Two way shape memory effect in NiTiHf high temperature shape memory alloy tubes

Accepted Manuscript Two Way Shape Memory Effect in NiTiHf High Temperature Shape Memory Alloy Tubes C. Hayrettin, O. Karakoc, I. Karaman, J.H. Mabe, R. Santamarta, J. Pons PII:

S1359-6454(18)30770-5

DOI:

10.1016/j.actamat.2018.09.058

Reference:

AM 14867

To appear in:

Acta Materialia

Received Date: 16 September 2018 Accepted Date: 25 September 2018

Please cite this article as: C. Hayrettin, O. Karakoc, I. Karaman, J.H. Mabe, R. Santamarta, J. Pons, Two Way Shape Memory Effect in NiTiHf High Temperature Shape Memory Alloy Tubes, Acta Materialia (2018), doi: https://doi.org/10.1016/j.actamat.2018.09.058. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Strain in Martensite 110 MPa

3

90 MPa

Thermomechanical training

2 45 MPa

1

2.5 2.0 1.5 1.0 0.5

21 MPa

0.0

0

0

150 200 Temperature (°C)

TE D

After training, oriented, thin martensite variants, and retained austenite

EP

Before training, large martensite variants and nano-precipitates

250

AC C

100

200

Ni50.3Ti29.7Hf20 HTSMA Torque Tubes Heat Treated at 550 °C for 3 Hrs

400

Thick tubes (#1 and 2) trained at 145 MPa Thick tubes (#3 and 4) trained at 200 MPa Thin tubes (#5 and 6) trained at 145 MPa

600 800 1000 Number of Cycles

1200

1400

1600

Two-way shape memory effect is thermally stable up to 450°C Two-Way Shape Memory Strain (%)

4

Strain in Austenite Actuation Strain

M AN U

Shear Strain (%)

145 MPa

3.0

RI PT

5

3.5

SC

Ni50.3Ti29.7Hf20 Polycrystalline Torque Tubes with nano-precipitates

Two-Way Shape Memory Strain (%)

ACCEPTED MANUSCRIPT

4

3 Before Annealing TWSMS

2

1

0 350

400

450 500 550 600 Annealing Temperature (°C)

650

700

ACCEPTED MANUSCRIPT

Submitted to Acta Materialia in September 2018

Two Way Shape Memory Effect in NiTiHf High Temperature Shape Memory Alloy Tubes

Department of Materials Science and Engineering, Texas A&M University, College Station, TX 77843, USA b The Boeing Company, St. Louis, MS 98124, USA c Departament de Fısica, Universitat de les Illes Balears, E07122 Palma de Mallorca, Spain *: Corresponding author: [email protected]

SC

a

RI PT

C. Hayrettina, O. Karakoca, I. Karamana*, J. H. Mabeb, R. Santamartac, J. Ponsc

Abstract

Two-way shape memory effect (TWSME) in nano-precipitation hardened, Ni50.3Ti29.7Hf20 high

M AN U

temperature shape memory alloy (HTSMA) thin walled tubes and its thermal stability were investigated. Torsional TWSME was induced in the thin wall tubes by repeated thermal cycling across their martensitic transformation under applied shear stress. The effects of training parameters and geometric factors, such as the number of training cycles, shear stress levels, and thickness of the tube walls, on the resulting TWSME were evaluated. Thermal stability of TWSME was characterized as a function of annealing treatments at elevated temperatures. It was found that under 200MPa, 600 thermal cycles were sufficient to reach a two-way shape memory strain (TWSMS) as high as 2.95%, which was shown to be stable upon

TE D

annealing up to 400°C for 30 minutes. This TWSMS was 85% of the maximum measured actuation strain under 200MPa. The microstructure after thermo-mechanical training was investigated using transmission electron microscopy (TEM), which did not indicate a significant change in precipitate structure and size after the training. However, small amount of remnant austenite was revealed at 100°C below the

EP

martensite finish temperature, with notable amount of dislocations. Overall, it was found that nanoprecipitation hardened Ni50.3Ti29.7Hf20 shows relatively high TWSMS and stable actuation response after

AC C

much less number of training cycles as compared to binary NiTi and nickel lean NiTiHf compositions. Tube wall thickness and training stress levels have been found to have negligible effect on shape memory strains and number of cycles to reach the desired training level, for the ranges studied.

Keywords: High temperature shape memory alloys; Two-way shape memory effect; Torque tubes; NiTiHf; Thermo-mechanical training.

1

ACCEPTED MANUSCRIPT

1. Introduction There is an increasing demand for lightweight and compact actuators in aerospace and automotive applications [1-6]. Solid state shape memory alloy (SMA) actuators are promising candidates as replacement for electrical motors or hydraulic systems in weight and space critical applications since they

RI PT

provide the highest energy density among all known active materials and systems [1, 6-9]. Additionally SMA actuators feature reduced complexity and part count. SMAs undergo large recoverable shape changes under high applied stress levels and generate comparable or higher work output than all other known solid state actuator materials [7, 10]. The majority of SMA actuators are designed to operate for one-way shape memory effect (OWSME) [6, 10, 11]. OWSME provides repeated actuation, only if there

SC

is an external biasing force to reset actuation after each cycle. However, two-way shape memory effect (TWSME) leads to repeatable actuation under no external stress, the internal stresses that lead to TWSME

M AN U

can also produce actuation work against an opposing external stress [12, 13]. TWSME eliminates the need for a biasing force to reset the actuator, as well as the components associated with the application of biasing force, providing simplicity, and reduced need for maintenance.

Binary NiTi SMAs are the most widely used and commercially available SMAs. One disadvantage of NiTi in emerging actuator applications is its low transformation temperatures, which may lead to unintentional actuation in service at warm climates. In addition, there is an increasing demand for high

TE D

temperature SMA actuators in hot environments, where no other actuation system could operate. One way to increase the transformation temperatures of NiTi is to add ternary alloying elements such as Zr, Pd, Hf, Pt, and Au [1, 4, 14, 15]. Among these ternary additions, Hf and Zr are the most affordable alternatives. Earlier studies on NiTiHf SMAs focused on compositions where Ni content kept below 50 at. % [15-17].

EP

Recently, the alloy systems higher than 50 at. % Ni have attracted notable attention because of the enhanced mechanical and thermal stability as a result of the formation of nano-precipitates in these compositions. Moreover, Ni-rich side of the stoichiometry makes it possible to tailor transformation

AC C

temperatures with precipitation heat treatments [15, 18, 19].

TWSME is not an intrinsic property of SMAs and it requires specific microstructural conditioning and procedures to generate internal stress storage. The details of the specific procedures depend on the type of SMA. Some of the procedures that provide this microstructural conditioning include aging under stress in precipitation hardenable SMAs [20], marforming [16, 17, 21], repeated mechanical deformation and subsequent low temperature annealing [3, 22], and thermo-mechanical training [23, 24]. These processes generate oriented internal stresses in the microstructure that bias the selection of specific martensite variants upon thermal cycling, resulting in reversible external shape changes without the need for the 2

ACCEPTED MANUSCRIPT

application of any external stress. So far, there exist some controversy about the exact microstructural origin of these internal stresses. However, the most accepted hypotheses involve one or a combination of the following: creation of dislocation substructures during the repeated motion of the martensite front; existence of remnant martensite pinned by point defects [25-27] or dislocations; or oriented precipitates to

RI PT

create remnant oriented stress around the precipitates [20].

While thermo-mechanical training and the resulting TWSME has been extensively studied in various low temperature SMAs, in particular for NiTi [3, 20, 28-31], the work on training and TWSME in high temperature SMAs (HTSMAs) is rather limited and relatively recent [22, 24, 32]. One of the reasons is

SC

the difficulty to obtain TWSME in HTSMAs because of their high transformation temperatures, which may be as high as the characteristic temperatures that can modify/eliminate the microstructural features

M AN U

responsible for TWSME. Among the published works on HTSMAs, Atli et al. studied the effects of ternary alloying additions of Pd [12, 13, 24] and Pt [24] on the TWSME in NiTi SMAs. Kockar et al. studied the effects of severe plastic deformation on the thermo-mechanical stability of the actuation behavior in a Ni49.8Ti42.2Hf8

HTSMA and also reported TWSME [16].

Although severe plastic

deformation can produce TWSME in NiTiHf, it is more appropriate to train the actuator components to obtain TWSME under the stress state that they would operate under in actual applications. This is because of the fact that TWSME is caused by oriented internal stresses [12] or oriented remnant martensite [33].

