SMART CONCRETE BRIDGE GIRDERS USING SHAPE MEMORY ALLOYS

Transcription

1 SMART CONCRETE BRIDGE GIRDERS USING SHAPE MEMORY ALLOYS Morgan State University The Pennsylvania State University University of Maryland University of Virginia Virginia Polytechnic Institute & State University West Virginia University The Pennsylvania State University The Thomas D. Larson Pennsylvania Transportation Institute Transportation Research Building University Park, PA Phone: Fax:

2 FINAL REPORT SMART CONCRETE BRIDGE GIRDERS USING SHAPE MEMORY ALLOYS Osman E. Ozbulut, Ph.D. Assistant Professor University of Virginia Reginald F. Hamilton, Ph.D. Assistant Professor Pennsylvania State University August 15, 2015

3 DISCLAIMER The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the U.S. Department of Transportation s University Transportation Centers Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof. 1

4 TABLE OF CONTENTS DISCLAIMER... 1 TABLE OF CONTENTS... 2 LIST OF FIGURES... 3 LIST OF TABLES... 5 ABSTRACT... 6 INTRODUCTION... 7 RESEARCH APPROACH... 8 SHAPE MEMORY ALLOY MATERIALS CHARACTERIZATION GROUT CHARACTERIZATION PULL-OUT TESTS DISCUSSION OF RESULTS CONCLUSIONS AND RECOMMENDATIONS REFERENCES

5 LIST OF FIGURES Figure 1. Self-post-tensioning process Figure 2. Phase transformation temperatures of SMAs Figure 3. Strain-temperature response of as-processed Ni 47.7Ti 43.5Nb 8.8 (at%) material at constant stresses of 150 and 300 MPa. A Ni 49.4Ti 50.2 (at%) result loaded to 120 MPa is included for comparison Figure 4. SEM images of the cast showing (a) an overview of the cellular arrangement, (b) the matrix encompassed by the eutectic in the region within the box in (a), and (c) is a 3D AFM image showing the varying surface topology Figure 5. SEM images showing (a) the deformation-processed microstructure with the Nb-rich particles oriented in the rolling direction, (b) a 3D AFM image of the smooth surface and (c) the cross-sectional view which is perpendicular to (a) Figure 6. High-magnification SEM images showing the Nb-rich particles in (a) cast and (b) deformation-processed microstructures Figure 7. Normalized heat flow vs temperature curves (i.e. thermo-grams) Figure 8. Strain-temperature responses for thermal cycling under constant stress for (a) the cast alloy and (b) the deformation-processed alloy. The single and double arrows depict the slopes during cooling and heating respectively. The symbols are defined within the text Figure 9. Characteristics parameters (a) forward T F and reverse T R transformation temperature intervals and (b) thermal hysteresis T H and recovery ratio for different levels of bias stress in Figure Figure 10. Scanning electron micrographs showing white Nb-rich particles in an as-cast microstructure of Ni 47.3Ti 44.1Nb 8.6(at%) material in (a) at 500 X magnification and (b) at 5000X magnification. (c) The microstructure of the as-processed material Ni 47.7Ti 43.5Nb 8.8 (at%), investigated in this work, shown in the rolling direction at 500X magnification and (d) at 5000X magnification. (e) The cross-section view of the rolled strip material at 10000X magnification. In the text, (d) and (e) are contrasted with the solution annealed microstructure in Figure Figure 11: SEM micrographs showing the white Nb-rich particles of the annealed material (900 C for 1hr) in the (a) rolling direction at 5000X and (b) cross-section at 10000X. These images are compared to the as-processed microstructure in Figure 10 (d) and (e) Figure 12: Region of interest covering the speckle pattern on the specimen surface. The ROI dimensions used in the DIC analysis are included. The dashed ovals at the top and bottom mark that the ROI extends into the specimen grip section. Consequently, those regions remain unchanged during deformation in Figures Figure 13. (a) The stress-strain response of as-processed material deformed at RT up to 8% average target strain and unloaded. (b) The strain-temperature response after unloading and during heating at null load. DIC strain contours are inset at selected points. The strain in the deformation fields is measured in the loading direction. In part (b), the scale has been adjusted to highlight the changes in strain contours during heating Figure 14: (a) The stress-strain response at 60 C, the material is deformed to 8% average strain and unloaded. (b) After unloading, the strain-temperature response during heating at null load. Inset images of strain contours are shown along selected points on both curves. The strain is in the loading direction. In part (b), the scale has been adjusted to highlight the changes in strain contours during heating Figure 15. The stress-strain response when annealed material is deformed at RT along with accompanying DIC strain fields at selected points. The strain in the deformation fields is calculated in the loading direction

6 Figure 16. SEM images showing (a) the microstructure and (b) the micro-constituents in a cast Ni 47.3Ti 44.1Nb 8.6 (at%) alloy and the microstructures of (c) the Ni 47.7Ti 43.5Nb 8.8 (at%) alloy sheet (d) theni 44.6Ti 42.8Nb 12.6 (at%) alloy rod materials Figure 17. The tensile stress-strain response of the NiTiNb alloys Figure 18. (a) The stress-strain curves for pre-straining of sheet and rod; (b) the subsequent strain-temperature ( ) response during shape memory recovery for sheet and rod materials Figure 19. The tensile stress-strain curves for the sheet pulled to increasing pre-strain levels at room temperature Figure 20. The stress-temperature response during constrained heating and cooling of specimens that have been pre-strained to (a) 5.4%, (b) 10.1% and (c) 12.2%. The double arrows indicate heating, and the single arrows cooling Figure 21. (a) Mixing grout with high shear Jiffler mixer. (b) Thermocouple attached to steel tendon. (c) Thermocouples on steel tendons and the data logger (d) Completed sample during temperature collection period Figure 22. Temperature measured in different commercially available grouts during curing Figure 23. (a) Pull-out test specimen and (b) test-set-up Figure 24. Bond stress-slip relationship for SMA bars at free end and loaded end for (a) Specimen 1 and (b) Specimen Figure 25. The variation of recovery ratio and recovery stress with pre-strain level for free and constrained recovery experiments * * Figure 26. The variation of T R and T R with pre-strain level for free and constrained recovery experiments Figure 27. Strain and stress recovered per unit temperature increase as a function of pre-strain level

7 LIST OF TABLES Table 1. Mechanical properties for the NiTiNb alloys. Note that the critical and yield stresses are based on a 0.2% offset Table 2. Summary of grout temperature test results Table 3. Characteristic metrics for pre-strain deformation of NiTiNb alloy sheet material

8 ABSTRACT With the increasing demands for high structural performance, the use of smart material have been considered in different engineering disciplines due to the appealing characteristics of these materials such as efficiency, self-actuation, adaptability, selfmonitoring and self-healing, and decision making. Recently, shape memory alloys (SMAs) have received considerable attention as a class of smart materials that can be employed in bridge engineering applications. SMAs have the ability to regain their original shape after being deformed up to 6-8% strain. This shape recovery is a result of an underlying reversible solid-solid phase transformation, which can be induced by either a stress or a temperature change. Several researchers have investigated the use of heatinduced SMAs to prestress concrete. However, the previous studies have been focused on thermally activating NiTi SMA tendons by electrical heating and have been mostly at the theoretical and laboratory study levels. This study investigates the feasibility of self-post-tensioned concrete elements by activating the SME of NiTiNb, a class of wide-hysteresis SMAs, using the heat of hydration of grout. First, the microstructure characterization of the NiTiNb widehysteresis shape memory alloys is discussed. Then, the tensile stress-induced martensitic transformations in NiTiNb SMA tendons are studied. Next, the temperature increase due to the heat of hydration of four commercially available grouts is investigated. Pull-out tests are also conducted to investigate the bond between the grout and SMA bar. Results show that the increase in temperature due to hydration heat can provide significant strain recovery during a free recovery experiment, while the same temperature increase only partially activates the SMAs during a constrained recovery. 6

