Brittle-ductile transition temperature of recrystallized tungsten following exposure to fusion relevant cyclic high heat load

The lifetime of tungsten (W) monoblocks under fusion conditions is ambivalent. In this work, the microstructure dependent mechanical behaviour of pulsed high heat flux (HHF) exposed W monoblock is investigated. Two different microstructural states, i.e. initial (deformed) and recrystallized, both machined from HHF exposed monoblocks are tested using tensile and small punch tests. The initial microstructural state reveals a higher fraction of low angle boundaries along with a preferred orientation of crystals. Following HHF exposure, the recrystallized state exhibits weakening of initial texture along with a higher fraction of high angle boundaries. Irrespective of the testing methodology, both the microstructural states display brittle failure for temperatures lower than 400 ◦C. For higher temperatures ( > 400 ◦C), the recrystallized microstructure exhibits more ductile behaviour as compared to the initial state. The observed microstructural state-dependent mechanical behaviour is further discussed in terms of different microstructural features. The estimated brittle-to-ductile transition temperature (BDTT) range is noticed to be lower for the recrystallized state as compared to the initial state. The lower BDTT in the recrystallized state is attributed to the high purity of the W in combination with its low defect density, thereby preventing segregation of impurities at the recrystallized boundaries and the related premature failure. Based on this observation, it is concluded that the common opinion of the aggravation of BDTT in W due to recrystallization is not unerring, and as a matter of fact, recrystallization in W could be instrumental for preventing the self-castellation of the monoblocks. © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license. ( http://creativecommons.org/licenses/by/4.0/ )


Introduction
The refractory bcc metal Tungsten (W) exhibits exceptional high temperature properties such as a high melting temperature, and a high thermal conductivity in combination with a low coefficient of thermal expansion. Hence, it is the material of choice for vital plasma facing components (PFCs) like the divertor, which extracts heat and helium ash from the burning plasma in future tokamak based reactors [1][2][3] . The W PFCs of the ITER project are based on a monoblock geometry brazed to an actively water cooled CuCrZr alloy heat sink, with an additional Cu interlayer to compensate for the thermal expansion mismatch between the W and the CuCrZr alloy [4] . Also, the microstructure of the monoblocks consists of elongated grains oriented parallel to the heat transfer direction (i.e. monoblock depth) to lower the risk of plasma contamination by tion [18][19][20][21] , self-castellation [19][20][21] and related mechanical property changes [18] . Furthermore, it is well known that W exhibits brittleness at temperatures below the brittle-to-ductile transition temperature (BDTT), which is a variant property, highly sensitive to the processing route and often named as one of the limiting aspects for a reliable operation of the reactor [22,23] . Additionally, the particle loads (neutrons and ions) could be detrimental by promoting defect generation and associated interactions, thereby leading to an increased BDTT due to irradiation hardening [24][25][26][27] , along with drop in thermal conductivity [28,29] , and gas atom retention [30][31][32] .
Fundamentally speaking, the BDTT phenomenon in W is linked to the mobility of dislocations, specifically the glide of the < 111 > screw dislocations and several strategies have been adopted to counter it, which includes alloying strategies, such as Re addition [33][34][35] as well as thermo-mechanical processing [23,[36][37][38][39] . Significant reductions in BDTT with increasing levels of deformation have been reported [38,40] , and the mechanism for reducing the BDTT has been further scrutinized at the level of microstructure, i.e. the grain boundary character, dislocation density, crystal orientation, grain size, as well the nature of mechanical test, specifically the loading/strain rate [41][42][43] . For example, the lower BDTT for heavily deformed W plates was attributed to the rolling induced increase of dislocation sources such as dislocation boundaries, boundary ledges, as well as the density of the edge dislocations [23,40] . Similarly, Ren et al. [44] concluded that the higher ductility of cold rolled W at lower temperatures is due to the presence of a higher fraction of low angle grain boundaries and edge type dislocations. Apart from the grain boundary and dislocation character, the influence of the grain size on the BDTT is well established, with an increase in BDTT with decreasing grain size at first (alike the Hall-Petch relation), reaching a maximum value, followed by a drop in BDTT for ultra-fine grains [23,36,41,45,46] . Recrystallization in W has been shown to promote brittleness at low temperatures (implying a higher BDTT as compared to the deformed state), quantified by a drop in the total elongation under tensile [47] or bending [44,48] loads as well as a reduction in fracture toughness [49,50] or impact energy [51] . This has been commonly referred to as recrystallization embrittlement [52][53][54] . Farrell et al. [36] , as well as Gorelik [53] , have attributed the origin of recrystallization embrittlement to the segregation of impurities (predominantly interstitial atoms) at the grain boundaries, thereby limiting transmission of slip and ultimately leading to crack nucleation and inter-granular fracture. Similar observations have also been reported by several other studies, while providing accurate details on the nature of impurity atoms such as carbon [55,56] , oxygen [54,55,57,58] , hydrogen [57] , phosphorous [59][60][61] , nitrogen [54,56] , fluorine [62] , nickel [63,64] , iron [65,66] and cobalt [65] . Moreover, numerically the influence of different atoms on the cohesion strength of different special grain boundaries in W has been explored by means of ab-initio simulations, distinguishing between the role of impurity atoms in either reducing or enhancing cohesion [67][68][69][70] . On the other side, Davis [65] has shown that in pure recrystallized W wires (as compared to iron or cobalt contaminated), reasonable ductility can be achieved at lower testing temperatures, with a total elongation exceeding 20%. Identical observations of ductile behaviour in a recrystallized state at lower temperatures have also been observed in high purity (ITER grade) forged W rods/rolled W plate [71][72][73][74] as well as W foils [51] .
The impact of the dynamic evolution of the microstructure of monoblocks (by recrystallization and grain growth mechanism) on their structural integrity has been a subject of major debate. This is due to observations of recrystallization assisted macro-crack formation in monoblocks following cyclic HHF exposure, ultimately posing a risk for the thermal fatigue lifetime of the monoblocks [75][76][77] . Panayotis et al. [ 21,78 ] have attributed the macro-crack formation in the HHF exposed monoblock to extensive plastic deformation of the recrystallized state. Conversely, Nogami and Guan [79] have concluded that plastic deformation during the cooling period could suppress the formation of macro-cracks, but they may re-appear in fatigue. Also, by means of FEA analysis, Durif et al. [80] have shown that the performance of recrystallized W with a higher BDTT promotes macro-crack formation. However, it is important to point out that there have been studies reporting no macro and/or micro-crack following HHF exposure of W monoblocks (ITER divertor qualification program [16] ) from different suppliers (Ansaldo Nucleare, Italy [16] ; ALMT, Japan and AT&M, China [17,18] ). The inhibition of macro/micro-crack formation in the recrystallized regime along with a superior thermal fatigue behaviour of W monoblocks (despite recrystallization) in these studies further elevates the question "Does recrystallization promote brittleness in pure W by raising the BDTT?" This research question constitutes the basis of the present paper. For this, mechanical tests in the form of tensile and small punch test on samples with an initial or recrystallized microstructural state, machined from high heat flux (HHF) exposed pure W monoblocks are performed. Additionally, microstructural characterization in terms of grain boundary character, grain size, grain shape as well as crystal orientation by electron back scatter diffraction (EBSD) technique is performed to examine the observed microstructure dependent mechanical behaviour. To the author's best knowledge, no studies have investigated the influence of the evolving microstructure by recrystallization under pulsed HHF loads, and, more importantly the role of purity of W monoblocks. The novelty of the paper involves new insights into the microstructure dependent mechanical behaviour of the pulsed HHF exposed pure W monoblocks.
The article is structured in the following manner. Section 2 highlights the monoblock details and the mechanical testing and microstructure characterization procedure. Section 3 presents and discusses the results of the microstructure characterization and microstructure dependent mechanical response. Section 4 provides estimates of the microstructure dependent BDTT reasoned based on microstructural features, whereas Section 5 presents the important conclusions resulting from the present work.