TE D

If the TWSME is in the same direction of stress applied during microstructural conditioning, it is called positive TWSME and is the most commonly observed TWSME. In few other cases, such as in single crystals with specific orientations, TWSME can be obtained in the reverse direction of microstructural

EP

conditioning; likewise, it is called negative TWSME [34].

Most studies investigate TWSME under uniaxial tension or compression stress states; the work focusing on the effect of shear stresses on the training performance and resulting TWSME is quite limited.

AC C

Contardo et al. studied various training methods to obtain TWSME in CuZnAl SMAs under shear [35], such as thermal cycling under constant stress, stress cycling at constant temperature above Af (austenite finish temperature), and severe plastic deformation at temperatures below Mf (martensite finish temperature). When stress cycling and thermal cycling were compared in the aforementioned study, it was found that thermal cycling produces TWSME more effectively. Thermal cycling required only 40 cycles under 16 MPa to obtain around 0.4 % TWSMS and resulted in 0.24 % accumulated plastic strain as measured in the austenite phase. On the contrary, stress cycling above Af required 200 cycles between 0 and 120 MPa to reach the same level of TWSMS as in the previous case, accompanied by 0.5 % accumulated plastic strain at the end of the training. Therefore, it was concluded that thermal cycling 3

ACCEPTED MANUSCRIPT

under stress is better than isothermal stress cycling in terms of the number of cycles to reach the desired TWSMS strain level, accumulated plastic strain level, and the average stress applied during the training. It has also been shown that TWSME is reduced down to the half of its initial value after the heat treatment at 140°C for 22 hours while the same takes 860 hours at 100°C. By extrapolation, it was concluded that at

RI PT

temperatures below 50°C, TWSME can exists for years.

Furthermore, Davidson et al. studied the required torque and strain levels to induce TWSME in solid Ni50Ti50 SMA cylinders after marforming, i.e. plastic deformation at temperatures below Mf [21]. It was suggested that elastic core of solid cylinder helps to establish TWSME in the skin by providing elastic

SC

bias force. Jardine et al. were one of the first to propose applications of hollow cylinders or tubes in aerospace industry [36]. Mabe et al. studied the TWSME as a result of the thermo-mechanical training on

M AN U

binary NiTi tubes under shear [37]. The amount of TWSMS obtained ranged from 2 to 4 %. TWSME was created almost immediately during training, but accumulated plastic strain significantly increased with the number of cycles. This behavior caused unstable actuation. The stable actuation behavior in binary equiatomic NiTi tubes, where per cycle plastic deformation is negligible, was only obtained after 1000 1500 heating-cooling cycles under constant shear stress. Recently Benafan et al. studied general characterization of stress-strain behavior of precipitation hardened Ni-rich NiTiHf torque tubes, including isobaric heating cooling response and TWSME on a range of tube sizes, thicknesses, and applied stress

TE D

levels [38]. Amount of actuation strain measured by digital image correlation was found to be on the order of 6 %. After a stepwise training procedure with 20 cycles in each stress level with the highest stress level of 500 MPa and total number of cycles of 100, TWSMS was found to be around 3%. In a follow-up study, Benafan et al. also examined the effect of loading configurations on TWSME response of Ni-rich

EP

Ni50.3Ti29.7Hf20 (at.%) alloy tubes [39]. Experimental results revealed that descending series (high-to-low stress from 500 to 0MPa) stabilized the shape memory response in fewer cycles than alternative loading configurations, while ascending series (low-to-high stress from 0 to 500MPa) resulted in highest and most

AC C

stable transformation shear strains. Most of the aforementioned studies on SMA torque tubes did not consist of detailed microstructural investigations.

To the best of the authors’ knowledge, there is no systematic combined study in the literature on the thermo-mechanical training under shear, governing microstructural mechanisms, and thermal stability of TWSME in HTSMAs. Therefore, the first objective of the present study is to investigate the evolution of TWSME in a nano-precipitation hardened NiTiHf HTSMA during thermo-mechanical training with various training parameters and geometric factors. The ultimate goal is to develop a fundamental understanding on how TWSME evolves with training in nano-precipitation hardened HTSMAs. The 4

ACCEPTED MANUSCRIPT

second objective is to investigate the thermal stability of TWSME at temperatures above its anticipated service temperatures (250°C). The ultimate goal for the latter is to understand the thermal stability of the defects responsible for TWSME at elevated temperatures (>300°C) and to determine the safe operation

RI PT

temperature ranges for this particular HTSMA.

2. Experimental Details 2.1 Materials

Ni50.3Ti29.7Hf20 (at. %) cylindrical bars were prepared using vacuum induction skull melting followed by hot extrusion at 900°C with an area reduction of 7 to 1. Extruded bars were machined using electrical

SC

discharge machining (EDM) to obtain torque tubes in two different geometries referred to as “thick” and “thin” for their two different wall thicknesses. Thick torque tubes had an outer diameter of 9.52 mm, an

M AN U

inner diameter of 5.79 mm, and a total length of 127 mm, while thin torque tubes had the same outer diameter of 9.52 mm, larger inner diameter of 6.86 mm, and longer total length of 140 mm. To connect the tubes to the test fixture (Fig. 1a) splines were machined on each end with length of 12.7 mm. The drawing of the splines can be seen in Fig. 1b. After the machining process, the samples were aged at 550°C for 3 hours in air to form coherent nano-precipitates, lenticular in shape with particle sizes varying from 8 to 25 nm in length and from 4 to 10 nm in width. Previous studies on this material under tensile stresses showed that the 550°C heat treatment for 3 hours results in a superior stability of actuation strain

TE D

and very small irrecoverable strain as a function of number of thermo-mechanical cycles [40-43]. This is due to the suppression of plastic deformation during martensitic transformation in the presence of nanoprecipitates. Differential scanning calorimetry (DSC), utilizing a TA Instruments Q20 apparatus, was used to determine the transformation temperatures for the heat treated samples. Martensite finish (Mf),

EP

martensite start (Ms), austenite start (As), and austenite finish (Af) temperatures were found to be 129, 143, 159, and 172 °C, respectively. A total of 6 torque tubes were tested in this study and the test

AC C

parameters for each tube is summarized in Table 1.

2.2 Experimental Setup

In order to mount the torque tubes to the thermo-mechanical cycling setup, custom made hex nuts with the across flats thickness of 19.05 mm were used. Female splines were machined into the internal surface. The specimens were mounted to a stationary phenolic base on one end and to a bearing, which was attached to a pulley, on the other end. The pulley diameter was 240 mm and constant loads were applied by hanging weights to the side of the pulley. The picture of the setup can be seen in Fig. 1a.

5

ACCEPTED MANUSCRIPT

Applied stresses were taken as the outer most surface stress of the tube and calculated using the following equation, ߬=

்௥

(1)

ഏ ሺை஽ ర ିூ஽ర ሻ యమ

where τ is the shear stress, T is applied torque, r is the radius of the point of interest, which is OD/2 in the

the outer surface strain of the tube and calculated using, ߛ=

ఏ௥ ௅

RI PT

present case, OD is the outer diameter, and ID is the inner diameter. Similarly, the strains were taken as

(2)

where γ is the shear strain, θ is the rotation angle, and L is the active length of the tubes. Equations 1 and

SC

2 were also used to calculate applied stress and strain of the torque tubes in similar studies by Mabe et.al. [37], Reedlun et. al. [44] and Davidson et. al. [21]. It is noted that there are some concerns on using thin walled cylinder and linear elastic approximations in these equations for SMAs, whose stress-strain

M AN U

behavior is significantly non-linear in nature. Keefe et al. [45] studied different methods to calculate the stress and strain on SMA torque tubes. They reported that linear model and non-linear models did not differ significantly in calculating strain. Therefore, in the present study, the linear model was chosen for the sake of simplicity.

Rotation angle θ was measured by a potentiometer attached to the loading wheel. Potentiometer was

TE D

excited to a known voltage up to 10 volts which was also monitored throughout the test for normalization purposes. Correlation between voltage and rotation was obtained by calibrating at known angles.

