Electron Beam Welding of Hot-Rolled Fe–Mn–Al–Ni Shape Memory Alloy Sheets

The present study focuses on the weldability of hot-rolled Fe–Mn–Al–Ni shape memory alloy sheets by vacuum electron beam welding. Tailored process-specific welding parameters, such as preheating with electron beam or beam oscillation during welding, allowed defect-free joining with very thin weld seams and heat-affected zones. By applying a post-weld cyclic heat treatment, abnormal grain growth can be promoted across the weld seams. However, regardless of the selected welding parameters, some specimens are characterized by the formation of smaller grains within the former fusion zone. In situ incremental strain tests reveal that the former fusion zone has only a minor influence on the functional properties and is not responsible for structural failure. Thus, electron beam welding is a promising welding technology for joining Fe–Mn–Al–Ni shape memory alloys.


Introduction
Shape memory alloys (SMAs) possess the unique ability to return to their original shape after large deformations [1]. The shape recovery, which is initiated upon the removal of applied stress, is called superelasticity (SE) [1]. In case of shape memory effect (SME), deformation initially remains after unloading and requires a subsequent heating for shape recovery [1]. These unique properties are based on a reversible, diffusionless solid-state martensitic phase transformation [1]. SMAs achieved their commercial breakthrough with the development of the Ni-Ti alloy by Buehler et al. [2] in 1962, which is still the most used SMA in industrial applications. There are numerous fields of application like actuators, seismic damping, mechanical joining, or applying prestresses [3]. Especially civil engineering opens up interesting fields for potential applications of SMAs [4]. However, the main drawback for an increased industrial use of Ni-Ti alloys is their high cost resulting from comparatively expensive alloying elements, high dependency of functional properties on compositional accuracy, challenging machinability [5], and weldability [6]. Therefore, low-cost iron-based SMAs, like Fe-Mn-Si-X (X = Cr, Cr-Ni, Cr-Ni-VC) [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] and Fe-Mn-Al-Ni-Z (Z = Ti, Cr) [23][24][25][26][27][28][29][30][31][32], have attracted pronounced attention in recent years. Particularly, the latter show promising functional properties. Superelastic strains up to 10% [25,27] and a low-temperature dependence of the critical stress needed for martensitic transformation [27,31] make the alloy promising for applications as damping elements. The thermo-elastic character of the phase transformation is based on the formation of the aluminum-and nickel-rich, nano-sized b-precipitates (B2, bcc), which are coherent and ordered to the disordered a matrix (A2, bcc) [27,33]. During martensitic transformation of a-phase into c'-phase (A1, fcc) they are elastically distorted but retain their coherency [33]. The size of the b precipitates is a major factor that influences the functional properties and can be tailored by an aging heat treatment at temperatures around 200°C [25,27,[34][35][36]. The modification of functional properties can be very beneficial for prestressing applications. Consequently, unaged condition may be used to apply a prestress upon heating by utilizing the shape memory effect [24]. Activation of the SMA at a temperature regime causing precipitate growth leads to a shift in transformation temperatures and, thus, to a change of functional properties from SME to SE [24]. Therefore, combined shape memory effects can be used in a single application, enabling prestressing elements with the ability to reduce prestress losses and multi-stage, tailorable prestress levels [24]. Furthermore, the superelastic properties are strongly influenced by grain orientation and grain size. Significant improvements in superelastic performance can be achieved when the grain size exceeds the cross-section of the specimens [29]. This is mainly attributed to the high anisotropy of transformation strains with respect to different grain orientations [37], which eventually lead to pronounced incompatibilities at grain boundaries of polycrystalline structures. These incompatibilities cause stress concentrations, especially in the vicinity of tripe junctions, impairing the transformation or even evoking intergranular fracture [38,39]. Therefore, the adjustment of a suitable microstructure is crucial. The grain size can be increased by applying a cyclic heat treatment between the a single-phase region and the a ? c two-phase region promoting abnormal grain growth (AGG) [40,41]. By an adjustment of the chemical composition, the grain growth rate can be directly influenced. Vollmer et al. [30] showed that the addition of Ti is very beneficial. Additionally, the formation of c-phase at the grain boundaries has a crucial influence on the structural integrity and can be tailored by the quenching medium as well as its temperature [39]. By exceeding a critical cooling rate, the formation of c-phase is completely inhibited, leading to crack formation at the grain boundaries [39]. Therefore, quenching into 80°C hot water was chosen in most of the studies.
