In situ synchrotron X-ray diffraction analysis of deformation behaviour in Ti–Ni-based thin films

Deformation mechanisms of as-deposited and post-annealed Ti_50.2Ni_49.6, Ti_50.3Ni_46.2Cu_3.5 and Ti_48.5Ni_40.8Cu_7.5 thin films were investigated using the in situ synchrotron X-ray diffraction technique. Results showed that initial crystalline phases determined the deformation mechanisms of all the films during tensile loading. For the films dominated by monoclinic martensites (B19′), tensile stress induced the detwinning of 〈011〉 type-II twins and resulted in the preferred orientations of (002)_B19′ parallel to the loading direction (∥ LD) and (020)_B19′ perpendicular to the LD (⊥ LD). For the films dominated by austenite (B2), the austenite directly transformed into martensitic variants (B19′) with preferred orientations of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD. For the Ti_50.3Ni_46.2Cu_3.5 and Ti_48.1Ni_40.8Cu_7.5 films, martensitic transformation temperatures decreased apparently after post-annealing because of the large thermal stress generated in the films due to the large differences in thermal expansion coefficients between the film and substrate.

1. Introduction

Ti–Ni-based shape memory alloys (SMAs) have been extensively investigated in the past three decades due to their excellent properties, including shape memory effect, superelasticity and biocompatibility (Miyazaki & Ishida, 1999; Fu et al., 2001, 2004; Miyazaki et al., 2009; Sun et al., 2012). Substitution of Cu for Ni in the Ti–Ni SMAs has been well known to reduce the composition sensitivity to martensitic transformation temperatures, narrow the transformation hysteresis, and change the transformation temperatures (Otsuka & Ren, 2005). Recently there have been extensive studies using the sputter-deposited Ti–Ni and Ti–Ni–Cu thin films to make microactuators and microelectromechanical system (MEMS) components (Miyazaki & Ishida, 1999; Fu et al., 2004). In order to make high-performance micro-actuators using the Ti–Ni-based films, a systematic study of their deformation mechanism is crucial.

Previous investigations (Miyazaki & Ishida, 1999; Fu et al., 2001; Chu et al., 2000; Du & Fu, 2004; Zhang & Qiu, 2006; Ishida et al., 2013) have been focusing on the deposition, microstructure and thermomechanical characterization of these thin films, and various methods have been used to study the stress-induced phase transformation of the Ti–Ni-based SMA films, such as tensile tests (Ishida et al., 1996), dynamic mechanical-thermo analysis (DMTA) (Zarnetta et al., 2009) and transmission electron microscopy (TEM) (Zhang et al., 2006; Ishida et al., 2008; Meng et al., 2008a,b, 2009, 2011). Compared with the above ex situ methods, in situ diffraction characterization techniques during thermomechanical loading are capable of providing in situ analysis of microstructure evolution, exactly corresponding to the stress-induced phase transformation processes (Malard et al., 2012). Among these in situ diffraction techniques, the in situ synchrotron X-ray diffraction (S-XRD) method has been used as a powerful tool to study the stress-induced phase transformation of the bulk SMAs (Kulkov et al., 2000; Wang et al., 2007). However, little work has been done for in situ S-XRD testing of the deformation behaviour of thin-film SMAs, mostly due to the difficulties in microscale testing of the free-standing films (Miyazaki & Ishida, 1999; Wang et al., 2012).

For the Ti–Ni–Cu films with Cu > 10 at.%, little work has been reported on their in situ deformation behaviour, although TEM has been applied to ex situ investigate their deformation behaviours (Ishida et al., 2008; Meng et al., 2008a,b, 2009, 2011). In our previous reported work (Wang et al., 2012), in situ S-XRD has been used to investigate the stress-induced phase transformation behaviour of a Ti_50.1Ni_40.8Cu_9.1 film. Results showed that the deformation mechanism of the Ti_50.1Ni_40.8Cu_9.1 film is significantly dependent on the film composition, microstructure and initial phase state (Wang et al., 2012). In this paper the deformation mechanisms of various Ti–Ni and Ti–Ni–Cu films (Cu < 9 at.%) will be systematically studied using the in situ S-XRD method, and effects of the film composition, microstructure and post-annealing condition on the deformation mechanisms will be discussed.

