Radiation induced hardening of beryllium during low temperature He implantation

The effect of ion irradiation on evolution of microstructure and hardening of beryllium with different impurity levels was investigated using TEM and nanoindentation. High purity S-65 nuclear beryllium grade and lower-pure S-200-F structural beryllium grade were implanted by helium ions at temperatures of 50oC and 200oC. 11 different energies were used, so as to create a quasi-homogeneous 3μm irradiated layer with average radiation damage of 0.1dpa and average He content of 2000 appm. At both temperatures in both grades, under TEM investigation the radiation damage appears as “black dots” (<10 nm in diameter) which are likely to be small dislocation loops with the number density of ~ 10 m. No bubbles were observed inside grains and at grain boundaries. Nanoindentation experiments demonstrated that the lower-purity S-200-F grade has higher average hardness (3.7±0.8 GPa) than the S-65 grade (3.4±0.8 GPa) in the as-received and irradiated states. After helium implantation of both grades the hardness increased by 60% for the 200oC irradiation and 100% for the 50oC irradiation. The higher purity S-65 grade showed smaller changes in hardness at 200oC than the less pure S-200-F. Use of EBSD to give the crystallographic orientation of indented grains revealed that in both grades in the as-received materials the hardness is about 2.5 times higher when the indentation direction is close to the [0001] c-axis of beryllium compared to indentation perpendicular to [0001]. Hardness anisotropy significantly decreased after irradiation: the “soft orientation” was most sensitive to radiation-induced hardening, with hardness increasing by about 140% after irradiation at 50oC and 100% after irradiation at 200oC, compared to about 15 20% for the “hard” orientation at both irradiation temperatures. Analysis of the possible hardening contribution demonstrated that the “back dots” should lead to about 0.85 and 1.7 GPa hardness increase, while the rest of the hardening should originate from helium bubbles with the size below the TEM resolution (at or below 1.5 nm).


Introduction
Due to a unique combination of mechanical and physical properties, beryllium is extensively used in a wide variety of nuclear facilities. For example, due to its high thermal conductivity, low chemical sputtering and oxygen gettering properties [1], beryllium is a plasma facing material (PFM) for the Joint European Torus (JET) tokomak [2,3] and the future ITER fusion test reactor [4,5]. Beryllium is also being considered as a neutron multiplier material in the helium-cooled pebble bed (HCPB) design concept for the tritium-breeding blanket of the demonstration fusion power plant (DEMO) [6,7]. Due to its low nuclear interaction cross-section, beryllium is an excellent material for particle-beam windows, and, for example, was successfully used as a primary beam window in neutrino production targets of the "Neutrinos at the Main Injector" (NuMI) beamline [8]. It is also being considered as a material for different target components in a new generation of proton accelerator driven particle sources such as, for example, the Long Baseline Neutrino Facility (LBNF) [9,10].
In comparison to other nuclear materials, beryllium experiences extremely high transmutant helium accumulation rate. For example, it is between 110 and 220 appm/dpa in the beryllium reflector of the ISIS neutron source (RAL, UK) and can reach 380 He-appm/dpa under neutron irradiation in fission reactors [11][12][13]. For comparison, a typical appm/dpa ratio for iron under neutron irradiation in fission pressurized-water reactor (PWR) is 0.35 appm/dpa [14] and for tungsten under neutron irradiation in fission high flux isotope reactor (HFIR) is 0.0008 appm/dpa [15]. In fusion reactors, beryllium will accumulate helium with the rate of 670 He-appm/dpa [14] that will lead to build-up of 1,500 He-appm at ITER end-of-life [16] and more than 15,000 He-appm at DEMO end-of-life (HCPB concept) [17]. Even higher helium accumulation, up to 4000 He-appm/dpa, takes place in beryllium in high energy proton irradiation environments, for example in neutrino sources [18]. Both helium accumulation and displacement damage effects in beryllium are known to be highly irradiationtemperature dependent: for temperatures less than 400ºC, radiation induced hardening accompanied with a drop in plasticity is usually observed, whereas severe non-hardening embrittlement is observed for high-temperature irradiation [19][20][21][22]. It is generally considered that the determining mechanisms here are strengthening of the beryllium matrix by point defect clusters and helium accumulation at low temperatures; and helium segregation to grain boundaries at high temperatures [12,[22][23][24].
