Thermomechanical analysis on Ti-Ni shape memory helical springs under cyclic tensile loads

Recent developments of smart actuators using the two-way shape memory effect (TWSME) for industrial applications are becoming more common in scientific research. Shape memory alloys (SMAs) present some characteristics, which make it unique material to be use in applications that require strength and shape recovery. Ti-Ni alloys are an important class of memory alloys due to the shape memory effect (SME) and superelasticity (SE), which are govern by the thermoelastic martensitic transformation (MT). This material was been used to manufacture smart actuators for mechanical industry devices and several other applications in areas as medicine, robotics, aerospace, petroleum and gas industries. In general, Ti-Ni SMAs undergo one-step (austenite  martensite) transformation, but in some cases, there exists intermediate R-phase giving rise two-step (austenite  R-phase  martensite) transformation. This investigation is interesting due to the importance of knowing the actuators response to external stimulus (heat source, electrical current and/or external stress). In this work it was investigated the mechanical behavior in helical actuators produced from Ti-Ni alloy commercial wires. Different characterization techniques were employ for analyzing the samples as: differential scanning calorimeter (DSC), scanning electron microscopy (SEM), optical microscopy (OM) and a non-commercial apparatus developed to apply an external traction stress in helical actuators during thermal cycles.


Introduction
Recently, many studies is being developed in the materials field with the aim to understand the transformation phenomena which involving shape memory smart materials.These materials present phase transformations that allow the research and development of actuator/sensor elements.Mechanical properties of Ti-Ni alloys are very interesting for developing smart actuators manufactured from these non-conventional materials.In many technological applications these actuators need to generate force and to avoid the degradation of the shape memory effect caused by the martensitic stabilization processes 1,2 .Shape memory phenomena and mechanical properties of the alloys may induce adequate conditions to obtain sensors/actuators capable to produce mechanical work to be used in several industrial applications such as: medicine, robotics, aerospace, petroleum and gas.Several shapes might be used to obtain smart sensors/actuators, however helical spring shapes are more interesting because of its characteristics of deflection, power generation and elastic constant 3 .
This research was developed as a mechanical and metallurgical study of commercial wires with a composition of Ti-50.4at%Ni.Several heat treatments were investigated in order to modify the critical transformation temperatures.The objectives were to shift the critical temperatures to values close to room temperature.These heat treatments can promote the presence of the R phase.This phase appears by several factors, which include the chemical composition of Ti-Ni and metal forming processes.
Despite several studies to understanding the R-phase formation in the Ti-Ni alloys, some applications could use the narrow ranges of shape recovery presented by this phase.Several researches indicate that the presence of the R-phase during the martensitic transformation hardens the matrix phase (two-steps, B2→R→B19´, where B2 and B19´ represent crystalline structures of the austenitic and martensitic phases, respectively).This hardening hinders Materials Research the generation of stress fields that are associated with the dislocation reconfiguration processes.The presence of these stress fields promotes the Two Way Shape Memory Effect (TWSME) 4,5 .Study of thermoelasticity in shape memory springs will allow understanding the behaviors of austenitic and martensitic phases with external loadings.
Initially, the wire was subjected to some heat treatment and characterized by differential scanning calorimeter (DSC), scanning electron microscopy (SEM), optical microscopy (OM) and Energy dispersive spectroscopy.Then two heat treatments were selected to obtain the helical actuators.The actuators were tested in an apparatus developed to apply an external traction stress in helical actuators during thermal cycles.Two wires were tested in a dynamic mechanical analyzer (DMA).