While several studies focus on the influence of the microstructure on the functional properties and the adjustment of the microstructure by different heat treatments, the weldability of the alloy has not been studied in depth to this date. For an industrial application, however, extended knowledge of weldability and the identification of suitable welding parameters is indispensable. According to the author's best knowledge, only three studies on welding of Fe-Mn-Al-Ni shape memory alloys were published up to now. Viebranz et al. [42] focused on tungsten inert gas welding and investigated the influence of the selected shielding gas composition as well as the arc linear energy on the microstructure in the fusion zone (FZ) and heataffected zone (HAZ). Vollmer et al. [43] studied induction butt welding of Fe-Mn-Al-Ni tubes and the effect of AGG, with a special emphasis on the welding zone, by applying a subsequent cyclic heat treatment. Another promising welding process for joining Fe-Mn-Al-Ni is vacuum electron beam welding. Owing to the high flexibility due to the nearly instantaneous beam deflection, the low heat input compared to other fusion welding processes and the optimal weld seam protection by the vacuum atmosphere, the process is often considered for the joining of materials that are generally difficult to weld [44,45]. Especially for SMAs, the concentrated energy input has the advantage that microstructural changes occur only in a very limited area and, thus, only a minor influence on the functional properties can be expected. Nevertheless, the only study on vacuum electron beam welding of Fe-Mn-Al-Ni is published by Krooß et al. [26]. The authors joined 5-mm-thick bars and observed the microstructural evolution and superelastic properties under compression load after cyclic heat treatment (CHT). The microstructure seems to be slightly influenced by the former weld zone due to grain boundaries still being present in this area after CHT. However, a detrimental influence on the superelastic properties was not observed under compression load.
The present study follows up on the findings of Krooß et al. [26] and extends them with the consideration of different welding parameters and in situ observations. The influence of welding parameters on weld seam geometry, defect formation, microstructure in FZ and HAZ, as well as microstructural evolution during post-weld cyclic heat treatment are investigated. Moreover, in situ incremental strain tests are carried out to reveal the impact of processinherent changes within the welding zone on martensitic transformation and the resulting effect on functional properties and structural failure under tensile load.

Material and Methods
Hot-rolled Fe-32.9Mn-14.1Al-7.3Ni (at. %) sheets with a thickness of 2 mm were produced on an industrial test line by thyssenkrupp Steel Europe AG (Duisburg, Germany). The final chemical composition of the sheet metal batch used in this work was determined by Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES). Sheet metal pieces of 48 9 10 9 2 mm 3 were extracted by means of water jet cutting with the long side perpendicular to rolling direction (RD). Afterward, the sheet metal pieces were ground down to 15-lm grit size to remove the surface layer and to allow a uniform surface finish for the welding studies. Finally, the sheet metal pieces were cut in the middle by means of water-cooled abrasive cutting. This enabled the preparation of a planar, deformation-free cut surface with low roughness for gap-free joining using a butt weld geometry.
