ExpErIMEntal dEtErMInatIon of phaSE EquIlIBrIa at 900 ° C and lIquIduS SurfaCE In thE Cu-niti SyStEM

Phase equilibria at 900 °C and the liquidus projection in the Ni-rich corner of the Cu-Ni-Ti ternary system were investigated by X-ray diffraction (XRD), scanning electron microscopy (SEM), electron probe microanalysis (EPMA) and differential scanning calorimetry (DSC) on the annealed and as-cast alloys. For the isothermal section at 900 °C, the hightemperature ternary compound (τ5) reported in the literature was not observed but the corresponding low-temperature compound (τ6) was identified to be stable at this temperature. For the liquidus projection, the primary phase τ4 was experimentally determined for the first time, and the related solidification paths were identified through the experimental data on the as-cast alloys. One solid-state invariant reaction, NiTi + τ4 → τ1 + Ni3Ti at 1098.3 °C and two ternary eutectic reactions, L → τ1 + τ4 + NiTi at 1126 °C and L → τ1 + τ6 + fcc(Cu,Ni) at 1069.5 °C were detected. Scheil reaction scheme was also presented accordingly. All obtained experimental results will provide reliable information for further thermodynamic optimization of the Cu-Ni-Ti ternary system.


Introduction
Shape memory alloys (SMAs) have been extensively investigated as an important category of functional materials for their potential applications in different engineering areas, such as composite materials, automotive, aerospace, mini actuators and micro-electromechanical systems (MEMS), robotics, as well as certain biomedical applications [1].Among various SMA systems, near-equiatomic NiTi alloys attract the highest technological interest because of their abilities to change shapes with temperature or load to provide controllable material and damping properties [2,3].The addition of Cu into the NiTi alloys can substitute for the Ni-sites in B2 phase (NiTi), which can considerably narrow the transformation temperature hysteresis [4,5], reduce the sensitivity of Ms to the Ni:Ti ratio [6], increase the strength difference between the parent and martensite phases and improve the transformational cyclic behavior [7].It was also reported that the amount of Cu substitution for Ni affects the transformation sequence [4].Moreover, the Cu-Ni-Ti alloys are good potential damping materials [8] and exhibit prominent glass forming ability through both rapid solidification processing and mechanical alloying [9].Therefore, in order to achieve the rational design of highperformance Cu-Ni-Ti shape memory alloys, the knowledge of phase equilibria and solidification sequence of the Cu-Ni-Ti system is of great necessity.
The phase diagrams of the three sub-binary systems have been well established, as shown in Fig. 1.The Cu-Ti and Ni-Ti phase diagrams are characterized with multiple binary compounds, whereas Cu-Ni is an isomorphous phase diagram with spinodal decomposition of the fcc(Cu,Ni) phase at low temperature [10,11].
According to review work by Schuster and Cacciamani [12], there are six ternary compounds stable in the Cu-Ni-Ti ternary system, which are summarized in Table 1.Several groups have contributed to the isothermal sections of the Cu-Ni-Ti system.The isothermal section of Cu-Ni-Ti at 800 °C over the whole composition range was investigated by Pfeifer et al. [13] using XRD, but the annealing time was too short (i.e. 2 h) in their work.Subsequently, Fedorov et al. [14] constructed the partial isothermal sections at 850 and 1000 °C in the Cu-rich corner by XRD method.It is worth mentioning that Fedorov et al. [14] have not observed τ 4 and τ 6 in their samples,
The experimental data on liquidus projection and invariant reactions are mostly available for the Curich and Ti-rich regions.In the Ti-CuTi-NiTi region, one reaction of transition type, Liquid + NiTi ↔ CuTi + NiTi 2 , and two ternary eutectic reactions, Liquid ↔ CuTi + CuTi 2 + NiTi 2 and Liquid ↔ bcc(Ti) + CuTi 2 + NiTi 2 were proposed by Gupta [39].However, these reactions are in contradiction to the review of Schuster and Cacciamani [12], in which CuTi cannot coexist with NiTi 2 on the liquidus surface.In the work by Schuster and Cacciamani [12], two U-type reactions, Liquid + bcc(Ti) ↔ CuTi 2 + NiTi 2 and Liquid + CuTi ↔ CuTi 2 + NiTi, and one ternary eutectic reaction as Liquid ↔ NiTi + CuTi 2 + NiTi 2 , were reported.This liquidus projection is also questionable because the range of the primary CuTi 2 phase is in contradiction to the liquidus in the Cu-Ti binary system [11].Furthermore, a ternary eutectoid reaction as bcc(Ti) ↔ CuTi 2 + NiTi 2 + hcp(Ti) at 738 °C was proposed by Yakushiji et al. [19] and accepted by other researchers [12].
In the CuTi-NiTi-Cu region, a series of invariant reactions were reported by Fedorov et al. [14], Yakushiji et al. [40], and Alisova et al. [41].These reactions were accepted in the critical review of Schuster and Cacciamani [12].However, a few modifications are required for some invariant reactions to achieve the consistency with the isothermal section at 870 °C, proposed by Loo et al. [15].For example, according to the reaction scheme [12] the three-phase CuTi + NiTi + τ 1 field should exist down to the room temperature, which is in contradiction to the experimental data [15,17,18].
For the Cu-Ni-NiTi region, the experimental information of the phase equilibria and invariant reactions is far from being established.Considering this situation, further investigations are needed to determine the phase relations and liquidus projection in this region.
The objective of the present work is to clarify the conflictions on the phase equilibria in literature and provide more accurate experimental data for further thermodynamic optimization of the Cu-Ni-Ti ternary system.With this purpose, we chose the isothermal section at 900 °C to investigate, and we tried to obtain the whole liquidus projection of the Cu-Ni-Ti system using both the annealed and as-cast alloys.