W monoblocks: Manufacturing, HHF exposure and machining
Tungsten monoblocks manufactured by AT&M China were obtained from RI Research Instruments GmbH after HHF exposure. As highlighted in [18] , the monoblocks were produced by sintering followed by hot rolling 2 , with a geometry as per ITER specifications [4] , i.e. dimension of 28 (width) × 28 (height) × 12 (thickness) mm and an armour thickness (distance between the surface and cooling tube, i.e. region with temperature gradient) of 8 mm as shown in figs. 1 a and 1 b (along with the monoblock reference scheme), with elongated grains oriented parallel to the monoblock depth (height direction) [5] . Additionally, the chemical composition of the W used in monoblocks conforms ITER specifications [4] and table 1 highlights the nature of impurities. The HHF exposure was performed using a electron beam at IDTF (ITER divertor testing facility) in St. Petersburg, Russia. Moreover, the ITER divertor qualification program scheme consisting of 50 0 0 cycles at 10 MWm −2 and 10 0 0 cycles at 20 MWm −2 was adopted [16] , ultimately resulting in substantial recrystallization in the top 5-5.5 mm of the Fig. 1. Sketch of the water cooled tungsten monoblock after high heat flux (HHF) exposure (a) isometric view (b) front view. All the listed dimensions are given in millimetres. The monoblocks have an armour thickness of 8 mm, with the recrystallization depth extending approximately up to 5 mm, i.e. zone A. Zone B refers to the initial microstructure. Mechanical testing specimens were machined in the form of tensile dog bone samples and small punch discs, from zone A as well as zone B (b,c) .  fig. 1 a), whereas the rest of the monoblock retained its initial 3 microstructure (indicated as zone B in fig. 1 b), as reported by Shah et al. [18] . It is important to point out that the evolution of the microstructure by recrystallization under HHF exposure is influenced by the stress, thereby displaying kinetics different from that of isothermal (static) conditions. To study the microstructure dependent mechanical behaviour, thin sheets (thickness varying between 0.5-1 mm) were machined via electro discharge machining (EDM) along the tube direction (i.e. WD-TD plane in fig. 1 b) of the monoblock, from the initial (Zone B) and the recrystallized (Zone A) microstructure. Next, from these thin sheets, specimens for tensile and small punch tests were processed by EDM. The processing induced size effect due to miniaturization of the tensile geometry is well known, whereby hardened edge regions dictate the macroscopic mechanical behaviour [81] . Based on the criterion established by Kohyama et al. [82] (critical thickness of 6-8 times the average grain size) for the unirradiated material state, the thickness of the tensile specimens was chosen such as to avoid these size effects. The tensile specimens had a dog bone geometry with a cross-section of 2 x 1 mm 2 , and the small punch specimens were disc shaped with a diameter of 8 mm and thickness of 0.5 mm. The tensile specimens had a surface roughness (R a ) of approximately 1.8 μm, whereas the small punch specimens had a lower surface roughness (R a ) of approximately 0.3 μm to minimize the effects of friction. The surface roughness of the small punch specimens is marginally higher than the upcoming standard specification of 0.25 μm [83] , but no major influence of this minor deviation is expected on the results. Additionally, for excluding the effects of warping, the thickness of the small punch specimens was dimensionally checked with a maximum tolerance of ± 10 μm. Figure 1 b shows the locations of the machined thin sheets in the monoblock along with a sketch of the machining locations of tensile and small punch specimens per sheet, whereas fig. 1 c provides a perspective view on the miniaturized specimen geometry and size adopted for the mechanical tests in this work. Note that for the recrystallized regime (Zone A) in the monoblock, the top ± 0.8 mm thick surface layer consists of secondary recrystallized large grains, while the rest of the regime displays a nearly identical homogenous microstructure of primary recrystallized grains [18] . Accordingly, to prevent the influence of the large secondary recrystallized grains on the mechanical behaviour, the top 1 mm surface area of the monoblock was excluded while machining the sheets.