Experimental Procedure
A cold-drawn wire of shape memory alloy (SMA) with chemical composition of Ti-50.4at%Ni and 0.89 mm diameter was used to manufacture a helical spring actuator.
Initially, wire samples were subjected to heat treatment to observe the behavior of the martensitic transformation critical temperatures (martensite start-M s , martensite finish-M f , austenite start-A s , austenite finish-A f , rhombohedral start or R-phase start-R s and R-phase finish-R f ).These critical temperatures were determined by differential scanning calorimetry (DSC).Thermal cycles were carried out in a range between -60 °C and 90 °C at a constant rate of 10 °C•min-1 (Mettler calorimeter, model 823 e ).Samples were homogenized in advance in a furnace at 400 °C (HT1) and 500ºC (HT2) for time of 1, 2, 4, 8, 12 and 24 hours, followed by quenching in water at 25 °C.
Two helical actuators were obtained by plastic deformation of the wire around a screw.Then, an assembly (wire/screw) was subjected to heat treatment at 400 °C for 24 hours, and then quenched in water at room temperature.The other set was treated at 500 °C.This process resulted in shape memory spring with outer diameter of 6.0 mm, 6.0 mm long and 4 active coils.
The internal stress and damping capacity was conducted in a dynamical mechanical analysis (DMA) system of the DMA Q800 da TA Instruments.The samples were submitted to a single cantilever fixation and a heating process was developed from 20 °C to 170 °C with rate of 5 °C•min -1 .Frequency and amplitude oscillation used were of 1 Hz and 5 μm, respectively.
The microstructure of the material was analyzed in order to identify precipitation and to visualize the martensitic variants.Scanning electron microscopy and energy dispersive spectroscopy were carried out to analyze samples submitted to HT1 and HT2 treatments.Samples were cold-mounted by a resin compound.Metallographic samples were sanded with sandpapers and then polished with aqueous solutions of alumina particles with size of 1 and 0.5 mm.The samples were attacked with HF-HNO 3 -CH 3 COOH (acide fluorrhydrique-acide nitrique-acide acétique) in the ratio of 2:5:5, for periods of 1 or 2 seconds.
Shape memory springs (SMS) were thermomechanically tested in a loading special apparatus.This device is constituted of a programmable silicon oil bath, a linear variation displacement transducer (LVDT), thermocouples and a computational data acquisition system.SMS were submitted to 40 thermal cycles under constant shear stress of 35, 70, 105, 135, 170, 200, 235 and 270 MPa in a temperature interval between 25 and 140 °C.The heating and cooling rate was estimated in 10 and 4 °C.min - , respectively 4,6,7 .
From data acquisitions, the curves of deformation versus temperature and temperature versus number of cycles were plotted.Figure 1 shows a typical graph obtained during thermomechanical cycling, as well as, the determination critical transformation temperatures under stress, by using of the tangent method, thermoelastic strains (ɛ t = difference between the largest and smallest displacement of LVDT), thermal hysteresis (H T ) and vertical shifts of hysteresis loops (X) 4,6 .

Influence of heat treatment on phase transformations
Figure 2a shows the curve of heat flux in function of temperature obtained by the DSC in a Ni-Ti wire sample in as-received conditions.DSC results do not shows any signal of transformation peaks for direct and inverse transformation.This can be explained from the plastic forming process to which the material was submitted (wire drawing process).During this process are introduced a large concentration of linear crystalline defects (dislocations).This large concentration of defects acts as an obstacle the occurrence of phase transformation through blocking of all variants of martensite 4,5,8,9 .
The calorimetric curves for the samples treated at 400 °C (HT1-24h) and 500 °C (HT2-24) for 24 hours are shown in Figure 2b, c.Table 1 shows the results of the calorimetric analysis of the samples treated at 400 °C and 500 °C for times of 1, 2, 4, 8, 12 and 24 hours.
Calorimetric analyzes show that heat treatments promoted changes in the martensitic transformation and changed their critical temperatures.All samples heat treated at 400 °C exhibit a transformation in two steps (B2→R→B19´) for all annealing periods.According to studies, two steps transformation occurs due to the formation of nickel-rich   precipitates during heat treatment in the matrix phase 10,11 .
Temperatures were dislocated in the sense of increasing the critical transformation temperature.Annealed samples at 400 °C showed A s variation from 33.8 °C for 1 hour to 48.5 °C for 24 hour ageing.M s showed variation from -21.2 °C for 1 hour to 0 °C for 24 hours ageing.
Samples heat treated at 500 °C showed two-step sequence: austenite, R-phase and martensite, indicating a trend of reduction between the peaks of the R-phase and martensite.In the test of sample aged for 12 hours shown that R-phase practically disappears, but present a single peak with large hysteresis thermal.
The modification observed on DSC results, where temperature and transformation phases are involved, this changes have been investigated in the literature and two main causes are reported: near-equiatomic Ti-Ni SMA lightly rich in nickel can form precipitates during heat treatment at high temperatures, such as Ti 3 Ni 4 , Ti 2 Ni 3 and TiNi 3 8,10,12 . These precipitates have influence on MT (martensitic transformation) because act as preferential sites for nucleation reaction, inhibiting and/or suppressing the appearance of R-phase.Precipitation occurs from diffusional processes which involve changes in local chemical composition, modifying transformation temperatures in such a way that increase of M s transformation temperature 7,[13][14][15][16] .A second important factor is about deformation stage or dislocation density configurations.This factor depends mainly of the material processing method, thermomechanical heat treatments and others.The wires used in this work were obtained by a cold-drawn process that results in a material with high dislocation density.Normally, alloys with high dislocation density show two transformation peaks during cooling which corresponding R and B19´phases 4,16,17 .

