SPIE Conf. 6170 Smart Structures and Materials 2006: Active Materials: Behavior and Mechanics - submitted Properties of a Ni Pd Ti high-temperature shape memory 19.5 30 50.5 alloy in tension and compression Ronald Noebe#, Santo Padula, II#, Glen Bigelow#, Orlando Rios+, Anita Garg#, and Brad Lerch# #NASA Glenn Research Center, Materials & Structures Division, Cleveland, OH 44135 +University of Florida, Dept. of Materials Science & Engineering, Gainesville, FL 32611 ABSTRACT Potential applications involving high-temperature shape memory alloys have been growing in recent years. Even in those cases where promising new alloys have been identified, the knowledge base for such materials contains gaps crucial to their maturation and implementation in actuator and other applications. We begin to address this issue by characterizing the mechanical behavior of a Ni Pd Ti high-temperature shape memory alloy in both uniaxial 19.5 30 50.5 tension and compression at various temperatures. Differences in the isothermal uniaxial deformation behavior were most notable at test temperatures below the martensite finish temperature. The elastic modulus of the material was very dependent on strain level; therefore, dynamic Young’s Modulus was determined as a function of temperature by an impulse excitation technique. More importantly, the performance of a thermally activated actuator material is dependent on the work output of the alloy. Consequently, the strain-temperature response of the Ni Pd Ti alloy under various 19.5 30 50.5 loads was determined in both tension and compression and the specific work output calculated and compared in both loading conditions. It was found that the transformation strain and thus, the specific work output were similar regardless of the loading condition. Also, in both tension and compression, the strain-temperature loops determined under constant load conditions did not close due to the fact that the transformation strain during cooling was always larger than the transformation strain during heating. This was apparently the result of permanent plastic deformation of the martensite phase with each cycle. Consequently, before this alloy can be used under cyclic actuation conditions, modification of the microstructure or composition would be required to increase the resistance of the alloy to plastic deformation by slip. Keywords: High-temperature shape memory alloy, NiTiPd, tensile properties, compression, specific work output, shape memory behavior, dynamic modulus, transformation strain, transformation temperatures. 1.0 INTRODUCTION Ever since the shape memory phenomenon was observed in NiTi alloys, the number of potential applications for this functional material has continued to grow. General applications include use in couplings, electrical interconnects, high- force actuators for automotive, aeronautic, and robotic applications, temperature sensitive flow control devices, and numerous other applications that can benefit from the unique properties of shape memory or superelastic behavior. Nevertheless, many desired applications for shape memory alloys never reach fruition because of the inherent temperature limitations of binary Ni-Ti alloys, which have a maximum martensite start temperature (M) of about 60 °C s [1]. In contrast to the relatively low transformation temperatures in Ni-Ti alloys, binary alloys based on precious metal- Ti compositions such as PdTi exhibit martensitic transformations with M temperatures exceeding 500 °C [2]. PdTi and s NiTi also form a continuous solid solution with a high-temperature B2 phase that transforms to a B19 (orthorhombic) or B19’ (monoclinic) low-temperature martensite phase with transformation temperatures between those of the binary alloys [3]. It has also been confirmed that the martensitic transformations over the entire range of ternary Ni-Pd-Ti compositions are thermoelastic in nature and that the alloys exhibit aspects of shape memory behavior similar to binary NiTi alloys [4,5]. Since then, work on ternary NiTiPd alloys has intensified because of their potential for use in high temperature applications. Shimizu et al. [6] examined the effect of stoichiometry ((Ni+Pd):Ti ≠ 50:50) on the transformation temperatures of a series of Ti Pd Ni alloys (where x = -0.6 to 1.5 at.%). An extremely steep drop in transformation temperature was 50-x 30 20+x observed for Ti-deficient alloys but only a moderate change in transformation temperature was observed on the Ti-rich side of stoichiometry where fine Ti Ni type precipitates were reported to form after aging at 500 °C. This effect of 2 stoichiometry on the transformation temperatures for NiPdTi alloys is consistent with binary NiTi alloys where there is a very steep drop in transformation temperatures for Ti-deficient compositions and nearly constant transformation temperatures with Ti content for stoichiometric and Ti-rich compositions [1]. The shape memory behavior of NiTi-30Pd (at.