Material characterization
To quantify the microstructure dependent mechanical behaviour, microstructural examination was carried out via EBSD. Specimens corresponding to the WD-TD plane of the monoblock from the initial (Zone B in fig. 1 b) and recrystallized state (Zone A in fig. 1 b) were prepared by mechanical grinding using SiC paper (up to 40 0 0 grit), followed by electropolishing with 1.5 wt.% NaOH (sodium hydroxide) solution at 15 V for approximately 45 seconds. For EBSD characterization, a FEI Sirion XL-30 ultra-high resolution scanning electron microscope (UHR-SEM) equipped with a field emission gun (FEG) and EDAX Hikari EBSD camera was used to acquire microstructure maps. An acceleration voltage of 25 kV along with a working distance of 10 mm was used for all mappings, while the step size was varied depending on the microstructural state. In case of the initial state, an area of 1.25 × 0.7 mm 2 was mapped with a step size of 0.5 μm, whereas for the recrystallized state, an area of 1.25 × 1 mm 2 was mapped with a step size of 1 μm. The EBSD map data was subjected to two clean-up procedures, i.e. grain dilation clean-up and neighbour confidence index (CI) correlation. Additionally, the indexed points with a CI less than 0.1 were ignored for post-processing, which was performed using a combination of TSL-OIM analysis TM and the MTEX toolbox [84] . The grains were identified using a minimum misorientation angle of 5 • and the grain diameter was determined from the grain area, assuming a spherical grain. The grain boundaries were defined by means of misorientation between the adjacent points, with boundaries having a misorientation between 2 • -15 • regarded as low angle grain boundaries (LAGBs) and boundaries with a misorientation larger than 15 • evaluated as high angle grain boundaries (HAGBs) [85] . For determining the average grain size from the mapped data, the area weighted average approach was used [86] . Moreover, to investigate the initial deformed state at the local and grain scale, as well as to obtain a qualitative comparison with the recrystallized state, different EBSD based criteria such as KAM (Kernel Average Misorientation), GROD (Grain Reference Orientation Deviation) and GOS (Grain Orientation Spread) were assessed [87] . The KAM value refers to the average misorientation between a pixel and its neighbours, restricted by a specific threshold value to avoid the influence of grain boundaries (predominantly HAGBs). It allows to get insight into the local plastic strain gradients arising from deformation assisted crystal lattice rotations [88] . The KAM maps in the present work were processed differently depending on the microstructural state. For the initial state, the KAM map was evaluated with respect to the second nearest neighbours (step size: 0.5 μm) with a threshold average misorientation of 5 • , whereas for the recrystallized case, the KAM map was processed with respect to the first nearest neighbours (step size: 1.0 μm) with a threshold average misorientation of 5 • . Furthermore, the GROD also provides insight into the local orientation field inside a grain and is usually defined as the angle difference between the pixel of interest and a fixed reference point. Mathematically, this is expressed as GROD m = α nm , where α nm is the angle difference between the pixel n and the reference orientation of grain m . The mean misorientation of the grains was used as the fixed reference point to evaluate the GROD, independent of the microstructural state. Based on the GROD criterion, the GOS provides information of the orientation gradients at a larger length scale, i.e. between grains and can be used to differentiate recrystallized grains from deformed grains. The GOS value of a particular grain is obtained by averaging the GROD of all the pixels belonging to the grain and can be expressed as GOS m = 1 P P i =1 GROD i , with P denoting the number of pixels inside the grain m . Irrespective of the studied microstructure in the present work, the recrystallized grains were determined using a threshold GOS value of 3 • [18,89] .
The assessment of the post mortem microstructure dependent failure mode mechanical testing was performed by examining the fracture surfaces of the tensile specimens. For this, a FEI Quanta F600 scanning electron microscope, equipped with an FEG was used. The fracture surface imaging was carried out with an acceleration voltage between 20-25 keV and a working distance of 10 mm. The fracture surfaces of the SPT specimens were not investigated because of the substantial fragmentation of some specimens, thereby limiting a thorough microstructural comparison.