Microstructural investigation
Figure 3 shows images of optical and electronic scanning micrographs for heat-treated samples with HT1-24 h and HT2-24 h.The electronic microscopy were carried out on the square dashed and are shown in the larger rectangle in the upper right side (darker).Acicular structures similar to martensitic variants can be visualized.The images revealed morphological distinction between them.Figure 3a shows a high number of well-defined acicular structures with different orientation.Compared to samples heat treated at HT1, samples subjected at HT2 shows less acicular form and larger grains due to a higher heat treatment temperature.According to literature M s transformation temperature increase with the increasing of grain size 18,19 .
Figure 4 shows energy dispersive spectroscopy (EDS) results for samples submitted to HT1-24h and HT2-24h.EDS was realized in a small region of the samples.Both samples show peaks of Ti and Ni elements.Results indicate in a qualitative way that the samples studied are Ni-rich.This technique is not suitable, but the results are consistent with the Ni-Ti wire manufacturer's information.

Thermomechanical analysis
Thermomechanical heat treatment process were performed in thermal silicone oil bath, known in literature as a training process 20 .The apparatus used for this work, was able to developing the heating and cooling on the actuators, which consequently induced phase transformation in the samples.A constant shear stress was applied in the actuator in order to induce martensite variants in a preferential direction in relation with the applied stress.During the process the training leads formation of preferential variants who induces shape modification and changes on the transformation temperatures 21 .
The obtained graphs from training process allowed the studied on several shape memory effect parameter modifications (transformation temperatures, thermal hysteresis and thermoelastic strain (Et)).Thermal hysteresis and thermoelastic strain are also important parameters on the actuators behavior.Thermoelastic parameters are capable to indicate the shape memory efficiency as a function of the linear displacement.Figure 5 shows curves of thermoelastic effect versus temperature during forty cycles for tensile stress of 105MPa for HT1-24 and HT2-24 heat treatments.
Among the most important information obtained during the training process, it was observed the strain variation due to temperature, the vertical displacement on hysteresis loops, the increasing on the thermoelastic strain (E t ) caused by stress increment, besides of the double "s" observed during cooling on samples heat treated at HT1-24.The vertical displacement on hysteresis loops might be attributed to transformation induced plastically (TRIP) due to accumulation of small plastic strain 22,23 .Other factors also may contribute, as for example: spring actuator displacement due to rotation, martensitic stabilization process and martensitic variants reorientation 6 .
Figure 6 shows thermoelastic strain (E t ) versus number of cycles curves for each applied shear stress.Initially, thermoelastic strains show an increase according with stress   Evaluating samples submitted to HT1-24 it was observed that for 70 and 105 MPa a gradual increase during training cycles until practically strain stabilization for the last cycles.It was also observed for results obtained with stress of 200 MPa a sharp fall on thermoelastic strain to levels near the obtained for 135MPa.It is possible that this particular stress might be involved with R-phase transformation suppression.In this case (200 MPa) it was observed during experimental procedure that the two steps transformation (B2→R→B19´) starts to be replaced by one step transformation (B2→B19´) 12 .
During the training process occurs martensitic variants reorientation where internal stresses generated by dislocation fields becoming preferential according to stress direction 20 .For low and intermediate stresses, internal stress fields are formed gradually in each cycle, increasing thermoelastic strain during training.In this sense, 105 and 135 MPa stress present the best results in comparison with tests developed, whereas 170 and 200 MPa stress produce a decrease of the thermoelastic strain due to the rapidly saturation of internal stress fields.These behaviors have been observed in cooper based shape memory alloy springs 5 .
Actuators heat treated with HT2-24 also exhibit an increase on the thermoelastic strain with stress increments; however, for low levels of stress, the actuators produced a better efficiency on the SME comparing to HT1-24 results.Stress of 105, 135, 170 and 200 MPa present almost the same thermoelastic strain near 40.0 mm (see Figure 5).The main difference from HT1-24 results is that for 235 and 270 MPa, the actuators exhibit a better SME.The main results suggest that due to R-phase transformation mechanism and the developing of preferential variants orientation for samples heat treated at HT1-24h has a better SME efficiency 24 .
The efficiency of the TWSME might be evaluated by temperature transformation evolution.Figure 7 shows the evolution of (M s ) for both heat treatments.Samples heat treated with HT1-24 shows for all studied situations an increase on transformation temperature when increasing cycles and stress, and a more homogeneous temperatures evolution.The actuators obtained used from samples heat treated at HT2-24 showed higher values of M s and a high level of permanent deformation for 235 and 270 MPa which did not allow the continuity of the test for those conditions.