Experimental procedure
Thirty-eight ternary alloys were prepared from Ni, Ti and Cu rods (Beijing KMT Technology Co. Ltd., China) of 99.99 wt.% purity in an arc melting furnace (WKDHL-I, Opto-electronics Co. Ltd., Beijing, Figure 1.Binary phase diagrams [10,11] of the Cu-Ni-Ti system along with the composition points of the alloys studied in the present work Composition (at.%)China) under high purity argon atmosphere using a non-consumable W electrode.The ingots were remelted 4 times to improve their homogeneity.No chemical analysis for the alloys was conducted since the weight loss of each alloy during the arc-melting process was less than 1 wt.%.Afterwards, the ingots were generally cut into two pieces.One was kept ascast and subjected to the microstructure observation to investigate the primary solidification phase and reaction type.The other was annealed and subjected to XRD, SEM, EPMA and DSC.These alloys were wrapped in Mo wires after mechanical polishing of the surfaces and then sealed in evacuated silica capsules for annealing in an L4514-type diffusion furnace (Qingdao Instrument & Equipment Co. Ltd., China) at 900 °C for 960 hours, followed by quenching in liquid nitrogen.The nominal composition for each alloy are shown in Fig. 1 and Tables 2 and 3. X-ray diffraction (XRD) measurements were performed on a Rigaku D-max/2250 VB+ X-ray diffractometer at 40 kV and 250 mA using a Cu-Kα radiation.Microstructure observations of the alloys in both as-cast or annealed states were conducted using scanning electron microscopy with energy dispersive X-ray analysis (SEM/EDX, Nova NanoSEM-230, USA) after standard metallographic preparation.The primary solidification phases and reaction types were determined from the backscattered electron (BSE) images of the as-cast alloys.The composition analysis of both as-cast and annealed specimens were performed by electron probe microanalysis (EPMA, JXA-8530, JEOL, Japan) at 15 kV with pure Ni (99.99 wt.%), Ti (99.99 wt.%) and Cu (99.99 wt.%) as standard materials.The phase transition temperatures were measured by differential scanning calorimetry (DSC) (DSC404C, Netzsch, Germany) of the annealed alloys 3#, 9# and 37#.The measurements were performed between 900 °C and 1300 °C with the heating and cooling rates of 5 K/min under a flowing argon atmosphere using the Al 2 O 3 crucible.The temperatures of invariant reactions were determined from the onset of the thermal effects on the heating curve, while those for liquidus were taken from the tangential intersection of the last peak of the heating curve.The analyses of relevant cooling curve are also enclosed for comparison, as listed in Table 4.