2. Experimental

Thin films of Ti–Ni and Ti–Ni–Cu with three different compositions were deposited onto four-inch silicon (100) wafers by co-sputtering of Ti–Ni, Ti and Cu targets using a DC/RF magnetron sputter (Coaxial MSS3A, UK) with a base vacuum of 5 × 10⁻⁷ mTorr, which was achieved by pre-sputtering of Ti film and then pre-heating the chamber at 450°C for half an hour before loading the Si wafers. The substrate temperature during the deposition was 450°C. The composition was determined by wavelength-dispersive X-ray microanalysis (WDX) using standard samples of pure Ti and Ni (purity, 99.99%). The X-ray diffraction (XRD) measurements of the Ti–Ni and Ti–Ni–Cu films were carried out using a Philips PW3719 X-ray diffractometer. The Ti–Ni and Ti–Ni–Cu thin films on the Si substrate were post-annealed at 650 and 750°C for one hour in a vacuum furnace (ULVAC, Vacuum Annealing Furnace VHA-1015), with a based vacuum of 3 × 10⁻⁵ Torr. The deposition and post-annealing parameters of the films as well as their composition are listed in Table 1. The transformation temperatures listed in Table 1 were determined using a differential scanning calorimeter (DSC; SETARAM instrument 131) with a heating/cooling rate of 5°C min⁻¹. The following notation for the transformation temperatures was used in this paper: A_s, A_f, R_s, R_f, M_s, M_f, in which A: austenite (B2); R: martensite (R-phase); M: martensite (B19′); subscript s: starting temperature; subscript f: finishing temperature. After annealing, the free-standing films were peeled off from the Si wafer for tensile testing.

Table 1
Film deposition parameters and characterization properties

	Target power (W)				Transformation temperature (^oC)
Film composition (at.%)	Ti–Ni (RF)	Ti (DC)	Cu (DC)	Post-annealing temperature (^oC)	Ms (Rs)	Mf (Rf)	As	Af	Film thickness (µm)
Ti–Ni (Ti50.2Ni49.6)	Ti50Ni50 400	70	0	As-deposited	38 (63)	19.4 (58)	73	84	4.5
				650	64	40	74	100
				750	64.5	53.2	80.2	95.4
Ti–Ni–Cu3.5 (Ti50.3Ni46.2Cu3.5)	Ti55Ni45 400	0	2.5	As-deposited	50	38	65	75.3	4.5
				650	22	0	25.8	36.2
				750	23.4	16.1	40	55.7
Ti–Ni–Cu7.5 (Ti48.5Ni40.8Cu7.5)	Ti60Ni40 400	0	7	As-deposited	58	34	65	75	3.5
				650	20	0	27	43
				750	21	11	14	54

The experimental set-up for the in situ S-XRD measurements was previously reported by Wang et al. (2012), and the configurations and samples dimension are shown in Fig. 1. A miniature mechanical tester was used for the tensile test. The uniaxial loading experiments at room temperature (RT) were performed with the strains increased stepwise from 0.2% to 1.3% along the specimen's longitudinal direction, with a strain rate of 1.165 µm s⁻¹. The strain was determined using the movement of the crosshead of the mechanical testing machine, and the mechanical error of the apparatus was less than 0.02 mm with a stretching range of 100 mm. The error of the force sensor was less than 1 N. The loading direction (LD) was parallel to the direction of 0° azimuth angle (φ) in the diffraction pattern. According to Bragg's law, the data at φ = 0° correspond to crystal planes whose normals are parallel to the loading direction [i.e. (hkl) ∥ LD], and those at φ = 90° are contributed by the crystal planes whose normals are perpendicular to the loading direction [i.e. (hkl) ⊥ LD]. The in situ S-XRD measurements during tensile tests were performed at room temperature at the synchrotron radiation X-ray scattering station at the National Synchrotron Radiation Laboratory in Hefei, China. A monochromatic beam of hard X rays (λ = 0.13 nm) was used with a beam size of 4 mm × 0.5 mm. The incident direction of X-rays is perpendicular to the LD of the specimen. The measurements were performed using a transmission geometry. A two-dimensional image-plate detector was used to measure the diffracted X-rays. The software package Fit2D (Hammersley et al., 1996) was used to analyze the two-dimensional diffraction patterns, and then the total intensity was obtained parallel or perpendicular to the loading direction.