The vacuum electron beam welding (EBW) process was carried out on a micro-electron beam machine SEM108, manufactured by pro-beam GmbH & Co. KGaA (Gilching, Germany) as well as JSC Selmi (Sumy, Ukraine). The welding machine can provide a beam current of up to 20 mA at an acceleration voltage of 60 kV. A small beam diameter of approximately 30 lm can be achieved for a beam current of 1 mA. Furthermore, the turbo pump vacuum system provides a low-pressure atmosphere as low as 10 -6 mbar to ensure an optimal protection of the weld seams. For welding, the sheet metal pieces were placed gap-free against each other and fixed by a clamping device. Four different parameter settings, shown in Table 1, were used for the welding process. The preheating, used for parameter set II and IV, was implemented by oscillating the electron beam in an area of 10 9 10 mm 2 with a beam current of 3 mA and utilizing the maximum deflection frequency of approximately 5 MHz until a temperature of 400°C was achieved. The temperature measurement was carried out by applying a thermocouple type K next to the heating area (position marked in Table 1 with a red cross). The oscillation during welding, used for parameter set III and IV, was performed with an amplitude of 1 mm perpendicular to welding direction with a deflection frequency of 1000 Hz. To decrease the size of the end crater, a slope out was used for all welding parameters by continuously reducing the beam current and defocusing across the last millimeter of the weld.
The following characterization steps are illustrated in Fig. 1. For microstructural characterization of FZ and HAZ in the as-welded condition, microsections were prepared longitudinal to welding direction (WD) and perpendicular to WD (cross-section). The specimens were ground down to a grit size of 5 lm and subsequently vibration polished using a colloidal SiO 2 suspension with 0.02-lm particle size. Afterward, specimens were etched using a solution of 3% nitric acid and 97% ethanol for the optical microscopy (OM) observations. In order to evaluate a possible loss of alloying elements due to welding process, chemical composition within the FZ as well as in base material was determined by means of energy-dispersive x-ray spectroscope (EDS) using an acceleration voltage of 20 kV.
To study the influence of the weld seams on AGG, a cyclic heat treatment (CHT) was applied. Therefore, tensile specimens as well as bar specimens were extracted by electric discharge machining (EDM), as shown in Fig. 1a.
To prevent oxidation during heat treatments, the specimens were sealed into quartz tubes under argon atmosphere. The CHT was carried out between the a single-phase region at 1225°C and the a ? c two-phase region at 900°C to  The light orange squares illustrate the areas of preheating. The red crosses mark the position of the thermocouple promote AGG. Three heat treatment cycles were conducted, each with a dwell time of 60 min at 1225°C and 15 min at 900°C. The heating and cooling ramps were 5 K min -1 . Quenching was performed in air for 40 s in order to avoid the formation of undesired martensite (cf. [43]), followed by an aging treatment for 5 h at a temperature of 225°C. Due to the direct aging heat treatment of the air cooled and still warm specimens, the martensitic transformation can be effectively impeded, because the growth of b-precipitates (B2) leads to a shift of the M s temperature to lower values. The bar specimens were used for microstructural characterization of the cross-section after CHT by OM. The preparation steps were the same as described above for the as-welded condition. To study the hardness distribution over the weld seams before and after CHT, Vickers microhardness measurements were performed with a load of 2.94 N (HV0.3).
The tensile specimen was ground again to a grit size of 5 lm and vibration-polished. The crystallographic orientations of the grains present in the gauge length were determined by electron-backscatter diffraction (EBSD). The applied acceleration voltage was 20 kV.
ISTs were carried out using a servo-hydraulic testing machine with a 5-kN load cell. The tests were performed with a constant crosshead displacement of 5 lm s -1 . Strains were measured using an extensometer with a gauge length of 12 mm directly attached to the specimens. The experiment was accompanied in situ by digital optical microscopy, so that images of the specimen surface were taken in the loaded and unloaded state. Figure 2 shows an overview of the microstructure after welding in the longitudinal section. The effect of preheating (II, IV) as well as oscillation (III, IV) of the electron beam during welding on the weld seam width as well as on the solidification microstructure is clearly visible. Both, preheating and oscillation, lead to an increase of weld seam width, grain size, and the amount of c-phase formed at the grain boundaries.