Phase equilibria at 900 °C
The selected annealed alloys along with the measured constituent phases and compositions by XRD and EPMA are listed in Table 2.In Fig. 2 3.The  5 phase, which was deduced considering a weak DSC thermal signal with a heating rate of 40 K/min by Zhu et al. [18] at 900 °C, was not observed in this work, but the τ 6 phase was found instead.

Liquidus surface
In order to understand the solidification sequence and reaction type, microstructure of the as-cast alloys in the Ni-rich corner was analyzed using the results from SEM, EPMA and XRD.In accordance with the primary phase, the investigated alloys could be classified into six groups: the first group including the alloys 1#, 2# and 8# exhibits the primary phase of NiTi, the second group including the alloys 9#, 14#, 19#, 25# and 26# exhibits the primary phase of τ1, the third group including the alloys 3#, 4#, 10#, 11# and 16# has τ 4 as the primary phase, the fourth group including the alloys 20#, 27#, 31#, 35# and 37# has τ6 as the primary phase, the fifth group including the alloys 5#, 6#, 12#, 13#, 17#, 18#, 22#, 23#, 28#, 29#, 32#, 33# and 36# exhibits Ni 3 Ti as the primary phase, and the sixth group including the alloys 7#, 24#, 30# and 34# has fcc(Cu,Ni) as the primary phase.According to the solidification paths, the as-cast alloys could be divided into 10 classes, and detailed analysis is given below.The identified phases and their compositions are presented in Table 3.The results for some representative alloys are discussed below.
Fig. 6 presents the microstructure and XRD pattern of the as-cast alloy 8#.As shown in Figs.6(a) and 6(b), two phases could be seen distinctly: the dark primary NiTi phase and the grey phase τ 1 .Besides, the eutectic NiTi + τ 1 structure could be seen in the vicinity of the primary NiTi.Therefore, a eutectic reaction L → NiTi + τ 1 could be asserted.Based on the analysis, the sequence of the phase formation from the liquid phase with decreasing temperature could be specified as L → NiTi, L → NiTi + τ 1 .
Microstructures and XRD patterns of the as-cast alloys 14# and 26# are presented in Fig. 7.As shown in the BSE images of these two alloys, primary phase τ 1 could be easily observed.Furthermore, for the ascast alloy 14#, the dark phase NiTi and many twophase eutectic structures of τ 1 + NiTi could be determined in the micrograph (Figs.7(a)-(b)), which means a reaction L → NiTi + τ 1 occurs during the solidification.Therefore, the solidification sequence of the phase formation from the liquid could be deduced in this alloy: L → τ 1 , L → NiTi + τ 1 .Figs. 7(c) and 7(d) display the micrograph of the as-cast alloy 26#.Different from the alloy 14#, the continuous phase τ 6 was formed adjacent to the primary phase τ 1 .It indicates that a eutectic reaction L → τ 1 + τ 6 exists.In addition, some fine three-phase structures τ 1 + τ 6 + fcc(Cu,Ni) could be observed, which suggests that the ternary eutectic reaction L → τ 1 + τ 6 + fcc(Cu,Ni) takes place at the final stage of solidification.Hence, the solidification path in the alloy 26# could be deduced as L → τ 1 , L → τ 1 + τ 6 , L → τ 1 + τ 6 + fcc(Cu,Ni).XRD patterns of the as-cast alloy 3#.From the BSE images of these alloys, it is evident that the primary phase is τ 4 in these alloys.In Figs.8(a) and 8(b) for the alloy 3#, two-phase eutectic τ 4 + NiTi structure could be observed in the vicinity of the primary phase τ 4 , which means that a eutectic reaction L → τ 4 + NiTi has taken place.Besides, the three-phase eutectic τ 4 + NiTi + τ 1 structure could be observed in some regions, which implies that the ternary eutectic reaction L → τ 4 + NiTi + τ 1 has taken place at the final stage of solidification.Therefore, the solidification sequence of this alloy could be deduced as: L → τ 4 , L → τ 4 + NiTi, L → τ 4 + NiTi + τ 1 .Compared with the microstructure of the as-cast alloy 3#, there is the grey τ 1 phase around the primary τ 4 phase for the alloy 10# in Figs.8(c) and 8(d), indicating that the reaction L → τ 4 + τ 1 could be asserted.In addition, the dark phase NiTi could be seen, which coexists with the τ 4 and τ 1 phases suggesting the occurrence of a ternary eutectic reaction L → τ 4 + τ 1 + NiTi.Hence, the solidification sequence of the alloy 10# could be deduced as: L → τ 4 , L → τ 4 + τ 1 , L → τ 4 + τ 1 + NiTi.Different from the alloys 3# and 10#, the τ 6 phase in the alloy 16# is encompassed by τ 4 phase, which suggests that the eutectic reaction L → τ 4 + τ 6 takes place during the solidification, as shown in Figs.8(e) and 