Figure 1 Experimental set-up for in situ S-XRD measurements and dimensions of the tensile specimen (Wang et al., 2012).

3. Results and analysis

3.1. Microstructure evolution after post-annealing

Fig. 2 compares the room-temperature XRD results of the as-deposited Ti–Ni and Ti–Ni–Cu films, and martensites are dominant in these films. For the Ti–Ni film, both the DSC and XRD results show that the transformation from R phase to B19′ phase has not finished at room temperature, thus the microstructure is a mixture of R phase and B19′ phase at room temperature, as shown in Fig. 2. For the Ti–Ni–Cu3.5 film, only the B19′ phase is obtained at room temperature. For the Ti–Ni–Cu7.5 film, both B19 phase and B19′ phase are observed. These are consistent with the results from the DSC measurement as listed in Table 1.

Figure 2 XRD results of the as-deposited Ti–Ni and Ti–Ni–Cu films at room temperature (X-rays, λ = 0.154 nm).

Fig. 3 shows typical TEM microstructures obtained from the Ti–Ni film. It shows a bright-field image of the as-deposited Ti–Ni thin film at room temperature, in which large twinned B19′ monoclinic martensite plates dominate. The martensite domains have fine striations in the Ti–Ni matrix. The diffraction patterns indicate that most of the martensites are 〈011〉 type-II twin plates. The grain size is about 0.5 µm from the TEM analysis.

Figure 3 Bright-field micrograph of as-deposited Ti–Ni film revealing that the monoclinic martensite plates are 〈011〉 type-II twins. Selected area diffraction patterns with incident electron beam along [101]M ∥ [−110]T.

In Fig. 4, fine striations of 〈011〉 type-II twins are also observed in the as-deposited Ti–Ni–Cu3.5 film. The twin widths vary from 0.01 to 0.1 µm, and the grain sizes are about 0.2–0.5 µm. The grain size of the films was found to slightly decrease with increasing Cu content (Ishida & Sato, 2011). However, in the Ti–Ni–Cu7.5 as shown in Fig. 5, (011) compound twins are the dominant plates of the B19 phase; the twin widths scatter over a range from 0.02 to 0.1 µm, with the average about 0.2 µm. Instead of 〈011〉 type-II twins, some random B19′ martensite variants are also found.

Figure 4 Bright-field TEM micrograph of as-deposited Ti–Ni–Cu3.5 film revealing the fine 〈011〉 type-II twins. Selected area diffraction patterns with incident electron beam along [101]M ∥ [−110]T.

Figure 5 Bright-field micrograph of as-deposited Ti–Ni–Cu7.5 film including the (011) compound twins. Selected area diffraction patterns with incident electron beam ∥ [100]M,T.

Figs. 6(a)–6(c) show room-temperature S-XRD results of the as-obtained Ti–Ni and Ti–Ni–Cu films post-annealed at different temperatures. After post-annealing, some martensites disappear, and their diffraction peaks shift to the lower angle side with the increase of annealing temperature. For the Ti–Ni film, as shown in Fig. 6(a), the peaks of the R phase disappear at room temperature because the finish temperature of the martensitic transformation increases with the increase of annealing temperature. This could be explained by the formation of Ni₄Ti₃ precipitates when the annealing temperature is increased to 750°C (Miyazaki & Ishida, 1999).

Figure 6 Room-temperature S-XRD results of Ti–Ni (a), Ti–Ni–Cu3.5 (b) and Ti–Ni–Cu7.5 (c) films post-annealed at 650°C and 750°C.

