Science and Technology of Welding and Joining ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ystw20 Dissimilar friction welding of NiTi shape memory alloy and steel reinforcing bars for seismic performance Matthieu B. Lezaack, Aude Simar, Yannick Marchal, Martin Steinmetz, Koen Faes & João Pacheco de Almeida To cite this article: Matthieu B. Lezaack, Aude Simar, Yannick Marchal, Martin Steinmetz, Koen Faes & João Pacheco de Almeida (2022) Dissimilar friction welding of NiTi shape memory alloy and steel reinforcing bars for seismic performance, Science and Technology of Welding and Joining, 27:6, 418-428, DOI: 10.1080/13621718.2022.2061692 To link to this article: https://doi.org/10.1080/13621718.2022.2061692 View supplementary material Published online: 12 Apr 2022. Submit your article to this journal Article views: 233 View related articles View Crossmark data Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ystw20 SCIENCE AND TECHNOLOGY OF WELDING AND JOINING 2022, VOL. 27, NO. 6, 418–428 https://doi.org/10.1080/13621718.2022.2061692 RESEARCH ARTICLE Dissimilar friction welding of NiTi shape memory alloy and steel reinforcing bars for seismic performance Matthieu B. Lezaacka , Aude Simar,a , Yannick Marchala , Martin Steinmetzb , Koen Faesc and João Pacheco de Almeidaa a Materials and Civil Engineering (iMMC), Institute of Mechanics, Materials and Civil Engineering (iMMC), UCLouvain, Louvain-la-Neuve, Belgium; b Architecture and Built Environment (LAB), Institute of Landscape, Architecture and Built Environment (LAB), UCLouvain, Louvain-la-Neuve, Belgium; c Technologiepark-Zwijnaarde, Belgian Welding Institute (BWI), Technologiepark-Zwijnaarde, Gent, Belgium ABSTRACT ARTICLE HISTORY Integration of NiTi superelastic bars with steel reinforcing bars is a promising path for increasing the seismic performance of reinforced concrete structures. Welding of NiTi alloy to steel is however not straightforward with standard welding techniques. Friction welding is investigated for performing this dissimilar junction using both regular friction (forced controlled) and milling (position-controlled) machines. When adequate shortenings and forging forces are selected, the welds can withstand the tensile strength of the steel rebars. Failure actually occurs in the heataffected zone of the steel bars. The intermetallic (IM) layer at the welded interface should be kept thin and cracks should be limited in the steel layer next to the interface in order to avoid early brittle interfacial cracking. Received 16 December 2021 Revised 12 March 2022 Accepted 14 March 2022 Introduction Reinforced concrete (RC) dominates the construction sector internationally. A key contribution to addressing the RC construction sector challenges, namely its carbon footprint, consists in designing structures for significantly longer service lives [1–3]. In seismic regions, this implies higher probabilities of sustaining stronger ground motions, which can cause excessive residual (or permanent) displacements, as shown after the NewZealand earthquakes of 2010–2011 [4]. To minimise the latter, harnessing the superelasticity of shape memory alloys (SMA) [5] has recently become an active topic of research[3, 6–14]. Since the response of RC structures is largely governed by the reinforcing bars (or simply rebars), the partial replacement of these ductile construction steel rebars – which depict large plastic strains – by SMA reinforcements is an attractive and simple solution to minimise residual structural displacements. The efficacy of this approach has been proven from investigations focusing on these regions of columns [9,10], beam-column joints [11], beams [12], and more recently from promising applications to RC walls [3,13,14] with Nickel-Titanium (NiTi) SMA rebars. One of the practical challenges for a more widespread engineering use of Nickel-based SMA rebars is the connection between the latter and standard deformed or ribbed steel rebars. Mechanical Nickelbased SMA-steel rebar connections are not desirable CONTACT Aude Simar KEYWORDS dissimilar welding; friction welding; shape memory alloy; steel reinforcing bars; force control; intermetallic layer; tensile strength for practical reasons as mechanical couplers are bulky, which leads to many problems during seismic response. A more fundamental connection between steel and NiTi SMA rebars is required for leveraging the full potential of NiTi-rebars reinforced structures: welding. Akselsen et al. [15] describe the state of the art in terms of welding shape memory alloys. They state that classical techniques such as arc or laser welding can affect drastically the SMA microstructure, by generating intermetallic (IM) phases or coarse dendrites. The mechanical properties of these classical welds are thus quite low, limiting the interest of welding SMAs [15]. Thus, it appears more appropriate to exploit the potential of friction welding to perform reliable welds between rod-shaped bars of SMAs. Friction welding is a solid-state joining technique, reviewed by Maalekia et al. [16] and Uday et al. [17]. Friction welding can be processed by rotary, linear or orbital friction of the components to be welded [16]. Mechanical energy is transformed into heat at the interface of the parts. Thus, energy losses are limited compared to fusion welding techniques, which makes friction welding an efficient process. Furthermore, the plastifying zone expells the initial interface material, allowing to remove interface oxides and impurities out of the weld core section [16]. Friction welding was successfully applied to materials hard to weld such as high carbon content steels [16]. Unlike fusion welding techniques, no shielding gas or aude.simar@uclouvain.be Supplemental data for this article can be