9 INTRODUCTION Post-tensioned (PT) structural elements are used quite often in structures such as office building, bridges, parking structures, and water tanks. In many cases, PT systems greatly increase structural capacities and provide an aesthetically pleasing structure. Although PT systems provide many advantages for designers and constructors, these systems have raised concerns regarding corrosion of the PT tendons. The degree of corrosion of PT tendons is critical to the structural performance of PT systems (Yuyama et al., 2007) and the cost to replace tendons can exceed several hundred thousand dollars per tendon (Evans and Bollmann, 2000). Over the past decade, there has been an increasing interest in the use of shape memory alloys (SMAs) for various civil engineering applications (Ozbulut et al., 2011a). SMAs are a class of metallic alloys that can remember their original shape upon being deformed. This shape recovery ability is due to reversible phase transformations between different solid phases of the material. The phase transformation can be mechanically induced (superelastic effect) or thermally induced (shape memory effect). Besides their ability to recover large strains with minimal residual deformations, SMAs possess excellent corrosion resistance, good energy dissipation capacity, and high fatigue properties (Dolce and Cardone, 2001; DesRoches et al., 2004). Superelastic SMAs can undergo large strains, in the order of 7 to 8%, and recover these deformations upon removal of stress. Due to their excellent re-centering and good energy absorbing capabilities in passive nature, the superelastic SMAs have been considered in a number of seismic applications (Saiidi and Wang, 2006; Alam et al., 2009; Speicher et al., 2011; Ozbulut and Hurlebaus 2011a-f; Bitaraf et al. 2010; Shook et al. 2008) SMAs that exhibit shape memory effect (SME) generate large residual deformations when the material is mechanically loaded over a certain stress level and unloaded. However, the SME SMAs recover those residual strains upon being heated. Several researchers have explored the use of SME SMAs as actuators for active vibration control (McGavin and Guerin, 2002; Aguilar et al., 2013). Since relatively large amount of material needs to be activated in a very short time to generate an active control force, the application of the SME SMAs for vibration control of civil structures has been mostly limited to theoretical studies. Several attempts have been made to use the SME SMAs for active confinement of reinforced concrete (RC) columns. Andrews et al. (2010) investigated the feasibility of using the SME SMA spirals to retrofit RC bridge columns through experimental and analytical studies. They found that high recovery stress associated with shape recovery of the SME SMAs can be utilized to apply an active confinement pressure on concrete columns to significantly improve the strength and ductility of the columns. In another experimental study, Choi et al. (2011) studied the bond behavior of concrete that is actively confined by the SME SMAs through 7

10 monotonic and cyclic tests. The potential use of thermally induced SMAs to prestress concrete has been another research topic. The use of post-tensioned SMA tendons in concrete elements can increase overall sustainability of structures by minimizing the susceptibility of post-tensioning tendons to corrosion and by enabling the adjustment of prestressing force during their service life. Maji and Negret (1998) were the first to utilize the SME in NiTi SMAs to induce prestressing in concrete beams. SMA strands were pretensioned into the strain-hardening regime and then embedded in small-scale concrete beams. Once the beams were cured, the SMA strands were activated by the applied heat. El-Tawil and Ortega-Rosales (2004) tested mortar beam specimens prestressed with SMA tendons. They considered two types of SMA tendons: 2.5 mm and 6.3 mm diameter wires. Test results showed that significant prestressing could be achieved once the SMA tendons were heat-triggered. Sawaguchi et al. (2006) investigated the mechanical properties of mini-size concrete prizm specimens prestressed by Fe-based SMAs. Li et al. (2007) examined the performance of concrete beams with embedded SMA bundles. Through an extensive experimental program, they studied the development of smart bridge girders that can increase their prestressing force to resist the excessive load as needed. In all of these studies, SMA tendons are triggered by an electrical source. RESEARCH APPROACH The key characteristic of SMAs is a solid-solid, reversible phase transformation between its two main microstructural phases, namely martensite and austenite. SMAs have four characteristic temperatures at which phase transformations occur: (i) the austenite start temperature As, where the material starts to transform from twinned martensite to austenite, (ii) austenite finish temperature Af, where the material is completely transformed to austenite, (iii) martensite start temperature Ms, where austenite begins to transform into twinned martensite, (iv) martensite finish temperature Mf, where the transformation to martensite is completed. If the temperature is below Mf, the SMA is in its twinned martensite phase. When a stress above a critical level is applied at a temperature below Mf, the twinned martensitic material converts into detwinned martensite phase and retains this phase upon the removal of the load. It can regain its initial shape when the SMA material is heated to a temperature above Af. Heating the material above Af results in the formation of the austenite phase and, in the ideal case, a complete shape recovery. By a subsequent cooling, the SMA transforms to initial twinned martensite phase without any residual deformation. Significant heat is generated during the hydration of cement products. Numerous factors such as the type and composition of cement, the proportion of the mix, and the ambient temperature affect the heat evolution during the hydration process. In concrete structures, internal temperatures of 70 C are not uncommon (Dwairi et al., 2010). Since 8

11 grout is generally composed of very high portion of cement, high temperature increases can be also observed during grouting applications. Therefore, hydration heat of grout can be used to trigger the SME of SMAs to obtain SPT concrete members. Figure 1 shows the process for development of the SPT concrete beams using SMAs. First, the SMA tendons, in the martensitic state, are pre-strained. Then, concrete is poured and the SMA tendons are installed in post-tensioning ducts after concrete hardening. The void between the duct and the SMA tendons is filled with grout. Due to the heat of hydration of grout, the temperature of the SMA tendons increases, which induces the transformation of the material to austenite when the temperature is over the As. A complete transformation to austenite phase occurs when the temperature reaches the Af. As the SMA tendons attempt to return back to their original shorter length, while being constrained at both ends, a tensile stress is produced in the tendons, causing pre-stress in the concrete beam. At T<As, Pre-stretch original SMA while in martensite phase STEP 1 Twinned Martensite Cast concrete, and install the tendon in post-tensioning duct STEP 2 Detwinned Martensite Fill the ducts with grout, and trigger the tendons using the heat of hydration STEP 3 Austenite Figure 1. Self-post-tensioning process. The conditions on the phase transformation temperatures and the required temperature window (service temperature) for self-stressing application are shown in Figure 2. First, the As should be larger than the highest possible ambient temperature as the pre-strained SMA tendons must stay in the martensite state at ambient temperature. This will prevent pre-stretched SMA tendons from recovering their deformations at the storage temperature or during the installation of tendons to the concrete member. Second, the Ms should be below the lowest possible ambient temperature. This will ensure that the heated SMA tendons maintain their recovery stress after cooling to the ambient temperature. If the temperature of the SMA tendons becomes lower than the Ms, the SMA 9

12 tendons will lose their recovery stress due to a phase transformation to martensite. This requirement for Ms coupled with the aforementioned requirement for As necessitate the use of the current NiTiNb class of wide-hysteresis (i.e. ) SMAs. The ternary alloying with Nb facilitates hysteresis C compared to 30 C in binary NiTi alloys (Zhang et. al., 1990; Zhao, 2000). Furthermore, the Af should be as close as possible to the As, which requires minimizing the differential TR=Af - As, to complete the phase transformation using the hydration heat. When the temperature rises over the As, the SMA tendons start to transform to austenite, and thus recovery stresses are induced. However, the maximum recovery stress will not be obtained until the microstructure is completely austenitic, at a temperature over the Af. Martensite fraction Should be over maximum ambient temperature 100% M f Martensite A s Austenite 0% Should be below minimum ambient temperature M s Service Temperature A f Temperature Should be close to the A s Figure 2. Phase transformation temperatures of SMAs. SHAPE MEMORY ALLOY MATERIALS CHARACTERIZATION Phase I: One-Way Shape Memory Effect During Temperature Cycling in Cast versus Deformation-Processed NiTiNb Alloys The transformation temperatures of NiTiNb alloys are well-suited for operating temperature ranges that are required for civil engineering applications, e.g., pre-stressing concrete beams (Choi et. al., 2010; Li et al., 2007). NiTi has been the most commercially successful SMA to date. Alloying NiTi with Nb has been shown to nearly double the thermal hysteresis (Siegert et al., 2002). A comparison of NiTiNb and NiTi thermal hysteresis (=ΔTH) is illustrated in Figure 3. The NiTiNb alloys are typically thermomechanically processed (e.g. rolling) after casting. 10

13 The shape memory effect (SME) response of deformation-processed NiTiNb shape memory alloys (SMAs) is benchmarked against the response of a cast alloy. The insoluble Nb element ternary addition is known to widen the hysteresis with respect to the binary NiTi alloy. Cast microstructures naturally consist of a cellular arrangement of characteristic eutectic microconstituents surrounding primary matrix regions. Deformation processing typically aligns the microconstituents such that the Figure 3. Strain-temperature response of asprocessed Ni47.7Ti43.5Nb8.8 (at%) material at constant stresses of 150 and 300 MPa. A Ni49.4Ti50.2 (at%) result loaded to 120 MPa is included for comparison. microstructure resembles discontinuous fiber reinforced composites. Processed alloys are typically characterized for heat-to-recover applications and thus deformed at constant temperature and subsequently heated for SME recovery, and the critical stress levels are expected to facilitate plastic deformation of the microconstituents. The current work employs thermal cycling under constant bias stresses below those critical levels. This comparative study of cast versus deformation processed NiTiNb alloys contrasts the strain-temperature responses in terms of forward ΔTF =Ms - Mf and reverse ΔTR =Af - As temperature intervals, the thermal hysteresis and the recovery ratio. The results underscore that the deformation-processed microstructure inherently promotes irreversibility and differential forward and reverse transformation pathways. The NiTiNb alloys are typically cast and subsequently thermomechanically deformation-processed into useful forms such as wires, rods, or sheets for practical application (Otsuka et. al. 1998). As cast microstructures generally consists of β-nb + eutectic NiTi (with dissolved Nb) in a cellular configuration surrounding primary NiTi (with dissolved Nb) matrix material, consistent with characteristic eutectic microconstituent phases (Zhao et. al. 1991; Zhao et. al. 1990; Zhao LC, 2000; Piao et. al. 1992; Sherif et. al. 2014; Xiao et. al. 2007; Prima et. al. 1996; Siegert et. al. 2002; Ying et. al. 2011; Wang et. al. 2012; Shi et. al. 2013; Schryvers et. al. 2013; Shi et. al. 2014; Tosetti et. al. 2014). Several investigations characterize the impact of deformation processing conditions on the microstructure (Wang et. al. 2014; Siegert et. al. 2002; Tosetti et. al. 2014; Kim et. al. 2014; Yan et. al. 2012; Yin et. al. 2012; Tong et. al. 2014). Ultimately, the findings show that the deformation-processed microstructure is made up of oriented β-nb particles that are dispersed throughout the matrix similar to aligned discontinuous fiber reinforced composites. 11

