Effects of neutron irradiation on the brittle to ductile transition in single crystal tungsten

Abstract Only limited data exist on the effect of neutron irradiation on the brittle to ductile transition (BDT) in tungsten. This work investigates the increase in brittle to ductile transition temperature (BDTT) following neutron irradiation to 1.67 displacements per atom, using four-point bend tests over a range of temperatures (623–1173 K) and strain rates (3.5 × 10−7 - 2.5 × 10−5 s−1). The BDTT was found to increase by 500  K after irradiation. The activation energy for the BDT was determined using Arrhenius analysis of the four-point bend tests. Nanoindentation strain-rate jump tests were used to characterise the activation volume for dislocation motion. These were quantified as 1.05 eV and 4.6 b3 respectively, very close to values found for unirradiated tungsten. This suggests that kink-pair formation is the controlling mechanism for the BDT before and after irradiation. This work also carries out a unique verification of inventory-code-modelling (via FISPACT-II) of transmutation of tungsten to rhenium and osmium under neutron irradiation using two independent techniques (X-ray and gamma-ray spectroscopy). These results show that modelling can correctly predict this transmutation, provided that an accurate neutron spectrum is used. This is a critical result given the widespread use of inventory codes such as FISPACT-II, and the associated nuclear data libraries, for modelling transmutation of tungsten.


a b s t r a c t
Only limited data exist on the effect of neutron irradiation on the brittle to ductile transition (BDT) in tungsten. This work investigates the increase in brittle to ductile transition temperature (BDTT) following neutron irradiation to 1.67 displacements per atom, using four-point bend tests over a range of temperatures (623e1173 K) and strain rates (3.5 Â 10 À7 -2.5 Â 10 À5 s À1 ). The BDTT was found to increase by 500 K after irradiation. The activation energy for the BDT was determined using Arrhenius analysis of the four-point bend tests. Nanoindentation strain-rate jump tests were used to characterise the activation volume for dislocation motion. These were quantified as 1.05 eV and 4.6 b 3 respectively, very close to values found for unirradiated tungsten. This suggests that kink-pair formation is the controlling mechanism for the BDT before and after irradiation. This work also carries out a unique verification of inventory-code-modelling (via FISPACT-II) of transmutation of tungsten to rhenium and osmium under neutron irradiation using two independent techniques (X-ray and gamma-ray spectroscopy). These results show that modelling can correctly predict this transmutation, provided that an accurate neutron spectrum is used. This is a critical result given the widespread use of inventory codes such as FISPACT-II, and the associated nuclear data libraries, for modelling transmutation of tungsten.

Introduction
Tungsten is the leading candidate as the plasma-facing material for first-wall and divertor applications in future nuclear fusion power plants, due to its high melting point (3695 K), low sputtering rates and good thermal conductivity [1]. However, there are concerns over its mechanical properties, in particular its low formability and high brittle to ductile transition temperature (BDTT) [2]. Work is ongoing for developing manufacturing routes for tungsten components such as cold-working [3] and tungsten composites [4] to improve ductility. These approaches have shown increasing promise in recent years [5,6].
There have been several studies of the fracture properties of tungsten [7,8] including its brittle to ductile transition (BDT), however there has been comparatively little work on the impact on neutron irradiation on the BDT. The BDT will be critical for determining the minimum operating temperature and lifetime of tungsten within a fusion power plant [9], where tungsten components may be exposed to >20 dpa during operation [10] at temperatures above 1073 K [9]. Even during the planned 14 years of ITER campaigns, tungsten components will be exposed to up to 0.5 displacements per atom (dpa) of irradiation damage [11]; the impact of this on the ductility of tungsten is not yet known. Any decrease in ductility as a result of irradiation may influence both operation of the device and end-of-life handling.