accessed here. https://doi.org/10.1080/13621718.2022.2061692 © 2022 Institute of Materials, Minerals and Mining. Published by Taylor & Francis on behalf of the Institute. SCIENCE AND TECHNOLOGY OF WELDING AND JOINING filler materials are required in friction welding. The limiting applicability of the friction welding process is related to the friction setup: at least one part should move and the materials should deform plastically to ensure a strong bonding [16]. Friction welding allows welding of dissimilar materials with rather highly efficient welds [18–20]. In particular, dissimilar welding is reviewed by Udey et al. [17]. Friction welding allows joining materials with highly different mechanical and thermal properties, which is often not successfull with classical fusion welding techniques [17]. Indeed, fusion welding techniques generate large and brittle intermetallic layers at the junction of dissimilar materials, or can lead to solidification defects, which are both avoided by the compressive forces during friction welding [17]. Concerning dissimilar friction welding of NiTi alloys, Fukumoto et al. [18] welded NiTi SMA alloy to stainless steel via friction welding. However, in NiTi SMA-steel friction welded samples, detrimental intermetallic layers of Fe2 Ti form during the process, leading to a brittle behaviour of the joint [18]. The intermetallic layer can be modified by adding Ni sheets between the rods before welding. This changed the intermetallic composition to TiNi3 at the NiTi – Ni interface while no reaction layer was observed at the Ni-steel interface. The resulting junction increased the weld tensile strength up to 500 MPa, 150% higher than the Ni-free interface version [18]. However, the addition of a Ni interlayer makes the process more complex, expensive and labour intensive, which is not suitable for our targeted application. Owing to the past limited investigations on welding SMA to steel using friction welding, the present work will propose a detailed analysis of NiTi – steel welds performed with this technique. One of the issues with dedicated friction welding machines is their limited availability for common civil engineering applications. Indeed, conventional milling machines are available in all machining workshops. The applicability of performing such welds on position-controlled machines is thus also investigated. Using suitable parameters allows the manufacturing of full-strength joints on both kinds of setups, opening the path towards widespread applications for minimal residual displacements in structures after earthquakes. Experimental methods Welding setups The individual steel rebar units are nominally of Class B according to Annex C of Eurocode 2 [21]. The nominal diameter is 12 mm, which is the equivalent plain bar diameter having the same weight per unit length as the deformed bar. The NiTi bars have a smooth surface and a diameter of 12.7 mm. The steel and NiTi bars are cut 419 in 120 mm length parts to obtain the test units. Friction welding is performed on two different machines: (i) a dedicated friction welding machine (Thompson, 40 tons axial force machine) presenting a possibility to control the applied force during the forging step (called ‘force control’, FC conditions); (ii) a position-controlled (called PC conditions) milling machine. This latest machine is actually a dedicated friction stir welding (FSW) machine (TRA-C industries) which, contrarily to a standard milling machine, presents the advantage of having a measurement of the forces during welding ideal for research purposes. However, this machine did not allow force control during the forging phase. On the PC machine, the NiTi bar is inserted into the rotating part while the steel rebar is clamped on the machine table by a chuck. On the dedicated FC machine, the NiTi part are inserted into the rotating head of the machine and the steel on the stationary side. The welding process is monitored in order to register force, shortening displacement, time and revolutions per minute (RPM). Characterisation methods Inductively Coupled Plasma Mass Spectrometry (ICP) is used to determine the chemical composition of the base materials. The latter, as well as the welds, are also characterised by optical microscopy after standard polishing. Microstructures are revealed by etching the polished surfaces. Nital etchant (ethanol + 2% vol HNO3 ) is used for revealing steel microstructure, while NiTi is etched by a solution of HF, HNO3 , and H2 O in volumetric proportions 1:2:10, respectively. Weld interfaces are further characterised by Scanning Electron Microscopy (SEM) for higher magnifications using a Zeiss-Ultra55 equipment. Chemical analysis of the weld interface is performed inside the SEM by Energy Dispersive XRay analysis (EDX). X-ray diffraction (XRD) was performed on base materials on a 5 × 10 mm⊃2 zone as well as on a welded specimen interface in order to identify the crystallographic phases. X-rays are obtained by Co Kα radiations, the diffraction lines were recorded from 2θ = 20° to 125° with 0.02° steps. The mechanical properties of the base materials and welds are also characterised. Micro hardness tests (HV0.5) were performed with an Emco-Test Durascan 5G machine. Weld cross-section interface hardness lines are performed with 250 μm spacing between indents. Uniaxial tensile tests on welds were performed with a MTS 100 kN machine. Steel and NiTi bars, 300 mm long, are directly tested without extra machining following ISO6892-1 standard test specifications [22]. Selected welded samples (see Section 3.2) are machined into cylinders 8 mm diameter before tensile 420 M. B. LEZAACK ET AL. Table 1. Chemical composition of base materials (steel rebars and NiTi) in wt-%. Material Steel NiTi Fe Ni Ti C Cu Mn