However, for the Ti–Ni–Cu3.5 film (Fig. 6b) and Ti–Ni–Cu7.5 film (Fig. 6c), both the start and finish temperatures of the martensitic transformation decrease apparently after post-annealing, and no significant amount of precipitates were observed. With the increase of annealing temperature of the Ti–Ni–Cu films, the intensities of austenite increase. This is consistent with the previous published results from the annealed Ti_50.1Ni_40.8Cu_9.1 film (Wang et al., 2012). Usually, with increasing annealing temperature of Ti–Ni–Cu films, the martensitic transformation temperature increases due to the formation of fine precipitates such as GP zones and Ti₂Cu precipitates (Ishida et al., 2008). In this paper, when heating the film and substrate from room temperature to annealing temperature, a large thermal stress generated due to the different thermal expansion coefficients of the film and substrate as well as evolution of film intrinsic stress may resist the formation of fine precipitates such as GP zones and Ti₂Cu precipitates. The thermal stress can be estimated from the following equation (Fu et al., 2005),

where α is the coefficient of thermal expansion, E is Young's modulus in austensite, ν is the Poisson ratio, the subscript `Si' indicates the silicon substrate and that of `f' indicates the film. We only considered the austensite for simplicity during the calculation, thus E_f was taken as 80 GPa (austenite), and α_Si and α_f (austenite) as 3 × 10⁻⁶ K⁻¹ and 15.4 × 10⁻⁶ K⁻¹, respectively, and ν_f as 0.3 (Fu et al., 2005). The calculated values of 650°C and 750°C annealed Ti–Ni–Cu3.5 samples are 893 MPa and 1034 MPa, respectively. If the martensitic transformation occurs during cooling, these stresses will be released significantly due to the self-accommodation of martensite (Grummon et al., 1999). However, it is possible that some small precipitates or defects formed during annealing, and they may have changed the intrinsic stress and cause a decrease of the transformation temperature.

3.2. In situ S-XRD results of Ti–Ni film

Fig. 7 displays the in situ S-XRD results of the 750°C annealed Ti–Ni film. Before the tensile deformation, the dominant phase in the Ti–Ni film is B19′. During tensile loading, along the LD, the peak intensities of (002)_B19′ keep increasing but those of (020)_B19′ decrease. The peak intensities of (−111)_B19′ and (110)_B19′ decrease along the LD, and those of the (111)_B19′ do not show significant changes. The Ni₄Ti₃ remains stable during loading. Whereas, perpendicular to the LD, all the changes in the peak intensities show reverse trends, although those of (−111)_B19′ do not show the apparent changes. The detailed re-orientation mechanisms of martensite variants during tension along two directions will be discussed in §4.

Figure 7 S-XRD results of 750°C annealed Ti–Ni film for (a) 0° azimuth angle and (b) 90° azimuth angle at different loading stages (at RT).

3.3. In situ S-XRD results of Ti–Ni–Cu3.5 film

Fig. 8 shows that, before the tensile test, martensite (B19′) is dominant for the 650°C annealed Ti–Ni–Cu3.5 film. During tensile loading, as shown in Figs. 8(a) and 8(b), the peak intensities of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD increase, but those of the same crystal planes decrease in the directions of (002)_B19′ ⊥ LD and (020)_B19′ ∥ LD. The intensities of the (−111)_B19′ and (111)_B19′ peaks do not show significant changes during the tensile process. This indicates that the orientations of certain martensite plates have been enhanced at the expense of the others, which is attributed to the detwinning or reorientation of martensite variants during uniaxial loading. Clearly, during loading, the applied stress induces changes of martensite variants and results in the preferred re-orientations of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD, which will be discussed in §4.

Figure 8 S-XRD results of 650°C annealed Ti–Ni–Cu3.5 film for (a) 0° azimuth angle and (b) 90° azimuth angle at different loading stages (at RT).

