Effects of composition on transformation temperatures and microstructure of NiTi-Hf shape memory alloys

High transformation temperature shape-memory alloys (HTSMA) usually present a martensitic transformation temperature (Ms) starting at 100 oC. That is the case of high nickel Ni-Ti-Hf alloys. This article presents experimental results obtained from arc melting of Ni50Ti50-XHfX .at% (X = 8, 11, 14, 17 and 20 .at%) alloys. This process homogenized every composition with similar relative crystallinity. Results confirm that transformation temperatures (TT) increase with increasing the amount of Hf. A martensitic matrix is formed by two metastable phases: R and B19’. From all the alloys studied, the B19’ phase presented the highest percent fraction. Gradually adding Hf3 .at% promoted a slow increase of crystalline fraction of R phase and a slow reduction of phase (Ti, Hf)2 Ni, located at grain boundaries. Coherent/semi-coherent interface between (Ti, Hf)2 Ni phase and the matrix may intensify the driving force for the formation of R phase, present on X-ray diffractograms.


Introduction
SMA are materials that have the ability to retrieve their original shape under appropriate thermal conditions.This is possible due to a solid state phase transformation named Thermoelastic Martensitic Transformation which restores great deformations (between 8 and 10%) (Besseghini et al., 1999).
Among these tertiary elements that favor high temperature applications of SMA, Hf is a chemical element economically accessible.There are three kinds of Ni-Ti-Hf SMA: interstitial (Ni 50-X Ti 50 Hf X .at%:rich in titanium), equiatomic (Ni 100-X/2 Ti 100-X/2 Hf X .at%(5 ≤ X ≤ 20)) and substitutional (Ni 50 Ti 50-X Hf X .at%:rich in nickel) (Zarinejad et al., 2008).The present paper studied the latter due to the fact that it has been broadly discussed in literature (Sanjabi et al., 2005), when compared to the former two kinds.The objective of this study was to perform a thermal, physical and structural characterization of Ni 50 Ti 50-X Hf X .at%shape memory alloys, adding Hf as a substitutional atom to Ti.This study contributes to practical applications of advanced materials such as sensors and actuators in the most varied sectors: automotive, oil and aerospace (Karaca et al., 2013).
The major difference of this article is the level of detail given to the crystalline fractions and their respective constituent phases, correlating them with the contents of added Hf and heat treatments, besides approaching them qualitatively and quantitatively based on nominal compositions not yet examined by literature.