Mo P Si Zn O 98.150 0.009 0.105 55.880 / 44.054 0.210 0.034 0.445 / 0.615 / 0.020 / 0.010 / 0.175 / 0.025 / / 0.023 Figure 1. (a) Uniaxial tensile response of base materials, (b) initial microstructure of NiTi, austenite phase is identified, (c) initial microstructure of steel: ferrite and pearlite phases are identified. testing. Broken samples are observed in SEM for fracture surface analysis. Results Base materials The chemical compositions of the base materials, obtained by ICP, are displayed in Table 1. Their mean hardness values are 235 ± 9 and 237 ± 11 HV0.5 for the steel and NiTi base alloys respectively. Figure 1(a) provides the tensile curves of the base materials, highlighting the pseudo-elastic plateau of the NiTi material at around 430 MPa. Steel has a higher yield stress of 530 MPa but a limited strain hardening in comparison with the NiTi. Indeed, NiTi reaches 770 MPa as ultimate tensile stress upon phase transformation to martensite. In Figure 1(b) the microstructure of the NiTi is displayed, revealing equiaxed grains identified as austenite phase. Figure 1(c) reveals the steel microstructure. Ferrite (note that the carbon content in Table 1 is 0.21 wt-%, hence inferior to the 0.76 wt-% eutectoid composition) and pearlite phases co-exist, and grains are significantly smaller than NiTi grains: around 10 μm for steel versus 50 μm for NiTi. Note that the steel rebars contain coarse voids elongated in the milling direction, up to 50 μm equivalent diameter. Selection of the studied welding parameters Welding parameters depend on the machine used. While the PC machine controls the rotational speed and the shortening displacement during welding, the FC machine controls the rotational speed and the axial force during welding. Welding trials are thus associated with a rotational speed, a maximum compressive axial force (or welding force), a total shortening (friction + forging shortening cumulated) and a total friction time. The interest of evaluating the welding feasibility on a position-controlled (PC) machine is the availability of such equipment for common civil engineering applications. Almost 25 welding tests have been performed, whose results are summarised in Figure 2 and in the supplementary material (Figure A.1). Figure 2 displays a cartography of evaluated parameters in terms of total welding shortening and welding force. The weldability window is identified as the welding conditions for which the welded joint maximum tensile stress is larger than the yield stress of the steel rebars ( > 530 MPa). This corresponds to the structural engineering design goal of enabling a pseudo-elastic cyclic response of the shape memory alloy rebars. This condition minimises the residual displacements of reinforced concrete (RC) members, before plasticity spreads to the adjoining steel rebar resulting in increasing member permanent deformations. It is interesting to note that a maximum force of about 50 kN is reached in this weldability window, requiring milling machines capable of holding such a large compressive force. For the following investigations, only four samples will be considered for illustrative purposes: two performed on the PC machine and two performed on SCIENCE AND TECHNOLOGY OF WELDING AND JOINING 421 Figure 2. Friction welding trials performed for this study. PC and FC setups are specifically identified. A weldability window was identified at high force, as the tensile strength of the obtained welds is significantly larger than what was obtained at lower welding forces. Four specific samples are discussed in this work, identified by black diamond. One of these samples is voluntarily outside the weldability window. the FC machine (indicated also in Figure 2). Their appearance just after welding is provided in supplementary Figure A.2. • A ‘PC brittle’ sample, outside of the weldability region, representative of brittle behaviour of the weld; • A ‘PC necking’ sample, at the maximum investigated shortening, representative of ductile behaviour of the weld and strain localisation followed by failure in the steel part of the weld; • A ‘FC necking’ sample, again representing ductile behaviour and strain localisation followed by failure in the steel part of the weld. • A ‘FC no necking’ sample, characterising ductile behaviour with fracture at the welded interface. between Figure 3(c,d). The ‘FC necking’ sample of Figure 3(d) has however a longer dwell time before releasing the force after welding. A consequence of the position control (PC) setup is that the compressive force attains peaks, corresponding to increasing shortening, followed by significant force relaxation. Indeed, the thermal retraction of the weld due to cooling is lowering the compressive force. The PC setup increases the shortening in two steps because of the sequential control that needs stopping the rotational speed before activating material forging. The maximum peak force of the ‘PC necking’ sample is around 50 kN, only slightly lower than the 54 kN force that the ‘FC necking’ sample attains at the end of the forging step. Comparing both ‘necking’ samples, the total shortening for the PC setup doubles with respect to the FC setup sample (for similar maximum applied force). Detailed welding conditions Figure 3 displays the processing curves associated to the welding of the four representative samples. FC samples are associated to classical friction welding curves, as shown in [16], while PC samples have rather unusual force evolution for friction welding. It shows that the ‘PC brittle’ sample reaches a low maximum welding force (26.9 kN) and limited shortening. In Figure 3(b), the ‘PC necking’ sample attains higher welding force and also higher shortening. In comparison, the FC samples in Figure 3(c,d) are subjected to