Results and Discussion
The most striking microstructural feature of parameter set I (Fig. 2a) is the relatively large grain size in the center of the weld seam, with elongated grains in weld direction (highlighted by white-dotted lines). Due to preheating up to 400°C (parameter set II), a decrease of the size of the elongated grains in the middle of the weld seam is visible (Fig. 2b), which may be attributed to changes of the prevailing temperature gradient, solidification rate, and heat flow direction. Oscillation of the electron beam (parameter set III and IV) results in more globular microstructure in the middle of the weld seam (Fig. 2c). This observation can be attributed to increased melt pool convection resulting from the high-frequency beam oscillation. Therefore, tips of growing columnar grains are detached and transported through the melt, acting as nuclei for subsequent grain growth. Consequently, the higher number of nuclei effectively retards the growth of columnar grains in the weld center. Similar observations have been documented for laser welding with oscillating beam guidance on finegrained steels [46]. The additional preheating (IV) does not result in any change of the grain shape but leads to a coarsening of the microstructure, which can be attributed to the reduced cooling rate (Fig. 2d).
The cooling rate also has a significant influence on defect formation, as shown in Fig. 3. Large cracks are visible for parameter set I, starting from the end crater (Fig. 3a). In comparison, crack formation can be completely prevented by preheating before welding, shown for parameter set II (Fig. 3b). Figure 4 shows a detailed view on the FZ and HAZ of each parameter set in the longitudinal section. Based on the different microstructural characteristics, the HAZ can be divided into two zones, which are referred as HAZ 1 and HAZ 2 in Fig. 4. The explanation of the origin of the separate zones is given below in the discussion of Fig. 7. Focusing on defect formation, Fig. 4a reveals that, in addition to the large crack starting from the end crater, numerous small cracks occur in the outer area of the FZ for parameter I (highlighted by dotted ellipse). Parameter II, III, and IV, on the other hand, are completely crack-free. In addition, numerous small martensite plates are visible in the FZ and HAZ 1 of parameter I (Fig. 4a). For parameter II these martensite plates only occur in the HAZ 1 (Fig. 4b) and for parameter III and IV ( Fig. 4c and d) there are no martensite plates either in the FZ or the HAZ 1. The  amount of c-phase at the grain boundaries increases from parameter II to IV, whereas c-phase is not present in parameter I. Based on previous studies, the formation of a thin film of ductile c-phase is necessary to prevent crack formation and can be achieved by a reduced cooling rate [39]. Accordingly, the absence of the c-phase is mainly responsible for cracking. The formation of stress-induced martensite leads to additional stress concentrations at the grain boundaries, which cannot be sufficiently accommodated without the presence of c-phase. Thus, parameter I pinpoints that the process window is limited in terms of welding speed by a critical cooling rate. However, preheating is an effective method to reduce the cooling rate and to prevent cracking without leading to a significant increase in weld width.
The differences in weld width become even more obvious in the cross-section (Fig. 5), reaching from 375 lm (I) to 1100 lm (IV). For all parameter sets, a small underfill at the top as well as a root sag at the bottom of the weld seam can be observed. This appearance becomes more pronounced with increased linear energy input (III, IV = 70 J mm -1 ; I, II = 21 J mm -1 ). An increasing linear energy is accompanied by a decreasing surface tension of the melt [47] so that molten metal can flow out of the opened vapor capillary easier. The flow-out is further enhanced by the oscillation of the electron beam during welding, as this widens the vapor capillary. Therefore, the underfill and the root sag of the weld seam are more pronounced for parameter set III compared to parameter set I.
The decrease of the solidification rate due to beam oscillation and increased linear energy input (III & IV vs. I & II) as well as a decreased temperature gradient due to preheating (I vs. II and III vs. IV) leads to a coarsening of the microstructure in the FZ (Fig. 6). This affects both the grain size and the size of the dendritic structures. Both look nearly equiaxial in cross-section (Fig. 6). However, if the longitudinal section of Fig. 4 is also considered, it can be seen that they are orientated columnar in the direction of occurring heat flow.