14 The functionality of deformation-processed NiTiNb materials is primarily investigated via isothermal pre-straining deformation of an initially martensitic or austenitic microstructure and subsequent assessment of SME recovery via heating ( Duerig et. al. 1989; Zhang et. al. 1990; Zhao LC, 2000; Piao et. a. 1993; Sherif et. al. 2014; Ying et. al. 2011; Wang et. al. 2012; Liu et. al. 2014). Either transformation strain recovery during heating are typically characterized without load or under displacement constraint to assess the recovery stresses. The β-nb particles are presumably soft and the critical stresses during isothermal pre-straining deformation are expected to plastically deform them, presuming that their flow stress matches that for pure Nb which is estimated between MPa (Duerig et. al. 1989; Melton et. al. 1988; Otsuka et. al. 1998). Plastic deformation of microconstituents as martensite deforms necessitates an increased thermal driving force for the reverse transformation that results in elevated As and Af temperatures and the wide hysteresis (Duerig et. al. 1989; Melton et. al. 1988; Zhang et. al. 1990; Piao et. al. 1993; Kusagawa et. al. 2001), commonly referred to as a stabilization effect (Zhang et. al. 1990; Zhao et. al. 2006; Hao et. al. 2011; Hao et. al. 2013; Kusagawa et. al. 2001). The current work is an original investigation of the one-way strain-temperature (ε T) response. Constant bias load levels are applied during thermal cycling and the levels are incrementally increased up to those reported to facilitate plastic deformation of particles. Moreover, this is the first comparative study of cast versus deformation processed NiTiNb alloy. Only the matrix undergoes the martensitic transformation (MT) and hence exhibits SME. In the cast alloy, large matrix regions exist without the obvious appearance of microconstituent phases within the regions. However, within the composite-like deformation processed microstructure, β-nb particles are dispersed throughout the matrix and presumably can interact differently with the MT morphology compared to the cast microstructure. The aim of this comparative study is to gain insights into the impact of β-nb particles using the cast material ε T response as a benchmark for the deformation processed material response, which is the prototypical NiTiNb SMA microstructure. The compositions of both alloys are nearly equal to Ni47Ti44Nb9 at.%, which is the recommended ternary composition for wide hysteresis applications (Otsuka et. al. 1998). Atlantic Metals and Alloys LLC supplied a cast alloy with the composition Ni47.3Ti44.1Nb8.6 at.%. Medical Metals LLC supplied a deformation-processed sheet with the composition Ni47.7Ti43.5Nb8.8 at.%. The thermo-mechanical processing methods for the strip are multiple thickness reductions using cold work via rolling and annealing near the recrystallization temperature (850 C). Tensile specimens with dog-bone geometry were electrical discharge machined (EDM) from the cast materials. The gage dimensions were length (l) = 10 mm, width (w) = 3 mm, and thickness (t) = 1 mm. The thickness of 12

15 the deformation-processed sheet material was t = 0.25 mm and EDM was utilized to micromachine dog-bone specimens with l = 10 mm and w = 3 mm. Specimens were mechanically polished for scanning electron microscopy (SEM) and atomic force microscopy (AFM) analysis. The materials were polished via SiC paper with the grit size decreasing from 180 to 1200 and finally polished using 0.02 μm colloidal silica. Microstructural images were taken at room temperature using a Philips XL30 ESEM scanning electron microscope. For higher magnification imaging, a FEI NanoSEM 630 scanning electron microscope was employed. SEM imaging was performed in back-scattered electron mode. A Veeco Metrology Autoprobe M5 AFM atomic force microscope was used in contact mode and in air. The contact force was maintained around 10~ 20 nn with an imaging frequency of 1 Hz and a minimum detectable surface feature height of 10-4 Å. Load-biased thermal cycling experiments were conducted using a MTS 810 servo-hydraulic load frame equipped with a customized temperature cycling set-up. Temperature gradients within the specimen were minimized and the heating and cooling rates were within 5 to 10 C/min. The specimens were first heated to 150 C, to ensure that the specimens were in the austenitic state. The desired bias load was then applied and held constant. The specimens were cooled to -90 C and then heated to 150 C. The external load for successive thermal cycles was increased incrementally between 10 and 300 MPa. The strain was calculated based on the displacement of the actuator. Figures 4-6 show the SEM images of the cast and deformation-processed microstructures. The microstructure in Figures 4(a) and 4(b) exhibits the hypoeutectic character; the characteristic eutectic microconstituent is arranged in a cellular configuration as boundaries encompassing regions of NiTiNb matrix. The AFM image in Figure 4(c) reveals topography of the matrix and cellular eutectic microconstituent. Locally between the matrix and eutectic well-defined boundaries exist and the cellular region are raised. The centers of NiTiNb matrix regions are the lowest height. Moving outward towards the eutectic, the height rises approaching the eutectic-matrix boundary. The height within the eutectic is relatively uniform. Figure 5(a) illustrates that the Nbrich particles are dispersed throughout the matrix and oriented in the processing directions, thus they appear as striations. The AFM images of the deformation-processed material are shown in Figure 5(b). The images reveal markedly refined topography that is relatively smooth compared to 4(c). Figure 5(c) shows the cross-section in which particles appear as speckles with spacing on the order of 100 nm. Figure 6(a) exposes the characteristic eutectic microconstituent with the lamellar and globular mixture of Nb-rich β-particles and α-nitinb that is typical of dissolved Nb (Zhao et. al. 1991; Zhao et. al. 1990; Zhao LC, 2000; Piao et. al. 1992; Sherif et. al. 2014; Xiao et. al. 2007; Prima et. al. 1996; Siegert et. al. 2002; Ying et. al. 2011; Wang et. al. 2012; Shi et. al. 2013; Schryvers et. al. 2013; Shi et. al. 2014; Tosetti et. al. 2014). The particles in the deformation- 13

16 processed material are shown in Figure 6(b). They are aligned and discontinuous yet the sizes of the particles remain consistent with those in Figure 6(a). Figure 4. SEM images of the cast showing (a) an overview of the cellular arrangement, (b) the matrix encompassed by the eutectic in the region within the box in (a), and (c) is a 3D AFM image showing the varying surface topology.. Figure 5. SEM images showing (a) the deformation-processed microstructure with the Nb-rich particles oriented in the rolling direction, (b) a 3D AFM image of the smooth surface and (c) the cross-sectional view which is perpendicular to (a). Figure 6. High-magnification SEM images showing the Nb-rich particles in (a) cast and (b) deformation-processed microstructures. The thermal induced martensitic transformation (TIMT) during thermal cycling without load was evident for the cast material in Figure 7. The TIMT brings about exothermic and endothermic events during cooling and heating, respectively, and thus peaks arise in the heat flow versus temperature thermo-grams measured using differential 14

17 scanning calorimetry (DSC) analysis. TIMT temperatures were Ms = C, Mf = C, As = C and Af = 11.4 C. For the deformation-processed material, however, evidence for the TIMT is not apparent in the DSC analysis. Figure 7. Normalized heat flow vs temperature curves (i.e. thermo-grams) The strain-temperature (ε-t) responses in Figure 8 show the one-way shape memory effect behavior for cast and deformation-processed alloys at increasing constant bias stress levels. A ε-t response for the 100 MPa bias load level is evident for the cast material in Figure 8(a). A bias load of 150MPa was needed for the processed material in Figure 8(b). Those bias stress levels were the minimum levels that brought about measurable transformation strain. The Ms temperatures for those bias stress levels for both materials are equivalent and approximately equal to - 67 C. For 100 and 150 MPa bias stress levels applied to the cast materials, the slopes for the heating and cooling segments of the ε-t curves are nearly equal. At 300 MPa, the slope for the heating segment differs from the cooling segment in Figure 8(a). For the deformation-processed alloy loaded at 150 MPa in Figure 8(b), the slopes for both curves are equivalent. The slopes of the 300 MPa cooling and heating ε-t curves, however, exhibit differential slopes. Moreover, each curve exhibits two slopes. An initial slope appears vertical and the stage is seemingly isothermal. A second different slope follows in the cooling ε-t curve. The heating curve exhibits multiple slopes albeit an isothermal stage is indiscernible. 15