Mechanical testing
The tensile tests were performed using a Hegewald & Peschke universal tensile machine (Inspekt retrofit) equipped with a furnace and a load cell of 20 kN. For each microstructural state, the tests were performed along the width direction (WD; principal stress direction in the monoblock) within a temperature range between 30 0 • C-60 0 • C, with one test per temperature. The lower limit for the testing temperature, i.e. 300 • C was evaluated up to the occurrence of brittle failure for both microstructural states. For all tests, irrespective of the testing temperature, the dwell time at the testing temperature was between 35-45 minutes and the temperature was monitored with a thermocouple placed close to tensile gauge section. The tests were performed using a preload of 25 MPa and a cross-head velocity of 0.1 mm/min (strain rate ≈ 10 −4 s −1 ), while the strains were measured using a ZwickRoell non-contact green coloured laser extensometer. For small punch testing, a custom setup mounted on a Hegewald & Peschke universal tensile machine (Inspekt 20) was used. The details of the setup are further elaborated in Appendix A . The tensile and small punch tests were performed at the Helmholtz-Zentrum Dresden-Rossendorf facility in Dresden, Germany. Figure 2 schematically shows the discrete microstructures observed in the HHF exposed W monoblock by means of an inverse pole figure (IPF) map obtained from EBSD. In fig. 2 a, it can be observed that the initial microstructure consists of relatively large grains elongated with respect to the width direction (WD) of the monoblock. The high angle grain boundaries ( > 15 • ) are further plotted in fig. 2 b to obtain a better picture of the elongated grains and their size. The large grains in the initial state exhibit substantial varying coloured streaks due to rolling process, resulting in grain sub-division and ultimately leading to the formation of smaller grains having a misorientation less than 15 • (not shown here). Also, a large fraction of the grains in the width direction (WD) IPF map ( fig. 2 a) are coloured green, thereby revealing a preferred crystal orientation of < 110 > ||WD.

Microstructural characterization
The width direction (WD) IPF map for the recrystallized microstructure is shown in fig. 2 c and a significant difference in the geometry of the recrystallized grains can be observed as compared to the initial grains. The recrystallized grains tend to exhibit a more equiaxed morphology, which is clear from the high angle grain boundary map shown in fig. 2 d. The grains in the recrystallized state tend to be relatively coarse with only a small fraction of fine grains, which further indicates the occurrence of substantial grain growth. Additionally, the fraction of grains shaded in green is considerable, thereby indicating a < 110 > ||WD type preferred crystal orientation.
Based on the EBSD data, fig. 3 presents a comparison between the two microstructures in terms of grain size, grain boundary character and grain shape. The grain size distribution ( fig. 3 a) for both the microstructural states (initial and recrystallized) shows a peak within the grain diameter range of 100 μm-150 μm. However, the fraction of finer grains ( < 50 μm), as well as larger grains ( > 250 μm) is higher in the initial state compared to the recrystallized state. The average grain size for the recrystallized state is approximately 139 μm, slightly larger than 116 μm for the initial state. As shown in fig. 3 b, the initial state is mainly characterized grain boundary character and grain shape analysis in terms of (c) grain aspect ratio (d) PARIS parameter of the initial microstructural state of the W monoblock as well as the recrystallized state following HHF exposure. Note that for evaluation of the grain aspect ratio and the PARIS parameter, the edge grains were excluded because of the uncertainty in determining their shape. by a higher fraction of LAGBs and a lower fraction of HAGBs. On the other hand, the recrystallized microstructure predominantly features boundaries of the high angle type, with a negligible fraction of LAGBs.
Meanwhile, as highlighted before, the geometry of the grains between the two microstructures differs significantly, thus the grain shape is further analysed using the grain aspect ratio and the percentile average relative indented surface (PARIS) parameter. The difference between the grain shape in the initial and the recrystallized state is illustrated in figs. 3 c and 3 d where the grain aspect ratio and PARIS parameter of a grain are expressed with respect to its diameter. In case of the initial state, the aspect ratio of the grains tends to be higher due to the presence of elongated grains, and consequently high PARIS values are also observed, as deformation assisted elongation of grains promotes lobateness of grains, ultimately leading to an increase in the sinuosity (concavity) of the grain boundaries. In contrast, for the recrystallized grains, irrespective of the grain diameter, most of the grains exhibit a relatively low aspect ratio, i.e. close to one. Moreover, the PARIS value of the grains is smaller than the initial state and close to zero, implying an increased convexity of the grains due to recrystallization.
The local (short range) and grain (long range) scale misorientation analysis for the initial and recrystallized state post-processed from the EBSD data is further shown in fig. 4 . The KAM map for the initial state overlaid with the HAGBs in fig. 4 shows that most of the area fraction of the microstructure displays a notable presence of plastic deformation assisted local orientation gradients (strain gradients and the related geometrically necessary dislocations (GNDs)) with misorientation values reaching as high as 3 • .
At the same time, for the recrystallized state in fig. 4 , the local orientation differences are relatively low with a maximum value of approximately 1.2 • , thereby revealing an overall lower defect density, as expected for a recrystallized microstructure.
At the grain scale, i.e. for the long range orientation variations, the GROD maps in fig. 4 clearly reveal the difference between the initial and recrystallized microstructures, with a comparatively higher orientation deviation from the mean grain orientation in the initial state as compared to the recrystallized state. These higher orientation deviations in the initial state arise from the substantial sub-grain formation due to plastic deformation. Also, the orientation deviations for the initial state tend to be higher near the HAGBs as compared to the grain interior (in fig. 4 ; as high as 45 • marked in red colour), revealing the localization of strain around the grain boundaries. For the recrystallized state, due to the overall lower defect density, the magnitude of the orientation deviations is relatively small.
Additionally, with respect to the long range deviations, the GOS approach can be used to examine the character of each grain, i.e. with or without deformation for both microstructures, as shown graphically through the GOS maps in fig. 4 . A comparison between the GOS maps of the initial and the recrystallized state reveals that the overall orientation spread for the initial state is relatively higher than for the recrystallized state. The majority of the grains in the recrystallized state display an orientation spread less than 3 • ( ≈ 96.5%), which is also the threshold spread used in the work and hence affirming the fully recrystallized state of the microstructure. In contrast, for the initial state, the majority of grains have an orientation spread exceeding 3 • ( ≈ 93%). The important   fig. 5 . In the initial state, the (110) pole figure shows a higher intensity along the width direction (WD), i.e. nearly 4 times the random intensity, implying that the majority of the crystals have their {110} normal to the width direction. Analogously, the orientations (001)||TD (tube direction) and (111)||TD (tube direction) also reveal intensities of nearly 3-3.5 times the random value. The texture components corresponding to the initial state examined in Euler space (not shown here) revealed three major texture components with the highest intensities for For the recrystallized state, the pole figures show a texture similar to that of the initial state, however substantial diminution of the texture intensity occurs, thereby indicating recrystallization assisted weakening of the deformation texture. The orientations (110)||WD, (001)||TD and (111)|| TD all show a texture strength of approximately 1.4 times the random value. Also, following recrystallization, the orientation of (001)||HD (height direction) reveals a higher intensity as compared to the other orienta-tions (1.9 times the random). This is due to the fact that < 001 > is the easy growth direction of the bcc crystals along the temperature gradient, which is in the height direction (HD) of the monoblock. The texture components for the recrystallized state are the same as for the initial state, i.e. {112} < 111 > ( ϕ 1 Figure 6 illustrates the microstructure dependent response of W specimens under uniaxial loading at different temperatures. Figure 6 a, i.e. for the initial microstructural state, reveals that at lower testing temperature (300 • C), the specimen fails at a relatively low stress of approximately 300 MPa in a brittle manner with negligible elongation. The elastic part of the stress-strain curve is not shown because of the negligible recorded elongation and only the failure stress is shown. As the testing temperature increases (to 500 • C), a marginal increase in the total elongation emerges; nonetheless, the specimens exhibit a negligible total strain ( ≈ 1-1.5%) and all fail in a brittle manner. In contrast to the lower testing temperatures, the specimen tested at 600 • C exhibits a substantial ductility with a total elongation reaching approximately 12%. Moreover, by examining the stress-strain curve of the 600 • C tested specimen, it can be observed that the onset of plastic instability (necking) occurs instantaneously after yielding, i.e.  negligible strain hardening, followed by a relatively stable post necking behaviour. Such a behaviour can be attributed to the increasing strain rate following necking, ultimately stabilizing the necking behaviour. The stress-strain curves of the initial state do not exhibit the commonly observed flat profile in the plastic regime of deformed W, more specifically for low temperature rolled W [38] . Additionally, irrespective of the testing temperature ( ≥ 400 • C), the ultimate tensile stress or failure stress (for brittle cases) tends to be within the range 420-450 MPa, which is considerably smaller than the reported values for low temperature rolled W [38,40,47] . This indicates a relatively higher rolling temperature and correspondingly overall lower defect density.