%) alloys has been studied in some detail. Tensile samples loaded to 2- 4% total strain in the martensitic condition were capable of recovering 100% of that strain through various mechanisms and almost 90% of the strain could be recovered in samples loaded to 6% total strain [5,6,7]. Similar shape memory behavior has been observed for samples deformed in compression [8] and torsion [9]. However, for alloys containing 40 at.% or more Pd including the TiPd binary alloy, only about 0.5% strain could be recovered in alloys loaded in tension to various strain levels [5,10]. This limited amount of strain recovery has been attributed to a low critical stress for slip (which is an irreversible process), such that the majority of the deformation is accommodated by slip rather than the more typical twin boundary motion or martensite reorientation (also referred to as detwinning) [10]. A number of studies have since been performed that have focused on improving the shape memory behavior of NiTiPd alloys through various mechanisms aimed at enhancing the alloys resistance to slip and thus promoting greater deformation by twin motion in hopes of enhancing the stress-free shape memory behavior of these alloys. Such studies have involved the use of solid solution strengthening [8,11], precipitate strengthening [6], and various thermomechanical processing routines [12,13], with little success in the former and moderate success with the other approaches. To the authors’ knowledge, all the work published to date on the functional behavior of NiTiPd alloys has involved recovery of strains under stress free conditions. However, the most likely applications for high temperature shape memory alloys involve their use as actuator materials, where the material recovers against an applied or biasing stress, thus performing work on a system. It should be noted that work can be extracted from NiTi shape-memory alloys at energy densities exceeding 10 J/cm3 [14], making binary NiTi alloys very intriguing for use as solid state actuators. However, recovery of NiTiPd alloys under stress has not been previously investigated and no direct studies comparing the general tensile and compression behavior of NiTiPd alloys have been reported. Consequently, a detailed study of the behavior of a Ni Pd Ti alloy under tensile and compression loading including the strain-temperature response of 19.5 30 50.5 the alloy under various constant stresses was performed to determine the work characteristics of this potential high temperature shape memory alloy. 2.0 MATERIALS AND PROCEDURES 2.1 Material Processing and Characterization An alloy of target composition Ni Pd Ti (all compositions are noted in atomic percent) was chosen for analysis, 19.5 30 50.5 because the slight Ti-rich composition would help ensure a high transformation temperature [6]. Ingots were prepared by vacuum induction melting of high purity elemental constituents (99.95 Ti, 99.995 Pd, 99.98 Ni) in a graphite crucible. The induction melter was equipped with tilt-pour capability and the melts were cast into a cylindrical copper mold 25.4 mm in diameter by 102 mm in length, with an appropriate hot top to accommodate shrinkage during solidification. Prior to extrusion, the ingots were homogenized in a vacuum furnace at 1050C °C for 72 h. The ingots were subsequently placed into mild steel extrusion cans, which were evacuated and sealed. The extrusion process was performed at 900 °C and a 7:1 area reduction ratio. The chemical composition was determined for all materials after extrusion by inductively coupled plasma spectroscopy and by conventional N/O and C/S determination. Rough blanks for cylindrical shaped compression samples were cut from the extruded bars by electrical discharge machining (EDM) and centerless ground to final diameter. The finished specimens were 5 mm in diameter by 10 mm in length. Cylindrical dog-bone-shaped tensile samples were cut from appropriate lengths of the extrusion rod using a computer controlled lathe. The total length of each sample was 50.8 mm with threaded button ends. The gage section of each of the tensile samples was 17.4 mm long by 3.81 mm diameter. Both tension and compression samples were tested in the direction parallel to the original extrusion axis. To complete the sample preparation phase, all samples were given a stress relief heat treatment at 400 °C for 1 hour followed by a furnace cool prior to testing. The function of this heat treatment was to relieve any residual stresses due to the extrusion and machining processes. The microstructure of the extruded Ni Pd Ti alloy after the stress relief treatment was examined by conventional 19.5 30 50.5 scanning electron microscopy (SEM) and transmission electron microscopy techniques (TEM). Samples for SEM were 2 mounted and metallographically polished prior to examination in the SEM. A JEOL 840 and Hitachi 6100 equipped with energy-dispersive spectroscopy (EDS) capability were used to determine the basic microstructure of the alloy and to obtain an indication of the volume fraction and composition of any additional phases present. TEM samples were prepared by cutting 3mm diameter by 0.5 mm thick disks by EDM. These disks were subsequently