18 Figure 8. Strain-temperature responses for thermal cycling under constant stress for (a) the cast alloy and (b) the deformation-processed alloy. The single and double arrows depict the slopes during cooling and heating respectively. The symbols are defined within the text. Metrics that characterize the ε-t response are plotted with increasing bias load in Figure 9. Figure 9(a) captures the effect of bias stress on the forward transformation temperature interval ΔTF =Ms - Mf and the reverse interval ΔTR =Af - As. For each material condition, the ΔTF is less than ΔTR. The deformation-processed material exhibits the narrowest ΔTF. The ΔTF for the cast material is over 30 C higher. The reverse transformation finish temperature Af exhibits a marked increase (greater than 80 C) when the bias load is increased from 150 to 300 MPa (see Figure 8). Consequently, for both materials, the ΔTR increases (by nearly 60 C) from the lowest to highest bias load. The dependencies of thermal hysteresis and recovery ratio on bias stress level are illustrated in Figure 9(b). The thermal hysteresis ΔTH is determined as the temperature differential at half the recovered strain during heating (see Figure 8). The hysteresis widens most when the stress is increased from 150 to 300 MPa. The recovery ratio equals [(εtr εirr)/εtr x 100], where εtr is the tensile strain accrued in the cooling ε-t curve and εirr is the unrecovered strain after heating (see Figure 8). The 150 MPa bias stress level facilitates a maximum recovery ratio for both materials and the ratio drops for the 300 MPa level. 16

19 Figure 9. Characteristics parameters (a) forward TF and reverse TR transformation temperature intervals and (b) thermal hysteresis TH and recovery ratio for different levels of bias stress in Figure 8. The current findings demonstrate that despite vastly different microconstituent morphologies, plastic deformation of β-nb particles can have similar impacts on the strain-temperature characteristic metrics for cast and deformation-processed materials. In the NiTiNb class of NiTi-based SMAs, the NiTi composition is expected to dictate the transformation temperatures (Duerig et. al. 1989; Melton et. al. 1988; Zhang et. al. 1991; Zhang et. al. 1990; Otsuka et. al. 1998). Indeed the current results for similar Ni47.3Ti44.1Nb8.6 at.% (cast) and Ni47.7Ti43.5Nb8.8 at.% (deformation-processed) compositions exhibit equivalent Ms temperatures, albeit only the cast alloy exhibits the MT during stress-free thermal cycling. A bias stress during thermal cycling was required to bring about measurable shape memory behavior for the deformation-processed material. The yield stress of the Nb-rich particles has been estimated around 200 MPa [23] and thus impacts on the metrics become apparent when the bias load is increased from 150 to 300 MPa. There is a drastic increase in Af for both alloys. Consequently, the reverse transformation temperature intervals ΔTR increase starkly (by comparison the forward transformation intervals ΔTF are relatively consistent). Though the cast material exhibits the largest hysteresis levels, the most marked increase in hysteresis with bias load occurs for the deformation-processed material. For both materials, a maximum is apparent in the recovery ratios and the ratios drop with that increase in hysteresis. In as-cast materials, microstructure analysis shows that the matrix regions are apparently free of particles whereas the particles are dispersed throughout the matrix for the deformation-processed materials. The matrix regions in the cellular cast microstructure can readily transform unabated by non-transforming β-particles. For the deformation-processed composite-like microstructure, the β-particles exist as closely spaced reinforcements within the matrix. The composite arrangement can constrain the MT. Hence, an external bias load is required to facilitate the one-way shape memory response and a measurable strain-temperature (ε-t) response. Note that recent studies observe sparsely distributed Nb-rich nano-precipitates with characteristic dimensions 17

20 on the order of 10 nanometers within the matrix of cast materials (Shi et. al. 2013; Lin et. al. 1991). Those particles do not exhibit orientation as shown for the deformationprocessed alloys. The slopes of the heating and cooling ε-t curves for cast and deformation-processed materials underscore that impact of the differential microconstituent morphologies on the energetics of the martensitic phase transformation. The ε-t cooling and heating curves exhibit differential slopes with increasing tensile bias load. The heating and cooling ε-t curves of the cast material exhibit a single slope, and thus a single stage at each stress level. When the heating and cooling ε-t segments exhibit similar slopes at the lower bias stresses, the responses are in accordance with crystallographic reversibility (Liu et. al. 1997; Liu et. al. 2000). At the highest stresses, the heating and cooling curves for the cast or deformation-processed alloy no longer exhibit equivalent slopes. The slope of the cooling curve reflects continuous undercooling, which overcomes elastic energy that otherwise resists the forward MT (Liu et. al. 1997). The cooling curves exhibit the steepest slopes for the deformation-processed material and hence elastic energy storage is not predominant. During the initial stage in the cooling curves for the process alloy, elastic energy storage is compromised which apparently facilitates the nearly isothermal growth of martensite (Liu et. al. 1997). The thermal hysteresis is directly related to energy that is irreversibly dissipated during the MT (Liu et. al. 1997; Liu et. al. 2000; Hamilton et. al. 2004). Stored elastic energy can be irreversibly dissipated during the forward MT due to plastic deformation and thus the hysteresis widens, as the stored elastic energy is not available to assist the reverse transformation (Piao et. al. 1993; Liu et. al. 1997). The drastic increase in Af, for both cast and deformation-process alloys reflects a martensite stabilization effect (Zhang et. al. 1990; Zhao et. al. 2006; Hao et. al. 2011; Hao et. al. 2013; Kusagawa et. al. 2001). Martensite can be stabilized when it is pinned or heavily dislocated such that the reverse transformation requires a higher driving force (Hao et. al. 2013; Kusagawa et. al. 2014; Salzbrenner et. al. 1979). It has been postulated that the β-phase can lock the martensitic phase in NiTiNb alloys (Lanba et. al. 2004). For the current results, the reverse transformation temperatures must increase greatly beyond the Af temperature for stress-free thermal cycling, as well as the temperatures at the lowest bias stress levels. The absences of a corresponding isothermal stage in the heating ε-t curves reveal differential transformation paths for the forward versus reverse MTs for the deformationprocessed alloys. Differential transformation paths can suggest non-thermoelastic MTs occur at the higher stress levels (Liu et. al. 2000; Lanba et. al. 2004; Liu et. al. 1998). The stark differential between forward and reverse transformation temperatures intervals can further point to a non-thermoelastic MT for both deformation-processed NiTiNb alloys as well as cast alloys (Liu et. al. 1997; Liu et. al. 2000; Shi et. al. 2012; Liu et. al. 18

21 1998). The widening thermal hysteresis and diminished recovery ratio at the highest bias stress imply marked irreversibility that is common for non-thermoelastic MTs. This comparative study of load biased thermal cycling of cast and deformationprocessed NiTiNb alloy microstructures supports the following conclusions Close inter-particle spacing and narrow spacing between the aligned reinforcements brings about a microstructure constraint that can suppress the thermal-induced MT. A minimum biasing stress overcomes the constraint. In the processed microstructure, the elastic energy storage is relaxed. At the highest levels of constant stress, the strain-temperature response of the deformation-processed alloys reflects that the forward and reverse MTs in processed alloys takes place in two stages. A single stage is observed in the MTs in the cast alloys. The slopes of the strain-temperature curves are steepest and the initial stage is seemingly isothermal for the processed alloy. The deformation-processed as well as the cast microstructures facilitate a stabilization effect that impacts the reverse transformation by increasing Af and diminishes the recovery ratio. The slopes of the heating and cooling segments are not equivalent. Plastic deformation associated with Nb-rich particles can readily occur at the 300 MPa bias stress level, considering that it exceeds the reported flow stress of Nb. Hence, the transformation may become non-thermoelastic. Phase II: Characterization of Martensitic Transformation Morphology for Prestrain Deformation and One-way Shape Memory Effect Recovery Authors have rationalized the wider hysteresis in the as- processed materials by assessing the impact of the deforming microstructure on the reverse transformation temperatures. Zhang et. al. (1990) proposed that plastic deformation of Nb-rich particles during the stress-induced martensitic transformation relaxes strain energy of martensite interfaces. This is accompanied by plastic deformation of the parent phase. Kusagawa et. al. (2001) found that after heating specimens that had been deformed in the austenitic state between Ms and As, residual strain was always present. Recently, Schryvers et. al. (2013)found evidence of plastic deformation in Nb-rich nano-precipitates after it had interacted with a growing martensite plate. A few works report on the fundamental thermo-mechanical deformation in order to understand microstructure-property relationships (Zhang et. al. 1990; kasagawa et. al. 2001; Schryvers et. al. 2013; Piao et. al. 1992; Wang et. al. 2012; Zheng et. al. 1997). The current work utilizes in-situ digital image correlation analysis in order to ascertain the surface morphology during the stressinduced MT and expound on the impacts of microstructure changes. 19