Tensile tests
In case of the recrystallized state, shown in fig. 6 b, the specimen tested at 300 • C shows a behaviour similar to the initial state, with negligible ductility and a relatively small failure stress of 30 MPa. For higher testing temperatures ( ≥ 400 • C), the ductility of the tested specimens increases with the testing temperature, reaching a total elongation of nearly 22% for 550 • C. Additionally, due to the recrystallized state, i.e. an inherently low defect density, the stressstrain curves show a substantial amount of strain hardening, which extends further with the increasing testing temperature. The dependence of the yield stress (proof stress) on the testing temperature in the examined range is found to be negligible, with a yield stress (proof stress) levelling around 120 MPa (identical to reported in Bonk et al. [47] ) as compared to the approximate 420 MPa of the initial state for the same testing temperatures. The complementary microstructural state dependent small punch test results are further detailed in Appendix A .   fig. 7 a at testing temperatures of ≤ 500 • C exhibit completely delaminated flat fracture profiles, revealing a typically brittle failure mode. The fracture profile of the specimen tested at 600 • C reveals substantial necking with fracture initiating at the centre of the cross-section and extending towards the surface at an angle of 45 • , identical to a slant fracture (not shown here).

Fractography: Tensile specimens
On the other hand, for the recrystallized state ( fig. 7 b), a delaminated flat type failure, i.e. brittle fracture, is only observed for the specimen tested at 300 • C. The fracture surface for the 400 • C specimen seems relatively flat at the macro-scale, however the appearance of the surface looks grainy and more dull as compared to the one tested at 300 • C, thereby demonstrating a quasi-brittle like failure. For higher temperature testing ( ≥ 500 • C), the fracture surfaces appear even more grainier with a significant amount of necking, as observed from the extensive plastic deformation of the crosssection. Additionally, substantial inter-granular cracking of grains can also be observed for the higher temperature tested specimens. Figure 8 provides a magnified picture of the temperaturedependent fracture mode of the two microstructural states. Figure 8 a reveals that the brittle fracture in the initial state for specimens tested ≤ 500 • C is a mixed-mode type of fracture, i.e. a combination of trans-granular and inter-granular. The majority of the grains fail by cleavage, visible as river like patterns on the fracture surface, thereby highlighting the trans-granular fracture mode (purple coloured triangle in the micrographs). Nonetheless, some areas do clearly depict cube facets, a typical characteristic of inter-granular fracture (indicated by red stars). Also, irrespective of the testing temperature, small residual pores from the sintering process can be observed. For the high temperature tested specimen (600 • C in fig. 8 a), the ductile fracture is mainly characterized by significant tearing, ultimately leading to the formation of a fibrous structure (green coloured squares) with elongated sintering pores. Similar to the initial state, the brittle failure in the recrystallized state ( fig. 8 b) also shows a mixed mode type fracture at lower testing temperatures, i.e. below 500 • C. However, one marked difference for the recrystallized state is the greater fraction of inter-granular type fracture mode as compared to the initial state. Although, the fracture features for the 500 • C tested specimen visually looks similar to the one tested at 400 • C, striations on the inter-granular faces along with a small fraction of fibrous structure (indicated by green square) also appear, characterising the transition from brittle-to-ductile failure. Additionally, the fraction of the ductile fibrous features is higher in the 550 • C tested specimen, thereby confirming the transition. The difference in the fracture mode between the two microstructural states, i.e. a relatively higher fraction of trans-granular cleavage in the initial state as compared to inter-granular in the recrystallized state can be attributed to the difference in the morphology of grains and the nature of grain boundaries. For the initial state, the grains tend to be elongated with a relatively high fraction of LAGBs. Thus, as a result, substantial crack branching occurs, leading to a transgranular mode. Unlike the initial state, the grains in the recrystallized state are close to a polyhedra equiaxed morphology with a higher fraction of HAGBs. Hence, the crack path along the HAGBs remains nearly straight, ultimately resulting in inter-granular fracture. Moreover, correlating the observed microstructure dependent fracture surfaces ( figs. 7 and 8 ) with the tensile curves ( fig. 6 ), an adequate agreement between the fracture mode and the mechanical behaviour can be identified, with the initial state exhibiting a lower global ductility, i.e. a predominantly mixed mode brittle fracture at lower temperatures as compared to the recrystallized state.

