Abnormal grain growth of 68Cu–16Al–16Zn alloys for elastocaloric cooling via cyclical heat treatments

Cu-based superelastic shape memory alloys are promising for low-stress elastocaloric cooling. We have synthesized bulk alloys of 68Cu–16Al–16Zn under different conditions in order to promote its grain growth and enhance its elastocaloric properties. High-temperature x-ray diffraction of untreated 68Cu–16Al–16Zn alloy showed that the phase boundary between the α + β mixed phases and the high temperature phase (β phase) was between 973 K and 1023 K. Based on this result, the 68Cu–16Al–16Zn alloy was heated and cooled in a furnace repeatedly between 773 K and 1173 K. The maximum grain size after heat treatment of the ingot rolled to 67% reached 11.1 mm. The latent heat of the martensitic transformation after grain growth was 6.3 J g−1, which is higher than the previously reported value for the compound. The stress–strain curve of 68Cu–16Al–16Zn rolled to 67% rolling with cyclical heat treatments showed a maximum stress of 106 MPa at 4.5% strain, with adiabatic temperature change of 5.9 K in heating during stress loading and 5.6 K in cooling in stress removal. Furthermore, no fatigue in the stress–strain behavior was observed up to at least 60 000 mechanical cycles at 2% strain.


Introduction
The history of shape memory alloys (SMAs) began with the discovery of the pseudoelastic effect in Au-Cd alloys by Ölander et al [1]. Then, Greninger et al reported the formation and disappearance of a martensitic phase in Cu-Zn alloys [2]. In 1963, Ti-Ni was reported to have shape memory, and these alloys came to be called SMAs [3]. Cu-Al-Ni was first discovered as a copper-based SMA by Kurdjumov et al in 1964. Since then, many copper-based alloys have been uncovered, and in 1974, Cu-Al-Zn was found to exhibit shape memory effects [4]. Currently, most of the SMAs in practical use are Ti-Ni, with a few Cu-Al-Ni and Cu-Al-Zn alloys. Ti-Ni is superior to Cu-Al-Ni and Cu-Al-Zn in many aspects, such as strength [5][6][7][8]. However, because Ti-Ni is difficult to process, other alloys have been pursued for applications [9], the most common ones of them being Cu-based SMAs. They have higher workability than Ti-Ni, and their cost is approximately 1/10th of that of Ti-Ni alloys [5-7, 10, 11]. The advantages of Cu-based SMAs over Ti-Ni include shape recovery strain.
We are interested in exploring different bulk SMAs for elastocaloric cooling, where the latent heat associated with the stress-induced transformation from austenite to martensite is used for pumping heat [12]. From this point of view, another major advantage of Cu-based SMAs is that their critical stress required for inducing the superelastic transformation is substantially lower than that for Ti-Ni. This makes it substantially easier to implement the actuators required to induce transformation in Cu-based SMAs. For the adiabatic superelastic transition in Ti-Ni, the required stress can be as large as >800 MPa, whereas for Cu-based SMAs, it can be much lower than 500 MPa. Here, we focused on Cu-Al-Zn SMAs whose critical stress was found to be lower than 100 MPa. Cu-Al-Zn alloys have the added advantage of relatively high ductility compared to other Cu-based SMAs [8].
Perkins reported the shape memory effect in 62∼62.5Cu-1.5∼2.0Al-35.5∼36.5Zn in 1974 [4], and by 1980, 65∼78Cu-2∼8Al-21∼36Zn [13][14][15] and 66∼72Cu-12∼17Al-12∼20Zn [13,16,17] alloys has been reported. In 2008, it was found that 68.13Cu-16.13Al-15.74Zn alloys exhibited elastocaloric effects [18]. One of the major issues in Cu-based SMAs is their cyclical fatigue property. The reason for the short fatigue life of Cu-based alloys is that their crystal grains are coarser than those of Ti-Ni owing to the tendency of thermal grain growth, which leads to grain boundary fracture when stress is applied [8][9][10][19][20][21]. Although attempts have been made to improve the crystal refinement and strength by adding Ti and V [21][22][23], there are drawbacks, such as smaller shape recovery properties [8]. The fatigue life is expected to be improved by using single crystal alloys [19,21]. However, the conventional Bridgman method of single-crystallization for Cu-based alloys is expensive [24]. In 2013, Kusama et al reported that in Cu-Al-Mn, crystal grains can grow from approximately 2 mm to as large as 50 mm by periodically repeating heating and cooling across the phase boundary temperature between the mixed phase of α (fcc) and β (bcc) and the single phase of β, that is, the high-temperature phase [24,25]. In the Cu-based alloy 68Cu-16Al-16Zn, grain growth is expected to occur by cyclical heat treatment following the same idea. Here, we report the development of large-grain bulk 68Cu-16Al-16Zn alloys and their elastocaloric cooling properties.