Brittle to ductile transition in unirradiated single crystal tungsten
The BDT in tungsten is controlled by the capability of dislocation motion to shield the crack tip. Below the BDTT, dislocation motion is limited and as a result the local stress can exceed the cohesive stress resulting in brittle fracture [12]. Above the BDTT, semi-brittle or ductile behaviour is observed with plastic deformation occurring before fracture.
Dislocation motion near a crack tip can be limited either by dislocation nucleation or dislocation mobility. Gumbsch et al. [13,14] showed that the BDT in single crystal tungsten was controlled by dislocation mobility around crack tips, rather than dislocation nucleation at crack tips. In this work the BDTT was measured as 370e470 K and the activation energy for the f110g110 crack system was found to be 0.2 eV. This low activation energy was a consequence of the specific crystal orientation used during these experiments, favouring control of the BDT by the rapid motion of pure edge dislocations. The same controlling mechanism has been observed in ultra-fine grain tungsten [15].
Giannattasio and Roberts [16] investigated the f100g001 crack system and determined a BDTT range of 390e500 K (strain rate dependent) and an activation of 1.0 eV. In this work the same BDTT range and activation energy was found for pure polycrystalline tungsten, showing that this crack system is more representative of bulk tungsten.
Dislocation-dynamics modelling by Tarleton et al. [17] indicated that in the general case the BDT in tungsten is controlled by kinkpair mediated glide of 1/2<111> screw dislocations. This modelling showed that the activation energy for the BDT observed by Giannattasio and Roberts (1.0 eV) is the result of the kink-pair nucleation energy of 1.75 eV being reduced by the internal resolved shear stresses in the crack-tip dislocation arrays (326 MPa with an inferred activation volume of 20 b 3 , where b is the magnitude of the Burgers vector).
An alternative interpretation has been provided by recent analysis by Swinburne and Dudarev [18] that predicted a singlekink formation energy is the characteristic activation energy for dislocation motion in microstructures with an obstacle spacing above a well-defined threshold value that is a function of temperature and stress. This threshold distance was shown to range from 10 À2 e10 1 mm for bcc iron (which shows similar BDT behaviour to that of tungsten) at realistic stresses and temperatures.
The material investigated by Giannattasio and Roberts was wellannealed single crystal. This suggests that the obstacle spacing should be above the threshold distance and so, according to Swinburne and Dudarev, a single kink energy close to 1 eV should be observed. This analysis agrees with the activation energy measured experimentally by Giannattasio and Roberts [16], if the energy contribution from stress acting on the dislocation is considered negligible. The relevance of this interpretation to this work is discussed further in section 1.4.

Effect of radiation damage on tungsten
One of the critical outstanding challenges for fusion materials is understanding the effects of neutron radiation on tungsten [5,19]. Within a fusion environment tungsten will be exposed to a high flux (up to 7 Â 10 18 m À2 s À1 ) of high energy neutrons (up to 14 MeV) in addition to extreme heat fluxes (with peak loads of 10e20 MW m À2 s À1 [20]) and implantation of hydrogen and helium ash ejected from the plasma. Interaction of the neutrons with tungsten nuclei causes transmutation of tungsten (predominantly to rhenium, osmium and tantalum [21]) and the kinetic energy of the neutrons causes collision cascades, resulting in the formation of excess vacancies and interstitials [22]. This excess of point defects results in the formation of dislocation loops and voids [23], and facilitates radiation enhanced segregation resulting in W-Re-Os precipitate formation [24,25]. This damage has been shown to result in significant hardening [26].
Due to the formation of W-Re-Os precipitates, accurate calculation of irradiation-induced transmutation is critical for understanding changes in properties under neutron irradiation [27]. The transmutation level is commonly calculated using FISPACT-II or similar inventory codes, but it is difficult and rare to obtain direct experimental validation of these calculations from neutron irradiation of tungsten (experimental validation of inventory calculations is most commonly done using experimental measurements of quantities derived from the nuclide inventory e e.g. from decay heat measurements [28]). This work will uniquely employ X-ray and gamma-ray spectroscopy in order to verify FISPACT-II modelling results. Table 1 summarises the available data for the effects of neutron irradiation on fracture properties in tungsten. All the experiments were carried out on well-annealed pure tungsten so the BDTT before irradiation would be expected to be around 400e500 K as established by Giannattasio and Roberts [16]. The tests were all carried out using tensile testing and the inferred BDTT was taken as the lowest ductile (>5% strain before fracture) test temperature. The results show a decrease in ductility and a significant increase in the BDTT with irradiation [29]. However, no clear dose dependence could be established and the impact of neutron irradiation would be expected to vary with irradiation temperature [30].