Microstructure dependent BDTT
In the previous section, the microstructural state dependent mechanical behaviour assessed by tensile tests were shown and briefly described. In this section, these results are further scrutinized in terms of the microstructure dependent BDTT as well as correlated to the corresponding microstructural characteristics. Taking into account the tensile curves shown in figs. 6 a and 6 b, the total energy absorbed prior to fracture, in other words, the strain energy density (modulus of toughness), calculated from the area under the stress-strain curve, provides a measure of the BDTT, which is schematically shown in fig. 9 a for the two microstructural states. For temperatures ≤ 500 • C, the initial state displays significantly lower toughness values, whereas for the recrystallized state, a steady increase in toughness with the testing temperature occurs, reaching nearly 4 times the toughness of the initial state (at 500 • C). For temperatures greater than 500 • C, regardless of the microstructural state, similar values of a relatively high toughness can be observed, thereby providing a measure of the ductile failure regime. Furthermore, assessing the temperature-dependent in-crease in toughness, it can be concluded that the BDTT of the initial state falls within the range 50 0 • C-60 0 • C, and similarly for the recrystallized state, where it is between 40 0 • C-50 0 • C. Also, the observed BDTT values in the initial state are significantly higher than reported by Yin et al. [90] under bending for CFETR grade tungsten (manufactured by AT&M, China). This can be explained based on the difference in processing, i.e. relatively finer grains compared to the present case.
In addition to the tensile curves, estimates of the microstructure dependent BDTT from the small punch test can be obtained as well by evaluating the fracture energy from the force-displacement curves ( figs. 10 a and 10 b in Appendix A ), i.e. the area under the curve evaluated with respect to maximum force, which is further illustrated in fig. 9 b for both microstructures. Note that for the force-displacement curves exhibiting significant pop-in events, a calculation procedure commonly used for the fracture energy was adopted, where a maximum load drop criterion of 20% (similar to Bruchhausen et al. [83,91] ) was used to define the maximum displacement, i.e. the areal integration limit. The fracture energy for the initial state in fig. 9 b typically shows a flattened sigmodial type function, with relatively low values at temperatures below 300 • C, followed by a small increase between 300 • C-350 • C, and ultimately resulting in plateau-like behaviour at temperatures above 450 • C. Contrarily, in case of the recrystallized state, fracture energy values alike that of the initial state can be observed up to 300 • C, whereas for high temperatures, the energy values tend to increase with increasing temperature (note that for some temperatures, the above observation is not completely correct as the energy values for the recrystallized state are similar to the initial state; nonetheless, the bounded sigmoid function shows the evolving trend). Moreover, scrutinizing the energy curves of the two microstructures in fig. 9 b more closely, a difference between the recrystallized state and the initial state can be clearly observed in the range 450 • C-500 • C, asserting the increased ductility of the recrystallized state as compared to the initial state in the small punch test. However, accurate quantitative determination and comparison of the BDTT between the two microstructural states cannot be obtained from the small punch energy curves. Nevertheless, both applied testing methodologies exhibit a pronounced temperature-dependent increase in the ductility of the recrystallized state as compared to the initial state, thereby qualitatively conforming the lowering of the BDTT for the recrystallized state following pulsed HHF exposure.
The relation between the observed mechanical behaviour and the microstructural features, specifically the difference in BDTT between the two microstructures, and the four main microstructural features, i.e. the grain size, grain shape, texture and the grain boundary character, are further assessed. Firstly, considering the influence of grain size, fig. 3 a showed that the average grain size of both microstructures is comparable (with small variations), and therefore its role in justifying the reported difference in BDTT (between the initial and recrystallized) is insignificant, and therefore not further discussed. Secondly, with respect to grain shape, specifically the aspect ratio, Reiser and Hartmaier [46] have shown that microstructures with elongated grains (higher aspect ratio) tend to have a higher toughness, i.e. lower BDTT. Taking this into account, the initial state (higher aspect ratio) should have a lower BDTT as compared to that of the recrystallized state. However, this is not noticed in the present case and thus, the role of the grain shape is judged to be less relevant. Thirdly, as shown previously in fig. 5 , the grains in the initial state do show an oriented crystallographic texture, but following recrystallization, the strength of the texture reduces. Although weak, a positive influence of the crystallographic texture on the BDTT, implying a reduction of BDTT with increasing strength of texture, has been observed experimentally [40,92,93] . Similar conclusions have also been drawn by Oude Vrielink et al. [94] , where a crystal plasticity based model was developed to in-  vestigate the influence of texture anisotropy on the BDTT in W, with the rolling direction (RD, tensile principal stress direction) exhibiting a marginally lower BDTT as compared to the transverse direction (TD, tensile principal stress direction). In this regard, it can be stated that the reduction of the BDTT in the recrystallized state cannot be attributed to the role of preferred crystal orientations due to its weakening. Finally, with respect to the grain boundary character, a higher fraction of LAGBs is observed in the initial state as compared to the recrystallized state, and theoretically, this should result in a lower BDTT for the initial state as compared to the recrystallized state [23,44] . However, this is not observed here and the relatively higher BDTT in the initial state (in spite of the dominant LAGB density) may arise from the hot-rolling induced higher density of dislocations leading to a greater intrinsic barrier for dislocation motion. As a result, higher flow stresses would be required to facilitate plasticity, which could be of the same order as the cleavage stress, thereby leading to brittle failure with no yielding. Also, as depicted by Ren et al. [44] , the character of dislocations plays a crucial role in governing the plasticity of W, where screw-type dislocations tend to have lower mobility as compared to edge dislocations.
For the recrystallized state, predominantly HAGBs are observed that can further act as a dislocation source, which in combination with a lower obstacle density, can assist in dislocation multiplication, thereby promoting ductility. Additionally, the increase in brittleness of the recrystallized state has often been linked with the segregation of impurities at the recrystallized grain boundaries and in this regard, preliminary energy dispersive spectroscopy (EDS) based characterization of the recrystallized grain boundaries was performed (not shown here) to investigate the impurity build-up. However, no noticeable amount of segregates were detected along the grain boundaries. Hence, we conclude here that the lower BDTT in the recrystallized state as relative to the initial state is not related to a specific microstructural feature, but is due to the negligible segregation of impurities at the grain boundaries in combination with an inherently low defect density, ultimately preventing the abrupt failure. This finding of lower BDTT and ductile behaviour of the recrystallized state is consistent with the observation of Davis [65] , where a fairly ductile behaviour of pure tungsten wires at 250 • C was observed in the primary recrystallized state. The lower BDTT of the recrystallized state in the present case in combination with the presence of LAGBs in the surface recrystallized grains (secondary recrystallized grains) [18] can further explain the observation of the absence of macro-crack formation (self-castellation) in some of the HHF exposed monoblocks [16][17][18] .
It also highlights the importance of purity control for manufacturing the monoblocks for ITER.