The torque tube samples were heated through conduction using custom heaters made out of nickel-

EP

chrome wires wrapped around the tubes. To compensate the heat loss at the ends of the samples where the grips are not heated, the frequency of the windings was increased towards the end of the samples. Temperature was constantly monitored and recorded using four thermocouples that are directly attached

AC C

to the tubes, evenly spaced throughout the tubes. An Omega N8200 proportional-integral-derivative (PID) temperature controller was used to control temperature and a variable a.c. power source (Variac) was used as the power source for heating. Two fans with the diameter of 5.5 cm were used to increase and control cooling speed. The PID controller was set for 4 minutes heating and 4 minutes cooling.

2.3 Experimental Procedures Among the methods mentioned in Section 1, the thermo-mechanical training route was chosen to obtain TWSME in the present study. Thermo-mechanical training is defined as thermal cycling between two specified temperatures under constant stress levels, until certain desired properties are reached, such as 6

ACCEPTED MANUSCRIPT

stable or zero irrecoverable strain per cycle or stable actuation strain, or until a specified number of cycles. It is known that martensitic transformation temperatures change as a function of applied stress in accordance with the Clausius-Clapeyron relation and they evolve with the number of cycles during extended thermal cycling. To ensure complete transformation, the upper cycle temperature should be

RI PT

above Af at all times. For this study, 50°C above Af temperature under 200 MPa shear stress, which is 240°C in the present case, has been chosen as the upper cycle temperature. Similarly, lower cycle temperature was chosen to be 80°C which is 50°C below Mf under 20 MPa which is the lowest shear stress level applied.

SC

The effects of test parameters and geometric factors, i.e. applied shear stress level, the number of training cycles, and tube wall thickness, on the evolution of TWSME have been investigated. For each of these

M AN U

parameters, two levels were chosen: the training shear stress levels were 145 MPa or 200 MPa, the number of training cycles were 600 or 1600 cycles, and the wall thicknesses were 1.33 mm or 1.87 mm. A full factorial test matrix would require 8 experiments without repetition, instead a parametric test matrix has been designed which can be seen in Table 1, due to the available limited number of tubes. The thin tube, high stress tests were removed from full factorial test matrix, and the number of cycles were kept 1600 except the thick tube, high stress tests. With this test matrix, the isolated effects of each parameter can be obtained keeping other parameters constant. The effects of stress can be revealed by

TE D

using the results of the tubes 1 and 2 versus that of the tubes 3 and 4; the effects of the wall thickness can be studied using the tubes 1 and 2 versus tubes 5 and 6; and the effects of the number of cycles can be deduced within these same tests.

EP

2.3.1 Thermo-Mechanical Training Paths

A training path here is defined as a pair of stress and the number of cycles. Several different training paths were used in this study. These paths can be classified based on the number of training cycles as “long” for

AC C

1600 cycles and “short” for 600 cycles, or based on the shear stress level under which the training is performed such as “high stress” for 200 MPa and “low stress” for 145 MPa. 600 cycles is chosen for short training path based on previous experience which shows less than 600 cycles is enough to fully train the present material. The low stress, 145 MPa, is chosen to be just above martensite reorientation stress for the

present material at room temperature (not shown here). Actuation behavior of the samples (actuation strain as a function of the applied stress) was determined by a procedure called “the characterization sequence”. The characterization sequence subjects the samples to 3 thermal cycles under a set of stress levels. Four to five stress levels were selected for 145 MPa and 200 MPa training stresses. These levels 7

ACCEPTED MANUSCRIPT

are summarized in Table 1. One example of the results from these characterization sequences is shown in Fig. 2a where only the 3rd cycles under each stress level are presented. Actuation strain is defined as the difference between strain in martensite at the lower cycle temperature and strain in austenite at the upper cycle temperature as shown in Fig. 2a. The evolution of strain vs. temperature response is presented as a

RI PT

function of the applied shear stress for the Tube-5 (Table 1) before thermo-mechanical training. Testing under 0 MPa was not performed due to the limitations of the experimental setup. Therefore, to report the actual TWSMS, which is the actuation response under no stress, first two actuation strain vs. stress data points were used to extrapolate and estimate the no stress behavior of the samples (Fig. 2b). These estimated points were marked with markers and connected with dashed lines in the plots, reported in the

SC

next section, for clear distinction from the experimentally measured data points. For the long route (1600 cycles), the characterization sequence was employed after the 100th, 350th, 600th, 1100th and 1600th cycles, while for the short route (600 cycles), the characterization sequence was applied after the 100th, 350th and

in plots like the one presented in Fig. 2c.

M AN U

600th cycles. The evolution of strain in martensite, strain in austenite, and the actuation strains are shown

2.3.2 Transmission Electron Microscopy Investigation

One sample (Tube-3) was selected for further microstructural investigation. Transmission electron microscopy (TEM) investigation was conducted in a JEOL 2011 high-resolution microscope (in phase

TE D

contrast mode) with a LaB6 filament, operated at 200 kV and equipped with an energy-dispersive X-ray (EDX) spectrometer. All the imaging has been done at room temperature. Specimens were machined by EDM into 3 mm diameter discs and mechanically polished to around 100 micron thickness. Lastly to obtain electron transparent zones, double jet electrochemical polishing used a 30% HNO3 / 70% ethanol

EP

solution at -20 °C and 12 V. Bright field and dark field images were obtained to demonstrate precipitate and dislocation distribution. Selected area electron diffraction patterns (SAEDP) were obtained to identify

AC C

phases and orientations of crystallites with each other.

2.3.3 Annealing Study

After the thermo-mechanical training and characterization sequences were completed, one sample (Tube2) was statically annealed, sequentially between 350 °C and 700 °C in 50 °C increments for 30 minutes at each temperature, to determine how TWSMS would evolve above the anticipated operating service temperatures. After each annealing step the sample was mounted to the test fixture again and the same characterization sequence was applied to monitor the evolution of TWSMS and the actuation response. If there was no reduction in the TWSMS, the sample was directly annealed for the next annealing step. However, if there was a reduction in the TWSM response, the sample was further trained under 145 MPa 8

ACCEPTED MANUSCRIPT

to reach the initial trained response and the same actuation strain level. To better follow and visualize these sequential procedures, a flowchart is constructed as shown in Fig. 3.

3. Results and Discussions

RI PT

3.1 Evolution of actuation strain and TWSME during thermo-mechanical training For the low stress (145 MPa), longer (1600 cycles) training path, both thick (Tube-1 and Tube-2) and thin (Tube-5 and Tube-6) torque tubes were tested. Two tubes were used for each condition to check the repeatability, and the reported results here are the averages from these two repetitions. As the results from the thick and thin tubes are almost equivalent, only those from the thick tubes are introduced in this

SC

section. Before any training, extrapolation of strains at two lowest applied stress levels reveals that the thick tubes exhibit -0.004 ± 0.053% strain upon thermal cycling under zero stress (Fig. 2b), which

M AN U

indicates there is no TWSME on the material before the training. The actuation strain on the un-trained tubes is about 3.107 ± 0.216 % under 145 MPa. After 100 training cycles, TWSMS increases to 1.907 ± 0.176 % and the actuation strain under the training stress only slightly changes to 3.283 ± 0.025 %. Percentage of TWSMS to the actuation strain under the training stress is about 60 % after 100 cycles. Upon further training, actuation shear strain vs. stress response (Fig 2b) becomes less dependent on the stress level. After 350 training cycles, TWSMS and the actuation strain under 145 MPa are 2.757 ± 0.131 % and 3.501 ± 0.004 % respectively. After 600 training cycles, both TWSMS and the actuation strain

TE D

under 145 MPa seem to saturate at about 3.013 ± 0.040% and 3.453 ± 0.016%, making the percentage of TWSMS to the actuation strain almost 84%. Further training cycles have negligible effect on TWSMS with a maximum value of 3.051 ± 0.028%. On the other hand, the actuation strain under 145 MPa decreases from 3.453 ± 0.016 % to 3.038 ± 0.024 % progressively by increasing the number of training

in Fig 4.

EP

cycles from 600 to 1600 cycles. Evolution of TWSMS as a function of the number of cycles can be seen

AC C

Evolution of the relevant strains in Tube-2 during the training is shown in Fig. 2c. Since the no load austenite is the reference state for the strain calculations, the strain values in austenite show the evolution of the remnant strain starting from a value of 1 % which includes elastic strain. The reduction in the actuation strain can also be seen between cycles 600 and 1600, similar to the results shown in Fig. 2b. It is noted that the accumulation of remnant strain is more significant in the first 400 cycles and decreases after that. Some small minima and maxima in the actuation strain can be observed in Fig. 2c. These small changes are caused by the experiments of the characterization sequence performed at those points, and they do not have a significant influence on the overall evolution trend of the actuation strain. Indeed, after about 10 training cycles, the actuation strain reaches the values observed before the characterization 9

ACCEPTED MANUSCRIPT

sequence. Initial strain in austenite was 1.00 % and after 1600 cycles, it reached to 1.55 %. This shows that even under notably high shear stresses, the actuation response is stable in the present nanoprecipitation hardened NiTiHf HTSMAs, i.e. remnant strain only evolves less than 1% in 1600 cycles.