After annealing at 750°C, the Ti–Ni–Cu3.5 film shows a different room-temperature microstructure from the 650°C annealed sample, which can be revealed from the S-XRD results shown in Figs. 2 and 6. Austenite becomes dominant in the film after annealing at 750°C as shown in Fig. 9.

Figure 9 S-XRD results of 750°C annealed Ti–Ni–Cu3.5 film for (a) 0° azimuth angle and (b) 90° azimuth angle at different loading stages (at RT).

Similar to the report by Wang et al. (2012), the differences in the initial microstructures result in significant differences in the deformation behaviour of films. For the 650°C annealed Ti–Ni–Cu3.5 film, the intensities of the austenite change slightly, and the main deformation mechanism is detwinning of martensite variants. With the increase of tensile stress, the peak intensities of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD increase, but those of the same crystal planes decrease along their perpendicular directions. For the 750°C annealed sample, with the increase of tensile stress, the intensity of the austenite peak (110)_B2 decreases and that of the martensite peaks slightly increases. Clearly the increase rates of the martensite peaks are quite different. The peaks of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD increase more significantly than those of the other oriented martensites. This clearly indicates that martensitic transformation and detwinning processes occur at the same time during deformation, and stress-induced martensitic transformation also occurs with (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD.

3.4. In situ S-XRD results of Ti–Ni–Cu7.5 film

Fig. 10 shows the tensile behaviour of in situ S-XRD results of the Ti–Ni–Cu7.5 film annealed at 750°C. The initial microstructure is dominant by austenite, and tensile-loading-induced martensitic transformation is preferred along (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD. Comparing the results in Fig. 10 with those in Figs. 7–9, it is concluded that the annealing temperature changes the microstructures of the Ti–Ni-based films before the deformation, and thus has a significant effect on the deformation behaviour of the films. During loading, it is found that the peak intensities of S-XRD for (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD increase, but this is opposite to those of the same crystal planes in their perpendicular directions. This is similar to the tensile behaviour of 750°C annealed Ti–Ni–Cu3.5 film.

Figure 10 S-XRD results of 750°C annealed Ti–Ni–Cu7.5 film for (a) 0° azimuth angle and (b) 90° azimuth angle at different loading states (at RT).

4. Discussion

After cooling the Ti–Ni-based films from high annealing temperatures, some twin-related neighbouring variants inside the films change their orientations to form martensitic structures in order to accommodate elastic strains applied to the martensites (Christian, 1982). In the Ti–Ni-based alloys, 〈011〉 type-II and (111) type-I twins were commonly observed (Otsuka & Ren, 2005), and these twinning structures were also reported to be dominant in Ti–Ni–Cu thin films with lower Cu content (less than 9%) (Miyazaki & Ishida, 1999). For the 〈011〉 type-II twins structure, its twin crystals are related to each other by a 180° rotational symmetry along the [0−11] axis (Liu & Xie, 2003). However, for the (111) type-I twins, the twin crystals will not rotate, but are symmetrically distributed as a mirror with respect to the (111) plane. The TEM results in Figs. 3–5 clearly show the large amount of martensitic twins in the Ti–Ni-based thin films.

During the tensile loading of the above-investigated Ti–Ni-based films (see Figs. 7–10), the peaks of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD increase, but the same crystal planes decrease perpendicularly to the loading directions. This trend could be well explained by reorientation of the martensitic variants (Krishnan, 1998; Ezaz & Sehitoglu, 2011), considering the existence of multiple martensitic twinning structures in these Ti–Ni–Cu thin films. For Ti–Ni and Ti–Ni–Cu3.5 films, when detwinning occurs during loading, the two twin-related variants would re-orientate and only one type of variant is left at the end of the detwinning process. Due to the rotation of the variants, the intensities of some crystal planes would decrease whereas the other planes increase, according to their orientations compared with the loading directions.