Preparation of 68Cu-16Al-16Zn alloy ingot in a vacuum melting furnace
The 68Cu-16Al-16Zn alloy was prepared in a vacuum melting furnace (Nisshin Giken Co., Ltd, NEV-M5V). The raw materials used were Cu (Rare Metallic Co., Ltd, 4 N, shot, 3φ × 3 mm), Al (Hirano Seizaemon Shoten Co., Ltd, 5 N, shot machined from ingot, 8φ × 5 ∼ 10 mm), and Zn (Kanto Chemical Co., Inc. 3N65, shot, approximately 5 mm). Because the boiling point of Zn at atmospheric pressure is lower than the melting point of Cu, Zn volatilization is more significant under vacuum. In this study, we attempted to suppress Zn volatilization as much as possible by reducing the pressure in the chamber to about 10 −1 Pa and then introducing Ar gas to a pressure of 0.09 MPa (≈0.9 atm). The total amount of raw material was calculated to be 0.5 kg. Cu and Al, weighed in stoichiometric ratios, were mixed in a carbon crucible to ensure that they were adequately dispersed. Then, 5% excess Zn shots, assumed to volatilize during dissolution, were placed on top. Figure 1 shows schematics and pictures of the raw metal are positioned in the carbon crucible. It is important to precisely control the amount of additional Zn to account for volatilization, placements of each metallic material in the crucible, and the position of the high frequency coil relative to the crucible. Thus, we were able to preferentially heat Cu and Al first, and then melt Zn. Three minutes after all the metals were expected to have dissolved, the alloy was poured into molds and cooled. This process allowed us to precisely control the compositions of the alloy samples.

Alloy ingot rolling
From the obtained ingot, three bulk pieces 24.5 mm in diameter and a height of 21 mm were cut to be rolled. A hot two-stage roll mill (Ohno Roll, 2/4DR-300S, maximum rolling load of 150 t) was used for rolling. Ingots were first heated in an electric furnace at 1073 K for 30 min and then rolled at a working ratio (reduction ratio) of 10%. The rolled ingot was again heated at 1073 K for 3-5 min and rolled at a working ratio of 10% of the thickness rolled in each successive pass, and this process was repeated until the thickness of 7 mm was reached.

X-ray diffraction for determining heat treatment conditions
To ensure that the cyclical heat treatment of the rolling ingots is effective, information on the crystalline phases at high temperatures is necessary. We confirmed the phase transition behavior by high-temperature x-ray diffraction using a multi-purpose x-ray diffractometer (Malvern Panalytical, Empyrean, Cu-Kα radiation, 45 kV-40 mA, mounted PIXel3D Detector with Medipix3) equipped with a high-temperature oven chamber (Anton Paar, HTK 1200N). Based on the results described below, cyclical heat treatment was performed in the range of 773-1173 K, followed by aging treatment at 373 K for 30 min.

Alloy composition analysis
The alloy composition was analyzed by ICP-AES (SHIMADZU ICPE-9000). Approximately 2 mg of alloy was dissolved in 25 ml of threefold diluted 60%-61% Nitric Acid (Kanto Chemical Co., Inc.). The alloy was completely dissolved after two weeks and then diluted four-fold with ultrapure water. Calibration curves based on 0, 7.5, 15, 22.5, and 30 ppm for Cu, 0, 0.75, 1.5, 2.25, and 3.0 ppm for Al, and 0, 1.8, 3.6, 5.4, and 7.2 ppm for Zn were prepared using standard solutions for Cu, Al, and Zn (Kanto Chemical Co., Inc.). The actual alloy composition was calculated by extrapolating the detected ppm concentrations of the dissolved alloys.

Etching process for grain size and grain boundary observation
Cross sections of the alloy pieces were polished with water-resistant abrasive paper with grain sizes of #100, #220, #600, #1000, and #2000 using a lapping tool (ML-110NT, Maruto Instrument Co., Ltd.). Next, the alloy was polished to a mirror surface with a lapping film grain size of #4000, followed by ultrasonic cleaning with distilled water. To etch the alloy after aging, a mixture of 7.5 ml of HCl (Kanto Chemical Co., Inc. 1.18 g cm −3 ), 2.5 g of FeCl 3 -6H 2 O (FUJIFILM Wako Pure Chemical Corporation) and 30 ml of distilled water were used. The alloy was then immersed in this mixture and sonicated for 90 s. 1 mol l −1 NaOHaq, 2.0 g of NaOH (FUJIFILM Wako Pure Chemical Corporation) dissolved in 50 ml distilled water, was used as a neutralizer to stop etching.