ground to a thickness of approximately 0.1 mm, dimpled, and then ion milled to final electron transparency. TEM analysis including electron diffraction and EDS analysis was performed on the second phase particles in order to conclusively identify all phases present using a Phillips 400T or CM200 microscope. X-ray analysis was used to determine the structure and lattice parameters of the martensite phase. Transformation temperatures and the temperature dependent electrical properties of the Ni Pd Ti alloy in the stress 19.5 30 50.5 free condition and in-situ during mechanical testing were determined by high resolution resistivity measurements. The samples were instrumented with a four point probe and a K-type thermocouple each of which were spot welded directly to the sample surface. A four point probe configuration was chosen in order to eliminate the effects of contact resistance. Pt or Ni wires were spot welded to the sample functioning as the voltage sensing leads. Current excitation was supplied by an Agilent power supply while the temperature and voltage signals were processed by a calibrated National Instruments DAQ system and analog signal conditioner. The digitized signals were used to calculate a resistance value for each specimen, which in turn allows for the calculation of the samples resistivity, after having appropriately accounted for changes in the sample dimensions due to thermal expansion. The transformation temperatures were also determined by measuring the strain-temperature response of the material under essentially zero load (stress free condition) using the mechanical test setup described below. 2.2 Mechanical Testing Mechanical testing was performed using an MTS servo-hydraulic test frame equipped with an MTS Flextest SE controller. Hydraulic collet grips with a hot grip insert were used to hold the threaded specimens. A 22 kip load cell was used, and strain measurements in tension were taken with a 12.7 mm gage length extensometer. The extensometer had a maximum range of +20/-10 % strain and was equipped with 85mm long alumina probes with a v-chisel edge. Strain measurements during compression testing were acquired using a laser extensometer. Both tension and compression specimens were induction heated using an Ameritherm Novastar 7.5 power supply. All temperatures were acquired through the use of K-type thermocouples, which were directly spot-welded to the sample. Temperature gradients across the gauge were calibrated to within +/-0.5% of the test temperature. Monotonic tensile tests were conducted in strain control at a rate of 1 x 10-4sec-1. Tensile specimens were run to failure or until the 20% strain limit of the extensometer was reached, in which case the samples were unloaded prior to fracture. Compression samples were run in displacement control at a rate which approximately yielded the same initial strain rate as the tension tests. Shape memory behavior was measured using unconstrained or load free strain recovery tests. During this test, specimens were deformed in tension in strain rate control at a rate of 1 x 10-4 sec-1 to the required strain level at 100 °C. At this point the controller was switched to load control and the load was reduced to 0 N. Load was held at 0 N while the specimens were thermally cycled to a temperature in the range of 400 ºC. Heating rates were maintained at 10 ºC/min after which the samples were allowed to air cool to 100 °C before repeating the process starting at the next larger strain increment. The work characteristics of the material were determined from constant load, strain versus temperature tests (referred to as load-bias tests). Specimens were strained at room temperature to the required load in strain control at a rate of 1 x 10-4 sec-1. At this point the controller was switched to load control to maintain the desired load. Specimens were then thermally cycled twice from room temperature to about 100 ºC above the austenite finish temperature. Heating rates were maintained at 10 ºC/min. In tension, the load bias tests were run in a series of increasing loads on the same sample, the specimen was unloaded at room temperature and then strained again to the next higher load level, at which point the thermal cycling process was repeated. The specific work output of the alloy was determined by measuring the resultant change in strain during the martensite-to-austenite transformation during the second heating cycle and multiplying by the stress applied during the thermal cycle. Compression load bias tests were run in a parallel procedure but with the loading phase in displacement rate control rather than strain rate control. Since it was very difficult to determine a linear elastic region in these samples, even with extensometry, modulus was determined by a dynamic technique using the impulse excitation of vibration method as described in ASTM Standard 3 E1876 [15]. A rectangular sample, 7.9 mm in width, 3 mm in thickness, and 49 mm long was heated in a furnace at a rate of 3.3 °C/min. Oxidation was minimized by backfilling the furnace with argon. The tests were conducted between 20 and 800 °C taking data every 5 °C during heating only. The samples were impacted with a ceramic projectile creating a flexural mode of vibration and the resulting frequency was measured using a microphone. This frequency was related to Young’s modulus by the equations given in reference [15]. 