RI PT

The stability of the strain in martensite is better than that in austenite: the strain in the first cycle was 4.36 % while the strain after 1600 cycles was 4.61 %. The difference in the increase of two characteristic strains yields a reduction in actuation strain. One possible reason of this reduction is the stabilization of austenite, martensite or both. It is known from the literature [16, 46-48] that plastic deformation or thermo-mechanical training can result in the stabilization of martensite above Af temperatures. This

SC

stabilization is due to the local defects pinning the martensite interfaces and preventing martensite to transform back to austenite. In such cases increasing the upper cycle temperature or low temperature

M AN U

annealing usually helps to relieve the remnant martensite. The DSC results shown in Fig. 5 demonstrate a reduction in the transformation enthalpy which is an indication of the reduction in the martensitically transforming volume. Transformation enthalpy dropped from 18.3 J/g to 13.1 J/g for the specimen thermo-mechanically trained under 200MPa for 600 cycles, corresponding to a 28% reduction. This reduction also points out the stabilization of martensite, austenite or both. However, the stabilization of austenite has not been reported in the literature, to the best of the authors’ knowledge. In the following sections, Transmission Electron Microscopy (TEM) investigations will show that the present torque tube

TE D

samples, trained under shear stress, contain retained austenite phase at 100°C below the Mf temperature, contributing to the observed reduction in the transformation enthalpies.

3.2 Effect of Training Stress Level on the Evolution of TWSME

EP

In order to determine the effect of applied training stress on the evolution of TWSME, two samples (Tube-3 and Tube-4) were trained under 200 MPa shear stress in addition to those trained under 145 MPa. Since it was shown above that 600 cycles are sufficient to saturate obtainable TWSME (see Fig. 4), the

AC C

tubes were trained up to this particular number of cycles. The evolution of TWSME of the samples trained under 200 MPa and 145 MPa can be seen in Fig. 6a. The results shown are the averages from the two identical tubes tested in each training stress level. The results from the initial characterization cycles before actual training showed notable difference in the actuation strains between the tubes trained under 200 MPa and those trained under 145 MPa. This is likely due to the more intermediate characterization sequence steps and more associated number of cycles in the samples for 145 MPa training than those for 200 MPa training (for example, 35 MPa and 90 MPa experiments with 3 cycles under each stress, Table 1), leading to a training effect before the actual training under 145 MPa or 200 MPa starts. For the trained samples, however, there is no notable difference in either actuation strain or TWSMS between the 10

ACCEPTED MANUSCRIPT

samples trained under two different shear stresses. The evolution of TWSME and the actuation strain during training is comparable for high and low shear stresses as can be seen in Figs. 6a and 6b. The evolution of TWSMS as a function of number of cycles has been presented in Fig. 4. In Fig. 4, the response of Tube-3 trained under 200 MPa is almost identical to other conditions. Tube-4, although

RI PT

identical to Tube-3, has slightly different response, namely the TWSMS is about 0.5% higher after 100 cycles, yet the overall trend in the dependence of TWSMS on the number of cycles is similar to other conditions. Therefore, it can be concluded that increasing training shear stress above 145 MPa does not

3.3 Effect of Sample Size on the Evolution of TWSME

SC

notably change the TWSMS and the number of cycles to reach a particular TWSMS.

In order to determine the effect of sample size on the training response and TWSME, in addition to the

M AN U

thick torque tube samples, thin torque tube samples (Tube-5 and Tube-6) were also trained. The evolution of TWSME and actuation strain in thin tubes can be seen in Fig. 7a and Fig. 7b, respectively, in comparison to the thick tubes. The results of the thick tube samples are included in Fig. 7a for the ease of comparison with the results of the thin walled tube samples. Clearly, there is no significant difference in the actuation strains and the evolution of TWSMS as a function of the number of cycles, as can be seen in Figure 4. The evolution of the critical strains with the number of cycles can also be seen in Fig. 7b. Note that there is also no significant difference in the accumulated plastic strain. This suggests that thin-walled

levels up to 145 MPa).

TE D

tube assumption and linear approximation are valid for the present tested conditions (under the stress

3.4 Microstructural Evolution

EP

TEM investigations were used to investigate the microstructural changes introduced by training in the NiTiHf HTSMA tubes. TEM of the samples before training show relatively large self-accommodated martensite variants containing a high density of spindle like H-phase precipitates (see [15, 49] for more

AC C

information on the H-phase), of about 8-25 nm in length and 4-10 nm in width, as shown in Fig. 8a and 8b. TEM investigations performed on the trained Tube-3 sample (the thick tube trained under 200 MPa for 600 cycles) reveal similar precipitate sizes and interparticle distances, as shown in Fig. 8c and 8d. Four variants of the H-phase can theoretically form in these materials [15, 49], which were observed in both trained and untrained specimens. The SAEDPs has predominantly shown the same twinning modes in both trained and untrained samples: (011)-Type I and (001)-compound twinning [49]. These observations indicate that the particular training process used here did not bring about any significant growth or realignment of the nanoprecipitates nor a change in the martensite twinning modes, as proven by the images in Fig. 8. 11

ACCEPTED MANUSCRIPT

On the other hand, the thermo-mechanical cycling notably changed the self-accommodated martensite morphology. In the untrained samples, the martensite variants are usually large and regularly spaced, and form different self-accommodating groups (Fig. 8b), whereas in the trained samples, there exist many

RI PT

regions showing a predominance of long and thin needles (less than 100 nm in width) of a few martensite variants with nearly-parallel intervariant boundaries. Fig. 9 shows an example of such microstructure with predominant aligned variants extending across the grain boundaries. Such microstructure is typical in low temperature SMAs, such as NiTi, with well-developed TWSME, which is a result of the defect structures

SC

evolved during the training process.

Furthermore, in the trained samples, there were many regions exhibiting (011)-Type I twins together with

M AN U

additional overlapped domains showing oblique interfaces with respect to the twinning planes (Fig. 10a). This particular microstructure differs from the typical arrays of (011)-twinned plates observed in edge-on conditions, in which all the twins show parallel interfaces (see [15] for the examples of the normal (011)Type I twinning morphology). The microstructure shown in Fig. 10a has been studied by SAEDP (Figs 10b-e) using very small selected area apertures. The position and approximate sizes of the apertures are indicated in Fig. 10a using white circles. The SAEDPs of zones V1 and V2 (Figs. 10c and 10d) reveal two B19’ variants with twinning relationship over the (011)B19’ planes, which corresponds to the normal

TE D

(011)-Type I twinning. However, the V3 zones, which are interspersed in between the twinned martensite variants, can only be indexed as B2 austenite. Moreover, the orientation between the B19’ variants and the V3 domains matches perfectly the orientation relationship between austenite and martensite. Therefore, Fig. 10 demonstrates the existence of the areas that have not been transformed into martensite

EP

even at room temperature, 100°C below the Mf temperature, i.e. remnant austenite, as a consequence of the training process. The presence of this remnant austenite correlates well with the observed decrease in the actuation strain upon prolonged cycling (Figs. 2b, 6b and 7b) as well as the decrease of the

AC C

transformation enthalpy after the training (Fig. 5).

Another important microstructural feature observed in the trained samples is the presence of dislocations interconnecting the nanoprecipitates as shown in Fig. 11a. They were not observed in untrained samples, thus the dislocations were created during the repeated martensitic transformation under stress. Fig. 11b presents an enlarged image taken under suitable conditions for imaging the (001)-compound internal twins, which are usually very thin in NiTiHf alloys containing nanoprecipitates. The dislocations appear with bright contrast in Fig. 11b. Interestingly, the image shows that the (001) twins are refined in the areas close to the dislocations and there is a discontinuity in the twin sequence at each side of the 12

ACCEPTED MANUSCRIPT

dislocation lines. Thus, these observations reveal the difficulty of martensite growth when the dislocations are present. The difficult martensite growth would explain the incomplete transformation on cooling and presence of remnant austenite in the trained materials at room temperature.