Measurement of the phase transition temperature and the latent heat
The transformation temperature and martensitic transition behavior of the alloys were characterized using differential scanning calorimetry (DSC, NETZSCH DSC3500 Sirius), and the latent heat was determined from the obtained DSC curves. The alloy was cut to fit into an aluminum cup for DSC measurements, and polished to make the contact surface between the bottom of the aluminum cup and the alloy sample smooth and uniform. The reference cell was an aluminum container. The rate of temperature change during the measurements was 10 • C min -1 .

Measurement of elastocaloric effects by stress-strain curves
Compression tests were performed on 67% rolled and heat-treated 68Cu-16Al-16Zn alloy. The sample was machined into a 2.4 mm × 2.4 mm × 15 mm bar shape. Temperature changes were recorded using an infrared camera. First, the maximum stress and strain were measured under isothermal conditions by repeatedly applying and removing stress every 60 s, with a 5 s waiting time between stress applications. Then, under adiabatic conditions, stress was applied to the sample to achieve a fixed strain in 1 s and maintained in compression for approximately 30 s to observe the temperature change of the sample. The stress was maintained until the sample returned to room temperature. The sample was then unloaded in 1 s. The temperature change in the sample was again recorded and maintained until it returned to room temperature.

68Cu-16Al-16Zn alloy preparation, phase transition behavior, grain growth by rolling and cyclical heat treatment
The compositions of Cu, Al and Zn in the fabricated ingots were 69.89(2) at%, 15.66(3) at%, and 14.45(1) at%, respectively. To determine the ideal heat treatment cycle conditions for the single crystal-like growth of the alloy, structural phase boundaries were investigated by high-temperature x-ray diffraction measurements from room temperature to 1073 K. Figure 2 shows the high-temperature x-ray diffraction patterns of the Cu-16Al-16Zn alloy from 923 K to 1073 K.
It exhibits the α phase at room temperature, α + β mixed phase from 473 K to 973 K, and the β phase at 1023 K, respectively. Based on these results, we surmised that there is a phase boundary between 973 K and 1023 K, which is necessary for the heat treatment cycle. Figure 3 shows a program summary of the heat-treatment cycles performed based on this finding. The temperature increase (decrease) rate in the range (I) in figure 3 was 1 K min −1 , and the temperature increase (decrease) rate in the range (II) was 0.5 K min −1 .
To confirm the effectiveness of the heat treatment cycles, formation of grains/grain boundaries was studied in 68Cu-16Al-16Zn alloys with 0%, 67%, and 83% rolling work rates. Figure 4 shows photographs of the alloys in their pristine state (left), after heat-treated at 1173 K for 10 min followed by quenching (middle), and cyclically heat-treated (right) for each rolling work rate after etching. The average grain size at 0% rolling was 0.74 mm for pristine, 1.75 mm for quenched only, and 3.33 mm after cyclical heat treatment. Similarly, the average grain size at 67% rolling was 2.21 mm for pristine, 2.87 mm for quenched only, and 11.1 mm after cyclical heat treatment. And, the average grain size at 83% rolling was 1.70 mm for pristine, 1.73 mm for quenched only, and 3.73 mm after cyclic heat treatment. Note that the sample on the right in figure 4 in the center row (rolling rate: 67%) is believed to be in a single-crystal state because no grain boundaries were observed after repeated etching.
To confirm whether the alloys in which no grain boundaries appeared were indeed single crystals or not, we performed rocking curve measurements with the detector fixed at 2θ = 29.9 • . Figure 5 shows the results of the rocking curve measurements obtained at five spots on the sample. The numbers in the photograph in figure 5 insert indicate the positions of the spots. Although diffraction peaks appeared at 25 • at all five locations, some diffraction peaks were separated at approximately 1 • . We believe that the orientation difference is small enough to prevent the appearance of grain boundaries due to internal stresses generated during heat treatment and other processes, or due to a shift in Zn composition at each measurement location. However, the results are similar, indicating that the heat treatment across the periodic phase boundary, in addition to the 67% rolling ratio, is effective for single crystalline-like grain growth.   Table 1 shows the transformation transition temperature and the latent heat as functions of the rolling working ratio and heat treatment conditions analyzed from the DSC curves. The latent heat does not change with the rolling work rate or heat treatment conditions, whereas the transition temperature shifts to a higher temperature with increasing cycles of heat treatment compared with when it is just quenched. This may be due to the fact that cyclical heat treatment increases the particle size and decreases the number of grain boundaries, thereby decreasing the energy barrier associated with the transition induced by temperature change. While outside the immediate scope of the present work, it would be interesting to systematically investigate how the change in the grain size stabilizes the transformation temperature. We plan to study this further in the future through more detailed microstructural investigation together with multi-cycle DSC.  Table 1. Transition temperature and latent heat of 68Cu-16Al-16Zn processed under different rolling work rates and heat treatment * * %: rolling work rate, CH: cyclical heat treatment, Q: only quenched, NA: non-aging treatment, A: aging treatment.

