed to cooling below the critical martensite start temperature, martensite can be formed as plates. 17 The structure of martensite is caused by martensite variants. In the beginning, the individual plates of martensite are very thin or parallel. Usually, martensite grows in broader units, which consist of two or four plates. It is known from literature 18 that Cu-Al-Mn alloys can undergo martensitic transformation from the β phase, which is stable at high temperatures. During quenching from β region transformation β (A2) → β 2 (B2) → β 1 (L2 1 ) occurs. In this case, three forms of martensite can arise from the β 1 phase: α'(3R), β 1 '(18R) and γ 1 '(2H), which depends on the aluminium and manganese content. At lower aluminium content, β 1 ' martensite is usually dominant. In contrast, higher aluminium content produces γ 1 ' martensite. For the range between higher and lower content of aluminium and manganese, there are β 1 ' and γ 1 ' martensite at the same time.
Mielzarek et al. 19 concluded that, below approximately 500 °C, Cu-Al-Mn has an ordered structure of L2 1 (bcc) that transforms into 18R, which can be observed as small acicular martensite or 2H, which can be observed as large acicular martensite. Hussain et al. 20 also investigated the Cu-Al-Mn-Ti alloy, and found the existence of a two-phase α + β region in casted state. By quenching these alloys, they observed a similar effect, i.e., the formation of martensite β 1 ', and γ 1 ', respectively. Gerdan et al., 21 in their investigation of martensite transformation, concluded that the microstructures of Cu-Al-Mn and Cu-Al-Mn-Zn alloys consisted of β 1 ' martensite (thinner phase), and γ 1 ' martensite (thicker phase), and in both alloys they observed no formation of precipitates. It is known that the addition of titanium to Cu-Al-Mn alloys results in grain refining. 20 By addition of titanium and with the formation of martensite after quenching, a slightly finer grain in Cu-Al-Mn-Ti alloys can also be observed.
The martensite microstructure can have two different morphologies: needle-like and plate-like. 22 β 1 ' and γ 1 ' phases are two types of thermal-induced martensites, and γ 1 ' phase forms as coarse variants, i.e., plate-like phase. In contrast, β 1 ' is a phase that forms as a needle-like phase and usually arises between γ 1 ' phase. Figs. 2 and 4 show that self-accommodated plate-like martensites nucleate and grow mostly in a different direction.  Table 2). In addition, similar values of EDS analysis were obtained for the quenched state of Cu-Al-Mn-Ti alloy (Table 3).
Figs. 5 and 6 show the XRD results of the investigated alloys before and after quenching. The major probable peaks were determined and marked in Figs. 5 and 6. The martensitic microstructures AlCu 3 (β 1 ') and Al 4 Cu 9 (γ 1 ') were clearly determined. The XRD results of the Cu-Al-Mn alloy before quenching showed that the microstructure consisted of austenite (β 1 phase) and two martensite (β 1 ' and γ 1 ') microstructures. In contrast, the Cu-Al-Mn-Ti alloy before quenching showed the presence of two phases in the microstructure, but without martensite. Thus, the microstructure of the Cu-Al-Mn-Ti alloy before quenching consisted of austenite (β 1 phase) and phase (Cu FCC). It is known that, at temperatures below 500 °C, Cu-Al-Mn has a two-phase microstructure (α phase with FCC, and β phase with L2 1 ). 23 At a temperature higher than 550 °C, only a single β phase is present in the structure. The absence of martensite in Cu-Al-Mn-Ti alloy before quenching was confirmed by both optical and scanning electron microscopy ( Figs. 1 and 3). The XRD results showed that there was no change in the microstructure of the Cu-Al-Mn alloy before and after quenching. The XRD analysis of Cu-Al-Mn alloy after quenching also showed the presence of β 1 phase (austenite) and β 1 ' phase (martensite), as in the case of the casted state. However, a more significant change could be observed on the X-ray patterns of the Cu-Al-Mn-Ti after quenching. The XRD analysis of quenched Cu-Al-Mn-Ti alloy showed the presence of three phases: β 1 phase (austenite), β 1 '  phase (martensite), and γ 1 ' (martensite). The presence of β 1 ' and γ 1 ' martensites was also determined in Cu-Al-Mn-Ti ingots after quenching by investigation of Mallik et al. 16 The presence of martensite β 1 '(18R) and γ 1 '(2H) coincides with the results of DSC analysis, where in Cu-Al-Mn-Ti alloy after quenching, the presence of two exothermic peaks was observed on the cooling curve. These peaks can be associated with the formation of β 1 ' and γ 1 ' martensites.