The effect of irradiation on the BDT in tungsten
These data are rather limited in several ways: firstly, all the irradiation temperatures are significantly below those expected in a fusion environment (>1073 K) [33]; secondly, only neutron fluences were considered and damage estimates in dpa are not possible from the data provided; finally, there is a lack of analysis of the BDT, only very approximate BDTT values could be inferred from the data and there is no analysis of the strain rate dependence. Only Steichen [29] carried out mechanical testing at multiple strain rates and the BDTT cannot be identified at most strain rates from the data in this work.
It is generally assumed that tungsten will become more brittle after even low dose irradiation [34]. While a small degree of embrittlement might not affect the operational lifetime of ITER, DEMO and future power reactors will have to operate reliably for much longer periods (decades) and to much higher radiation doses [35]. For example, the "second" tritium breeder blanket of DEMO is required to last for almost 15 years and withstand up to 50 dpa [36] (in steel; the value for tungsten would still be of the order of 30 dpa). A greater understanding of the impact of neutron irradiation on the BDT is required if tungsten is to be used as a plasma-facing material.
The experiments described in this paper are the first to characterise fully the increase in BDT following neutron irradiation and its underlying mechanisms.

Modelling the irradiation induced increase in BDTT
As discussed above, irradiation damage will introduce voids, precipitates and dislocations loops in tungsten. These will act as obstacles to dislocation motion. The model by Swinburne and Dudarev predicts that if this obstacle spacing decreases sufficiently then a doubling of the BDTT would be expected following irradiation (from around 400 Ke800 K), along with a doubling of the activation energy from a single to double kink formation.
An alternative explanation for the increase in BDTT in Fe-2.5Cr (also BCC) was proposed by Yi and Robertson [37] based upon dislocation dynamics modelling. Under this model, cross-slip of dislocations is required to unpin dislocations from irradiationinduced obstacles. This significantly reduces the effective dislocation mobility. Since dislocation mobility is temperature dependent this decrease in dislocation mobility can be described as a defect induced apparent temperature shift (DDIAT). DDIAT was shown to match the change in BDTT for several experiments on FeeCr.
Similar to the Swinburne and Dudarev model the BDTT would be expected to increase with irradiation induced defect density. However, the activation energy for the BDT in tungsten would not be expected to change following irradiation, as thermally activated slip of 1/2<111> screw dislocations (controlled by double kink formation) remains the controlling mechanism.
The work in this paper aims to establish the activation energy of the BDT after irradiation and so will provide evidence to distinguish between these models.

Sample preparation
Single crystals of commercially pure W were sourced from Metal Crystals and Oxides Limited, Cambridge UK. Frost these, samples (1 Â 1 Â 12 mm bars) were prepared for four-point bend tests, as described in previous work [16], with sharp pre-cracks of 60 mm depth and a {100} <001> crack system. The cracks produced by this preparation technique were characterised by Murphy et al. [38]. The samples were irradiated in the High Flux Reactor at Petten under Extremat II [39] (as described by Klimenkov et al. [40]). The irradiation took place at 900 C and lasted for 8 cycles (208 full power days) in positions C7 and C3, in Extremat II drums 3 and 4. The neutron flux was 6.8 Â 10 18 m À2 s À1 (3.6 Â 10 18 m À2 s À1 , E > 0.1 MeV) for the first location (148 full power days) and 6.6 Â 10 18 m À2 s À1 (3.4 Â 10 18 m À2 s À1 , E > 0.1 MeV) for the second location (60 full power days), giving a total fluence of 1.21 Â 10 26 m À2 (6.5 Â 10 25 m À2 , E > 0.1 MeV).