Evaporation of alloying elements, especially of manganese, due to welding processes of high manganese steels   Fig. 6 in detail was reported in numerous studies. In particular for additive manufacturing of Fe-34Mn-15Al-7.5Ni-at.% [48] and Fe-17Mn-5Si-10Cr-4Ni-1(V,C)-wt.% [49] by laser powder bed fusion high losses of manganese are reported. Moreover, Günther et al. [50] show significant evaporation of manganese with increasing energy input during electron beam melting of Cr-Mn-Ni TRIP steel. A reduction of manganese content was also observed after tungsten inert gas welding of Fe-30Mn-6Si [51]. Therefore, in the present work, EDS measurements in the fusion zones as well as in the base material were carried out ( Table 2). For all welding parameters used, almost no evaporation of alloying elements is observed, which is in agreement with previous results of Krooß et al. [26] for the EBW process. Only in the FZ of III and IV, a slight reduction of the manganese content is measured, probably related to the higher linear energy input used as well as prolonged and widened appearance of the vapor capillary due to beam oscillation. The measured Si content is most likely caused due to residues of the SiO 2 polishing suspension on the surface of the specimens.
The presence of two microstructurally distinct HAZs, which are already mentioned in Fig. 4, is also visible in the cross-sections (Fig. 5 and highlighted in Fig. 7). The formation of two zones can be explained by the local peak temperatures occurring during the welding process. In HAZ 1, an almost entirely a single-phase microstructure is visible, resulting from heating above the c solvus temperature (& 1150°C [30,41]) into the a single-phase region followed by rapid cooling after the welding process. In contrast to the FZ, no dendritic structures are visible in HAZ 1, clearly indicating that this area was heated above c solvus temperature, but has not been melted (cf. Figure 7). Continuing into HAZ 2, a slight chance of c-phase morphology can be observed, which is visible in the lighter coloration in the etched micrograph. It can be assumed that this area was heated up to temperatures above 200°C but below 1150°C, where diffusion processes occur. Especially for heating into the a ? c two-phase region (& 800°C-1150°C [52]), a change of c-phase morphology is to be assumed. Additionally, at temperatures above 400°C, b-Mn phase is likely to precipitate [53]. At lower temperatures, growth of b-precipitates may still occur. However, in addition to the aging temperature, the aging time has a crucial impact on the growth of bprecipitates [34,35]. In consideration of the fast cooling rates obtained by high-power beam welding processes, this effect should only play a subordinate role for the investigated specimens.
The different zones are also distinguishable by hardness measurements as can be derived from Fig. 8a. The hardness values of the indents, marked in Fig. 7, are highlighted with the identical color in Fig. 8a and c. The FZ is characterized by the highest hardness values (413 HV for the red-circled indent). In HAZ 1, a hardness of about 369 HV can be measured. For both zones, an a-phase microstructure with nearly the same grain size and a thin film of c-phase at the grain boundaries is visible. However, dendritic structures within the grains are exclusively present inside the FZ, probably causing the higher hardness value. A very high c-phase fraction is present in HAZ 2 and in the base material, which explains the low hardness values. Since the c-phase fraction in the base material is already very high due to the hot rolling process, only a minor microstructural change is observed in HAZ 2. The c-phase fraction seems to be slightly reduced and the c-phase precipitates are arranged into larger structures, resulting by the heating into the a ? c two-phase region and explaining the slightly increased hardness compared to the base material.