8(f).In addition, the tiny phase τ1 at the border between the τ 4 and τ 6 phases, could be observed.It means this threephase eutectic structure τ 4 + τ 6 + τ 1 formed in the final solidification.Therefore, the solidification sequence of this alloy could be deduced as: L → τ 4 , L → τ 4 + τ 6 , L → τ 4 + τ 6 + τ 1 .Fig. 9 exhibits the microstructure and XRD pattern of as-cast alloy 31#.The matrix phase τ 6 could be From the microstructure and XRD pattern for the as-cast alloy 24#, as shown in Fig. 11, we can see that fcc(Cu,Ni) precipitated as primary phase and the liquid phase composition moving along the fcc(Cu,Ni) phase surface during the solidification process.Therefore, the solidification path of alloy 24# could be determined as L → fcc(Cu,Ni).
DSC results of the selected annealed alloys are presented in Table 4 and Fig. 12.For the annealed alloy 3#, two peaks during heating and three peaks during cooling could be seen in Fig. 12(a).As shown in Fig. 4(c) and Figs.8(a-b), the alloy annealed at 900 °C has a three-phase τ 1 + Ni 3 Ti + NiTi microstructure for 960 h while τ 4 , NiTi as well as τ 1 phases are observed in the as-cast state.Comparing the BSE images in the same magnification for the alloys in annealed state and the as-cast state, we can see that the amount of NiTi is reduced while that of τ 1 is increased after annealing.According to the isothermal section at 900 °C (Fig. 5), the two-phase region of τ 1 + Ni 3 Ti exists at this temperature suggesting that the solidstate invariant reaction NiTi + τ 4 → τ 1 + Ni 3 Ti can be assigned to the first peak at 1098.3 °C on the heating curve.The peak at 1102 °C on the cooling represents the invariant ternary eutectic equilibrium L → τ 4 + NiTi + τ 1 , which is also observed for the annealed alloy 9#, as shown in Table 4.And the peak at 1135.4 °C on the heating represents the monovariant transition of eutectic type L → τ 4 + NiTi at 1135.4 °C.The peak at 1213.6 °C on the cooling curve of the sample 3# corresponds to the primary phase formation reaction L → τ 4 .However, no thermal effect was detected on the heating curve.The DSC curves of the annealed alloy 37# show three peaks during heating and cooling in Fig. 12(b).Three phases, τ 6 , τ 1 and fcc(Cu,Ni), were observed in both as-cast and annealed alloys, as shown in Table 3 and Fig. 2(e).
The liquidus projection was drawn to be consistent with the experimentally determined primary crystallization phases, solidification sequences and DSC results from this work.The ternary alloy compositions in this work are displayed and marked in Fig. 13.In the present work, we focused on the Nirich corner (more than 32 at.%),so the liquidus surface beyond this composition was from the literature [12,17,18].The primary phase regions of τ 1 and NiTi meet with the L + τ 1 + NiTi monovariant line as shown in Fig. 13.And the primary phase regions of τ 1 and τ 6 merge with the L + τ 1 + τ 6 monovariant line.This implies that after the primary nucleation of the τ 1 phase, the solidification course could pass along the L + τ 1 + NiTi monovariant line (corresponding to τ 1 + NiTi eutectic microstructure) or the L + τ 1 + τ 6 monovariant line (corresponding to τ 1 + τ 6 microstructure).Microstructures of the alloys 14# and 26# in Fig. 7 and the DSC result of alloy 9# (Table 4) confirmed this conclusion.As shown in Fig. 13, the primary phase regions of τ 4 and NiTi merge with the L + τ 4 + NiTi monovariant line, while the primary phase regions of τ 4 and τ 1 or τ 4 and τ 6 meet with the L + τ 4 + τ 1 or L + τ 4 + τ 6 monovariant line respectively, which indicates that the solidification path could pass along the L + τ 4 + NiTi monovariant line or the L + τ 4 + τ 1 monovariant line, or the L + τ 4 + τ 6 monovariant line after the primary nucleation of the τ 4 phase.There will be three cases of microstructure with solidification paths: when passing along the L + τ 4 + NiTi line, the primary τ 4 could be surrounded by the eutectic τ 4 + NiTi; when passing along the L + τ 4 + τ 1 line, it should be surrounded by τ 1 phase; when passing along the L + τ 4 + τ 6 line, it should be surrounded by the τ 6 phase, which is consistent with the observation of BSE images (Fig. 8) and the DSC result of alloy 3# in Fig 12(a).As shown in Fig. 13, the primary phase regions of τ 6 and τ 1 meet with the L + τ 6 + τ 1 monovariant line, which suggests that the solidification path could pass along this monovariant line after the primary nucleation of the τ 6 phase.Microstructure of alloy 31# (Fig. 9) and DSC result of alloy 37# (Fig. 12