Damage and transmutation
Modelling of the irradiation was carried out using the FISPACT-II inventory code [41,42], which simulated the change in composition (transmutation) and predicted the damage (dpa) levels within the sample during the 208 days of exposure [40,43]. This was carried out using a calculated spectrum, provided by NRG, Petten [44], which considered nearby thermal neutron absorbing experiments. The calculated transmutation level was compared to that measured using energy dispersive X-ray spectroscopy (EDS) using a Mira3 XMH scanning electron microscope with an Oxford Instruments X-Max 80 EDS detector at the Materials Research Facility at CCFE. The EDS spectra were taken from a neutron irradiated sample and a standard W -5 wt% Re sample. Fig. 1a shows the recorded raw EDS spectrum. The rhenium and tungsten La peak sizes were compared to calculate the level of rhenium present in the neutron irradiated samples. Fig. 1b shows the fitted Gaussians to the tungsten and rhenium peaks after subtraction of the X-ray background from the raw data.
Damage calculations (obtained from FISPACT-II using a threshold displacement energy E d ¼ 55 eV [45]) indicated that the average across the samples was 1.67 dpa, at a dose rate of 9.3 Â 10 À8 dpa/s. The calculated final composition of the samples was 98.5 wt% W, 1.4 wt% Re and 0.1 wt% Os. The EDS analysis indicated a composition of 1.2 ± 0.1 wt% Re and 0.1 wt% Os, in good agreement with the FISPACT-II results. No segregation of Re or Os could be detected using SEM techniques; however, previously published analysis of identically irradiated W samples using EDS in the TEM has shown the presence of Re precipitates following the irradiation [40]. Note that the FISPACT-II calculations do not currently propagate the inherent nuclear data (TENDL-2017 [46]) uncertainties to transmutation rates, although the typical % errors on the reactionrates of the key neutron capture and multiplication reactions (see Ref. [43]) are 5e10%.
An additional comparison between the inventory simulations and the experiment was obtained via two gamma spectroscopy measurements of the samples carried out in the ADRIANA laboratory at CCFE using the Canberra Broad Energy Germanium (BEGe) and Small Anode Germanium (SAGe) well-type high purity germanium detectors [47]. Canberra's Genie 2000 software [48], was used to perform photopeak analysis on two gamma lines; one from 186 Re at 137 keV and another from 186m Re at 99 keV. Fig. 1c shows the gamma spectrum measured using the SAGe detector approximately 9 years after the end of the irradiation at Petten. The sample geometry was explicitly modelled for each measurement, and validated MCNP [49] models of the detectors were used to calculate the photopeak efficiencies for each gamma line studied. A 15 cm mounted detection geometry was used with an acquisition time of 474 h on the BEGe detector, and an in-well geometry was used on the SAGe detector, with an acquisition time of 235 h. Specific activities for identified nuclides were then calculated and compared to results from FISPACT II modelling. The activity measured for the primary detected isotope, 186 Re, via the 137 keV peak, was 187 ± 2 Bq g À1 (the average of the two detector measurements, with uncertainties based on counting statistics). Measurement of the 99 keV peak for 186m Re was only possible in the low-background, high-geometric efficiency conditions of the SAGe detector; the half-life of 186m Re is 2 Â 10 5 years compared to only 3.7 days for 186 Re, and so high detection efficiency and low signal-to-noise ratio are required to produce a meaningful result. The direct 186m Re activity was found to be 200 ± 9 Bq g À1 .
The decay-corrected (to account for the~9 years of cooling) activity calculated by FISPACT-II for both 186 Re and 186m Re was 95 Bq g À1 , with a nuclear data uncertainty of around 7% (this does not include uncertainties associated with the MCNP modelling or the experiment). Note that after such a long period of cooling the two nuclides are in secular equilibrium since all the original (created under irradiation) 186 Re has long since decayed away and only the residual activity feeding from 186m Re remains e hence they have identical activity, as predicted in FISPACT-II simulations. Fig. 1d shows the nuclear reaction cross section (xs) data associated with the neutron capture cross sections to produce 186 isotopes from 185 Re (one of the main stable isotopes produced via transmutation of W). The figure shows that the available experimental data (from EXFOR) is well captured by the reaction associated with the production of 186 Re. However, this is not important in this case because only the 186m Re produced during irradiation remains. Clearly, the cross section associated with producing this metastable radioisotope (the blue curve in Fig. 1d) is more uncertain e there is very little experimental data measuring this minor channel, which is understandable given the difficult-to-measure long half-life of this nuclide and the fact that more than 99% of the neutron capture (n,g) on 185 Re is associated with 186 Re at sub-keV neutron energies. As the cumulative reaction rate (RR) plot in Fig. 1e shows, the majority of the 186 production is governed by the giant resonance in the cross section at around 2 eV (around 70% of the total RR comes from neutron at this narrow energy). Thus, the results are extremely sensitive to the precise attribution of the cross section in this resonance to the two 186 radioisotopes, but nonetheless the agreements (within a factor of 2) is reasonable given the calculation uncertainties.