Summary and conclusions
In the present work, the mechanical behaviour of W monoblocks exposed to a few thousand cycles of pulsed HHF was examined to understand the extent of the mechanical property changes, specifically the BDTT, as a consequence of the dynamic microstructure evolution by recrystallization and grain growth. For this, two different microstructural states, i.e. the initial state and recrystallized state from the HHF exposed monoblocks, were investigated by uniaxial tensile and small punch testing. The following conclusions are drawn from this work: • The initial microstructural state characterized by a high density of LAGBs, and preferred crystal orientation reveals a BDTT in the temperature range 50 0 • C-60 0 • C. The observed BDTT values for the initial state tends to be higher in comparison to the values reported in literature, specifically for the deformed microstructural state. This higher BDTT may be explained on the basis of the processing route of the monoblocks, where a higher rolling temperature results in minimal grain refinement, relatively lower defect density as well as a different dislocation character as compared to lower temperature rolling. • The recrystallized microstructure developed after pulsed HHF exposure, characterized by a relatively lower orientation spread and a high density of HAGBs reveals a BDTT within the temperature range 40 0 • C-50 0 • C, marginally lower than the initial state, thereby affirming that the recrystallization process does not escalate the low temperature brittleness of W. • While comparing the two microstructural states, the recrystallization assisted reduction of BDTT in W (as compared to the initial state) observed in the present work can be attributed to the purity of the W samples, thereby reducing the extent of impurity based segregates and the related weakening of the recrystallized grain boundaries, ultimately preventing brittle failure. Additionally, as the average grain size for both states is comparable, the HAGBs in the recrystallized microstructure act as dislocation sources, which in combination with the overall lower defect density results in an increased global ductility. Also, the commonly reported enhancement of the low temperature ductility due to the deformation induced defects serving as new dislocation sources, was not observed in the present case. • Although the two mechanical test methods, i.e. tensile and small punch tests, used in the present work differ significantly in terms of the stress state inside the specimen, the observations on the microstructure dependent mechanical behaviour obtained from these methodologies are in qualitative agreement. • The lower BDTT of the recrystallized state combined with the higher fraction of LAGBs in the recrystallized surface regime of the monoblock can explain the negligible macro-crack formation (self-castellation) in some of the HHF W monoblocks. This highlights the added value of the recrystallization mechanism. Nonetheless, the significant reduction of the yield stress due to recrystallization, and consequently, the accumulation of plastic strain at longer time scales could negatively affect the thermal fatigue resistance of the monoblocks.
In addition to the influence of the heat loads, the actual load on the W monoblocks in a fusion reactor also entails plasma ions and neutrons. Thus, for quantitative and realistic estimates on the microstructure dependent BDTT, it is important to account for particle loads as well. Preliminarily, the influence of neutrons can be understood through the induced generation of lattice defects in the W monoblock. This can further assist in recrystallization and grain growth or increase the BDTT by irradiation hardening, depending on the irradiation temperature. Simultaneously, the formation of helium bubbles may lower the thermal conductivity, thereby raising the temperature gradient in the monoblock and the stresses, which may again promote recrystallization. Also, these bubbles can also retard recrystallization by exerting a drag force on the moving grain boundaries (as a function of irradiation temperature) [95][96][97] . Moreover, considering the fact that the major part of the bulk monoblock experiences temperatures in the range 80 0 • C-10 0 0 • C, the recovery effects in the microstructure need to be systematically evaluated to assess the durability of the monoblocks, by performing (difficult) long term heat as well as particle exposure experiments.