Comparing the maximum hardness values in the FZ of the different welding parameters, it can be seen that they  reach a similar level for II, III, and IV. Only parameter I differs from the other samples with a maximum value of 468 HV. As is already evident from Figs. 4a and 6a, in this condition almost no c-phase is present in the FZ. In addition, the finest dendritic structures are present and a partial martensitic transformation was observed. These findings can explain the increased hardness value in the FZ of parameter set I. After CHT, cooling at air for 40 s and subsequent aging heat treatment at 225°C for 5 h, a homogeneous hardness distribution is visible over the whole specimen for all welding parameters (Fig. 8b). A characteristic microstructure of this condition is shown in the inset of Fig. 8b. Due to the fast cooling rate after CHT, a nearly complete a-single-phase microstructure is present. The inset also visualizes some remaining subgrain structures after CHT. The subgrains provide the driving force for the AGG and are formed by the precipitation and subsequent dissolution of the c-phase during cyclic heat treatment [41]. The small dark spots are most likely c-phase, which is formed during cooling at air due to the reduced cooling rate compared to water quenching. Since, parameter I exhibited cracks in the Fig. 7 Schematic representation of the peak temperature distribution during electron beam welding across the fusion and heat-affected zones explaining the resulting microstructures. The lower image shows the cross-section of parameter set III with the hardness indents used in Fig. 8a. The circled hardness indents of the different zones are shown in Fig. 8c in detail Fig. 8 Hardness measurement of the as-welded condition (a) and after cyclic heat treatment, quenching at air for 40 s and subsequent aging at 225°C for 5 h (b). The inset of (b) shows the characteristic microstructure present for all specimen after CHT. The microstructures of the marked indents in (a) are shown in (c) FZ (cf. Figure 3); this parameter set was not considered further for the heat treatment studies.
It is worth to note that the hardness value in the aswelded condition is already higher than the aged condition after CHT. This was not to be expected since the b precipitates (B2), which significantly influence the hardness, should be distinctively smaller in the as-welded condition compared to the aged condition. On the other hand, the small grain size as well as the dendritic structures contribute to an increase in hardness and can, thus, probably compensate the decreased hardness by the smaller b precipitates. A TEM analysis of the precipitates into FZ will be subject of future work. Figure 9 shows the cross-sections of some specimens after CHT. The former FZs are centrically aligned in the depicted micrographs (highlighted with white-dotted boxes).
It is obvious that CHT results in considerable grain growth and in grain sizes of several millimeters. However, it is noticeable that in the area of the former weld seam, smaller grains are often still present. No dependence of the grain growth behavior on the welding parameters used can be seen. In the specimen shown in Fig. 9d, one grain even seems to have grown across the entire weld seam. Similar observations of a hampered AGG in the welding zone have been made by Krooß et al. [26] and Vollmer et al. [43]. However, the basic mechanisms leading to impeded grain growth in the FZ remain unclear until now and will be focus of further studies.
In situ incremental strain tests with specimens of the CHT conditions were carried out to study the influence of the former weld seam on the functional and structural properties. Since the microstructure and mechanical properties after CHT proved to be independent from the welding parameters employed, only the results of one sample of parameter II are shown below. Figure 10 shows the EBSD orientation mapping of the tested specimen plotted in load direction. A bamboo-like structure with three large grains of several millimeters' length is visible. In the former weld zone two small grains are present. According to the color gradient, subgrain structures can be seen in the h217iand h323i-oriented grain. During the first loading cycle up to 2% strain, the martensitic transformation occurs at a stress level of about 310 MPa resulting in a stress plateau (Fig. 10b). Stress drops during loading are related to the transformation events of new martensite plates [29,35,39,54]. These martensite plates can be observed by the dark contrast at the surface of the specimen resulting from the occurring topography (Fig. 10c). The first transformation takes place in the h415i grain and is completely reversible after unloading (Fig. 10d). When the habit plane of the martensite plate in the h415i grain reaches the grain boundary of the small grain in the center (marked with a red arrow in Fig. 10e), an increase of the stress can be observed in the stress-strain curve of the second load cycle (Fig. 10b red arrow). The following stress drop results from the transformation in the right h217i grain. This transformation behavior was already observed in a previous study of bamboo-like structures by some of the current authors [23]. However, in contrast to the aforementioned study [23], the martensite plate does not cross the grain boundary. After unloading, an irreversible strain