Four-point bend tests
The BDT was characterised using four-point bend tests carried out using a molybdenum rig within a vacuum furnace in the Fusion Materials Laboratory at Karlsruhe Institute of Technology (KIT). The temperature was monitored throughout the tests using thermocouples placed close to the samples. Surface tensile stresses s were calculated using equation (1): where P is the applied load, L and a are the outer and inner load spans (10 mm and 5 mm respectively), and w and t are the width and thickness of the beam (both 1 mm). The geometry of the samples is shown in Fig. 2. for production of Re186 and its metastable state Re186 m. The data is compared to the available experimental measurements from the EXFOR international database and the simulated HFR flux spectrum experienced by the samples is shown for reference using the right-hand y-axis scale. "per leth." or per lethargy refers to the standard practice of dividing the flux by the logarithm of the energy bin width to remove bias caused when bins have large variation. (e) shows the cumulative reaction rate (RR) of the two neutron capture reactions under the HFR spectrum. See main text for more details. Surface strain rates were calculated from loading rates in the elastic portion of the loading curve, assuming an elastic modulus of 400 GPa. The tests were carried out under cross-head displacement control at four constant displacement rates, shown in Table 2.
Samples that showed >10% strain before fracture were classified as ductile. This was used instead of a fracture toughness criterion as the limited number of samples available meant that accurate fracture toughness values could not be determined. The fracture surfaces from the four-point bend tests were characterised using SEM at KIT. This allowed confirmation of a sample as either brittle to ductile, giving further confidence in the transition temperature for each strain rate, and allowed comparison of the fracture surfaces from irradiated and unirradiated samples.
The results from four-point bend tests at each strain rate are shown in Fig. 3. At each strain rate, with increasing temperature, a distinct transition from brittle to ductile behaviour was observed. The brittle-ductile transition temperatures (BDTT), taken to be the lowest ductile temperature for each strain rate, are shown in Table 2; BDTT increased with increasing strain rate. The largest uncertainty in the measurements was the temperature range between testing temperatures and so the uncertainty in the BDTT has been taken as the gap between the highest brittle and lowest ductile test temperatures. Fig. 4 compares the BDTT values from these experiments on irradiated tungsten single crystals with those found for unirradiated W single crystal tested using the same method [16]. Where data are available at a comparable strain rate (z7 Â 10 À6 s À1 ), neutron irradiation to 1.67 dpa has caused the BDTT to increase by z 500 K.
For unirradiated tungsten, the transition temperature (T BDT ) e strain rate ð_ εÞ relationship was found to follow an Arrhenius relation: equation (2), where A is a constant and k is the Boltzmann constant; the associated activation energy, E BDT , was found to be 1.0 ± 0.05 eV [16,50].
The variation of BDTT with strain rate for irradiated tungsten was also found to follow this relation (Fig. 5), with an activation energy of 1.05 ± 0.3 eV, similar to the value for unirradiated tungsten.
Figs. 6 and 7 show SEM micrographs of fracture surfaces are shown for unirradiated [16] and irradiated samples. In both specimen types, at temperatures far below the BDT, fracture is by cleavage on the {001} plane, with little visible deformation, whereas at temperatures close to the BDT rough fracture surfaces with characteristic river lines are observed. Above the BDT, fracture surfaces show significant macroscale deformation; here fracture occurred at strains greater than 10%. The notch and pre-crack are labelled in Fig. 6b and were measured as having depths of 25 mm and 35 mm respectively, agreeing with the values measured during the testing of unirradiated material.