The in situ S-XRD results (Figs. 7–9) could explain the sequences of the detwinning process. The presence of (020)_B19′ and (002)_B19′ gives indirect evidence of the existences of 〈011〉 type-II twins or (111) type-I twins. Because all the above diffraction results reveal the coexistence of (020)_B19′ and (002)_B19′ planes, this clearly indicates that there are multi-variants of martensites in these annealed films. Furthermore, the interplanar angle between (020)_B19′ and (002)_B19′ is 90° in a martensitic unit-cell, the (020)_B19′ plane is obtained by rotating the (002)_B19′ plane by 180° along the [0–11]_B19′ direction as shown in Fig. 11(a), and the (011)_B19′ plane is the symmetrical plane of (002)_B19′ and (020)_B19′ in geometry as shown in Fig. 11(b). The twinning elements of the 〈011〉 type-II twins are as follows: twinning plane K₁ = (0.7205, 1−1), shear direction η₁ = [011], K₂ = (011) and η₂ = [1.5727 1−1]. Thus, during the detwinning of 〈011〉 type-II twins, the intensities of (002)_B19′ and (020)_B19′ change inversely, but those of (011)_B19′ more or less remain at similar intensities. Although the (011)_B19′ plane could not be clearly identified from the S-XRD results, another two planes of (−111)_B19′ and (111)_B19′, which both have equal dihedral angles with the (002)_B19′ and (020)_B19′ planes, do not change apparently during the deformation of the films (Liu & Xie, 2003). This suggests that the symmetrical planes of (002)_B19′ and (020)_B19′ planes would not be significantly rotated. Therefore, we could conclude that 〈011〉 type-II twins exist in these Ti–Ni and Ti–Ni–Cu films, and the deformation of martensite is contributed to by the detwinning of 〈011〉 type-II twins. This results in the preferred orientations of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD. There could be other modes of twinning structures, but it is difficult to estimate the twinning mode using the S-XRD method. What we could confirm is that the occurrence of detwinning corresponds to the intensity changes of these martensitic peaks.

Figure 11 Stereographic projections of B19′ martensite showing the crystallographic relationship of the (002) and (020) planes: (a) 180° rotational symmetry about the [011] direction; (b) mirror symmetry about the (011) plane.

It was reported that when the Cu content is larger than 15% the Ti–Ni–Cu thin film undergoes a one-step martensitic transformation from B2 to B19 (Miyazaki & Ishida, 1999). However, when the Cu content of the Ti–Ni–Cu films is lower than 9%, the path of transformation would be B2 → B19 → B19′. When the Cu content is low and Ti is rich in the film, the transformation is only from B2 to B19′. The maximum Cu content of the films in this paper is about 7.5%. It is found that both B19 and B19′ martensite are dominant in Ti–Ni–Cu7.5. However, for the Ti–Ni–Cu films with lower Cu content, most are dominated by the B19′ martensite at the initial state, and the martensite variants are formed as 〈011〉 type-II twins which are often observed in Ti–Ni alloy.

5. Conclusions

Effects of film composition, microstructure and post-annealing condition on the deformation mechanisms of the Ti–Ni and Ti–Ni–Cu films were investigated using in situ S-XRD. The following conclusions can be obtained.

(i) With increasing annealing temperature, the growth of Ni₄Ti₃ in Ti_50.2Ni_49.6 film increases the transformation temperature, but the stress field of Ni₄Ti₃ could usually suppress the martensitic transformation.

(ii) For the Ti_50.3Ni_46.2Cu_3.5 and Ti_48.1Ni_40.8Cu_7.5 films, the martensitic transformation temperatures decrease with the increase of annealing temperature because of the thermal stress generated due to the different thermal expansion coefficients of the film and substrate as well as evolution of film intrinsic stress.

(iii) The deformation behaviours of the Ti–Ni or Ti–Ni–Cu films are determined by the initial state of these films. When the sample is dominated by the B19′ phase, detwinning of 〈011〉 type-II during tension would result in the preferred orientations of (002)_B19′ ∥ LD and (020)_B19′ ⊥ LD.

(iv) When annealing the films at high temperature, the B2 phase becomes dominant, and the austenite would be directly converted into martensites during loading with the same texture distributions for the situation where the B19′ phase is dominated.