RI PT

As it was already noted, TWMSE has been traditionally associated with the defects generated, or modified, during training. For instance, growth of aligned precipitates by aging under stress can generate TWSME by means of the aligned internal stress created by the precipitates, which select certain variants over the other. The dislocations created by the austenite-martensite front moving back and forth during a training process based on the repeated transformation cycling, also create internal stresses which favors

SC

certain variants over others, yielding to TWSME [50]. Finally, retained martensite plates of suitable variants can also contribute to TWSME by promoting growth of the retained plates instead of nucleation

M AN U

of other variants [25-27]. In the present work, the TEM results rule out any precipitate growth or reorientation as a possible factor responsible for the TWSME observed. On the other hand, the TEM results clearly demonstrate dislocation formation during training, which may contribute to the observed TWSME. It is interesting to note that TWSMS saturates at ~600 cycles in the present NiTiHf material and further training does not significantly increase the amount of TWSMS for most of the samples (Fig. 4). Therefore, the dislocation density observed in the present work (Fig. 11) should be close to the maximum that can be reached in this alloy and initial thermal treatment. Such dislocation density is rather low in

TE D

comparison with the dense dislocation tangles typically observed in solution heat treated (precipitate-free) Ni-Ti alloys after a few transformation cycles [51-53]. Moreover, other alloys in precipitate-free condition, such as CuZnAl [35] and NiTiPd [12], do not require this many cycles to reach the TWSMS saturation. These observations demonstrate the strengthening effect caused by the H-phase

EP

nanoprecipitates in Ni50.3Ti29.7Hf20, which act as barriers to the dislocation formation and motion, and may justify the need for prolonged training treatments to induce the TWSME in this material.

AC C

3.5 Stability of TWSME at Elevated Temperatures Since TWSME in SMAs is a result of defects and internal residual stresses, it is important to understand the evolution of TWSME as a function of temperature and to determine its stability at high temperatures, especially above 0.4TH (TH=homologous temperature), where diffusional mechanisms start playing significant role in materials response. This is particularly critical for HTSMAs due to their high operating temperatures as compared to conventional SMAs. For this purpose, a thick walled torque tube sample (Tube-2) was annealed at a set of temperatures after the long training procedure (Fig. 2) and the characterization sequence was utilized after each step of annealing to evaluate the evolution of TWSMS and actuation strain under 145 MPa. Fig. 12 summarizes the results after selected annealing treatments. 13

ACCEPTED MANUSCRIPT

As can be seen in Fig. 12, after 350 and 400°C annealing treatments, there is no significant change in TWSME response. The slight increase in the TWSMS after these annealing treatments as compared to the TWSMS right after the training is probably because of the reverse transformation of remnant martensite and thus increase in the transforming volume. On the other hand, above 450°C up to 600°C annealing

RI PT

treatments, there was a sequential reduction in TWSMS. In these cases, the sample was further trained for 100 cycles before the next annealing step to stabilize the actuation strain under 145 MPa again. After annealing at 650°C for 30 min, TWSMS was reduced down to 0.49 % and actuation shear strain vs. stress response was almost identical to the initial behavior before the training. This indicates that the annealing at 650°C for 30 min. is sufficient to relieve all the internal stresses stored during the thermo-mechanical

SC

training under 145 MPa. In contrast, after the annealing at 700°C for 30 min, actuation strain levels at a given stress are lower (shown in Fig. 12) than those before training, indicating that the microstructure

M AN U

changed significantly after this annealing treatment. These results indicate that TWSME in nanoprecipitation hardened NiTiHf HTSMAs is stable up to 400°C above which the internal residual stresses start relaxing, because of the annealing of the defects responsible from these stresses.

In order to better assess potential microstructural changes that may occur during the static annealing experiments performed here, we have conducted DSC experiments on the annealed trained samples to determine the transformation temperatures. In the present alloy, there should be few competing changes in

TE D

the microstructure: new formation or growth of nano-precipitates and relaxation of internal stresses through annihilation of dislocations, diffusion of point defects, and transformation of remnant martensite into austenite. Since transformation temperatures in SMAs are quite sensitive to even small microstructural changes [10, 12, 24], DSC can sometimes provide powerful evidence of the type of

EP

microstructural changes. Figs. 5 and 13 show the DSC results of the samples before and after training and after various annealing heat treatments.

The results indicate that no significant change occurs on

transformation temperatures after annealing up to 600°C for 30 minutes. However, above that

AC C

temperature, the transformation temperatures notably change. This indicates that precipitate structure in the material may have changed.

Using the data in Fig. 12, the evolution of TWSMS as a function of annealing temperature is plotted in Fig. 14, which shed light onto possible mechanisms responsible for the origin of TWSME in these materials. As mentioned earlier, the three possible reasons of TWSME are aligned nano-precipitates, remnant martensite, and dislocations and point defects or some combination of these mechanisms. The microstructural investigations with TEM demonstrated that there is no significant change in the precipitate structure such as size and orientation after training suggesting that possible internal stress field 14

ACCEPTED MANUSCRIPT

around any aligned precipitates cannot be a mechanism responsible for TWSME here. Annealing can induce precipitate growth, and since the precipitates grow under no stress in the present case, this should decrease TWSMS. Moreover, the change in the precipitate size and fraction should also alter the composition and transformation temperatures of the matrix, and since the transformation temperatures did

RI PT

not significantly change after annealing below 600°C, the contribution of the precipitate growth to the change in TWSME should be insignificant for the present cases. If the main reason of TWSME in present material would have been remnant martensite, a sudden drop in TWSMS would be expected above a critical annealing temperature. However, the gradual decrease in TWSMS with increasing annealing temperature in Fig. 14 suggests that the main source of TWSME in the present materials is dislocations

SC

and point defects that they gradually annihilate, possibly following an Arrhenius type temperature

4. Summary and Conclusions

M AN U

dependence.

The work herein on the two-way shape memory effect (TWSME) evolution as a function of training stress, number of cycles and post-training annealing in nano-precipitation hardened, Ni50.3Ti29.7Hf20 high temperature shape memory alloy (HTSMA) thin walled tubes revealed that under 200 MPa, 600 thermal cycles are enough to reach a two-way shape memory strain (TWSMS) as high as 2.95 %, which was shown to be stable upon annealing up to 400°C for 30 minutes. This TWSMS is 85 % of the maximum

TE D

measured actuation strain under 200 MPa. The microstructure after thermo-mechanical training was investigated using transmission electron microscopy (TEM), which did not indicate a significant change in precipitate structure and size after the training. Small amount of remnant austenite was revealed 100 °C below the martensite finish temperature, with notable amount of dislocations. Overall, it is found that

EP

nano-precipitation hardened Ni50.3Ti29.7Hf20 shows relatively high TWSMS after much less number of training cycles as compared to nickel lean NiTiHf compositions. Although binary NiTi shows TWSME after less number of training cycles, plastic strain continues to accumulate (unstable) up to 1000 -1500

AC C

cycles, where precipitation hardened NiTiHf has stable TWSMS after 600 cycles. Tube wall thickness and training stress levels have been found to have negligible effect on shape memory strains and number of cycles to reach the desired training level, for the ranges studied. In the light of the results obtained, the main findings can be summarized as follows: 1. TWSME can be induced in the precipitated hardened Ni-rich NiTiHf HTSMAs by repetitive thermo-mechanical cycling. TWSME saturates in about 600 cycles and further cycling reduces the actuation strain under stress. 2. The chosen training shear stress levels (145 MPa and 200 MPa) and the tube wall thicknesses have not produced notable differences in TWSME. The latter can be attributed to validity of thin 15

ACCEPTED MANUSCRIPT

walled tube assumption and linear approximation to calculate stresses and strains in the present conditions. 3. TEM observations have demonstrated the existence of remnant austenite after the training, which leads to the reduction in actuation strain with thermo-mechanical cycling after training. TEM

RI PT

images have also revealed a moderate level of dislocation accumulation interconnecting the nanoprecipitates. The dislocation arrays formed during cycling and, eventually, retained martensite are considered to be the origin of TWSME in these alloys.

4. TWSME is thermally stable upon annealing at temperatures up to 400°C after 30 min heat treatments. The thermal treatments from 450°C to 650°C gradually decrease the two-way shape

SC

memory strain. Annealing at 650°C for 30 min yields to actuation response close to that of the untrained material whereas the annealing at 700°C for 30 min notably changes the microstructure

M AN U

and the transformation temperatures, resulting in partial martensitic transformation during cycling between the selected temperature range for testing in the present study.