Experimental procedure
Five nominal compositions of Ni-50 Ti 50-X Hf X .at%alloy were prepared with pure Ni (99.98%),Ti (99.99%) and Hf (99.95%) in an arc melt furnace, as shown in Figure 1(d).The elements were placed accordantly to its melting temperature, in a way that elements with higher melting point stayed on top of the others, as presented in Figure 1(a).This procedure ensures that all elements will melt.The preparation of the metal portions was done by using a method that starts with the calculation of the Proportional Masses of the "n" chemical elements involved in the nominal composition (Soares, 2016), as explained in Table 1.The elements used were previously weighed (analytical scale with a precision of 0.0001g) considering 10 g of alloy and pickled (chemical process of superficial oxidation removal) with acid solutions.Next, the samples were cleaned with acetone or ethylic alcohol under ultrasonic agitation.The elements were then melted to a single ingot and remelted three times taking care to flip the ingot over between each melt in order to achieve a higher composition homogeneity.The chamber of arc melting shown in Figure 1(c) was evacuated to 10 -5 Torr and filled with ultrapure argon at 0.5 atm; this purge process was repeated three times.At the last purge, the chamber was pressurized with argon at 0.9 atm.The residual oxygen was eliminated by melting and cooling a titanium getter (figure 1(b)) for at least two times.This is extremely important because the slightest change in composition and the presence of a few ppm of oxygen may change the properties of the alloy.The maximum torch temperature in the furnace is about 3600 o C and time of incidence on the metals was around 15 seconds (each melting).
Thermal characterization of the alloys was performed by differential scanning calorimetry (DSC) using equipment 404 from Netzsch with a heating rate of 10 ºC/min in a cop-per crucible in an ultrapure argon (99.998%) atmosphere.
An X-ray Diffraction (XRD) technique was used to identify phases and compounds in the alloys aided by X'Pert HighScore Plus V3.0 software from PA-Nanalytical.The equipment employed was a SHIMADZU X-ray diffractometer model XRD 6000, operating with CuKα radiation (λ = 1.54056Å), at a voltage of 40 kV and current of 30 mA.The 2θ scan (12º-90 º) was from 78 º and speed of 2 º/min.Slits used were:: (i) divergence = 1 º; (ii) dispersion = 1º; e (iii) reception = Ø 0.3 mm.Data collected were enough for structure refinement by the least-squares method (Rietveld, 1967).It was also possible to perform the indexation of crystallographic planes.Results obtained presented chi-squared X 2 = (R wp /R exp ) 2 or GOF (goodness of fit) lower than three (X 2 < 3), which is a value considered acceptable by the international crystallographic union.
The relative degree of crystallinity X c = I calc /( Icalc + I back ).100 was calculated considering the sum of integrated areas, of peaks (calculated intensities) and of base line (background intensitie) (Colman et al., 2014).Icalc were adjusted by the Pseudo-Voigt function (gaussian + lorentzian) and Iback was modeled by Sonnevelds's empiric polinomy (Sonneveld e Visser, 1975), composed by two coefficients.
For microstructural characterization and determination of chemical composition, samples were ground, polished and etched with a solution of H 2 O:HNO 3 :HF and then observed by a Scanning Electron Microscope (SEM), model VEGA 3 from TESCAN, with EDS module, which does not have a heating/cooling chamber.(Rietveld, 1969).Results are presented in Table 2 and show that the volume of the crystalline cell (space group P 1 21/m 1) evolves as the content of Hf increases.The characterization of B19'phase, refined using Card 99-003-0004, allowed the calculation of Miller indexes (h k l), as shown in Table 3.The number of indexed peaks with only the B19'phase was 77 (at the Hf8 .at%alloy), 73 (at the Hf11 .at%alloy), 20 (at the Hf14 .at%alloy), 233 (at the Hf17 .at%alloy) and 194 (at the Hf20 .at%alloy).As an example, the nominal composition Ni 50 Ti 36 Hf 14 .at%alloy was chosen to show the indexation of crystalline planes in Table 3.The adjustment of the values obtained reveals that there was a great fitting of experimental data in relation to standard CIF values, besides demonstrating a good level of repetition of this structure in the crystalline reticulate.The characterization of B19'phase, refined using Card 99-003-0004, allowed the calculation of Miller indexes (h k l), as shown in Table 3.The number of indexed peaks with only the B19'phase was 77 (at the Hf8 .at%alloy), 73 (at the Hf11 .at%alloy), 20 (at the Hf14 .at%alloy), 233 (at the Hf17 .at%alloy) and 194 (at the Hf20 .at%alloy).As an example, the nominal composition Ni 50 Ti 36 Hf 14 .at%alloy was chosen to show the indexation of crystalline planes in Table 3.The adjustment of the values obtained reveals that there was a great fitting of experimental data in relation to standard CIF values, besides demonstrating a good level of repetition of this structure in the crystalline reticulate.The calculated relative crystallinity was 58.16% (Hf8 .at%),63.88% (Hf11 .at%),63.58% (Hf14 .at%),67.47% (Hf17 .at%)and 65.08% (Hf20 .at%).Close values confirm the uniform homogeneity of the 5 compositions, besides expressing a good level of crystallization during processing.