higher forces, around 55 kN, which are sustained during the following time interval, corresponding to classical friction welding curves [16]. There is no significant difference in terms of friction time, axial force and shortening Microstructural characterisation of the welded joints Figure 4 shows the microstructure of the four investigated welds observed with scanning electron microscopy (SEM) at the weld centre. In Figure 4(a), the brittle joint interface is highlighted. The ∼ 50 μm thick interface is composed of various intermetallic (IM) layers, cracks and porous zones. The coarse IM layer thickness of Figure 4(a) (weld centre) is not constant throughout the radial direction, as also observed by [18]. In Figure 4(b), the ‘PC necking’ sample is showing a much thinner interface. A few cracks aligned in the radial direction, resulting of high deformations under welding, are still visible on the steel side. In Figure 4(c,d), 422 M. B. LEZAACK ET AL. Figure 3. Welding curves: compressive axial force, shortening and RPM of the bars during friction welding; (a) ‘PC brittle’ sample, (b) ‘PC necking’ sample, (c) ‘FC no necking’ sample, (d) ‘FC necking’ sample. Note that the ‘FC necking’ sample has a holding time twice longer than sample ‘FC no necking’ after forging, i.e. the welding force is maintained constant during cooling of the weld. On all samples, the friction time is highlighted, defined as RPM > 0 and Force > 0 simultaneously. Friction time is longer for PC samples (40 s and 50 s, respectively) than for FC samples (10 s in both cases). Figure 4. (a) ‘PC brittle’ weld; a 50 µm thick reaction layer composed of intermetallic phases (IM) is formed at the interface, (b) ‘PC necking’ sample; the reaction layer is reduced to less than 1 µm, (c) ‘FC no necking’ sample; the reaction layer is small but defects (cracks and porosities) are visible at interface vicinity, (d) ductile ‘FC necking’ sample; reaction layer is even lower than 1 µm. Cracks and coarse porosities are identified by white arrows on steel parts of (b) (c) and (d). the FC samples have also thin interfaces, with similar aspect. The thin IM layers seem homogeneous along the radial direction at the scale of these SEM measurements, keeping a sub-micron size. Porosities and cracks are observed on the steel part of ‘PC necking’ and FC samples. Note that the interface of Figure 4(a) will be further analysed via EDX, see Figure 7. Other samples present too thin IM layers to be observable with EDX. Figure 5 shows a detailed SEM characterisation of the three interfaces corresponding to ductile failure. SCIENCE AND TECHNOLOGY OF WELDING AND JOINING 423 Figure 5. Detailed interface microstructure: (a) PC necking sample, showing refined steel grains around the interface compared to initial microstructure and martensitic NiTi structure (b) FC no necking sample, presenting finer steel grains and martensitic NiTi structure, (c) FC necking sample, the finest steel grains are observed near the interface, a refined austenitic microstructure is visible 3 µm away from the weld interface as a martensitic layer had formed next to the interface. The intermetallic layer is thinner than 1 µm in all conditions. Microstructures of the two FC samples are not equivalent despite similar welding conditions in terms of force and friction time in Figure 5(b,c). On the ‘FC necking’ weld, NiTi grains are austenitic but refined compared to the initial grain size (see Figure 1(b)) near the weld interface. For that sample, a thin martensitic layer of 3 μm has formed at the junction, oppositely to the ‘PC necking’ and ‘FC no necking’ samples, where martensitic layers develop along tens of micrometres thickness. This martensitic layer was indeed identified by XRD as will be shown next on the sample ‘PC necking’ presenting the thickest layer. Concerning the steel part of the weld, equiaxed and refined grains are observed compared to the initial structure (see Figure 1(c)). Note that steel grains are larger for the ‘PC necking’ sample. Figure 6 shows the XRD analysis of base materials and weld interface. The peaks for the steel material correspond to ferrite and body-centred cubic BCC microstructures in accordance with Zhang et al. [23]. Peaks in NiTi material correspond to austenitic structure, as shown by Dudr et al. [24] and Fukumoto et al. [18]. The weld interface XRD measurement reveals that a new phase had formed after friction welding, as a new rising peak is visible. This new peak is related to Martensitic NiTi, as observed by Dudr et al. [24]. Figure 7 shows the EDX line analysis of the chemical composition of the intermetallic layer of the ‘PC brittle’ sample. The approximatively 40 μm thick interface is a mix of the main chemical elements: Ni, Ti, and Fe in varying proportions depending on the position. On the steel side, two or three Fe-rich intermetallics are generated with various Fe content. On the NiTi side, a highly porous IM layer presents a reduced Fe content. Mechanical characterisation of the selected welded joints Figure 8 displays the hardness of the weld mid crosssection for all samples. Figure 8(a) displays the hardness profile at a 10 mm distance from the weld interface on both sides, while Figure 8(b) highlights the interface Figure 6. DRX results on base materials and weld interface of ‘PC necking’ sample; steel base alloy is identified as ferrite and body-centred cubic (BCC) Fe + C phases, NiTi base alloy is identified as austenitic NiTi, while the welded interface shows a rising peak corresponding to martensitic NiTi. hardness variation. Left side of Figure 8(a) reveals a softening of the steel, reaching a minimum hardness at 4 mm (‘PC brittle’) or 8 mm away from the interface (all other samples). Right side of Figure 8(a) shows that NiTi hardness