Training Process versus Internal Friction
It was already observed that internal friction in near equiatomic Ti-Ni alloys is related to stress fields for martensite reorientation.These fields are strongly related with the annealing temperature and their interaction with other lattice defects 25 .Heat treatment and training process seriously affects the internal friction behavior in Ti-Ni alloys 26,27,28 .These events have an importance on the reconfiguration of Ti-Ni alloy defects, inducing modification that might change the material properties such as memory effect, damping capacity, strength, hardness, and others 26,27,28 .Martensitic transformations are associated to the shear stress direction applied during the procedure.According to studies Ni/Ti ratio has influence on internal stress degree in the parent and martensite phases 26 .Internal defects propagation hinders the movement of martensite-austenite interface, inducing thermoelastic transformation to require more energy to overcome internal stress between phases.The increase on transformation energy can be visualized on the M s behavior, which increases with increasing applied stress and number of cycles (Figure 7) 6,20 .Figure 8 shows internal friction (tan δ) (Figure 8a) and storage modulus (Figure 8b) as function of temperatures during cooling step by DMA technique.Figure 8a indicates that the peak temperature for austenite to R-phase and R-phase to martensite transformation occur at 115 °C and 73 °C, respectively for HT1-24 heat treatment.Figure 8b indicates austenitic transformation to occur at 100 °C for HT2-24.Figure 8 also exhibit a high level of internal friction on samples submitted to HT1-24 where internal friction reaches 0.1528 (tan δ).Same studies developed a relation between R-phase transformation and damping capacity where the occurrences of R-phase significantly soften the storage modulus and thus increase the internal friction [29][30][31] .Samples heat treated with HT1-24 exhibit a higher internal friction, compared with HT2-24.
Figure 9 shows M s versus stress for 1, 5, 10, 15, 20 and 25 cycles.For both heat treatments M s increase when stress increase.For samples heat treated at HT1-24 (Figure 9a) was detected two points under the linear fit.These points refer to stress of 70 and 200 MPa, exactly the same values who exhibit a fall on thermoelastic strain (200 MPa -Figure 7).Changes observed for 200 MPa might be due to interaction of stress fields on martensitic and R-phase during training process.
Figure 9b shows samples heat treated at HT2-24 where M s for 35 MPa has been reduce when the number of cycles is increasing.For stress between 70 and 170 MPa, the M s exhibits a very narrow variation, being this temperature parameter very close for different cycles.For stress between 200 and 270 MPa, M s becomes very wide for increments with the number of cycles.

Conclusions
In this work we found that heat treatments cause in Ti-Ni alloys, slightly rich in nickel, an increase in critical transformation temperature.In the case of heat treatment at 500 °C (HT2-24), the R phase was apparently removed.In colorimetric tests thermal hysteresis shows a small increase of about 4 °C compared to the other treatments (A s -M s = 14 °C).In thermomechanical cycles test hysteresis loop was about 32 °C, which is much lower than samples with 1-24 HT treatment that exhibit hysteresis of about 50 °C due to the two steps transformation.
Internal friction shows that the HT1 treatment leaves samples more resistant.This fact can be verified by thermomechanical cycles tests.Probably the HT2 treatment promotes reduce on stress fields (dislocation), allowing greater thermoelastic deformations for small loads.HT1 treatment actuators have more strength and can be used for larger loads.However, there is a load limit.In this work we found for HT1 treatment, for 200 MPa a drop in thermoelasticity that should be related to interaction between the martensitic and rhombohedral transformations.

Figure 1 .
Figure 1.The typical curves of thermoelastic strain versus temperature and determination of thermoelastic properties.

Figure 9 .
Figure 9. Variations in Ms temperatures due to the stress applied during the thermomechanical cycles.

Figure 8 .
Figure 8. Internal friction tan δ and storage modulus as a function of temperature during step cooling for Ti-Ni alloy by DMA.

Table 1 .
Summary of results of the calorimetric properties of the samples treated at 400 °C and 500 °C.