5. Acknowledgments

This study was supported by the Boeing Company. Additional support was received from the US Air Force Office of Scientific Research, under Grant no. FA9550-18-1-0276, the NASA University Leadership Initiative Grant Number NNX17AJ96A, and Spanish MINECO and FEDER under Project

AC C

EP

TE D

Number MAT2014-56116-C4-1-R.

16

ACCEPTED MANUSCRIPT

References:

[1] J. Ma, I. Karaman, R.D. Noebe, High temperature shape memory alloys, International Materials Reviews 55(5) (2010) 257-315.

RI PT

[2] D. Dye, Shape memory alloys: Towards practical actuators, Nature Materials 14(8) (2015) 760. [3] Y. Liu, Y. Liu, J. Van Humbeeck, Two-way shape memory effect developed by martensite deformation in NiTi, Acta Materialia 47(1) (1998) 199-209. [4] J. Van Humbeeck, High temperature shape memory alloys, Journal of Engineering Materials and Technology 121(1) (1999) 98-101.

SC

[5] G. Firstov, J. Van Humbeeck, Y.N. Koval, High-temperature shape memory alloys: some recent developments, Materials Science and Engineering: A 378(1) (2004) 2-10.

M AN U

[6] J.M. Jani, M. Leary, A. Subic, M.A. Gibson, A review of shape memory alloy research, applications and opportunities, Materials & Design (1980-2015) 56 (2014) 1078-1113. [7] D.C. Lagoudas, Shape memory alloys, Science and Business Media, LLC (2008). [8] S. Hirose, K. Ikuta, Y. Umetani, Development of shape-memory alloy actuators. Performance assessment and introduction of a new composing approach, Advanced Robotics 3(1) (1988) 3-16. [9] A. Nespoli, S. Besseghini, S. Pittaccio, E. Villa, S. Viscuso, The high potential of shape memory alloys in developing miniature mechanical devices: A review on shape memory alloy mini-actuators, Sensors and Actuators A: Physical 158(1) (2010) 149-160.

TE D

[10] K. Atli, I. Karaman, R. Noebe, G. Bigelow, D. Gaydosh, Work production using the two-way shape memory effect in NiTi and a Ni-rich NiTiHf high-temperature shape memory alloy, Smart Materials and Structures 24(12) (2015) 125023.

EP

[11] D. Canadinc, W. Trehern, H. Ozcan, C. Hayrettin, O. Karakoc, I. Karaman, F. Sun, Z. Chaudhry, On the deformation response and cyclic stability of Ni50Ti35Hf15 high temperature shape memory alloy wires, Scripta Materialia 135 (2017) 92-96. [12] K. Atli, I. Karaman, R. Noebe, Work output of the two-way shape memory effect in Ti50.5 Ni24.5Pd25 high-temperature shape memory alloy, Scripta Materialia 65(10) (2011) 903-906.

AC C

[13] K. Atli, I. Karaman, R. Noebe, H. Maier, Comparative analysis of the effects of severe plastic deformation and thermomechanical training on the functional stability of Ti50.5Ni24.5Pd25 high-temperature shape memory alloy, Scripta Materialia 64(4) (2011) 315-318. [14] A. Evirgen, I. Karaman, R. Santamarta, J. Pons, R. Noebe, Microstructural characterization and superelastic response of a Ni50. 3Ti29. 7Zr20 high-temperature shape memory alloy, Scripta Materialia 81 (2014) 12-15. [15] R. Santamarta, R. Arróyave, J. Pons, A. Evirgen, I. Karaman, H. Karaca, R. Noebe, TEM study of structural and microstructural characteristics of a precipitate phase in Ni-rich Ni–Ti–Hf and Ni–Ti–Zr shape memory alloys, Acta Materialia 61(16) (2013) 6191-6206.

17

ACCEPTED MANUSCRIPT

[16] B. Kockar, I. Karaman, J. Kim, Y. Chumlyakov, A method to enhance cyclic reversibility of NiTiHf high temperature shape memory alloys, Scripta Materialia 54(12) (2006) 2203-2208. [17] X.-L. Meng, Y.-F. Zheng, Z. Wang, L. Zhao, Shape memory properties of the Ti36Ni49Hf15 high temperature shape memory alloy, Materials Letters 45(2) (2000) 128-132.

RI PT

[18] A. Evirgen, F. Basner, I. Karaman, R.D. Noebe, J. Pons, R. Santamarta, Effect of aging on the martensitic transformation characteristics of a Ni-rich NiTiHf high temperature shape memory alloy, Functional Materials Letters 5(04) (2012) 1250038. [19] G. Bigelow, A. Garg, S. Padula, D. Gaydosh, R. Noebe, Load-biased shape-memory and superelastic properties of a precipitation strengthened high-temperature Ni50.3Ti29.7Hf20 alloy, Scripta Materialia 64(8) (2011) 725-728.

SC

[20] M. Nishida, All-round shape memory effect in Ni-rich TiNi alloys generated by constrained aging, Scripta Metallurgica 18(11) (1984) 1293-1298.

M AN U

[21] F.M. Davidson, C. Liang, D.W. Lobitz, Investigation of torsional shape memory alloy actuators, Smart Structures and Materials 1996: Smart Structures and Integrated Systems, International Society for Optics and Photonics, 1996, pp. 672-683. [22] X. Meng, W. Cai, Y. Zheng, Y. Rao, L. Zhao, Two-way shape memory effect induced by martensite deformation and stabilization of martensite in Ti36Ni49Hf15 high temperature shape memory alloy, Materials Letters 57(26-27) (2003) 4206-4211. [23] T. Tadaki, K. Otsuka, K. Shimizu, Shape memory alloys, Annual Review of Materials Science 18(1) (1988) 25-45.

TE D

[24] K. Atli, I. Karaman, R. Noebe, D. Gaydosh, The effect of training on two-way shape memory effect of binary NiTi and NiTi based ternary high temperature shape memory alloys, Materials Science and Engineering: A 560 (2013) 653-666. [25] J. Perkins, Residual stresses and the origin of reversible (two-way) shape memory effects, Scripta Metallurgica 8(12) (1974) 1469-1476.

EP

[26] J. Guilemany, J. Fernandez, On the mechanism of two way shape memory effect obtained by stabilized stress induced martensite, Scripta Metallurgica et Materialia;(United States) 30(3) (1994).

AC C

[27] J. Guilemany, J. Fernández, Mechanism of two way shape memory effect obtained by stabilised stress induced martensite J, Journal de Phys. IV. Coll. C 5 (1995) 355. [28] R. Lahoz, L. Gracia-Villa, J.A. Puértolas, Training of the two-way shape memory effect by bending in NiTi alloys, Journal of Engineering Materials and Technology 124(4) (2002) 397-401. [29] C. Maletta, A. Falvo, F. Furgiuele, A Phenomenological approach for real-time simulation of the two-way shape memory effect in NiTi alloys, Journal of Engineering Materials and Technology 130(1) (2008) 011003. [30] C. Urbina, S. De la Flor, F. Ferrando, TWSME improvement by thermal cycling at zero stress in NiTi shape memory alloys, Second International Conference on Smart Materials and Nanotechnology in Engineering, International Society for Optics and Photonics, 2009, p. 74930L.

18

ACCEPTED MANUSCRIPT

[31] C.-Y. Chang, D. Vokoun, C.-T. Hu, Two-way shape memory effect of NiTi alloy induced by constraint aging treatment at room temperature, Metallurgical and Materials Transactions A 32(7) (2001) 1629-1634. [32] X. Meng, W. Cai, Y. Fu, Q. Li, J. Zhang, L. Zhao, Shape-memory behaviors in an aged Ni-rich TiNiHf high temperature shape-memory alloy, Intermetallics 16(5) (2008) 698-705.

RI PT

[33] J. Perkins, R. Sponholz, Stress-induced martensitic transformation cycling and two-way shape memory training in Cu-Zn-Al alloys, Metallurgical Transactions A 15(2) (1984) 313-321. [34] I. Kaya, H. Tobe, H. Karaca, M. Nagasako, R. Kainuma, Y. Chumlyakov, Positive and negative twoway shape memory effect in [111]-oriented Ni51Ti49 single crystals, Materials Science and Engineering: A 639 (2015) 42-53.