3.0 RESULTS AND DISCUSSION 3.1 Material Characterization A total of six ingots were cast, homogenized, and extruded for use in this and follow on studies. The average composition of the six extrusions was 50.5 ± 0.1 at.% Ti, 19.4 ± 0.2 at.% Ni, 29.4 ± 0.2 at.% Pd, 0.44 ± 0.03 at.% C, and 0.28 ± 0.03 at.% O. Within the resolution of the chemical measurement techniques, none of the individual ingots were discernibly different in composition from each other or the aim composition. In addition, oxygen and carbon (due to the use of graphite crucibles) were present in essentially equal levels within all the extrusions. The microstructure of the extruded and stress relieved material is shown in Figure 1. The alloy was martensitic at room temperature and predominantly single phase containing a small volume percent of second phase particles. Both x-ray and electron diffraction confirmed that the martensite phase was the B19 (orthorhombic) phase typically observed in ternary alloys with greater than 10 at.% Pd [3,5]. The lattice parameters were a = 0.277 nm, b = 0.447 nm, and c = o o o 0.474 nm. There were two types of second phases present, an intermetallic phase containing Ti, Ni and Pd and the other containing a high amount of Ti tied to the interstitials, C and O. Between electron diffraction and energy dispersive spectroscopic (EDS) analysis, it was determined that the former phase was Ti (Ni,Pd), which is isostructural to the face- 2 centered-cubic Ti Ni phase commonly observed in Ti-rich NiTi alloys. The Ti (Ni,Pd) phase had a lattice parameter a = 2 2 o 1.16 nm. Ti Ni has a high solubility for interstitially located oxygen [16] and given the high oxygen content in this 2 material the Ti (Ni,Pd) probably contained some oxygen as well. The other phase was a typical fcc structured Ti(C,O) 2 with a = 0.44 nm. The oxycarbide particles ranged in size from about 0.5 to 2 µm in diameter. The Ti (Ni,Pd) phase o 2 was blocky in morphology and ranged from 2-6 µm in size. This phase formed interdendritically during melting and then was broken up during the extrusion process. The total volume fraction of second phase particles was less than 5%. It should be noted that no fine precipitate phase was observed in this alloy as reported by Shimizu et al. [6], nor were any phase transformations observed during differential thermal analysis of the alloy up to 1000 °C, other than the martensitic transformation. The temperature dependence of the resistivity and the strain temperature response of the material determined under zero load conditions are plotted in Figure 2. The transformation temperatures were determined from this data by constructing linear, polynomial curve fits through the low temperature, intermediate temperature, and high temperature portions of the heating and cooling curves, respectively. Through the utilization of the equations for these best fits, start and finish temperatures were determined by interrogating the intersection points. The resulting transformation temperatures determined by the resistivity and strain measurements can be found in Table 1. There was excellent agreement between the two techniques. Given the substantial differences in the electrical properties between the parent B2 and martensite phases, it is apparent why resistance measurement is a prominent characterization technique in the study of shape memory alloys. The technique can be used as an in-situ probe for determining the microstructural changes that occur in a material not only as a function of temperature but also as a function of stress as demonstrated later in this paper. The design of resistance heated SMA actuated devices also requires an understanding of the temperature dependence of the electrical resistance of these alloys in order to determine power requirements for actuators. The thermal expansion behavior of the martensite and austenite phases were similar with the measured coefficients of thermal expansion being 16 x 10-6 and 11 x 10-6 °C-1 for the martensite and austenite phase, respectively. The volume change associated with the martensite-to-austenite transformation resulted in an axial strain change along the length of the sample of nearly 0.2%. 4 Table 1: Transformation Temperatures and Hysteresis for Ni Pd Ti 19.5 30 50.5 Transformation Measurement Technique Temperature (°C) Dilatometry Resistance A 250 256 s A 257 260 f M 249 249 s M 238 239 f Hysteresis (A - M) 8 11 f s 3.2 Isothermal Stress-Strain Behavior in Tension and Compression Representative true stress - true strain curves for the Ni Pd Ti alloy in tension and compression are shown in 19.5 30 50.5 Figures 3 and 4. All isothermal stress strain curves are shown in true stress - true strain, thus allowing for the direct comparison of the mechanical properties in tension and compression. Figure 3 is a plot of the stress strain relationship of the austenite and martensite phases well above or below the transformation temperatures, while Figure 4 shows the isothermal mechanical behavior of the Ni Pd Ti alloy between or very near the transformation temperatures. 