Industrial purity beryllium grades used in nuclear facilities contain a wide variety of trace impurities, typically oxygen, iron, aluminium, nickel, copper, silicon, carbon and magnesium. The majority of these impurities have very limited solubility in beryllium, and have strong tendencies to create precipitates and to segregate to defects such as grain boundaries or dislocations [18,25]. While the effects of different impurities on mechanical properties of beryllium have been extensively studied in the past (see reviews [7,25] and references sited), the mechanisms underlying these processes are not always well understood. Moreover, data about the effects of impurities on evolution of mechanical properties under radiation is very limited, especially under the extreme conditions relevant to beryllium components of fusion reactors and particle accelerators.
The work reported here compares hardening introduced by helium implantation in two industrial beryllium grades: the high purity nuclear grade S-65 and the lower-purity structural grade S-200-F. The S-65 grade is being used as a plasma-facing component in Joint European Torus fusion experiment [3] and was selected as the reference material for use in ITER [7]. The S-200-F grade is also used in nuclear facilities, for example for beryllium frames and reflector elements in the Japan Materials Testing Reactor (JMTR) at JAEA [26].
We report data on irradiation hardening effects in beryllium at 50ºC and 200ºC. The temperatures represent the lower boundary of the expected irradiation temperatures in Be for ITER first-wall applications, and typical irradiation temperatures of the beryllium in the operating NuMI neutrino target [18] and the currently designed LBNF [9,10]. Considering the literature data [19][20][21][22], for both temperatures the "radiation induced hardening" regime should dominate. Available experimental data on radiation damage effects in beryllium are mostly collected from materials irradiated by fission reactor neutrons. However, the applicability of the data for fusion reactor applications and accelerator driven particle sources is uncertain, since fission-spectrum neutrons do not induce sufficient levels of transmutation-produced helium [11][12][13]. In the current work, an ion beam accelerator was used to introduce radiation damage and about 2000 appm He, relevant to both fusion and accelerator applications and giving also the advantage of fast and relatively inexpensive materials irradiation without activating the sample.
Nanoindentation is widely used for evaluation of local mechanical behaviour of materials at micro-scale and is particularly useful for ion irradiated materials where only a shallow damaged layer is available. It has been successfully used for screening of radiation induced hardening in different nuclear materials, including: ferritic-martensitic [27][28][29], austenitic [30] and RPV [31] steels, ODS alloys [32] and tungsten [33,34]. No nanoindentation data on radiation hardening of beryllium is available in the literature; however the applicability of the technique to examine the effects of impurities on the hardness of beryllium was recently demonstrated [35].
Micro and macro-hardness examinations of beryllium often show significant scatter of indentation results [35,36] that may complicate the determination of radiation-induced hardening effects. This originates from the high elastic and plastic anisotropy of beryllium [25,37]: the room temperature (RT) critical resolved shear stresses for <a> basal slip (the primary slip system at RT) is 15 MPa, for the <a> prismatic systems is 68 MPa and the secondary <c + a> pyramidal systems is 2000 MPa [37,38].
Previous experiments [35,39,40] showed that indentation hardness coupled with crystallographic data from EBSD or XRD is a useful approach for comparing the effects of different treatments on beryllium.
This approach is used here, with special attention being paid to the mitigation of any artefacts originating from localised plastic deformation of the surface around the indents (so-called pile-ups and sink-ins) which can significantly affect interpretation of nanoindentation data in beryllium [35].