Thermal properties
Thermodynamic calculation for Cu-Al-Mn was carried out by Thermo-Calc software based on the CALPHAD method and minimisation of Gibbs energy. Fig. 7 shows the vertical section of calculated phase diagram with Cu 85 wt.%-Mn 7 wt.%-Al. Calculation of phase diagram was performed according to optimised parameters for ternary system Cu-Al-Mn by Miettinen 24 in Cu-rich region. The disorder-order phase transitions of β-phase A2 → B2 → L21 were not taken in the Miettinen calculations due to lack of data. It is well known that, under fast cooling, metastable martensitic phases β'(18R), γ 1 '(2H) or α'(3R) can be formed depending on the component composition. In Fig 7,   The peaks are more pronounced in some transformations and less pronounced in others. A more detailed analysis of the DSC thermogram of the Cu-Al-Mn alloy before quenching showed two exothermic peaks on the cooling curve (Fig. 8a). Both of these points represent a martensite transformation, but it can be concluded that two different types of martensite occurred. The characteristic temperatures of transformation (A s , A f , M s , and M f ) were determined from Figs. 8 and 9, and are given in Table 4 (Table 4). In the case of the quenched Cu-Al-Mn alloy, only one exothermic peak can be observed on the cooling curve, which represents a martensitic transformation (Fig. 9a). The austenite formation start temperature for the Cu-Al-Mn-Ti alloy before quenching was 11.31 °C, and the transformation finished at 39.13 °C. The start temperature of the martensite formation was 37.23 °C, and finished at -12.10 °C, which is a similar behaviour with the casted Cu-Al-Mn alloy. The Cu-Al-Mn-Ti alloy before quenching also exhibited two exothermic peaks on the cooling curve (Fig. 8b, Table 4). It is probably also the case of the formation of two different forms (structures) of martensite. DSC analysis showed that, despite low intensity, martensite still formed in the as-cast state, with M s at 37.23 °C and M f at -12.10 °C. Since the formation of martensite structure had only just begun, that is probably the reason why martensite was not observed in SEM and OM micrographs. The quenched Cu-Al-Mn-Ti alloy achieved the start temperature of austenitic trans-formation at 8.33 °C and finished at 34.08 °C. The start temperature of the martensite formation was at 37.57 °C, and the martensite finished at -9.57 °C (complete martensite microstructure). In contrast to the quenched Cu-Al-Mn alloy, in the case of the quenched Cu-Al-Mn-Ti alloy, two types of martensite are visible on the cooling curve (Fig. 9b). These results indicate that the values of transformation temperatures for austenite and martensite microstructures vary with addition of titanium. In other investigations 25-28 of ductile Cu-Al-Mn shape memory alloys, the authors mentioned that transformation temperatures can be in a range from -200 to 150 °C; e.g., Zhou et al. 26 in their investigation of Cu-8.7Al-10.8Mn alloy determined that M s , M f , A s , and A f at the top of the sample were 7.0, -11.4, 10.9, and 25.8 °C, respectively. The temperature hysteresis for the Cu-Al-Mn and Cu-Al-Mn-Ti alloys before and after quenching is shown in Table 4, and can be ranged The change in temperature hysteresis is important to determine SME, and whenever the hysteresis is narrow, SME is high. According to Canbay et al., 12 temperature hysteresis proceeds from the absorption of the internal energy described by three mechanisms: (i) dissipation of energy due to internal friction, (ii) storage of energy, and (iii) heat transfers due to the latent heat of phase change.
The research of electrical resistance was performed up to 200 °C. Electrical resistance versus temperature for Cu-Al-Mn alloy before quenching is shown in Fig. 10. The Cu-Al-Mn alloy before quenching shows changes in the electrical resistance curve with temperature change, i.e., during the heating and cooling processes. These changes of electrical resistance are visible as deviations of linearity, and are observed as endothermic and exothermic peaks for Cu-Al-Mn before quenching, i.e., only in Fig. 10a. These peaks can be caused by phase transformation. On the heating curve, one temperature interval of microstructural change is seen between 80 and 120 °C with peak maximum at 104.2 °C. This microstructural change can be caused by formation of austenite (A s and A f temperatures). On the other side, the cooling curve shows one exothermic peak, and this peak represents a martensitic transformation with M s = 105 °C and M f = 75 °C with peak maximum at 86.4 °C. The temperatures at which deviations of linearity appeared are only visible on the curve for the as-cast Cu-Al-Mn alloy (Fig. 10a). Other studied samples show behaviour similar to Fig. 10b, i.e., phase transformations are not visible on the curves. This is expected and in accordance with DSC results. DSC results ( Table 4) show temperatures of phase transformations for the Cu-Al-Mn-Ti alloy before and after quenching below 40 °C. The device used for electrical resistance measurement cannot detect martensite transformation at such low temperatures.

Conclusions
The influence of quenching on microstructure and thermal properties of ductile Cu-Al-Mn and Cu-Al-Mn-Ti shape memory alloys were investigated. The obtained results suggest the following conclusions: -Optical micrographs of the Cu-Al-Mn alloy before and after quenching showed the presence of a martensitic microstructure, while in Cu-Al-Mn-Ti alloy, the formation of martensite before quenching was not detected. Optical micrographs of Cu-Al-Mn-Ti alloy after quenching show that martensitic microstructure was achieved. -SEM and XRD analysis of Cu-Al-Mn alloy before and after quenching showed the presence of the lath type (β 1 ') martensite with thin plates, and partial regions of the untransformed β phase. Before quenching, Cu-Al-Mn-Ti alloy consisted of a two-phase region of α + β 1 . After quenching, a formation of martensites (β 1 ' and γ 1 ') and austenite (β 1 ) was observed.