19.5 30 50.5 Isothermal tension and compression tests revealed that the yield stresses in both modes were comparable over the range of temperatures investigated and that the general characteristics of the stress-strain curves were similar except for samples tested below the martensite finish temperature (M) (Figure 3). The room temperature (RT) and 200 °C tensile f curves (Figure 3a) exhibited a “pseudo-stress” plateau or region of low hardening rate at stress levels starting about 250 MPa. While this pseudo-stress plateau was not nearly as dramatic as that observed in binary NiTi alloys [17], it was still attributed to the reorientation of favored martensite variants, since deformation in this regime was thermally recoverable (as described in section 3.4). In contrast, no discernable stress plateau was observed in compression and a high work hardening rate was observed from the onset of yielding out to strains of about 10% indicating a much greater resistance to deformation. Deformation by detwinning mechanisms is polar in nature, in contrast to slip behavior, such that a reversal in the shear direction will not produce twin movement in variants favorably aligned for operation in the forward direction [18]. In other words, twins aligned favorably for operation in compression will not operate under a tensile stress and vice versa. Apparently the manner in which the material was processed (extrusion in this case) resulted in a more favorable variant structure for detwinning in tension than compression. Additional testing and microstructural characterization of deformed samples are planned to elucidate the actual deformation mechanisms operating in the martensite phase in compression and after the stress plateau in tension. Well above the transformation temperature, the mechanical properties of the austenite were comparable in tension and compression (Figure 3a & b). After yielding, the material exhibited a low work hardening rate that decreased with increasing isothermal test temperature. In the 400 °C tensile and 500 °C compression tests the Ni Pd Ti alloy 19.5 30 50.5 exhibited a slight work softening due to recovery processes that become dominate at elevated temperatures. For the results shown in Figure 4a and 4b, the samples were heated to well above the austenite finish temperature (A) f and allowed to cool back to the test temperature before applying load to the samples. The process of heating significantly beyond the A and subsequently cooling back to test temperatures of 255 and 272 °C ensured that the f resulting stable phase was that of austenite, since the martensite start temperature (M) was 249 °C. In-situ resistivity s measurements and the occurrence of a well defined stress plateau indicated that upon loading in either tension or compression, stress induced martensite is formed. This behavior can be seen in detail in Figure 5, which is a superimposed plot of resistivity, determined in-situ during compression testing at 255 °C, with the stress - strain data. The resistivity initially increased with strain during elastic deformation of the austenite, and then immediately decreased as stress induced martensite was formed (the resistivity of austenite is larger than that of the martensite as shown Figure 2, hence the decrease in resistivity). For both tension and compression, the relative amounts of stress induced martensite as well as the stress at which the transformation occurred were comparable. For tensile samples deformed below the M s (225 and 245 °C), deformation was similar to that of martensite at other temperatures, shown previously in Figure 3a. Finally, a tensile ductility minimum was observed in the region where the stress induced martensite occurred, similar to the behavior reported in NiTiPt shape memory alloys [19]. 5 ) m 140 c 1.0 * m 120 h Resistivity o 0.8 o- 100 ) TTii ((NNii,,PPdd)) cr 0.6 % 22 mi 80 n ( y ( Dilatometry 0.4 rai vit 60 St TTii((CC,,OO)) ti 0.2 s si 40 e 0.0 R 0 100 200 300 400 500 Temperature (oC) F igure 1. Backscattered electron image of the m icrostructure of the extruded Ni19.5Pd30Ti50.5 alloy. At Figure 2. Electrical resistivity and strain versus r oom temperature, the alloy is an orthorhombic structured temperature for Ni19.5Pd30Ti50.5. Corresponding martensite, which contains a small volume fraction of TiC transformation temperatures determined by these two (fine black particles) and Ti(Pd,Ni) (larger gray phase). techniques are listed in Table 1. 2 2000 2000 (b) Compression (a) Tension a) a) RT MP 1500 RT MP 1500 200 oC ( s ( ress 1000 300 oC tres 1000 350 oC t S S e 500 400 oC ue 500 365 oC u r r 200 oC T 500 oC T 0 0 0 5 10 15 20 0 5 10 15 20 True Strain (%) True Strain (%) Figure 3. Representative true stress - true strain behavior for Ni Pd Ti at temperatures significantly above or below the 19.5 30 50.5 transformation temperature of the alloy determined in a.) tension and b.) compression. 