The paper is structured as follows. First, the investigated beryllium grades, irradiation conditions, the nanoindentation hardness experiments and data treatment approaches are described. Then, the microstructure of the implanted layers are shown and discussed. Before hardness values are calculated, indentation contact areas for different samples and different crystallographic orientation of grains are analysed, taking into account any pile-up and/or sink-in, and the tip-indented surface contact area coefficients are reported. The resulting correction parameters are then used to calculate the nanoindentation hardness of beryllium grains of different purity and different crystallography before and after irradiation. After, the load -indentation depth curves from different beryllium states (material grade, irradiation condition) are presented and the pop-in events (horizontal plateaux observed on the load-displacement at some at some critical loads) are characterised to highlight differences between different states of the materials. This is followed by a general discussion of the origins of the observed irradiation induced hardening and the applicability and significance of the obtained data for engineering design considerations for nuclear facilities with beryllium elements under irradiation.

Materials
Two vacuum-hot-pressed industrial beryllium grades were investigated in this work: a high purity nuclear grade S-65 (Be > 99.2%) [41] and the lower purity structural grade S-200-F (Be > 98.5%) [42]. Microstructures of both types are reported in Ref. [35]. The grades have similar average grain diameters of 7.2± 4.4 µm [35]. Our previous STEM/EDS studies of beryllium [18] showed that grain boundaries of both grades are decorated by numerous precipitates enriched by oxygen, aluminium, iron, silicon, magnesium, titanium and other impurities, while the grain cores are mainly free of precipitates.

Irradiation conditions
Before helium implantation, surfaces of samples were mechanically ground and polished with SiC paper, diamond paste and finally colloidal silica. To avoid sputtering of beryllium during implantation, a 1 µm layer of aluminium was deposited by sputter coating at the surface of samples. The aluminium layer was removed after implantation by polishing in colloidal silica.
Implantation was done at the National Ion Beam Centre (NIBC) at the University of Surrey.
Sequential irradiation with He + ions with eleven energies from 1.2 MeV down to 0.2 MeV with steps of 0.2 MeV was used to create approximately a 3 µm deep damaged layer, as calculated with SRIM (Stopping and Range of Ions in Matter) [43,44]. Totally, 8.810 20 (He)atoms/m 2 were implanted. The radiation dose calculation was done using a displacement threshold energy of 31eV [45] and the ''Quick'' Kinchin and Pease option. The calculated average damage is 0.1 dpa with an average dose rate of 1.710 -4 dpa/s. The average helium content in the damaged layer is 2000 appm; however there are wide oscillations around the average value (see Figure 1).

2.3.Experimental techniques
Indentations were conducted using a Keysight G200 nanoindentation system, with a diamond pyramid Berkovich tip, using the continuous stiffness measurement (CSM) technique [46] with a 42 Hz and 2 nm oscillation. The frame stiffness of the instrument and indenter tip was calibrated using a fused silica sample. An atomic force microscope Dimension D3100 was used for topographical investigation of the prints after indentation. The surface profiles were then obtained using the open source software Gwyddion [47]. These profiles were used to calculate the contact area between the indenter tip and the sample surface.
The nanoindentation hardness was calculated as where A is the projected contact area between indenter tip and the sample at load P [48]. Two approaches were used for quantification of the contact area A: i) the method proposed by Oliver and Pharr [46] which forms the basis of the indentation testing standard [48] and ii) the method which uses the actual tip-surface contact area, corrected for effects due to pile-ups and/or sink-ins [49]. In the standard method proposed by Oliver and Pharr [46] the contact area is determined from the tip area function (ℎ ) that expresses the indenter's cross-sectional area in terms of the contact depth, ℎ . The contact depth, ℎ = ℎ − ℎ , is the difference between the measured total penetration of the tip into the surface, ℎ , and the elastic surface displacement, ℎ . The elastic displacement is calculated using the Oliver-Pharr model [46]: ℎ = , where is the measured maximum load, S is the contact stiffness, which is computed as the slope of the unloading curve continuously evaluated in the CSM mode, and is the tip geometry coefficient (0.75 for a Berkovich tip) [50].