Sample
Ms 0% The latent heat values observed here for the large grain 68Cu-16Al-16Zn alloys are consistently higher than the previously reported values of 2.3 J g −1 [26] by Manosa et al and 4.3 J g −1 [27] by Qian et al for similar composition alloys, and the largest value observed here was 6.3 J g −1 . On the other hand, this value is in good agreement with 6.2 J g −1 derived from indirect measurements [18,28]. Figure 6 shows the stress-strain curve of a 68Cu-16Al-16Zn alloy specimen under isothermal compression to 4.4% strain. The required stress is 109 MPa, which is considerably lower than that of any other elastocaloric material based on SMAs. Figure 7 shows IR images displaying the temperature change of the specimen under adiabatic compression of 4.4% strain. From left to right, the images show the specimen at room temperature prior to stress application, the specimen immediately after the stress was applied, and after the stress was unloaded. Figure 8 shows the time-temperature profile analyzed from the same data taken with the IR camera during this process. We have performed this experiment using different stress values. Figure 9 shows the maximum and minimum temperatures observed at different strain levels. The largest adiabatic ∆T observed here are 5.9 K and 5.6 K for exothermic process in compression and endothermic process in removal of the stress, respectively, for 4.4% strain. Using 6.3 J g −1 and the measured transformation temperature (together with the specific heat of the alloy), we arrive at ideal ∆T of 16.5 K. It is well known that experimentally measured adiabatic ∆T can be much lower than the ideal ∆T, the maximum  . Infrared camera image of 68Cu-16Al-Zn16 alloy (processed with 67% rolling process and cyclical heat treatment) under compression to 4.5% strain (left: before stress applied, center: immediately after stress is applied, right: immediately after stress is unloaded). possible ∆T, due to a number of reasons including parasitic heat loss and internal loss in the material as reflected in the hysteresis in the stress-strain curves.

Elastocaloric effect of the 68Cu-16Al-16Zn alloy
Finally, we examined the fatigue behavior of the specimen. We monitored the stress-strain curve for adiabatic compression in cyclical processes, where the stress was applied for 1 s, held for 5 s, and unloaded for 1 s. To date, we have performed this measurement up to 60 000 cycles with 2% strain (figure 10). Thus far, the curves so far indicate that there is a minimal change in the mechanical properties. While the measurement will continue to larger numbers of cycles and a future plan includes testing to higher strain levels, the result is promising, perhaps indicative of the robust long-term performance of the alloy developed in this study. Figure 8. Time-temperature data for 68Cu-16Al-Zn16 alloy (processed with 67% rolling process and cyclical heat treatment) measured during an adiabatic compression to 4.5% strain followed by unloading. Figure 9. Adiabatic temperature changed (∆T) observed for 68Cu-16Al-Zn16 alloy (processed with 67% rolling process and cyclical heat treatment) for strains ranging from 3-4.5%. Figure 10. Comparison of stress-strain curves of 68Cu-16Al-Zn16 alloy (processed with 67% rolling process and cyclical heat treatment) after mechanically cycled at 2.0% set strain.

Conclusions
Grain growth processes were performed via cyclical heat-treatment on 68Cu-16Al-16Zn alloys, which exhibit elastocaloric effects. The ingots were rolled to 67% and subjected to cyclical heating and cooling across the structural phase boundary between the α + β mixed phase and the high-temperature β phase. This process was successful, and resulted in a maximum grain size of 11.1 mm. The alloy exhibited a relatively large latent heat of 6.3 J g −1 from the DSC curve, which is higher than the previously reported values for alloys of the same composition. The stress-strain curve of 68Cu-16Al-16Zn up to 4.5% stress showed a maximum required stress of 106 MPa, substantially lower than that of any other SMA-based elastocaloric materials. Adiabatic heating and cooling ∆T of 5.9 K and 5.6 K were observed for 4.5% strain, respectively. We have performed continuous compression cycle tests on the alloy with 2.0% strain, and thus far, the material has exhibited minimal fatigue after 60 000 cycles.

Data availability statement
The data supporting the findings of this study are available upon reasonable request from the authors.