1000 2000 (a) Tension (b) Compression ) ) a 800 a P P1500 M M ( ( s 600 s s s re 225 oC re1000 St 400 272 oC St ue 200 245 oC ue 500 272 oC 300 oC Tr 255 oC Tr 255 oC 0 0 0 2 4 6 8 0 5 10 15 20 True Strain (%) True Strain (%) Figure 4. Representative true stress - true strain behavior for Ni Pd Ti at temperatures within or near the transformation 19.5 30 50.5 hysteresis in a.) tension and b.) compression. 6 The temperature dependence of the yield stress in compression and tension, determined from the proportional limit of the isothermal stress strain curves, is shown in Figure 6. Generally, the yield stress in tension and compression were similar at a given temperature. The yield stress of the martensite decreased with increasing temperature and reached a minimum near the A. Beyond this minimum the alloy was austenitic (though the formation of stress-induced martensite f was possible) and the yield strength increased significantly with increasing test temperature until a peak near 350 °C. At temperatures above 400 °C the austenite weakened and the onset of slip occurred at lower stresses. It is important to note that the yield stress for the martensite represents the stress at which detwinning begins, while yielding of the austenite phase is due to plastic deformation by dislocation processes. 3.3 Dynamic Modulus It was very difficult to define a linear elastic region from the tensile and compression tests, especially at temperatures near or below the transformation temperature of the alloy as the elastic modulus was very dependent on strain level. Therefore, in order to have an accurate representation of the elastic modulus of these materials as a function of temperature a dynamic modulus test was conducted. The modulus - temperature relationship is plotted in Figure 7, and has three main regions corresponding to the material in the fully martensitic condition, the two-phase martensite plus austenite region that exists between A and A, and the fully austenitic phase. The dynamic modulus of the martensite at s f room temperature was just over 20 GPa higher than that of the austenite at A, but decreased with temperature. In f contrast, the modulus increased slightly with temperature for the fully austenitic material suggesting that the stability of this phase was increasing with increasing temperature up to about 750 °C. The region bound by the transformation A s and A temperatures exhibited a stronger temperature dependence than did either of the single phases. The modulus f within this regime was proportional to the effective modulus of the resulting transformation products in the two-phase martensite plus austenite region, terminating at a minimum of 59 GPa at the A. At this point the material was fully f austenitic and the modulus value was that of the austenite at the transformation temperature. It should be noted that the temperatures in Figure 7 were corrected to account for the fact that the actual sample temperature was not measured during the test. Because the martensite-to-austenite transformation is very endothermic, it suppresses the actual temperature of the sample compared to the furnace temperature. The thermocouple used to record the sample temperature for the modulus measurements was near but not touching the sample, since a directly attached thermocouple would interfere with the modulus measurements. In contrast, thermocouples were spotted welded directly to the resistivity and dilatometry samples used to generate the data in Figure 2. 3.4 Shape Memory Behavior Shape memory behavior is the ability of a material to recover twinning-induced deformation that is introduced at temperatures below the martensite finish temperature by heating the material through its transformation temperature. In practice [6-13], it is simply determined by measuring the amount of strain introduced into a sample during deformation below the M, heating the sample through the transformation above the A, and cooling back usually to room temperature f f where the final dimensions of the sample are measured and used to determine the amount of strain recovered. However, if one were to monitor the strain - temperature response of the material continuously during this process, as shown in Figure 8a, it is apparent that a number of different mechanisms contribute to the overall strain recovery of NiTiPd alloys as first reported by Lindquist [20]. Figure 8a shows the strain temperature response of the Ni Pd Ti alloy as it was 19.5 30 50.5 loaded to approximately 2% total strain and unloaded at 100 °C, resulting in approximately 1% “plastic” deformation. This was followed by heating to 400 °C and cooling back to 100 °C in an unconstrained condition. Three contributing factors to strain recovery can be identified in Figure 8a: (1) Continuous shape recovery began on heating up to the A s temperature and the amount of recovery was estimated from the differences between the deformed martensite strain – temperature heating curve and the original undeformed martensite thermal expansion behavior taken from Figure 2. This difference may be the result of the rearrangement of twins in the material due to the initial deformation, such that the expansion characteristics of the detwinned material along this one direction may no longer be the same as measured in a sample with a more randomly oriented martensite. However, no attempt was made to confirm this through microscopic observations of the sample structure. (2) The majority of the strain recovery occurred between A and A s f but there were really two competing effects. There was a volume increase due to the transformation from martensite to the austenite phase and there is the typical strain recovery due to the shape memory effect, which was generally the overriding factor for strain change in samples deformed to greater than 1% plastic strain. (3) Finally, there was the strain change attributed to the volume decrease as the material transformed back from austenite to martensite. However, this strain does not necessarily have to be equivalent to the strain change on heating of the deformed sample for essentially 7 2000 105 500 Stress Compression a) 1600 Resistivity 100 nit) Pa) 450 Tension MP 95 u M 400 ( b. ( s 1200 90 ar ss 350 s ( e re 85 y tr 300 t 800 t S S i v ue 80 sti eld 250 Tr 400 75 si Yi 200 e R 0 70 150 0 5 10 15 20 0 100 200 300 400 500 600 True Strain Temperature (oC) Figure 5. Stress - strain behavior and corresponding in-situ Figure 6. Yield strength (proportional limit) determined in resistivity - strain measurements for Ni Pd Ti tested at tension and compression as a function of temperature for 19.5 30 50.5 255 °C. The stress plateau and resistivity behavior are Ni Pd Ti . 19.5 30 50.5 indicative of stress induced martensite at this temperature. a) 120 P G s ( 100 u dul 80 Figure 7. Dynamic modulus for o M Ni Pd Ti as a function of temperature 19.5 30 50.5 ungs 60 AsA dteemtepremraintuerde o ann dh e7a5ti0n g°C b.e tween room o 40 f Y c mi 20 a n y D 0 0 100 200 300 400 500 600 700 800 Temperature (oC) (a) 100 (b) % Total Strain Recovered 2.0 After Complete Hysteresis m) %)80 Strain (cm/c 011...505 Thermal Expans12ion Marte3AnssitAef raction of Initial ain Recovered (246000 c(tohnetrrmibault imona rdc(tseuohnenas ttrpioitbee mu rrteeeiocccnooh vvadeenurrieyys mtfuorpo (m m1to)e AcAshs a)ton iAsfm) (2) 0.0 Ftr S contribution due to mechanism (3) 0 (volume change manifested in the axial direction) 50 100 150 200 250 300 350 400 450 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Temperature (oC) Initial Plastic Strain (%) Figure 8. a.) Unconstrained strain - temperature response of Ni Pd Ti deformed to 2% strain at 100 °C unloaded and heated 19.5 30 50.5 through the transformation temperature and cooled back to 100 °C. The various mechanisms that contribute to the total strain recovery of the material are defined on the plot. b.) Percent plastic strain recovered after thermal cycling versus initial plastic strain introduced in the sample and each of the components that make up the total strain recovered. 8 the same reasons as discussed for thermal expansion differences that occur in region 1. Furthermore, as the initial strain introduced in the sample increased, the strain measured along the deformation axis, due to the volume decrease on cooling from austenite to martensite, was found to decrease significantly, as shown in Figure 8b. This may be attributed to nucleation of a preferred orientation of martensite on cooling due to the dislocation structure generated in the material during plastic deformation of the martensite at the larger initial strains, but was not confirmed experimentally. The overall strain recovery as a percent of the initial plastic strain introduced into the sample at 100 °C is plotted as a function of initial plastic strain in Figure 8b. The Ni Pd Ti alloy is capable of recovering 100% of the initial 19.5 30 50.5 plastic strain introduced into the sample up to an initial plastic strain of 1% and is capable of recovering over 80% of the initial plastic strain at levels up to 3.5%. Different research groups plot this type of recovery information in slightly different forms but our results are consistent with those previously published for NiTi-30Pd alloys [5-7, 10-12]. The individual contributions of the total strain recovery are also shown in Figure 8b as a function of initial plastic strain and are labeled (1)-(3), respectively, corresponding to the discussion of these mechanisms in the previous paragraph and as defined in Figure 8a. 3.5 Constant-Load, Strain-Temperature Tests and Work Output Given the significant amount of research performed on unconstrained shape recovery of NiTi-30Pd alloys [5-13], it is surprising that no investigation of constrained recovery has been reported. This is especially astounding given the fact that the primary application for high-temperature shape memory alloys (HTSMA) such as NiTiPd would be for use as solid state actuator materials where work output (or the ability of the material to recover