22 Fundamental thermo-mechanical characterization of SMAs reports on the shape memory effect, i.e. deformation of thermal induced martensite and subsequent recovery during heating, which is illustrated in Figures 4 and 8. The characterization may also report on the superelastic/pseudoelastic effect as a result of the stress-induced reversible A M transformation at a constant temperature. The digital image correlation (DIC) analysis has primarily been used to characterize the superelastic effect in-situ, i.e. throughout deformation (Hamilton et. al. 2008; Hamilton et. al. 2011; Daly et. al. 2007). The authors have observed general trends as DIC strain contours evolve during the pseudoelastic shape memory response. A summary of the trends with respect to the macro-scale measurements for the stress-strain response follows. The elastic response facilitates homogeneous strain fields and localization is not observed in the full-field DIC strain maps. Upon further loading, a critical value of stress is reached, the stress-strain response deviates from linearity, and a stress plateau develops. That critical stress brings about the stress-induced A M transformation and transformation strain is accrued over a constant stress plateau. DIC measurements illustrate the transformation morphology by delineating a boundary between high strains that appear in local regions corresponding to martensite from strikingly lower strain levels corresponding to not yet transformed austenite. Multiple and differently oriented boundaries may exist. Single bands of transforming material may be evident between boundaries or multiple bands can interact and overlap. After the plateau, the stress-strain response may transition to a second linear-elastic type response, which corresponds to elastic deformation of a fully martensitic microstructure. The high, local strain regions are no longer present and the appearance of the contours is similar to the initial linear-elastic response of austenite, albeit the strain levels are much higher. An ideal pseudoelastic stress-strain unloading curve will follow the elastic loading curve. However, the reverse transformation A M, takes place over a lower constant stress plateau and thus a hysteresis exists. Finally, the initial deformed austenite recovers elastically. It is well understood that the reverse A M MT morphology is the reverse of that for the forward A M. Hence, the DIC strain fields during unloading disappear in the reverse sequence that they appear during loading. Based on the aforementioned fundamental understandings, the reversible pseudoelastic shape memory alloy response can be related to the martensitic transformation morphology using DIC surface strain measurements. Utilizing in-situ DIC, the current work elucidates the martensitic transformation morphology of a Ni47.7Ti43.5Nb8.8 (at%) SMA. The bulk material was supplied after thermo-mechanical deformation into a rolled strip. Compared to the as-cast material, the microstructure becomes highly oriented after rolling. The evolution of full-field strains is assessed throughout the stress-strain response with the temperature held constant. Furthermore, after unloading the strains are assessed during heating under null load. It is well known that the transformation stress is temperature dependent. The dependence is modeled 20

23 using a Clausius-Clapeyron relationship and the stress increases with increasing constant temperature. Thus we contrast DIC measurements for different test temperatures. The influence of microstructure is studied by characterizing the thermo-mechanical response for material in the as-processed (or rolled strip) state and after solution annealing. A NiTiNb rolled strip with composition Ni47.7Ti43.5Nb8.8 (at%) was obtained from Medical Metals LLC. The strip was 6 mm wide and 0.25 mm thick. Dog-bone specimens with a gage length of 10 mm and gage width of 3 mm were micro-machined from this strip. Some specimens were solution annealed at 900 C for 1 hr in an Ar atmosphere within a Lindberg/blue-box furnace (BF51732 series) and then quenched in room temperature water. Scanning electron microscopy (SEM) was performed at room temperature in a Philips XL30 ESEM in back-scattered mode. Alloying NiTi with Nb leads to the formation of Nb-rich precipitates embedded in the NiTi matrix. The lower magnification image of an as-cast Ni47.3Ti44.1Nb8.6 (at%) material is shown in Figure 10(a). The constituents are a eutectic-like microstructure (light inclusions) and matrix (darker background). Zooming into an area reveals that the eutectic structure consists of lighter Nb-rich particles, which are shown in Figure 10(b). Processing leads to deformation of these particles. For the Ni47.7Ti43.5Nb8.8(at%) rolled strip investigated in this work, the microstructure is shown in Figure 10(c). The lighter striations along the rolling direction (direction the white arrow) result from the particles elongated in the rolling direction. The micrograph in Figure 10(e) shows a cross-section view and these particles speckle the matrix. Only the matrix material participates in the MT. SEM micrographs of the as-processed material in Figure 10(d) and (e) can be compared to the annealed material in Figure 11. Figure 11(a) shows the rolling direction and the crosssection is in Figure 11(b). After annealing, the long striations of Nb-rich particles break up into particles that are aligned in the rolling direction. The cross-sectional images reveal that the Nb-rich particles are larger after solution annealing. 21

24 Figure 10. Scanning electron micrographs showing white Nb-rich particles in an ascast microstructure of Ni47.3Ti44.1Nb8.6(at%) material in (a) at 500 X magnification and (b) at 5000X magnification. (c) The microstructure of the as-processed material Ni47.7Ti43.5Nb8.8 (at%), investigated in this work, shown in the rolling direction at 500X magnification and (d) at 5000X magnification. (e) The cross-section view of the rolled strip material at 10000X magnification. In the text, (d) and (e) are contrasted with the solution annealed microstructure in Figure 11. The thermomechanical experiments were conducted on an MTS 810 servo hydraulic load frame. The heating was accomplished via induction heating of the grips and the specimen. The average Figure 11: SEM micrographs showing the white Nb-rich strain is based on actuator particles of the annealed material (900 C for 1hr) in the displacement. The (a) rolling direction at 5000X and (b) cross-section at temperature was measured 10000X. These images are compared to the as-processed via a thermocouple attached microstructure in Figure 10 (d) and (e). to the specimen. For the mechanical loading/unloading at a constant temperature, the specimen was first heated under zero load to the test temperature. It was then loaded in displacement control mm/s, which corresponds to an average strain rate of 1.7 X 10-4 /s. After loading, the specimen was unloaded in force control at an equivalent stress rate of 1.2 MPa/s. After unloading, the specimen was heated in force control at a rate of C/min with the load set to zero in order to try and recover residual deformation. 22

25 In-situ full-field deformation fields were obtained along the gage length using digital image correlation (DIC). DIC is an optical technique that measures displacement fields by tracking a high intensity pattern on the specimen surface. This pattern is applied by painting a thin, uniform white coating of Golden Airbrush Titanium White (#8380) on the surface, followed by a speckle pattern of Golden Airbrush Carbon Black (#8040). The pattern was applied using an IWATA Micron- CMB airbrush in order to paint micron size speckles. In-situ images of the specimen surface were captured using a Grasshopper GRAS-20S4M/C CCD camera (1600X1200 pixels). The image capture was synched with the data capture and was collected using Vic- Snap software from Correlated Solutions. The images and data were captured at a frequency of 1 Hz. The images that were captured have a spatial resolution of 18 μm/pixel. The correlation is performed in an region of interest (ROI) that is selected in a reference image. The definition of the ROI is shown in Figure 12. This region is divided into subsets and the pixel intensity of each subset is calculated and tracked in deformed images by optimizing the displacements and displacements gradients (strains). Each subset is a square region that is approximately 450 μm 2 in size. The DIC was performed via the software Vic-2D from Correlated Solutions, Figure 12: Region of interest covering the speckle pattern on the specimen surface. The ROI dimensions used in the DIC analysis are included. The dashed ovals at the top and bottom mark that the ROI extends into the specimen grip section. Consequently, those regions remain unchanged during deformation in Figures Inc. The general theory of DIC is explained in the works of Sutton et al. (1983, 1986) and Chu et al. (1985). Figure 13(a) shows the stress-strain response for the as-processed Ni47.7Ti43.5Nb8.8 (at%) specimen. The specimen is deformed to 8% average strain at room temperature (RT=21 C). Inset images show strain contours (calculated based on full-field DIC measurements) at selected points. The contours are the normal component of strain, which coincides with the loading direction. Initially, the specimen exhibits a linear elastic response. At an average strain of 1.4%, the stress drops and the response transitions onto a stress plateau. At an average strain of 5.8%, the response becomes linear, with the stress increasing until the average strain reaches the 8% target value. 23

26 The inset DIC strain contours show that during the stress drop, a localized band of high strain appears at the top of the image. As the specimen is further deformed, the lower boundary of the band travels towards the bottom of the image and the localized region of strain grows across the gage length. The maximum localized (or micro-scale) strains within the band remain unchanged at around 9.5% as it grows across the gage length. Note that serrations appear in the plateau, which are delineated within a dashed box. At the onset of the serrations, the distinct local small bands appear which are circled: one with the same orientation as the original band and another with a different orientation. The original band and two auxiliary bands grow together until the average strain reaches 5.8%. Within the bands the DIC strain measurements approach 10%. The band stops growing throughout the image at the end of the plateau and the onset of the increase in load. Throughout the linear hardening-like response, the length of the corresponding DIC strain contour remains uniform. The results stress-strain curve exhibits a strain hardening like increase while the load increases; however, the slope is nearly linear as opposed to the non-linear strain hardening response observed in conventional ductile metals. Though the length of the DIC strain contour remains uniform, the strain reaches around 11% despite the average strain only reaching the 8% target value. The DIC strain fields during unloading indicate that the length of the high strain contours does not shrink. Instead, localized strain values within the region decrease to around 7%. Once the specimen is completely unloaded, it is heated under null load. Figure 13(b) shows the strain-temperature response of the specimen during heating. Upon heating, average strain recovery ensues immediately at 21 C and recovers 2.6% until the temperature reaches 51 C and average recovery strain saturates. Inset DIC images illustrate that strain recovery is highly localized. During this recovery, the length of the region of high strain does not decrease in size. Locally, however strains decrease from 7% to nearly 2.5%. The recovery of strains reflects that stress-induced martensite is stabilized during loading. Heat input was necessary to facilitate the reversion of that martensite via the shape memory effect. As permanent residual strain remains after heating, we presume that it is due to irreversible plastic deformation. 24