has a minimum value at 3 mm away from the interface for both ‘FC’ samples before a slight increase up to ∼ 275 HV0.5, 10 mm away from the interface. The ‘PC’ samples NiTi side have more homogeneous hardness, around 230 HV0.5 (which is the initial hardness of the material). The full weld section hardness map of ‘FC necking’ sample is available in supplementary Figure A.3. In Figure 8(b), the hardness near interface region is highlighted. Steel hardness is increasing when approaching the interface, as does the NiTi hardness (except for the ‘PC brittle’ sample that shows homogeneous 240 HV0.5 hardness). Local peak hardness near the weld interface is not related to IM layer for ‘FC’s and ‘PC necking’ samples, as the indents are at least 424 M. B. LEZAACK ET AL. Figure 7. EDX measurements on ‘PC brittle’ sample interface. A 40 µm reaction layer generates intermetallic phases containing Fe, Ni and Ti in varying proportions. Figure 8. (a) Overview of the mid-section hardness profile through the four studied welds, (b) zoom on the interface region. Indent spacing is 250 µm. 50 μm away from the IM layer. However, the proximity of the IM layer could constrain plasticity during indentation and cause the appearance of an additional parasitic hardening effect. Oppositely, the peak value at interface for the ‘PC brittle’ sample is related to partial indentation in the thick ( ∼ 40 μm) IM layer. Figure 9 reveals the macroscopic tensile mechanical behaviour of the four selected welds. While the ‘PC necking’ sample has a ductile deformation, leading to 9% engineer fracture strain, the ‘PC brittle’ sample and all similar samples of the non-weldable zone of Figure 2 break under elastic deformation regime, at extremely low stresses and strains ( < 200 MPa and < 0.5%, respectively). The two samples obtained with the FC machine show a ductile tensile response. The ‘FC necking’ sample has a similar deformation capacity compared to the ‘PC necking’ sample, whereas the ‘FS no necking’ sample breaks earlier without significant necking, which is apparent from the absence of a final unloading branch. Both necking samples failed in the steel part of the weld outside the central 25 mm gauged by the extensometer, while the ‘PC brittle’ and the ‘FS no necking’ samples broke at the welded interface. Base material curves are provided for comparison purposes, evidencing the attainment of the ultimate steel rebar strength, as originally desired. In Figure 10, the broken surfaces of the four samples are investigated. In Figure 10(a), the weld interface failed, revealing a flat surface composed of the intermetallic layer (Fe-Ni-Ti). Figure 10(b) reveals the broken surface of sample ‘FC no necking’, showing the influence of the intermetallic layer on fracture path. In that case, the intermetallic layer is more dense than what can be observed in Figure 10(a) and leads to a flat decohesion surface. In some regions of the fracture surface of the ‘FC no necking’ sample the steel material is reached, as proven by Supplementary Figure A.4. In Figure 10(c), the ‘FC necking’ sample and ‘PC necking’ both reveal a very similar fracture surface: a ductile failure in the steel part of the weld. Ductile damage is visible, as a wide population of cavities and porosities are visible on the fracture surface. Discussion Effect of the welding parameters/conditions on the joint quality Many welding trials were performed on both PC and FC machines (see Figure 2 and supplementary Figure A.1). The PC machine allowed the control of the shortening at the friction stage (i.e. while the parts are SCIENCE AND TECHNOLOGY OF WELDING AND JOINING 425 Figure 9. Tensile response of ‘PC brittle’, ‘PC necking’, ‘FC necking’ and ‘FC no necking’ welded samples in contrast with the base material tensile curves; ‘PC brittle’ sample breaks at low stress and low strain, in the elastic regime. Oppositely, both ‘PC necking’ and ‘FC necking’ samples show the same behaviour and fail by necking at a strain corresponding to the crossing of the base materials tensile curves, around 9% strain. Figure 10. (a) ‘PC brittle’ sample failure surface, porosities are observed, (b) ‘FC no necking’ sample: fails at the interface but no more porosities on fracture surface, some regions of the crack surface reach the steel material (see Supplementary Fig A.4), (c) ‘FC necking’ or ‘PC necking’ sample fracture surfaces, showing ductile damage by void nucleation growth and coalescence. rotating) and at the forging stage. The very first trials were performed at limited shortening during the friction stage, without forging shortening. The resulting welds were brittle and resulting tensile strength was lower than 40 MPa. Iterations were done in order to increase progressively the total shortening, by adjusting the friction shortening as well as the forging shortening (see supplementary Figure A.1). Below 20 mm of total shortening, none of the obtained welds were successful in rising the strength above 150 MPa. The examination of these welds after fracture revealed a thick intermetallic layer, of the same nature than the displayed interface in Figure 4(a). The thick intermetallic layer is inhomogeneous along the radial direction, in terms of existence of several IM compositions and distributions, as well as in terms of total IM layer thickness. It was observed that limited shortening of the bars is associated to lower compressive forces during the forging steps on the PC machine, as seen for the ‘PC brittle’ sample in Figure 3. Low shortening and low forces result in thicker intermetallic layers. This layer is detrimental to the mechanical properties of the weld as also demonstrated by Fukumoto et al. [18]. The intermetallic layer is a preferential path for crack propagation as it already contains cracks