Results and discussion
As an example, Figure 3 presents in detail the form by which X c (%) was calculated in the alloy with 11 .at% of hafnium.by (Wu et al., 2010) that precipitation of Ti 2 Ni may also induce formation of the R phase.Herein, formation of the R phase is related to precipitation of the (Ti, Hf) 2 Ni phase (Yi et al., 2017).Figure 2 presents X-ray diffraction patterns showing the coexistence of the R and (Ti, Hf) 2 Ni phases.Hence, precipitation of (Ti, Hf) 2 Ni may play an important role in the sequence of transformation in Ni-Ti-Hf alloys.According to literature (Suresh et al., 2014), an increase in the amount of Hf in the alloy favors a decrease in the amount of (Ti, Hf) 2 Ni phase, as shown in Figure 4.This precipitation at the grain boundary creates a coherent/ semi-coherent interface between the precipitate (Ti, Hf) 2 Ni and the matrix, which may intensify the driving force for formation of the R phase.According to Figure 4, considering the relative crystallinity (X c ) of each composition, an increase in the amount of the R phase was experimentally observed and is related with the addition of Hf and Ni (Karaca et al., 2011).(Suresh et al., 2014).
Considering the martensitic matrix (R + B19'), according to Figure 6, percentage values of both instable phases are inversely proportional and establish logarithmic relationships.Decreasing values of total hysteresis establish a linear function and indicate that the R phase (rhombohedral) is more unstable than B19' (monoclinic).
The presence of precipitates inhibits the transformation B2 → B19' while favoring the transformation B2 → R (Moshref-Javadi et al., 2013).As Hf is added, the alloys get richer in nickel and consequently reduce the formation of precipitates at grain boundaries, as shown by diffractometry results.Hence, it is possible to conclude that in Ni-Ti-Hf-based alloys, the transformation B2 → R is more affected by the amount of nickel than by the effect of precipitates (Moshref-Javadi et al., 2013).
Transformation temperatures are summarized in Table 4.The increase of TT in HSTMA alloys of Ni-Ti systems are probably also related to the fact that the third element atomic radius and melting point are higher than Ni's and Ti's.In our case, it is observed once the hafnium atomic radius is 1.55 Å and its melting point is 2232.85ºC, while nickel has r = 1.24Å and T m = 1454.85ºC, the same way Ti presents r = 1.40 Å and T m = 1667.85ºC (Lide, 1999).Furthermore, some authors (Potapov et al., 1997;Firstov et al., 2004;Evirgen et al., 2015) attribute the increase of transformation temperature in Ni-Ti-Hf to the increase in volume of the unit cell on martensitic transformation B19', which increases the stored elastic energy.Table 4 Transformation Temperatures (TTs) measured for Ni 50 Ti 50-X Hf X .at%alloys.
Total enthalpies (μ = 12 J/g), shown in the last column of Table 2, have close values which indicates energy densities consistent with those that are normally required in the shear of atomic blocks of the crystalline network (>10 J/g).
It can be seen in the SEM micrograph of Figure 7 that the microstructure consists of precipitates of the (Ti,Hf) 2 Ni phase formed preferentially at grain boundaries of the Ni-Ti-Hf matrix.Grain boundaries act like preferential nucleation sites for the precipitated phase because they minimize the interfacial energy between matrix and precipitates.Heterogeneous lenticular precipitation of the (Ti, Hf) 2 Ni phase occurs due to its nucleation being preferentially at grain boundaries.While diffusivity at the grain boundary is larger than the diffusivity at crystalline network, precipitation at the grain boundary is expected.Because of the concentration gradient at grain boundaries, resultant of the solute rejection during formation of (Ti, Hf) 2 Ni phase, solute atoms diffuse through the crystalline network toward the grain boundaries.Coherent or semi-coherent planar interfaces are formed when interface energy between matrix and precipitate is minimum.Therefore, the occurrence of these interfaces is strongly dependent of the orientation of grain boundaries.However, the presence of planar interfaces may intensify a driving force to promote the transformation of the R phase in the alloy (Suresh et al., 2014).
Reduction of transformation hysteresis due to redistribution of Ni, has been reported in literature (Meng et al., 2001).However, for the alloys of the present study, a compositional change does not seem to be the only reason for low values of hysteresis observed in Table 4. Enrichment of the alloys by Ni have a greater effect on Ms temperature than on As or Af temperatures (Meng et al., 2002), as shown in table 4.
The (Ti, Hf) 2 Ni precipitate is richer in Ti compared to Hf.However, the amount of Hf in (Ti, Hf) 2 Ni phase, present in the matrix, is always higher than the amount of Ti, according to Figures 8(a) and 8(b).Thus, considering only the precipitated phase (Ti, Hf) 2 Ni, the higher amount of Hf in the matrix compensates the Ti loss.The atomic quantification of these elements is organized in the first two lines of Table 5.In relation to the martensitic matrix Ni-Ti-Hf phase, the amount of the elements were quantified by EDS.Results are shown in Table 5 and are close to the values adopted as nominal.This reinforces the idea that the preparation of the alloys had an acceptable level of precision due to the proximity of the values measured by EDS to the theoretical values.

Conclusions
The influence of the amount of Hf on transformation temperatures and on microstructural evolution of Ni 50 Ti 50- X Hf X .at%shape-memory alloys was studied.Based on experimental results, correlated with literature, it may be concluded that: 1.The microstructure consists of (Ti, Hf) 2 Ni precipitates formed prefer-entially at the grain boundaries of the Ni-Ti-Hf (R + B19') phase; 2. The increase in the amount of Hf in the alloy favors reduction in the amount of (Ti, Hf) 2 Ni phase, which precipitates at grain boundaries; 3. The (Ti, Hf) 2 Ni phase forms a coherent/semi-coherent interface between the (Ti, Hf) 2 Ni precipitate and the matrix, which may intensify the driving force to formation of the R phase; 4.The increase in the amount of Hf in Ni-Ti-Hf alloys increases phase transformation temperatures as well as the volume of monoclinic structures (B19' phase), but also favors the reduction of total hysteresis.
Figure 1 (a) chemical elements Ni, Ti and Hf; (b) open chamber of the arc melting furnace and arrangement of metals in bas-relief; (c) bench with closed chamber for evacuation, purge and melting; (d) external elements of the furnace and gauge detail.

Figure 2
Figure 2 presents normalized room temperature diffractograms of the alloys studied.Diffracted peaks

Figure 2
Figure 2 CuKα1 radiation diffractograms of the five Ni 50 Ti 50-X Hf X at.% alloys.B19' phase is predominant throughout the entire martensitic matrix.Lattice parameters of the monoclinic structure Figure 3 Visual separation of the 3 Intensities: observed (experimental), calculated and background.

Figure 4
Figure 4Crystalline fractions of the three identified phases.

Figure 5
Figure 5 presents DSC thermograms of studied alloys.The sorting by composi-

Figure 5
Figure 5 DSC results for the 5 nominal compositions of Ni 50 Ti 50-X Hf X .at%alloys.

Figure 8
Figure 8Punctual measurements of (Ti, Hf) 2 Ni phase present at grain boundary (spectrum 2) and inside the grain (spectrum 4).

Table 2
Unity cells parameters of Ni 50 Ti 50-X Hf X .at%alloys.