27 Figure 13. (a) The stress-strain response of as-processed material deformed at RT up to 8% average target strain and unloaded. (b) The strain-temperature response after unloading and during heating at null load. DIC strain contours are inset at selected points. The strain in the deformation fields is measured in the loading direction. In part (b), the scale has been adjusted to highlight the changes in strain contours during heating. The stress-strain response for a specimen deformed to 8% average strain at a constant temperature of 60 C is shown in Figure 14. Initially, the stress-strain response exhibits a linear elastic type of response. Upon reaching an average strain of 1.7%, there is a transition from linearity to a non-linear response, similar to a strain-hardening type response. At 2.7% average strain, there is a stress-drop. Relative the room temperature response in Figure 13(a), the slope of the stress-strain curve rises linearly. The accompanying DIC strain contour measurement that corresponds to the transition from the initial linear elastic response (i.e., at the average strain of 1.7%) exhibits a uniform blue color that implies a uniform increase in strain. When the stress drops, a local region (outlined with a dashed white circle) of high strain appears at the center of the full-field contour. In the following image, the local strain within the dashed boundary increases further. As the specimen is further loaded, the upper and lower boundaries of the central region, respectively, grow in the upward and downward vertical directions. The bold black arrow mark when the length of the deformed region becomes constant. In subsequent images the local strains increase in a diffuse manner. As the specimen is further loaded to 8% average strain, the length of the high strain region does not grow in size, however, the strains increase uniformly to around 12.5%. The DIC strain fields during unloading indicate that the length of the high strain contours does not shrink. Instead, localized strain values within the region decrease from about 13% to around a minimum of 8%. 25

28 After complete unloading, the specimen was heated under null load. The straintemperature response is shown in Figure 14(b). Instead of local strains decreasing immediately upon heating, the specimen must be heated by about 10 C in order to initiate strain recovery. The specimen only recovers 0.2 % average strain between the temperatures of 70 and 100 C. This recovered strain is over an order of magnitude less than the average strain of 2.6% that was recovered for the specimen deformed at RT. Furthermore, the evolution of the full field strains during heating is much more diffuse and miniscule compared to the recovery during heating of the specimen initially deformed at RT. The accompanying DIC deformation fields indicate that the strains within the deformed region decrease uniformly by a small amount from 8 to 7%. The stark contrasts in deformation and recovery suggest that plastic deformation is dominant throughout loading rather than the reversible martensitic deformation. Figure 14: (a) The stress-strain response at 60 C, the material is deformed to 8% average strain and unloaded. (b) After unloading, the strain-temperature response during heating at null load. Inset images of strain contours are shown along selected points on both curves. The strain is in the loading direction. In part (b), the scale has been adjusted to highlight the changes in strain contours during heating. The RT response of a solution annealed specimen is shown in Figure 15. The accompanying DIC deformation fields along selected points during the deformation are also shown. During the initial elastic response, the strains in the DIC deformation fields increase homogeneously. The curve deviates from linearity around an average strain of 1% and strains continue to increase uniformly across a large region in the gage section till an average strain of around 1.6%. At an average strain of 1.6%, there is a stress drop and a localized band appears towards the bottom of the specimen. As the specimen is loaded, 26

29 the uppermost boundary of the band grows towards the top of the specimen. The strains outside the bands remain unchanged, while the strains within the band increase slightly to around 10%. This band grows as the deformation proceeds throughout a stress plateau. At an average strain of 3.2%, the specimen fractures. The fracture initiates at the uppermost boundary of the high strain region. The effects of test temperature on the full-field deformations of Ni47.7Ti43.5Nb8.8 (at%) rolled strip material have been characterized. Both the RT and 60 C stress-strain loading responses are accompanied by a stress-drop and a transition onto a stress plateau via the formation of a high strain band that grows. The behavior can be attributed to the initiation of the stress-induced forward MT. Throughout the stress plateau, at both temperatures, the local transformations measured using DIC analysis reach about 13%. On the other hand, the average strain determined from the crosshead displacement is set to reach a target value of 8%. Contrasts in the evolution of the full-field strain contours throughout the plateau suggest that the large local strain levels arise from different deformation Figure 15. The stress-strain response when annealed material is deformed at RT along with accompanying DIC strain fields at selected points. The strain in the deformation fields is calculated in the loading direction. mechanisms. At room temperature, multiple bands appear that do not exist during deformation at the elevated temperature. Upon heating, both specimens recover local strains uniformly across the deformed region. However, the specimen deformed at RT recovers a larger amount of strain via the shape memory effect. The much lower level of local recoverable strains at 60 C as well as the slight diffuse change of stress contours suggest that shape memory effect recovery is miniscule. Thus plastic deformation dominates the martensitic transformation at the elevated temperature. In the case of the solution annealed response, there is a region between the deviation from linearity and the stress-drop where the strains increase homogeneously across a large area of the gage section. The stress-strain response for the as-processed and solution annealed specimen at RT show that the stresses at which the response deviates from linearity are similar. The solution annealed specimen fractures during the stress plateau region at 3.2% strain. Also, the solution annealed specimen fractures at the top of the 27

30 high strain region that grows during the stress plateau. Recalling the microstructures of the as-processed and annealed specimens in Figures 10 and 11, the deformation results suggest that the breaking up of the elongated particles and the increase in size of the particles after annealing weakens the material. Phase III: Microstructure Characterization and Thermo-mechanical Experimentation for Feasibility of Self-Prestressing Concrete Members Using NiTiNb Shape Memory Alloys For materials characterization, the microstructure is reported along with the characteristic thermally-induced martensitic transformation temperatures and the mechanical properties. Then, the NiTiNb SMAs are pre-strained and subsequently heated in order to assess them for the self-post-tensioning application. The SME recovery during heating without constraint is described with respect to the influence of prestrain/pre-stress on the reverse transformation temperatures ( As and A * f * ). The recovery is constrained during heating in order to assess the influence of pre-strain/pre-stress on * stress generation as well as the reverse transformation temperatures ( As and A * f ). The phase transformation temperatures As and Af and their differential TR=Af - As depend on the microstructure of the NiTiNb alloy, i.e. the composition and the microconstituent morphology. The influence of Nb addition has been systematically investigated with respect to the microstructure and transformation temperatures (Piao et al., 1992; Siegert et al., 2002). A common ternary alloy composition for widening the thermal hysteresis (Ms - As) while providing useful shape memory effect recovery behavior is Ni47Ti44Nb9 (at%) (Otsuka and Wayman, 1998). Tailoring the microconstituent morphology via deformation processing is the fundamental means to control the phase transformation temperatures (Siegert et al., 2002). The alloys are typically cast and further cold- or hot-worked into final forms for practical application. In this work, the microstructure of a cast and deformation-processed (sheet) alloy with similar compositions are reported. Atlantic Metals and Alloys LLC supplied a cast alloy with the composition Ni47.3Ti44.1Nb8.6 at.%. Medical Metals LLC supplied a deformation-processed sheet that was 6 mm wide and 0.25 mm thick with the composition Ni47.7Ti43.5Nb8.8 at.%. The compositions of both alloys are nearly equal to Ni47Ti44Nb9 at.%, which is the recommended ternary composition for wide hysteresis applications above. The grain sizes for the cast and sheet material were determined as 300 μm and 300 nm respectively using the Intercept Procedure from ASTM E Texture is rarely reported (Yan et. al., 2012) and it is likewise beyond the scope of this work. The impact of differential thermo-mechanical processing was contrast by studying an extruded rod material (3.45 mm diameter) provided by. Memry Corporation with the composition Ni44.6Ti42.8Nb12.6 (at%). Specimens for mechanical testing and 28

31 microstructure analysis were wire electro-discharge machined from the rod material with an 8 mm gage length and 1.1 x 0.5 mm 2 cross-section and from the sheet material with a 10 mm gage length and 3 x 0.25 mm 2 cross-section. The microstructure is characterized along with the characteristic thermally-induced martensitic transformation temperatures and mechanical properties. The cast microstructure consists of a net-like arrangement of a characteristic eutectic microconstituent encompassing a NiTiNb matrix and a scanning electron microscopy (SEM) image is shown in Figure 16(a). The martensitic transformation occurs in the matrix regions. The as-cast microstructure is representative of the microstructure prior to thermo-mechanical processing. The SEM image in Figure 16(b) shows the typical eutectic micro-constituent, which is made up of β Nb-rich particles and α-nitinb matrix (Zhao 2000; Siegert et al., 2002; Wang et al., 2014). Deformation processing breaks up the net-like structure (Siegert et al., 2002; Wang et al., 2014; Yan et al., 2012). The micro-constituent morphology for the sheet material is shown in Figures 16(c). The image reveals a composite-like microstructure with β Nb-rich particles (appearing as the lighter streaks) that are elongated and discontinuous fiber-like reinforcements aligned in the primary processing direction within the NiTiNb matrix. It is well known that the unique stabilization of martensite, which is the cornerstone of NiTiNb shape memory behavior, is attributed to the microconstituent morphologies (Melton et. al., 1988; Duerig et. al., 1989; Cai et. al., 1994; Zhao, 2000; Kusagawa et. al., 2001; Seigert et. al., 2002; Yan et. al., 2012; Wang et. al., 2014). The microstructure images in Figure 16(c) and 16(d) underscore differential microconstituent morphologies resulting from different thermomechanical processing done by the different companies, which show the sheet and rod materials respectively. The images reveal a composite-like microstructure with β Nb-rich particles (appearing as the lighter streaks) that are elongated and discontinuous fiber-like reinforcements aligned in the primary processing direction within the NiTiNb matrix. The area fractions of the second particles were estimated using a digital image pixel thresholding technique (Sahoo et. al., 1988; Pal and Pal, 1993), which takes advantage of the dark and light contrasts of the matrix and particles respectively, and uses the average of several SEM images. The particle average area fraction for the rod microstructure was 18%. Consistent with the similar compositions, the fractions for the cast and sheet materials were 10%. The average interparticle spacings were determined from cross-sectional SEM images. For the sheet material, the inter-particle spacing was about 100 nm and it was 500 nm for the rod material. These findings convey a refined microconstituent morphology for the sheet compared to the rod. Moreover, the microstructure characterization illustrates that deformation-processing after castings affords the ability to tailor the microstructure, via orienting the microconstituents and inter-particle spacing. 29