after forging (Figure 4(a)) as well as porosities. The intermetallic zone is mainly composed of Fe2 Ti brittle phase and TiNi3 phase, according to Fukumoto [18], which is in good agreement with the results of Figure 7, as the Ni, Ti, and Fe are coexisting inside the reaction layer. Increasing drastically the shortening improves the mechanical properties of the weld on the PC setup, as proved by the ‘PC necking’ sample, see Figure 9. In that case, the brittle intermetallic layer is driven out of the interface during forging [16], following metal plastic deformation. Expelled material generates a curved flash around the welded section (see Supplementary Figure A.3). Larger shortenings necessarily increase the amount of expelled material. The significant reduction of this interface intermetallic layer due to large plastic deformations was not observed by Fukumoto et al. [18], who obtained larger intermetallic layers in 426 M. B. LEZAACK ET AL. their reference welds, i.e. without using an interlayer Ni sheet. Furthermore, the intermetallic layer at the interface obtained in the current welds inside the weldability window is homogeneous in the radial direction, which contrasts with the results by Fukumoto et al. [18], where significant thickness variations were reported along the radial direction. Interestingly, the total shortening of ‘FC’ samples during friction welding is about half of the ‘PC necking’ shortening (Figure 2) with equivalent peak force (around 50 kN). Friction time is defined as the duration of the processes under which the force is positive (Force > 0 kN) and the parts are in rotation (RPM > 0). Figure 3 suggests that, provided that sufficient heat is generated during the friction stage, similarly high weld strengths can be produced. As it can be seen for the samples ‘PC necking’ and ‘FC necking’, this heat can be produced through longer friction phases at lower compressive forces (‘PC necking’), or shorter friction phases at higher compressive forces (‘FC necking’). According to Figure 3, the friction time of FC samples is close to 10 s at high compressive force (15–20 kN), while the PC samples have approximately 40–50 s friction time at much lower compressive forces. The RPM is also higher on the PC machine, 1500 rev min−1 oppositely to the FC machine that worked under 1250 rev min−1 . The increased duration and higher RPM could heat up the material to higher temperatures at the vicinity of the weld interface despite the lower compressive forces. This could facilitate the deformation of the interface region during forging for the PC samples, leading to larger deformation at lower compressive force. Effect of the welding conditions on the microstructure The plastic deformation of the interface during welding (mostly in the forging step) has a strong influence on the initial materials’ microstructures. It is obvious in Figure 5 that steel grains (of all samples) are much smaller than the initial grains of Figure 1(b). The heat and plastic deformations led to dynamic recrystallisation (DRX) of the grains near the interface, as observed by Satyanarayana et al. [20] on steel to steel friction welded joint. Furthermore, the steel part of the weld contains cracks and porosities, resulting from the intense deformation (indicated by the white arrows in Figure 4). Cracks may have formed by the collapse of initial porosities contained in the steel bar, in addition to nucleation of new cracks by intense shearing in the radial direction. Despite these relatively coarse defects ( ∼ 10 μm long, and many smaller sizes), failure occurs 20 mm away from the interface on the steel side of the ‘FC necking’ sample, (see Supplementary Figure A.5). Oppositely, the ‘FC no necking’ sample is probably affected by the presence of these coarse defects as the fracture surface (Figure 10(b)) reaches the pure steel material in some regions, as demonstrated in Supplementary Figure A.4. A critical size of defect could thus explain the difference between the ‘FC no necking’ and the ‘FC necking’ samples. Longer dwell time after forging could be related to better closure of the cracks near the interface in the ‘FC necking’ welding conditions. Similarly to the steel part, NiTi initial austenitic grains appear to be significantly refined in the sample ‘FC necking’ of Figure 5. It was shown that Tisteel joints caused DRX in the Ti part of the weld, as described by Dey et al. [19]. The mechanism is thus related to this recrystallisation under high temperatures and strains. It was also demonstrated by Mani Prabu et al. [25] that high temperatures and strains initiate dynamic recrystallisation when welding NiTi plates via friction stir welding. According to the properties of NiTi alloy, high temperature and high strains play opposite roles in the microstructure crystallography. High temperature favours austenite formation, while high strain favours the formation of detwinned martensite, as described by Xie et al. [26] and Jacobus et al. [27]. Furthermore, strain rates influence phase transformations: it was shown by Xie et al. [26] that the stress needed for austenite to martensite transformation is higher under higher strain rates. Related to our work, the strain rates obtained during friction welding could rise the austenite to martensite transformation stress to higher levels, reducing the final amount of martensite near the interface, even if this aspect was not further investigated. Focusing again on the weld manufacturing, friction welding induces first high temperatures before applying large strains during the forging step. The NiTi material close to the interface is thus austenitic before welding. Increasing the temperature is not affecting the austenitic nature of the NiTi but once the forging starts, the NiTi layer