Summary
Phase equilibria at 900 °C and liquidus projection in the Ni-rich corner of the Cu-Ni-Ti ternary system were experimentally investigated through XRD, SEM, EPMA and DSC measurements.Six two-phase and five three-phase regions at 900 °C were experimentally determined by annealed samples.Experimental results from the annealed samples indicate that the low-temperature ternary compound (τ 6 ) remains stable up to 900 °C, while the reported high-temperature compound (τ 5 ) was not observed at this temperature.By analyzing the phase composition and microstructure of the as-cast alloys, the primary phase τ 4 was experimentally determined, and the corresponding solidification paths were identified.Based on the experimental results of this work as well as the reliable literature information, the isothermal section at 900 °C and liquidus projection of Cu-Ni-Ti system as well as the reaction scheme are constructed, which shall be used in the following thermodynamic modeling of the Cu-Ni-Ti ternary system.
s strong and clear signal; w weak signal; Not detec.Not detected According to Fig. 4(a), there are two phases in the annealed alloy #1.According to the XRD results shown in Table 2, the matrix is the NiTi phase and the second phase is  1 distributed randomly.In Fig. 4(b) of the annealed alloy 22#, there are two phases -Ni 3 Ti wrapped in the net of fcc(Cu,Ni).It was also established that the microstructure of the alloy 3# is  1 + NiTi + Ni 3 Ti (Fig. 4(c)). 1 + Ni 3 Ti was observed in the annealed alloy 10# (Fig. 4(d)), and the NiTi + CuTi + CuTi 2 structures was observed in the annealed alloy 38# (Figs.4(e)).

Figure 14 .
Figure 14.Reaction scheme for liquid in the Cu-Ni-Ti system with temperature in °C.The invariant reactions among solid phases are not included

acknowledgement
The financial support from the project of Innovation-driven Plan in Central South University (Grant No. 2015CX004), Ministry of Industry and Information Technology of China (Grant No. 2015ZX04005008) and the National Key Research and Development Program of China (Grant No. 2017YFB0701700) are acknowledged.references

Table 1 .
Crystal structures and lattice parameters of the phases in the Cu-Ni-Ti system

Table 2 .
Summary of the phases and their compositions in the equilibrated Cu-Ni-Ti alloys after heat treatment at 900

Table 3 .
Summary of the primary phases and solidification paths of the Cu-Ni-Ti as-cast alloys

Table 4 .
Temperatures extracted from the DSC curves in the Cu-Ni-Ti system and their interpretation