Nanoindentation: strain-rate jumps
Nanoindentation of both unirradiated and neutron irradiated tungsten was carried out with a Berkovich tip using a Nanoindenter XP (MTS Systems Corporation, Oak Ridge, TN) within the Materials Research Facility at Culham Centre for Fusion Energy. The continuous stiffness measurement technique was used with an oscillation of 2 nm at 45 Hz. Indents were made to a depth of 2000 nm at a target strain rate of 0.05 s À1 . Hardness values (averaged over the indentation depth range 800e1800 nm) were 5.0 ± 0.1 GPa for unirradiated tungsten and 7.2 ± 0.1 GPa for irradiated tungsten.
The indentation strain rate sensitivity was measured using the method described by Maier et al. [51]. Four different strain rates were used between 0.01 and 0.1 s À1 at indentation depths between 1200 and 2400 nm. A typical hardness response curve is shown in Fig. 8. Fig. 9 shows hardness results from strain-rate jump tests. These results were analysed using equations (3) and (4) Table 3. The activation volume for unirradiated tungsten of 5.6 b 3 agrees well with results from other nanoindentation experiments [52]. The activation volume does not appear to have changed significantly following irradiation.

Transmutation
There is good agreement between the measured rhenium content from EDS, and the level predicted from FISPACT-II inventory modelling. This provides evidence that, with an accurate spectrum, computational modelling can be reliably used to predict transmutation. However, it is essential that an accurate spectrum is used. The initial modelling for this work used an average spectrum for the reactor, resulting in a large error in estimation of transmutation [43]. Only once a location specific spectrum was used, was good agreement between experiment and modelling achieved.
There is a factor of two difference between the simulationpredicted activity of 186 Re/ 186m Re and that measured by gammaray spectroscopy. While this at first appears concerning in light of the excellent agreement with the simulated and measured total Re production, there are a number of possible explanations. Firstly, the concentrations of these 186 radioisotopes in comparison to the main Re transmutation products are very low; for example, around 10 5 times more atoms of 187 Re are predicted compared to the slowly decaying 186m Re. Thus, a discrepancy between the model and the gamma spectroscopy result does not necessarily correlate to a significant error in overall transmutation. Secondly, and perhaps more importantly, there is large uncertainty in the nuclear reaction data to produce 186m Re, which is the nuclide being measured (regardless of production rate for the short-lived 186 Re). As Fig. 1d and e showed, there is considerable reason to doubt the amount of 186m Re predicted by the simulations and thus a factor of two is not unreasonable. Fig. 3. Four-point bending stress-strain curves for neutron-irradiated single crystal W at a range of temperatures and strain rates. The transition from brittle to ductile behaviour with increasing temperature can be clearly observed at each strain rate. Fig. 4. Comparison of BDTT at a range of strain rates for unirradiated W [7] and neutron irradiated W. Error bars were calculated as the temperature gap between highest brittle test temperature and lowest ductile test temperature.