close to the interface turns into detwinned martensite phase under high stresses. The NiTi martensitic zone that develops near the interface is larger in the ‘PC necking’ and ‘FC no necking’ samples compared to the ‘FC necking’ sample, according to Figure 4. The existence of the martensitic NiTi in ‘PC necking’ sample was shown by X-ray Diffraction analysis in Figure 6. The thinner martensitic zones of the ‘FC necking’ sample could be related to microstructural rearrangements under high pressures and temperatures, as the holding time after forging is longer. As shown in Figure 5(c), austenitic refined grains are visible near the weld interface, finer than the initial NiTi grains. Recrystallisation of the microstructure is thus obvious and could have transformed the martensite straininduced zones into austenitic grains, as observed in other thermomechanical process at high strains and temperatures [25]. SCIENCE AND TECHNOLOGY OF WELDING AND JOINING Despite recrystallisation, a martensitic layer survives at the interface vicinity, as displayed in Figure 5. The martensitic layer thickness could explain the difference between the ‘FC necking’ and ‘FC no necking’ samples tensile behaviours (in addition to the defects in steel part, as discussed above). Indeed, the larger martensitic layer of ‘FC no necking’ sample is associated to significantly higher hardness at the weld interface (see Figure 8), as referred with respect to other dissimilar welds by Uday et al. [17]. Oppositely, the thinner martensitic layer of ‘FC necking’ sample gives a smoother hardness profile when crossing the interface (see Figure 8). Reducing hardness gradients lowers the strain localisation at the interface, which could prevent brittle failure under straining. Links between microstructure and mechanical properties According to the discussed effect of welding parameters on the microstructure, the highest compressive forces during welding lead to the finest interface layers. It is then proved on Figure 9 that the coarse intermetallic layer of sample ‘PC brittle’ results in poor mechanical behaviour, breaking in the elastic regime at 120 MPa. All other investigated PC samples that do not reach high forces, i.e. above 30 kN (see Figure 2 and supplementary Figure A.1), show brittle behaviour and limited tensile strength as the described ‘PC brittle’ sample. Oppositely, the thin intermetallic layers of ‘PC and FC necking’ samples allow the joints to reach the ultimate tensile strength of the steel, as the interface does not break under loading, hence attaining the original goal of this research. The effect of intermetallic layers on tensile behaviour was investigated in the literature (Refs. [28, 29]). It is thus admitted that the tensile stress rises with decreasing intermetallic layer thickness under a specific size threshold. Full strength weld is thus achieved, activating the pseudo-elasticity effect in the NiTi bar, at least a few mm away from the weld interface where the material is unaffected by the welding process temperature and pressure. However, despite similar intermetallic layer thickness (see Figure 5), the ‘FC no necking’ and the ‘FC necking’ samples have different behaviour. The former breaks at the interface, meaning that another feature is affecting the fracture location. A first explanation could be related to the existence of remaining defects after welding close to the interface, as some cracks are already visible in Figure 4(b–d). The presence of a larger crack could increase the risk of interface failure, as shown in Supplementary Figure A.4, where the crack surface crosses through the steel material in some regions. According to the hardness measurements in Figure 8, the high temperatures and strains near the welded interface affect the hardness of the material. 427 The weld interface has hardened in both NiTi and steel vicinities compared to the initial hardness of both base materials. This localised hardening is mainly explained by the dynamic recrystallisation in steel regions and martensitic transformation in the NiTi part of the weld. However, steel hardness decreases within a minimum 8 mm from the weld interface, due to the progressive reduction of the amount of recrystallised grains. This softening in the steel affects the strain localisation under uniaxial loading and favours necking in steel part of the weld, as observed for both ‘FC necking’ and ‘PC necking’ samples. Conclusions The present work demonstrated that welding NiTi bars to steel rebars is feasible. Full strength welds were obtained using both a dedicated friction welding machine (force-controlled FC setup) and a millingtype machine (position-controlled PC setup). Shortening during welding is the first order parameter in the PC machine setup, allowing for high welding forces and high strains. The PC machine has a longer friction time, leading to extensive softening of the base materials around the weld interface. Such effect is limited in the FC machine, allowing smaller shortenings for identical peak forces. In the analysed samples failing by necking, the interface between steel and NiTi is similar independently of the machine used for welding. The thin interfaces, less than 1 μm thick, are associated to a strong bonding and excellent tensile properties. In some cases, the welded sample breaks after necking in the steel part of the weld, while similar welding conditions in terms of friction time, shortening, and forces, can break at lower strains at the welded interface, without necking. This could be explained by the presence of coarser defects in the steel part of the ‘no necking’ welds and the smoother hardness profile near the weld interface, limiting the localised straining at the brittle interface layer. However, full strength welds can