SC

[35] L. Contardo, G. Guenin, Training and two way memory effect in CuZnAl alloy, Acta Metallurgica et Materialia 38(7) (1990) 1267-1272.

M AN U

[36] A.P. Jardine, J.N. Kudva, C.A. Martin, K. Appa, Shape memory alloy TiNi actuators for twist control of smart wing designs, Smart Structures and Materials 1996: Smart Structures and Integrated Systems, International Society for Optics and Photonics, 1996, pp. 160-166. [37] J.H. Mabe, R.T. Ruggeri, E. Rosenzweig, C.-J.M. Yu, NiTinol performance characterization and rotary actuator design, Smart Structures and Materials 2004: Industrial and Commercial Applications of Smart Structures Technologies, International Society for Optics and Photonics, 2004, pp. 95-110. [38] O. Benafan, D. Gaydosh, High temperature shape memory alloy Ni50. 3Ti29. 7Hf20 torque tube actuators, Smart Materials and Structures 26(9) (2017) 095002.

TE D

[39] O. Benafan, D.J. Gaydosh, Constant-torque thermal cycling and two-way shape memory effect in Ni50. 3Ti29. 7Hf20 torque tubes, Smart Materials and Structures 27(7) (2018) 075035. [40] A. Evirgen, I. Karaman, R. Santamarta, J. Pons, R. Noebe, Microstructural characterization and shape memory characteristics of the Ni50.3Ti34.7Hf15 shape memory alloy, Acta Materialia 83 (2015) 48-60.

EP

[41] H. Karaca, S. Saghaian, G. Ded, H. Tobe, B. Basaran, H. Maier, R. Noebe, Y. Chumlyakov, Effects of nanoprecipitation on the shape memory and material properties of an Ni-rich NiTiHf high temperature shape memory alloy, Acta Materialia 61(19) (2013) 7422-7431.

AC C

[42] O. Karakoc, C. Hayrettin, M. Bass, S. Wang, D. Canadinc, J. Mabe, D. Lagoudas, I. Karaman, Effects of upper cycle temperature on the actuation fatigue response of NiTiHf high temperature shape memory alloys, Acta Materialia 138 (2017) 185-197. [43] O. Karakoc, C. Hayrettin, D. Canadinc, I. Karaman, Role of applied stress level on the actuation fatigue behavior of NiTiHf high temperature shape memory alloys, Acta Materialia 153 (2018) 156-168. [44] B. Reedlunn, S. Daly, J. Shaw, Tension-torsion experiments on superelastic shape memory alloy tubes, ASME 2012 Conference on Smart Materials, Adaptive Structures and Intelligent Systems, American Society of Mechanical Engineers, 2012, pp. 213-222. [45] A.C. Keefe, G.P. Carman, Thermo-mechanical characterization of shape memory alloy torque tube actuators, Smart materials and structures 9(5) (2000) 665.

19

ACCEPTED MANUSCRIPT

[46] I. Karaman, H.E. Karaca, Z. Luo, H. Maier, The effect of severe marforming on shape memory characteristics of a Ti-rich NiTi alloy processed using equal channel angular extrusion, Metallurgical and Materials Transactions A 34(11) (2003) 2527-2539.

RI PT

[47] B. Kockar, I. Karaman, A. Kulkarni, Y. Chumlyakov, I. Kireeva, Effect of severe ausforming via equal channel angular extrusion on the shape memory response of a NiTi alloy, Journal of Nuclear Materials 361(2) (2007) 298-305. [48] B. Kockar, K. Atli, J. Ma, M. Haouaoui, I. Karaman, M. Nagasako, R. Kainuma, Role of severe plastic deformation on the cyclic reversibility of a Ti50.3 Ni33.7 Pd16 high temperature shape memory alloy, Acta Materialia 58(19) (2010) 6411-6420.

SC

[49] F. Yang, D. Coughlin, P.J. Phillips, L. Yang, A. Devaraj, L. Kovarik, R.D. Noebe, M. Mills, Structure analysis of a precipitate phase in an Ni-rich high-temperature NiTiHf shape memory alloy, Acta Materialia 61(9) (2013) 3335-3346.

M AN U

[50] Z. Wang, X. Zu, L. You, X. Feng, C. Zhang, Investigation on the two-way shape memory effect and alternating current electrothermal driving characteristics of TiNiCu shape memory alloy, Journal of Materials Science 39(10) (2004) 3391-3395. [51] D. Norfleet, P. Sarosi, S. Manchiraju, M.-X. Wagner, M. Uchic, P. Anderson, M. Mills, Transformation-induced plasticity during pseudoelastic deformation in Ni–Ti microcrystals, Acta Materialia 57(12) (2009) 3549-3561.

TE D

[52] T. Simon, A. Kröger, C. Somsen, A. Dlouhy, G. Eggeler, On the multiplication of dislocations during martensitic transformations in NiTi shape memory alloys, Acta Materialia 58(5) (2010) 18501860.

AC C

EP

[53] A. Pelton, G. Huang, P. Moine, R. Sinclair, Effects of thermal cycling on microstructure and properties in Nitinol, Materials Science and Engineering: A 532 (2012) 130-138.

20

ACCEPTED MANUSCRIPT

Tables and Figures

Tube-2 Tube-3 Tube-4 Tube-5

Characterization Shear Stress Levels (MPa)

Number of ThermoMechanical Cycles

145

20, 35, 55, 90, and 145

145

20, 35, 55, 90, and 145

1600

200

20, 55, 110, and 200

600

20, 55, 110, and 200

600

20, 45, 90, 110, and 145

1600

20, 45, 90, 110, and 145

1600

200 145 145

1600

AC C

EP

TE D

Tube-6

Training Shear Stress (MPa)

SC

Tube-1

Wall Thickness Thick (1.86 mm) Thick (1.86 mm) Thick (1.86 mm) Thick (1.86 mm) Thin (1.33 mm) Thin (1.33 mm)

M AN U

Name

RI PT

Table 1. The list of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes used in this study. Thick or thin walled tubes were tested under constant torsional loading and thermally cycled (i.e. trained) across the martensitic transformation temperatures. The training shear stress levels and number of cycles applied are summarized in the table. Thick tubes had the wall thickness of 1.86 mm while thin walled tubes had a thickness of 1.33 mm.

21

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

Figure 1. a) Picture of the thermo-mechanical cycling setup for the torque tube actuators, indicating relevant components such as variable alternating current power supply (Variac) (1), phenolic base (2), cooling fans (3), temperature controller (TC) (4) and data acquisition board (DAQ) (5). b) The drawing of the spline design used to attach the torque tubes to the base and the load.

22

ACCEPTED MANUSCRIPT

a

5 4

Strain in Austenite Actuation Strain

Strain in Martensite 110 MPa

3

90 MPa

RI PT

Shear Strain (%)

145 MPa

2 45 MPa

1 21 MPa

100

150 200 Temperature (°C)

b

4 3 2 1 0

0

c

20

40

EP

5

Ni50.3Ti29.7Hf20 HTSMA Torque Tubes Heat Treated at 550 °C for 3 Hrs After Third Training (600 Cycles) After Fourth Training (1100 Cycles) After Fifth Training (1600 Cycles)

60 80 100 Stress (MPa)

120

140

Strain in Martensite

4 3

AC C

Shear Strain (%)

250

M AN U

Tubes 1 and 2 Before Training After First Training (100 Cycles) After Second Training (350 Cycles)

TE D

Actuation Strain (%)

5

SC

0

Characterization Sequence was performed

Actuation Strain

2

Strain in Austenite

1 0

0

400

800 1200 1600 Number of cycles Figure 2. a) A representative shear strain vs. temperature response of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tube samples, presenting the definition of the critical strains (strain in martensite, strain in austenite, and the actuation strain) under various stress levels in the “characterization 23

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

sequence”; see the text for details. This particular response is from the Tube-5 sample (Table 1) before thermo-mechanical training. Only the 3rd cycles under each stress level are shown. b) Shear strain vs. applied shear stress response of the NiTiHf torque tube samples (Tube-1 and Tube-2 in Table 1) during the “characterization sequence” under a training shear stress of 145 MPa after different number of thermo-mechanical training cycles. Note that TWSMS was determined by extrapolating the strain values under the two lowest stress levels. The figure shows the average from the two identical tubes tested. c) Evolution of the strain in martensite, strain in austenite, and actuation strain in Tube-2 under 145 MPa for the long training path (1600 cycles). The strain drops at certain number of cycles indicates that the characterization sequence was applied to monitor the TWSMS evolution.