Data availability
The raw or analysed data reported in this study are available upon reasonable request.

Declaration of Competing Interest
This work and the manuscript present no conflict of interest. of Mario Houska and Wolfgang Webersinke (Helmholtz-Zentrum Dresden-Rossendorf, Germany) in performing the high temperature tensile and small punch test is greatly acknowledged. Additionally, the assistance of Marc van Maris from the Multi-Scale lab (Eindhoven University of Technology, The Netherlands) is highly appreciated.

A1. Setup details
The setup consisted of a punch with a ball diameter of 2.5 mm and a receiving hole with a diameter of 4 mm. The sample was clamped in the setup between the upper and lower die, and heated by means of an electric heater. The entire setup, i.e. comprising the punch and the die, was thermally insulated. More details on the geometry of the setup including the schematic of the setup can be found in the work of Altstadt et al. [98] . For higher temperatures ( ≥ 400 • C), a ceramic ball (Al 2 O 3 ) was used because of its high thermal stability. The force was recorded with the help of a load cell having a resolution of ± 5 N, and the displacement of the punch was recorded via an inductive sensor with a resolution of ± 1 μm. The displacement of the punch was corrected for the machine compliance to retrieve the displacement of the sample. For both microstructural states, the specimens were punched along the height direction (HD) with a loading rate of 0.5 mm/min and temperatures between room temperature and 650 • C, with one test per temperature. Moreover, the temperature was recorded by a thermocouple placed under the specimen (i.e. within the lower die).

A2. Test results
The microstructural state dependent force-displacement curves obtained from the small punch tests at different testing temperatures are shown in fig. 10 . The force-displacement curves for the initial state at temperatures ≤ 550 • C ( fig. 10 a) exhibit significant pop-in events, characterized by sudden load drops. Also, the amplitude of the pop-ins is higher for the lower temperature regime ( ≤ 350 • C) and decreases as the testing temperature increases. Similarly, for the recrystallized state in fig. 10 b, the low temperature ( < 350 • C) force-displacement curves also display a significant amount of pop-in events with a relatively high amplitude, whereas at higher temperatures ( ≥ 350 • C), the amount of pop-ins tend to reduce, with nearly smooth curves for temperatures > 400 • C. An exception to the above observation is the presence of small popin events in the force-displacement curve measured at 600 • C. Furthermore, independent of the microstructural state, the presence of pop-in events in the force-displacement curves at lower temperatures confirms the brittleness of W, as these pop-ins are caused by unstable cracking events such as the initiation of a crack and sudden growth. Note that, during testing, the pop-ins were associated with intense acoustic emissions. Moreover, for both microstructural states, the initial part of the force-displacement curves at higher temperatures show a change in slope, with overall a smaller force at equivalent deformation as compared to lower temperatures. This again indicates a decreasing yield stress, which accompanies the increasing ductility of W at increasing temperatures. Nonetheless, upon comparing the microstructural states, i.e. fig. 10 a and fig. 10 b, it can be observed that the overall force required for the initial state at lower temperatures is higher as compared to the recrystallized state. Also, irrespective of the testing temperature, the amount of pop-ins for the initial state tends to be larger as compared to the recrystallized state.   11. EBSD based local and grain scale orientation analysis and comparison between the initial and recrystallized microstructural state represented by means of (a) KAM distribution, (b) areal GOS, (c) grain based orientation spread relative to grain diameter, (d) correlation between the grain aspect ratio and orientation spread; The edge grains are not considered here, given the incomplete information of their morphology.

Appendix B. Supplementary EBSD analysis
In fig. 11 a, the extent of short range orientation gradients between the two microstructures is accurately depicted by the KAM distribution, where 48% of the total pixels for the initial state have a misorientation greater than 1 • as compared to that of 8% for the recrystallized state. The areal GOS distribution is shown in fig. 11 b and clearly reveals that the majority of the grains in recrystallized state have a spread less than 3 • , while the majority of the grains in the initial state have a spread value exceeding 3 • . Additional analysis of the microstructure dependent orientation spread and grain morphology is shown in fig. 11 c and fig. 11 d. In fig. 11 c, the orientation spread of each grain (for both microstructures) is plotted against its diameter along with the recrystallization threshold spread (black dotted line). This plot reveals that for the initial state, larger grains always exhibit an orientation spread greater than 3 • due to the deformation assisted sub-grain formation, whereas in case of the recrystallized state, irrespective of the grain diameter, the orientation spread is almost below the threshold spread value. Additionally, as the rolling process changes the morphology of the grains to more elongated ones, fig. 11 d provides insight into the aspect ratio of grains with respect to their orientation spread. The grains in the recrystallized state tend to have a lower aspect ratio with a lower orientation spread. On other hand, the grains in the initial state show a relatively high aspect ratio along with a higher orientation spread.