be achieved in both cases, independently of the failure mode. This welding process can be safely performed on classical position-controlled milling machines. Under loading of the composite NiTi-steel structure, the excellent bond will activate the pseudo-elasticity of the NiTi parts, limiting plastic irreversible deformation. The present work thus highlights the applicability of welding NiTi to steel rebars to improve the seismic response of reinforced concrete structures. Future studies will address the response of these welds under cyclic and dynamic loading. Acknowledgements The authors are grateful to the technical staff of the LACAMI and LEMSC technological platforms of iMMC. M.B.L. acknowledges the support of a FRIA grant, Belgium. This 428 M. B. LEZAACK ET AL. work was supported by the Fonds de la Recherche Scientifique – FNRS under Grant(s) n° F.4501.21. [13] Disclosure statement No potential conflict of interest was reported by the author(s). [14] Funding This work was supported by Fonds De La Recherche Scientifique - FNRS [grant number F.4501.21]; Fonds pour la Formation à la Recherche dans l’Industrie et dans l’Agriculture FRIA. [15] [16] [17] References [1] Angst UM, Hooton RD, Marchand J, et al. Present and future durability challenges for reinforced concrete structures. Mater Corros. 2012;63(12):1047–1051. [2] Flatt RJ, Roussel N, Cheeseman CR. Concrete: An eco material that needs to be improved. J Eur Ceram Soc. 2012;32(11):2787–2798. [3] Abdulridha A, Palermo D. Behaviour and modelling of hybrid SMA-steel reinforced concrete slender shear wall. Eng Struct. 2017;147:77–89. [4] Marquis F, Kim JJ, Elwood KJ, et al. Understanding postearthquake decisions on multi-storey concrete buildings in Christchurch, New Zealand. Bull Earthq Eng. 2017;15(2):731–758. [5] Jani JM, Leary M, Subic A, et al. A review of shape memory alloy research, applications and opportunities. Mater Des. 2014;56:1078–1113. [6] Tazarv M, Saiid Saiidi M. Low-damage precast columns for accelerated bridge construction in high seismic zones. J Bridg Eng. 2016;21(3):4015056. [7] Alam MS, Youssef MA, Nehdi ML. Exploratory investigation on mechanical anchors for connecting SMA bars to steel or FRP bars. Mater Struct. 2010;43(1):91–107. [8] Muntasir Billah AHM, Shahria Alam M. Seismic performance of concrete columns reinforced with hybrid shape memory alloy (SMA) and fiber reinforced polymer (FRP) bars. Constr Build Mater. 2012;28(1):730–742. [9] Saiidi MS, Wang H. Exploratory study of seismic response of concrete columns with shape memory alloys reinforcement. ACI Struct J. 2006;103(3):436–443. [10] Saiidi MS, O’Brien M, Sadrossadat-Zadeh M. Cyclic response of concrete bridge columns using superelastic nitinol and bendable concrete. ACI Struct J. 2009;106(1). [11] Youssef MA, Alam MS, Nehdi M. Experimental investigation on the seismic behavior of beam-column joints reinforced with superelastic shape memory alloys. J Earthq Eng. 2008;12(7):1205–1222. [12] Saiidi MS, Sadrossadat-Zadeh M, Ayoub C, et al. Pilot study of behavior of concrete beams rein- [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] forced with shape memory alloys. J Mater Civ Eng. 2007;19(6):454–461. Tolou Kian MJ, Cruz-Noguez C. Reinforced concrete shear walls detailed with innovative materials: seismic performance. J Compos. Constr. 2018;22(6):4018052. Pacheco de Almeida J, Steinmetz M, Rigot F, et al. Shape-memory NiTi alloy rebars in flexural-controlled large-scale reinforced concrete walls: experimental investigation on self-centring and damage limitation. Eng Struct. 2020;220:110865. Akselsen OM. Joining of shape memory alloys. Shape Mem Alloy. 2010;9:183–210. Maalekian M. Friction welding – critical assessment of literature. Sci Technol Weld Join. 2007;12(8):738–759. Uday MB, Fauzi MNA, Zuhailawati H, et al. Advances in friction welding process: a review. Sci Technol Weld Join. 2010;15(7):534–558. Fukumoto S, Inoue T, Mizuno S, et al. Friction welding of TiNi alloy to stainless steel using Ni interlayer. Sci Technol Weld Join. 2010;15(2):124–130. Dey HC, Ashfaq M, Bhaduri AK, et al. Joining of titanium to 304L stainless steel by friction welding. J Mater Process Technol. 2009;209(18–19):5862–5870. Satyanarayana VV, Reddy GM, Mohandas T. Dissimilar metal friction welding of austenitic–ferritic stainless steels. J Mater Process. Technol. 2005;160(2):128–137. CEN (European Committee for Standardization). (2004). Eurocode 2: design of concrete structures – Part 1-1: General rules and rules for buildings. ISO 6892-1. (2016). Metallic materials – Tensile testing – Part 1: Temperature, Method of test at room. Zhang B, Rasmussen B, Jorissen A, et al. Comparison of vibrational comfort assessment criteria for design of timber floors among the European countries. Eng Struct. 2013;52:592–607. Dudr M, Drahokoupil J, Heller L. (2015). Study of nitinol wires using X-ray diffraction and optical methods. In: Struktura 2015. Prabu SSM, Madhu HC, Perugu CS, et al. Microstructure, mechanical properties and shape memory behaviour of friction stir welded nitinol. Mater Sci Eng A. 2017;693:233–236. Xie Z, Liu Y, Van Humbeeck J. Microstructure of NiTi shape memory alloy due to tension–compression cyclic deformation. Acta Mater. 1998;46(6):1989–2000. Jacobus K, Sehitoglu H, Balzer M. Effect of stress state on the stress-induced martensitic transformation in polycrystalline Ni-Ti alloy. Metall Mater Trans A. 1996;27(10):3066–3073. Tanaka T, Morishige T, Hirata T. Comprehensive analysis of joint strength for dissimilar friction stir welds of mild steel to aluminum alloys. Scr. Mater. 2009;61(7):756–759. Jimenez-Mena N, Jacques PJ, Ding L, et al. Enhancement of toughness of Al-to-steel friction melt bonded welds via metallic interlayers. Mater Sci Eng A. 2019;740-741:274–284.