Figure 3. Flowchart of the annealing procedure used in the present study to investigate the thermal stability of two-way shape memory effect (TWSME). Rather than using a different sample for each annealing temperature, the experiments were designed to make sure that the initial trained states before each annealing temperature were similar. When needed, 100 additional training cycles under the training stress were sufficient to restore the as-trained condition prior to each annealing step.

24

3.5 3.0 2.5

RI PT

2.0 1.5

Ni50.3Ti29.7Hf20 HTSMA Torque Tubes Heat Treated at 550 °C for 3 Hrs

1.0

Thick tubes (#1 and 2) trained at 145 MPa Thick tubes (#3 and 4) trained at 200 MPa Thin tubes (#5 and 6) trained at 145 MPa

SC

0.5 0.0 0

200

400

M AN U

Two-Way Shape Memory Strain (%)

ACCEPTED MANUSCRIPT

600 800 1000 Number of Cycles

1200

1400

1600

Heat Flow (A.U.)

TE D

Figure 4. Evolution of TWSMS as a function of the number of cycles for the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes. It can be seen that TWSMS saturates after 600 cycles in all conditions. There is no significant effect of wall thickness or training stress level for the conditions studied in the present work.

Ni50.3Ti29.7Hf20 HTSMA Heat Treated at 550 °C for 3 Hrs Before training After 600 cycles under 200 MPa

EP

Cooling

AC C

Heating

50

100

150 Temperature (°C)

200

250

Figure 5. DSC results for the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes before and after thermo-mechanical training under 200 MPa shear stress for 600 cycles. While there is not a significant change in transformation temperatures, there is a reduction in the enthalpy of transformation which can be related to the stabilization of martensite and austenite. (HT: Heat treatment).

25

ACCEPTED MANUSCRIPT

a

3 2 High Stress 200 MPa

1

0

50

100 Stress (MPa)

150

200

M AN U

b

Low Stress 145 MPa Before Training After First Training (100 Cycles) After Second Training (350 Cycles) After Third Training (600 Cycles)

SC

0

5 4

Strain in Martensite

3

Actuation Strain

2

TE D

Shear Strain (%)

RI PT

Actuation Strain (%)

4

1 0

100

EP

0

Low Stress145 MPa High Stress 200 MPa

200 300 400 Number of Cycles

Strain in Austenite

500

600

AC C

Figure 6. a) Shear strain vs. applied shear stress response of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes trained under low stress, 145 MPa (Tube-1 and Tube-2 in Table 1) and high stress, 200 MPa (Tube-3 and Tube-4 in Table 1) during the “characterization sequence” after various number of thermo-mechanical training cycles demonstrating how actuation strains and TWSMS evolve with the number of training cycles. The results shown are the average from the two identical tubes tested in each training stress level. b) Evolution of critical strains with the number of cycles for high (200 MPa, Tube 3) and low (145 MPa, Tube 2) stress training.

26

ACCEPTED MANUSCRIPT

a

3 2 Thin Tubes 1.33 mm

1

20

40

60 80 100 Stress (MPa)

120

140

M AN U

5

SC

0

Strain in Martensite

4 3

Actuation Strain

Thick Tube Thin Tube

2 1 0

400

EP

0

TE D

Shear Strain (%)

Thick Tubes 1.86 mm Before Training After First Training (100 Cycles) After Second Training (350 Cycles)

After Third Training (600 Cycles) After Fourth Training (1100 Cycles) After Fifth Training (1600 Cycles)

0

b

RI PT

Actuation Strain (%)

4

Strain in Austenite

800 1200 Number of Cycles

1600

AC C

Figure 7. a) Shear strain vs. applied shear stress response of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes trained under 145 MPa with two different wall thicknesses. Thick walled tubes (Tube-1 and Tube-2 in Table 1) and thin walled tubes (Tube-5 and Tube-6 in Table 1) during the “characterization sequence” after various number of thermo-mechanical training cycles demonstrating how actuation strains and TWSMS evolve with the number of training cycles. The results shown are the average from the two identical tubes tested in each training stress level. b) Evolution of critical strains with the number of cycles for thick (Tube-2) and thin (Tube-5) walled tubes.

27

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

Figure 8. Bright field TEM images of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes before and after thermo-mechanical training under shear stress. a) Precipitate structure of the untrained material which was precipitation heat treated at 550°C for 3 hrs. Precipitates are lenticular shaped with dimensions of 8 to 25 nm in length and 4 to 10 nm in width. b) Martensite structure of the untrained material shows large martensite variants in self-accommodated morphology, absorbing the precipitates. c) Precipitate structure of the trained material after 600 cycles under 200 MPa (Tube-3), basically showing that there is no notable change in precipitate size and morphology after training. d) Martensite morphology of the trained material (Tube-3).

28

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 9. Bright-field TEM image of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes after thermo-mechanical training under shear stress exhibiting large and thin martensitic plates, which leads to nearly parallel plates, even in different crystallographic grains.

29

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

Figure 10. a) The bright field TEM image of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes after thermo-mechanical training under shear stress showing different martensite variants and remnant austenite at room temperature. b) Selected Area Electron Diffraction Pattern (SAEDP) of the entire region shown in (a). c) SAEDP from the region marked as V1, showing a martensite variant (B19’); d) SAEDP from the region marked as V2, showing a martensite variant (B19’); and e) SAEDP from the region marked as V3, showing remnant austenite (B2).

30

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 11. Bright field TEM images of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy torque tubes after thermo-mechanical training under shear stress showing dislocations. a) Dislocations between precipitates and b) refined martensite twins interrupted by dislocations.

TE D

Before Training After Training (1600 Cycles) After 1st Annealing (350 ºC 30 min) After 3rd Annealing (450 ºC 30 min)

4

EP

3 2 1

AC C

Actuation Strain (%)

5

After 5th Annealing (550 ºC 30 min) After 7th Annealing (650 ºC 30 min) After 8th Annealing (700 ºC 30 min)

0

0

20

40

60 80 100 Stress (MPa)

120

140

Figure 12. Two-way shape memory strain (TWSMS) and actuation strain under various stresses in thermo-mechanically trained Ni50.3Ti29.7Hf20 high temperature shape memory alloy tube (Tube-2) after static annealing treatments at various temperatures for 30 min. The TWSMS is stable at temperatures up to 400°C and completely deteriorates above 600°C. For the sake of clarity, the responses from annealing steps 400, 500 and 600 are not shown. 31

ACCEPTED MANUSCRIPT

a

After trained 700 °C 30 min annealed

Cooling

After trained 350 °C 30 min annealed

After trained 700 °C Heating 30 min annealed

100

120

140 160 Temperature (°C)

180

200

M AN U

b

RI PT

Before trained 550 °C 3 Hrs aged

SC

Heat Flow (A.U.)

Annealing Temperature Increases With 50 °C Increment Starting From 350 °C

180 170 160 150

TE D

Temperature (°C)

Austenite Peak Temperature (Ap)

Martensite Peak Temperature (Mp)

140 130

400

450 500 550 600 Annealing Temperature (°C)

EP

350

650

700

AC C

Figure 13. a) DSC results of the untrained and trained + annealed samples of the Ni50.3Ti29.7Hf20 high temperature shape memory alloy tubes. The most inward pair is for the as precipitation heat treated, untrained material. The shifted curves to outside are for annealing steps. Annealing was done for 30 minutes for each step and started from 350°C up to 700°C with 50°C increments. b) Martensite (Mp) and austenite (Ap) peak temperatures are plotted as a function of the annealing temperature. Note that there is no significant change up to 600°C annealing step. At and above 700°C heat treatment, the transformation temperatures reduced to previous levels but it is suspected that microstructure has already been changed.

32

4

3

RI PT

Before Annealing TWSMS

2

SC

1

0 350

M AN U

Two-Way Shape Memory Strain (%)

ACCEPTED MANUSCRIPT

400

450 500 550 600 Annealing Temperature (°C)

650

700

AC C

EP

TE D

Figure 14. Change in Two-Way Shape Memory Strain (TWSMS) as a function of the annealing temperatures for the Ni50.3Ti29.7Hf20 high temperature shape memory alloy tubes after the thermomechanical training. Each annealing step is for 30 minutes at given temperature. As trained sample’s TWSMS has been added with blue dashed line to be a reference. There is a slight increase after first annealing which may be associated with the dissolution of retained martensite, thus liberating more material to go through the phase transformation.

33