strain against some biasing force) is the primary consideration. Consequently, load bias testing was performed on the Ni Pd Ti alloy in order 19.5 30 50.5 to quantify the work output of the material as a function of the applied stress and mode of application (uniaxial tension or compression). For the constant-load, strain-temperature testing, the load was applied to the sample at room temperature and then held constant as the sample was heated and cooled through the transformation. During this process, the strain in the sample due to the initial applied load at room temperature, prior to any thermal cycling, was always much smaller than when the sample was cooled through the transformation while under load. As a result, the transformation strain during the initial heating cycle was also much smaller than in successive cycles. This behavior is clearly demonstrated in Figure 9, which shows the initial loading and then two complete heating and cooling cycles for the Ni Pd Ti alloy under a tensile 19.5 30 50.5 stress of 99 MPa. This type of behavior is quite common in conventional NiTi alloys [21] and is also observed in NiTiPt shape memory alloys [19]. There are probably several contributing factors: 1) The yield strength of the martensite phase decreases with increasing temperature reaching a minimum at the transformation temperature for the alloy. This effect would be expected to contribute to the increased strain by the unfavorable mechanism of plastic deformation. Evidence for the operation of this mechanism can be seen in the fact that the strain - temperature hysteresis loops do not close, with the strain on cooling through the transformation regime always being larger than the strain on heating. This results in permanent deformation of the alloy. 2) The dynamic Young’s modulus for the martensite phase decreases significantly with increasing temperature from approximately 82 GPa at room temperature to a minimum of about 58 GPa at the A temperature. 3) The largest contribution to the enhanced strain is probably due to the nucleation of f preferred variant orientations of martensite during cooling under load, which favor the applied strain. In other words, cooling under load begins to “train” the material, resulting in a greater fraction of favorably oriented martensite variants that can accommodate a larger strain in the sample for a given stress. Because of the difference in the level of strain developed in the material during loading at room temperature versus cooling through the transformation temperature under load, the first complete heating and cooling cycle at each stress level in the load-bias tests were ignored and only the second complete cycle or hysteresis is plotted in Figures 10a & b. Figure 10 shows the results of the strain – temperature behavior of the HTSMA at various stress levels for both tensile and compression loading. It is interesting to note that the strain - temperature response of the Ni Pd Ti under load 19.5 30 50.5 (Figure 10) was actually much simpler than that observed under stress free conditions (Figure 8a), with essentially all the strain change (due to shape recovery and plastic deformation) occurring within transformation region of the alloy. In addition, the actual transformation temperatures were found to increase with applied stress. At 393 MPa, the transformation temperatures were A = 285, A = 317, M = 299, and M = 267 °C, or an average increase of about 42 °C. s f s f 9 3.0 Second Cooling Cycle 2.5 Second Heating Cycle Figure 9. Initial loading and first two First Cooling Cycle ) 2.0 complete constant stress (99 MPa), strain- % temperature cycles for Ni Pd Ti . 19.5 30 50.5 ( First complete heating and cooling cycle n 1.5 i is represented by solid lines and the a second complete heating and cooling r St 1.0 cycle by dotted lines. 0.5 First Heating Cycle 0.0 Room Temperature Loading (99 MPa) 0 100 200 300 400 Temperature (oC) 12 (a) Tension -14 (b) Compression 10 -12 393 MPa ( 9.6 J/cm3) 687 MPa (6.4 J/cm3) -10 %) 8 %) -8 n ( 295 MPa ( 7.5 J/cm3) n ( -6 517 MPa (8.5 J/cm3) Strai 46 197 MPa ( 4.3 J/cm3) Strai --42 393 MPa (9.2 J/cm3) 295 MPa (7.7 J/cm3) 99 MPa ( 1.5 J/cm3) 0 197 MPa (4.5 J/cm3) 2 2 99 MPa (1.6 J/cm3) 0 100 200 300 400 100 200 300 400 500 Temperature (oC) Temperature (oC) Figure 10. Load biased, strain – temperature response of the Ni Pd Ti alloy at different stress levels in a.) compression and b.) 19.5 30 50.5 tension. Solid lines are heating curves and dashed lines represent the material response on cooling. The constant stress level and resulting work output are indicated for each set of curves. )12 ) 3.0 3 (a) % (b) m Compression Compression /c10 Tension n ( 2.5 Tension e i l a u r o 8 St (j n 2.0 k 6 o r Wo mti 1.5 4 a c r ifi 2 fo 1.0 c s e n p a S 0 r 0.5 0 100 200 300 400 500 600 700 800 T 0 100 200 300 400 500 600 700 800 Stress (MPa) Stress (MPa) Figure 11. a.) Specific work output for Ni Pd Ti as a function of applied stress loaded in both tension and compression and b.) 19.5 30 50.5 The corresponding transformation strain versus applied stress. 10