of about 1.8% remains. Figure 10f shows that the martensite plate in the h415i grain almost Fig. 9 Etched micrographs depicting the microstructure after cyclic heat treatment with subsequent aging at 225°C for 5 h in cross-sectional view. The former weld zone is located in the middle of the specimens and marked with a white-dotted box. Grain boundaries are highlighted with black-dotted lines completely retransforms, whereas some martensite plates remain in the h217i. This orientation-dependent reversibility of martensite is known from literature [23,28,37,55,56]. The loss of reversibility is related to a local detwinning of the martensite and occurs more likely at grain orientations near h001i in tension [23]. After loading up to 6%, the h217i grain is almost completely transformed. The further transformation in the h415i grain seems to be restricted. Moreover, a martensite plate with deviating orientation appears (Fig. 10g, white arrow), probably as a result of a changed local stress field during loading. After unloading the formed martensite in the h217i grain is almost completely pinned and functional degradation is started in the h415i grain, i.e., the formed martensite is partially no longer reversible. During the subsequent loading cycle, structural failure occurs at a strain of 6% within the h415i grain at the place where the martensite plate with deviating orientation occurs at the former load cycle (Fig. 10i). The pinned martensite plates visible adjacent to the fracture surface in Fig. 10i can be observed as well in the upper area of the fracture surface in Fig. 11. Noticeable is the very smooth fracture surface in the lower left area, which is shown magnified in Fig. 11c and d. Probably, the smooth surface is the grain boundary of an additional grain located underneath the EBSD characterized surface of the specimen. It is known from literature [38,39,57] that triple junctions lead to stress concentrations during martensitic transformation. This could also be an explanation for the occurring martensite plate with deviating orientation in Fig. 10g, as well as for the brittle failure at 6% strain. However, it is obvious that the former weld zone is not responsible for structural failure and only has a minor influence on the superelastic properties. These results are in good agreement with observations made by Vollmer et al. [43] for induction butt welded Fe-Mn-Al-Ni. Fig. 10 In situ incremental strain test after cyclic heat treatment with subsequent aging at 225°C for 5 h. EBSD orientation mapping of the tested tensile specimen plotted in load direction with the corresponding inverse pole figures of the transforming grains (a). Stress-strain diagram up to failure at 6% (b). In situ optical microscopy images (ch) of the area marked in (a) with a white-dotted box. Postmortem optical microscopy image revealing the location of failure (i). The former weld zone is highlighted with orangedotted boxes

Conclusion
In the present study, vacuum electron beam welding was used for joining hot-rolled Fe-Mn-Al-Ni sheets. The influence of different welding parameters on defect formation, microstructural evolution, and abnormal grain growth was investigated. The effect of the weld zone after post cyclic heat treatment on the functional properties was analyzed by applying incremental strain tests accompanied by in situ digital optical microscopy. Based on the results reported in the present work, the following conclusions can be drawn: • Electron beam welding enables robust joining of Fe-Mn-Al-Ni shape memory alloys with very thin weld seams (width \ 500 lm) and small heat-affected zones • High cooling rates, e.g., by a comparatively high welding speed, promote crack formation. The insufficient formation of ductile c-phase at the grain boundaries eventually leads to cracking. • Preheating up to 400°C by an oscillating electron beam as well as beam oscillation transversal to the welding direction during welding reduces the cooling rate, promotes the formation of c-phase and, thus, can successfully prevent cracking. • Abnormal grain growth across the weld seam can be promoted by cyclic heat treatment. However, a few smaller grains remain in the former weld zone in some specimens, regardless of the selected welding parameters. • The functional properties are almost unaffected by the welding process, as the transformation takes place in the larger grains outside the weld seam and structural failure occurs outside the weld zone.
Author Contributions AB contributed to conceptualization, investigation, methodology, visualization, and writing of the original draft preparation. MW contributed to conceptualization, investigation, and writing, reviewing, & editing of the manuscript. PW contributed to investigation, visualization, and writing, reviewing, & editing of the manuscript. MV contributed to conceptualization and writing, reviewing, & editing of the manuscript. NS contributed to conceptualization and writing, reviewing, and editing of the manuscript. SB contributed to resources and writing, reviewing, and editing of the manuscript. TN contributed to conceptualization, supervision, funding acquisition, and writing, reviewing, & editing of the manuscript.
Funding Open Access funding enabled and organized by Projekt DEAL.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.

Data Availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.