32 Figure 16. SEM images showing (a) the microstructure and (b) the micro-constituents in a cast Ni47.3Ti44.1Nb8.6 (at%) alloy and the microstructures of (c) the Ni47.7Ti43.5Nb8.8 (at%) alloy sheet (d) theni44.6ti42.8nb12.6 (at%) alloy rod materials. Differential scanning calorimetry (DSC) analysis was carried out to determine transformation temperatures. DSC was carried out using a power compensated Perkin- Elmer DSC8500. The temperature scan rate was 40 C/min. The methodology was as follows: (i) heat from 50 C to 100 C, (ii) hold for 1 minute at 100 C, (iii) cool to -120 C, (iv) hold at -120 C for 1 minute, (v) reheat to 200 C, (vi) hold at 200 C for 1 minute, (vii) cool to 50 C. The transformation temperatures for the cast alloy were Ms = -64 C, Mf = -106 C, As = -81 C and Af = 11 C. For deformation-processed NiTiNb alloys endothermic and exothermic events did not arise in DSC measurements down to liquid nitrogen temperature. Thermal cycling under constant bias load is typically employed to determine characteristic transformation temperatures for processed NiTiNb materials (Melton et al., 1988; Kim et al., 2011). Constant bias loads were increased from 10 MPa up to a load that facilitated measurable transformation strain. Biasing with a constant stress of 150 MPa did reveal the thermally-induced transformation. The transformation temperatures for the sheet were Ms = -64 C, Mf = -75 C, As = -29 C and Af = -6 C, and those for the rod were Ms = -52 C, Mf = -71 C, As = -29 C and Af = -1 C. The strength properties were determined from uniaxial tension loading until failure and the stress-strain responses are shown in Figure 17. The uniaxial tension test for each material was determined at room temperature (~23 C) with the material in the austenite state and thus the martensitic transformation is stress-induced. The tests were conducted 30

33 in displacement control using an equivalent strain rate of 2.0 x10-4 /s. Strain was measured via a miniature extensometer within the gauge length and computed via digital image correlation (DIC), using the Inspect Extensometer (IE) tool from the DIC software Vic-2D. The IE gage length of the matches the gage lengths of the specimens. The material properties are summarized in Table 1. The moduli for the deformationprocessed materials are greater than that for the cast material. Deformation processing improves ductility and the fracture strain increases compared to the cast material. For the cast and rolled sheet materials with similar composition, deformation processing improves the mechanical strength. Stress-strain curves for processed composite microstructures evolve as follows: austenite linear-elastic response, stress drop/softening and plateau (indicative of the phase transformation), linear-elastic response of martensite, non-linear strain-hardening response, and fracture. Upper critical stress (referred to as Austenite Critical Stress in Table 1) levels for the rod and sheet materials reach about 500 and 640 MPa respectively and lower plateau stress levels are 460 and 590 MPa. The strain throughout the plateau response is slightly larger for the rod (7.5 %) compared to the sheet (7.0 %). The cast netlike microstructure facilitates an initial linear-elastic slope followed by a deviation from linearity, and then a critical stress that precedes the transformation. Moreover, a plateau is apparent for the processed materials with strain accruing throughout the plateau. Rather than a plateau, the transformation for the cast material exhibits a hardening-like response. The martensitic sheet exhibits higher strength properties and failure strain. Figure 17. The tensile stress-strain response of the NiTiNb alloys. Table 1. Mechanical properties for the NiTiNb alloys. Note that the critical and yield stresses are based on a 0.2% offset. 31

34 Composition (at%) Cast Ni47.3Ti44.1Nb8.6 Sheet Ni47.7Ti43.5Nb8.8 Rod Ni44.6Ti42.8Nb12.6 Modulus (GPa) Austenite Critical Stress (MPa) Martensite Yield Stress (MPa) Ultimate Tensile Strength (MPa) Failure Strain (%) The pre-strain experiments were conducted on an MTS 810 servo hydraulic load frame and at room temperature, which averaged about 23 C and was well above the Af temperatures so the starting microstructure was austenitic. Pre-strain for binary NiTi SMAs is typically carried out in the martensitic state and martensite reorientation takes place rather than the stress-induced austenite-to-martensite transformation (Duering and Melton, 1989; Cai et al., 1994; Zhao, 2000; Wang et al., 2014). In this work, the martensite was stress-induced, which is made possible by the Nb addition in the ternary NiTiNb SMAs. The specimens were loaded in displacement control at an average strain rate of about 2.0 x10-4 /s and they were unloaded upon reaching the desired pre-strain level. Residual strain remained after unloading. To assess the recovery ratio of residual strain, specimens were heated at zero load (referred to as free recovery). Thermomechanical experiments were conducted for the current work on an MTS 810 servo hydraulic load frame equipped with a custom thermal-cycling set-up. The specimens were heated via induction heating. The temperature was measured via a thermocouple affixed to the specimen. The induction coil design (Semiatin and Zinn, 1988) minimized thermal gradients in the specimen. Heating and cooling rates were controlled so that they were maintained around C/min. Specimens were allowed to cool in the ambient back to room temperature. Preliminary free-recovery experiments were conducted on both deformation-processed material in order to assess the material response of the differential composite-like microstructures. For the sheet and rod materials, free recovery is contrast for the pre-straining stress-strain response shown in Figure 18(a). After unloading, there is residual strain. * A The strain recovery begins at the s temperature in Figure 18(b). The strain saturates at a * Af temperature when the reverse transformation is complete. Since saturation is achieved, * * * the reverse transformation temperature interval T = A A fully activates the shape R f s memory effect. As shown in Figure 18(b), not all the strain is recovered during heating ε res 32

35 and there is permanent irrecoverable strain heating is shown as ε perm. The strain that is recovered during. The percentage of residual strain that is recovered via free SME full ε recovery is defined as the recovery ratio ( rec *100% ). The recovery ratio (58%) ε res for the sheet material is higher than that of the rod (49%). The recovery will be * T incomplete if the temperature is raised by a fraction of R ; therefore, the material will be partially activated. In order to achieve full activation, as well as maximize the * recovery ratio, the material microstructure must be designed such that T R matches the heat of hydration of the grout. (a) (b) Figure 18. (a) The stress-strain curves for pre-straining of sheet and rod; (b) the subsequent strain-temperature ( ) response during shape memory recovery for sheet and rod materials. The recovery behavior can be further characterized based on the heating straintemperature (ε-t) curves. A dotted line is drawn tangent to the curves in Figure 18(b) and demonstrates the extent of strain recovery within a select temperature range. The strains for the sheet and rod respectively recover with temperature at 0.12 %/ C and 0.09 %/ C. Contrasting the initial slopes, the sheet exhibits a higher recovery ratio over a smaller temperature interval, which may better match the possible heat of hydration. After the * initial slope the strain recovers gradually and the sheet alloy exhibits a T R = 49 C and 33

36 * T R = 57 C for the rod. However, it can be seen that when the temperature is increased to 50 C, which corresponds to a 26 C increase from the pre-straining temperature, 91% of for the sheet and 77% of for the rod is achieved. The lower temperature interval and higher recovery ratio of the sheet as compared to the rod confirms that extent of deformation experienced by sheet material resulted in a more promising microstructure for activation via heat of hydration of the grout. Thus, understanding the microstructure property relationships can enable tailoring the eutectic microconstituent orientation (giving rise to an apparent elongation in Figures 16(d) compared to 16(c)) and interparticle spacing in order to tune the activation strain and TR = Af As. The following * * * section focuses on stress generation during heating with a fixed displacement constraint for the sheet material, as the martensite exhibited higher strength and fracture strain and the recovery ratio is substantial over a lower temperature interval. The stress-strain responses for the pre-straining of the sheet material are shown in Figure 18. Pre-strain levels of more than 12% are commonly suggested (Otsuka and Wayman, 1998) and in this work pre-strain levels approach that value and exceed it. The pre-strain levels were 5.4, 8.8, 10.1, 12.2 and 16.1% in order to deform martensite to different extents of the stress-strain response. During loading, the curve exhibits an initial linear-elastic response up to a stress peak, followed by a stress drop. Then a stress plateau indicative of the phase transformation is observed. The evolution of the morphology depends on whether the pre-strain in Figure 18 is within the plateau region, at the completion of the plateau, within the linear elastic response of martensite after the plateau, or within the strain-hardening type behavior of martensite after the elastic response. For the 5.4% pre-strain level, the microstructure is a mixture of martensite and austenite. The martensite volume fraction increases throughout the plateau and the material should be completely martensitic at the 8.8% pre-strain. Beyond the plateau, the martensite deforms for the 10.1% pre-strain and the stress-strain response exhibits a linear-elastic type response. The highest pre-strain levels (12.2 and 16.1%) are within the non-linear response and the martensite likely yields. 34