Irradiation hardening
The results from nanoindentation in this experiment are compared to the previous results from neutron irradiation of single crystal tungsten at similar temperatures in Table 4. Similar irradiation hardening was observed in this experiment as in previous experiments. The irradiation hardening measured in this work is slightly lower than that measured following irradiation to 0.7 dpa in HFIR. This is most likely caused by differences in irradiation spectrum, which have previously been shown to have a strong impact on the hardening observed in tungsten [53]. This makes measurements and calculations of transmutation as carried out in   6. SEM fractography showing (a) Low ductility observed in unirradiated sample fracture at 77 K with a strain rate of 7 x 10 À4 s À1 [16], (b) Low ductility observed in neutron irradiated sample fracture at 473 K with a strain rate of 7 x 10 À7 s À1 , (c) River lines observed in sample fractured at 478 K with a strain rate of 7 x 10 À4 s À1 (close to BDT) [16], and (d) River lines observed in sample fractured at 723 K with a strain rate of 3.5 x 10 À7 s À1 (close to BDT). this work critical for analysing mechanical properties after neutron irradiation.
The neutron spectrum in HFIR has a strong thermal component and so causes significantly more transmutation (initial rate of 3.4 % Re dpa À1 ) than that characterised in this work (0.8 %Re dpa À1 ). For comparison transmutation in DEMO is predicted at a rate of 0.1 %Re dpa À1 [54]. As a result a greater density of rhenium precipitates is expected in the material analysed by Fukuda et al. [55] causing greater hardening. It appears that any solution softening from rhenium, as reported for tungsten-rhenium alloys [56], is lost due to precipitate formation under neutron irradiation.
The high temperature of this irradiation may also be affecting the extent of irradiation hardening. Hasegawa et al. [27] showed that temperatures above 700 K enable precipitate and void formation in tungsten under neutron irradiation. This has been backed up by TEM studies of annealing radiation damage in tungsten, which have shown that above 1073 K interstitial dislocation loops grow rapidly and vacancy clusters form voids [57]. Given that the results compared below are all close to or above this temperature, it is likely that the differences in irradiation hardening are caused by the variation in dose and neutron spectrum.
Results from ion irradiation of tungsten from Armstrong et al. [58] are also included for comparison in Table 4. The material investigated in that work was composed of large grains (>50 mm) and was analysed using nanoindentation. As a result, this material is largely equivalent to single crystal. It is clear from the comparison of these results that neutron irradiation has caused significantly greater hardening than ion irradiation. This effect has been highlighted previously [19] and is likely to be associated with the lack of transmutation from ion irradiation.
However, ion irradiation of W À 5Re to 1.2 dpa at 573 K produced only 0.85 GPa of irradiation hardening [58]. That work showed that rhenium clusters were not formed until 33 dpa under ion irradiation, whereas they are commonly observed under neutron irradiation below 1 dpa [19]. This difference in irradiation effects requires further investigation if ion irradiation is to be used to replicate neutron damage in tungsten.

Brittle to ductile transition
The brittle to ductile transition temperature for single crystal tungsten increased by around 400 K after irradiation to 1.67 dpa. This result correlates well with large increases in hardness observed in this work and elsewhere after neutron irradiation. It also agrees with the limited results available from previous experiments as presented in Table 1. This has significant connotations for future reactor designs as such a large increase in BDTT will limit the operating temperature window for tungsten. Further investigations into the BDTT increase in polycrystalline tungsten are required to understand the full impact of neutron irradiation on the use of tungsten in fusion reactors.
Neutron irradiated BDTT data are only available for a small range of conditions for BCC metals. The largest experimentally observed increase in BDTT for each metal as a proportion of meting point are compared in Fig. 10. These results show that generally unirradiated BCC metals have a BDTT of around 0.1e0.15 T m that increases to a  In the experiments report here on initially pure tungsten, neither the activation energy for the BDT, nor the activation volume for dislocation motion during indentation appear to have changed significantly following neutron irradiation. This strongly suggests that the activation mechanism for dislocation mobility has not changed following neutron irradiation. This is despite the presence of voids (diameter z 5 nm, number density z 2.5 Â 10 21 m À3 ) and precipitates (diameter z 3e15 nm, number density z 5 Â 10 21 m À3 ) observed within the microstructure by Klimenkov et al. [40].
It has been shown elsewhere that adding rhenium to tungsten lowers the BDTT, even with rhenium contents as low as 1.9% [62]. This effect has clearly not been seen here following transmutation to 1.4 wt% Re. The work by Klimenkov et al. [40] showed that a high density of rhenium precipitates had formed. However, a volume fraction could not be calculated so it is unknown how much rhenium remains in solution. It is clear that any decrease in BDTT from rhenium in solution has been overwhelmed by other changes in the microstructure during irradiation.
The approximate doubling of the BDTT from z450 K to z900 K (depending on strain rate) following neutron irradiation initially appears to agree with the prediction from the model proposed by Swinburne and Dudarev [11]. However, an increase in activation energy from a single kink to double kink energy would be expected and this has not been observed experimentally. Additionally, the activation volume for dislocation motion has remained constant following irradiation, suggesting that the kink mechanism remains the same.
An alternative explanation is that a double kink activation mechanism is acting in the unirradiated and irradiated material. This agrees with the dislocation dynamics modelling carried out by Tarleton and Roberts [17], with the effect of stress acting on the activation volume of the dislocation accounting for the difference between the observed activation energy (1.05 eV) and the theoretical value of 1.75 eV.
In this case, an explanation for the large increase in BDTT without an increase in activation energy is required. The TEM evidence published by Klimenkov et al. [40] shows that a high density of voids and precipitates are present. These defects would significantly reduce the effective dislocation velocity, according to the model suggested by Yi and Robertson [37]. This would result in an increased BDTT, while retaining the kink-pair activation energy. However, further modelling is required to understand whether this mechanism can explain the large increase in BDTT observed in this  work.
Understanding the underlying BDT mechanism is critical for predicting how the BDTT will change under irradiated to greater dpa. Data from irradiation of Eurofer-97 shows that the increase in BDTT does not plateau until >10 dpa [59]. This suggests that under fusion conditions, the BDTT of tungsten could increase beyond the values measured in this work, further limiting the operating window for tungsten in a fusion reactor.

Conclusions
The level of transmutation in tungsten after neutron irradiation to 1.7 dpa was measured at 1.2 wt% Re, in good agreement with FISPACT-II models, which also matched gamma spectroscopy results to a reasonable degree of accuracy. This is a useful and rare validation of inventory simulations with FISPACT-II and gives confidence in a computational approach that is used widely for analysing transmutation in fission irradiation experiments and for predicting the behaviour of tungsten in future fusion devices.
Four-point bend tests over a range of temperatures and strain rates show that neutron irradiation to 1.7 dpa increases the BDTT relative to that for unirradiated tungsten by approximately 500 K at a strain rate of 7 Â 10 À6 s À1 . This is a critical result for the design of future fusion devices.
Arrhenius analysis of these results shows that the observed activation energy associated with the BDTT has not changed significantly due to irradiation, remaining close to 1.0 eV. This has been identified as the activation energy for motion of 1/2<111> screw dislocations controlling the BDT. Nanoindentation strain-rate jump tests show that the activation volume for dislocation motion has not changed significantly following neutron irradiation.
These results strongly suggest that the controlling mechanism for the BDT in tungsten (the mobility of dislocations near the crack tip) has not changed following neutron irradiation, despite the presence of very fine cavities and precipitates observed by TEM [40]. These results disagree with the latest modelling of the changes in brittle to ductile transition following neutron irradiation [18], which predict a doubling of the activation energy. Instead, a decrease in effective dislocation velocity due to the increased obstacle density is a possible explanation for the large increase in BDTT but further modelling is required.

Data availability
The raw data required to reproduce these findings are available to download from https://doi.org/10.17632/xdftk2np78. The processed data required to reproduce these findings are available to download from https://doi.org/10.17632/8xxw6xtfgr.