37 Figure 19. The tensile stress-strain curves for the sheet pulled to increasing pre-strain levels at room temperature. Residual strain remains after unloading in Figure 19 and the corresponding displacement is fixed during heating which constrains shape memory effect recovery and generates recovery stresses. The stress generation experiments utilize the heating set-up described in the previous section. The displacement sensor of the MTS machine measures the change in actuator position and thus the change in length of the specimen and the grips. For the experiments in this work, an extensometer would not easily mount to the thin sheet specimens. Hence, the displacement was used for stable feedback that maintained the constant constraint and avoided damaging the extensometer. Note that ASTM standard E328 outlines Standard Test Methods for Stress Relaxation for Materials and Structures at constant temperature and recommends mounting an extensometer within the gage section to fix/constrain the strain. Using the extensometer measurement to maintain a fixed strain constraint can fix the sample length, as the grips would be excluded from the constraint, and the impact on the results will be considered in future work. The current experiments best represent the constraining conditions for prestressing via heat of hydration of the grout. During heating, the displacement was programmed so that is was fixed at the residual strain after pre-straining and the stress generation results are expected to be reliable for the context of the discussion in this work. In Figure 20, the recovery stress is plotted as it evolves throughout heating. The stress generation begins at the onset of heating and thus close to the test temperature. For the lowest pre-strain in Figure 20(a), the recovery stress reaches a maximum (or peak) and drops. For the higher pre-strain levels in Figures 20(b) and 20(c), the recovery stress increases to a maximum and decreases slightly up to the maximum temperature. The 35

38 temperature at the maximum stresses generated during heating should correspond to the temperature at which martensite reversion ceases. Hence, in Figure 20, that temperature ** A f is designated as (associated with constrained residual strain recovery). Recovery stresses reaching maximums during heating have been observed for NiTiNb (Wang et al., 2014) and a Fe-based SMA (Dong et al., 2005) For each pre-strain level, recovery stress accrues after heating is complete as the temperature decreases to room temperature. Recovery stresses increased during cooling for the Fe-based SMA (Dong et al., 2005). The observations that the generated stress reaches a maximum during heating and that it increases during cooling merit further study beyond the scope of the current work. The findings pertinent to the pre-stressing application, which are discussed later in the discussion, are that the maximum recovery stresses are generated under constraint after straining to 12.2% and the resulting transformation temperature interval T = A A ** ** ** R f s. Figure 20. The stress-temperature response during constrained heating and cooling of specimens that have been pre-strained to (a) 5.4%, (b) 10.1% and (c) 12.2%. The double arrows indicate heating, and the single arrows cooling. GROUT CHARACTERIZATION Portland cement, potable water along with any admixtures to obtain required properties are the basic grout materials. The chemical reaction between Portland cement and water is exothermic, i.e. producing heat. This heat is called the heat of hydration. In order to determine the temperature increase during grouting post-tensioning ducts, four commercially available tendon grouts were tested. All grouts were prepackaged and approved by Virginia Department of Transportation for post-tensioning applications. The water-to-grout ratios for each commercial grout were set per manufacturer s direction and given in Table 2. To prepare test specimens with each grout, a mixing cylinder was 36

39 cleaned and a bag of selected grout and the required water were placed in the cylinder. The contents were mixed in the cylinder for 3 minutes using a high shear Jiffler mixer attached to a Milwaukee Magnum Drill at 2500 RPM. Timing started as soon as water was added into the mixing cylinder. The resulting grout mixture was poured into a mm ( inch) cylinder with a thermocouple attached to a single tendon placed in the center. The thermocouple was then connected to a Campbell Scientific data logger which monitored the temperature of the curing grout every minute for 48 hours. This process, which is illustrated in Figure 21, was completed for each grout brand. (a) (b) (c) (d) Figure 21. (a) Mixing grout with high shear Jiffler mixer. (b) Thermocouple attached to steel tendon. (c) Thermocouples on steel tendons and the data logger (d) Completed sample during temperature collection period 37

40 The time versus grout temperature plots for each specimen as well as ambient temperature are given in Figure 22. Three specimens were prepared and tested for Grout I and Grout II on three different days. The results for Grout I are consistent for each specimen. The highest temperature recorded during curing is 41 C, which indicates a temperature increase of 19 C to 20 C from initial temperature of 21 C to 22 C due to the heat produced by the cement hydration. The temperature of Grout II reaches 48 C for two specimens and 51 C for one specimen. At three different tests of Grout II, the average temperature increase is 28 C. The peak temperature and average increase in temperature for Grout III and Grout IV are similar to the results obtained from Grout I. For Grout IV, two samples at two different water-to-grout ratios were tested. It is observed that the peak temperature is slightly higher and occurs a few hours earlier when a lower water-to-grout ratio is used (Figure 22d). The grout temperature reaches its peak value at 10 to 18 hour after casting for Grout I, Grout II and Grout IV whereas the peak temperature occurs at 2.5 hour after casting for Grout III. For all specimens, the grout temperature reduces to values between 22 C and 24 C near 30 hour after casting and remained almost constant thereafter. (a) (b) 38

41 (c) (d) Figure 22. Temperature measured in different commercially available grouts during curing. The results of experimental tests conducted to characterize the grout temperature during curing are summarized in Table 2. These results suggest that a commercially available tendon grout (Grout II) can provide an average of 28 C increase in temperature during the hydration, process, which can be used to activate SMA tendons. Table 2. Summary of grout temperature test results Specimen Grout Water-to- Grout Ratio Initial Temperature ( C) Maximum Temperature ( C) Temperature Increase ( C) S1 Grout I S2 Grout I S3 Grout I S4 Grout II S5 Grout II S6 Grout II S7 Grout III S8 Grout III S9 Grout IV S10 Grout IV S11 Grout IV S12 Grout IV

42 PULL-OUT TESTS To investigate the bond behavior of SMA bars with grout, pull-out tests were conducted. SMA bars with a diameter of 3.5 mm were cut into 220 mm segments using a cutoff wheel. Two mm ( inch) cylindrical molds were used to manufacture pull-out specimens. Holes with a diameter of 3.5 mm were drilled at the center of the top and bottom of the molds to allow SMA to pass through. The SMA bar was secured at the bottom hole of each mold and Grout II was poured inside the mold. The specimens were left to cure for 3 days. Figure 23(a) shows a schematic diagram of the specimen used for the pull-out test. Since it is difficult for standard grips to fully hold on to SMA bars because of their small diameter size, special aluminum sleeves were fabricated. Two 50 mm-long sections were cut from a 10-mm aluminum rod. These sections were then placed on a lathe, and a 3.5-mm hole was drilled all the way through. The sleeve was then attached to the specimen by means of twisting since the aluminum sleeve hole was a little bit smaller than the diameter of the SMA bar. The tight fit was useful to establish a mechanical interlock to help the SMA resist slippage out of the sleeve during testing. The pull-out tests of the SMA bar conducted using an MTS servo-hydraulic load frame. The specimen was held in place by a testing cage attached to the top head of the load frame by a large bolt. The load was applied to the SMA bar at a rate of mm/s and measured by a built-in load cell of the load frame. The slip of the SMA bar relative to grout were measured using Digital Image Correlation (DIC) method at the loaded and free ends of the specimen. Two cards with a speckle pattern were attached to the loaded and free ends of the SMA bar. The bottom card was attached directly at the end of the grout cylinder to reduce any errors in the calculation of slippage due to strains in the SMA. The optical system capture the movement of speckle patterns on the cards and provide an output of the average vertical movements at each time step. Figure 23(b) shows pull out test set-up. 40

43 (a) Figure 23. (a) Pull-out test specimen and (b) test-set-up. (b) Two specimens were tested and the applied tensile force and the slip of the bar were recorded. Bond strength is defined as the shear force per unit surface area of the bar and calculated by the following equation: where T is the tensile load on the SMA bar, db is the nominal bar diameter, and lb is the embedment length of the bar. Figure 24 illustrates bond stress-slip curves both at the loaded and free ends of the SMA bar. The bond behavior is characterized by an initial increase in the bond stress up to 1.2 MPa for the first specimen and up to 1.4 MPa for the second specimen, and with insignificant slippage and a softening thereafter. Since SMA bars had a very smooth surface, mechanical bearing forces were very low and the load transfer was primarily provided by friction. 41