In the second method, the standard Oliver-Pharr algorithm for contact area calculation was modified in order to take into account the influence of the plastic deformation on the sample surface around indents during loading. The method is based on the approach proposed by Kese [49] and modified to include sink-ins effects [35]. The method requires experimental measurement of the indent profile through the central point of indentation and the centre of each edge. Where pile-up is observed, the contact between the indent and the pile-up is determined and is then projected onto the direction of the free surface normal, giving the pile-up height. The contact area of each pile-up is then approximated as a semi-ellipse, with the major axis equal to the length of the side of the projected triangular area of the indent print, and the minor axis being measured on the indent profile image as the projected distance of the pile-up contact perimeter [49]. Similarly, to account for sink-ins, the reduction in contact area was also approximated as a semi-ellipse, with a major axis calculated in the same way, whereas the minor axis was measured on the indent profile image as the projected distance between the sink-in contact perimeter [35]. For the investigated indents the corrected projected area, corr , was calculated as the contact area determined by the Oliver-Pharr analysis [46], O-P plus (or minus) the area due to out-of-plane deformation near each edge for pile-ups (or sink-ins). In this paper HO-P refers to hardness values obtained with the Oliver-Pharr algorithm, whereas HC refers to hardness values obtained after the contact area correction.
At least 200 indents were made for each sample with a target maximum indentation depth of 400 nm. Surface position correction was made for each indentation before data analysis. For consistency, load data averaged between 320 and 360 nm indentation depth were used for indentation hardness calculations, avoiding the depth range within which there is a strong indentation size effect [27,50]. Crystal Plasticity FEM calculations for beryllium [35] predict that the plastic zone beneath indents extends to a depth ~7 larger than the Berkovich tip penetration depth. The selected penetration depth for hardness calculations is almost 10 times smaller than the thickness of the implanted layer, allowing minimisation of any possible influence of the non-irradiated substrate material on the radiation-induced hardening data.
Because of the high hardness anisotropy of beryllium [35], in this paper indentation data are presented and discussed according to the angle between the indentation direction and the [0001] axis of the indented grain, denoted as . To investigate the crystallographic dependence of nanoindentation hardness, EBSD analyses of the indentation arrays were made, using a JEOL 840A scanning electron microscope equipped with the EDAX-TSL EBSD; OIM TSL software [51] was used for analysis of the EBSD data. Indents on or near to grain boundaries (no more than 2µm away) were not considered in analysis.
To investigate residual strains in samples after implantation and possible effects on nanoindentation data, a confocal Raman imaging microscope (WITec alpha300 AR) has been used to acquire Raman spectra and perform large area scans (3 mm × 3 mm). The analysis is based on determining the peak-shift corresponding to a 2 stretching band [52] in beryllium. The peak shift is caused by small changes to the short-range order in the atomic arrangement of the implanted material.
Therefore, it is indicative to the micro-strains created by implantation. During the experiments firstly, the structural and strains heterogeneity evidenced by a change in the spectral pattern have been determined, and then the main spectral components in homogeneous regions have been calculated.
Thus, by accumulating a large number of pixels (600 × 600 = 0.36 × 10 6 ), the position measurement error of the beryllium main peak is greatly reduced for the averaged spectral components. The TrueSurface feature enabled dynamic focusing of the confocal laser during scanning. Data acquisition was carried out with the conventional charge-coupled device (CCD) camera in the low-noise highintensity detection mode with a dwell time of 0.5 s per pixel using a green laser (λ = 531.95 nm with an estimated penetration depth of about 30 nm [52]), a 1800 g/mm grating and a Zeiss EC Epiplan-Neofluar 50× long-distance objective to accommodate for topography variation during scans. The laser power was set at 15 mW. The system was pre-calibrated to maximise the intensity of the silicon main peak on a single crystal silicon reference sample. Resulting spectra were treated with a cosmic-ray removal procedure and a shape-based background subtraction algorithm in WITec Project FIVE 5.2 Plus post-processing software. Where applicable, a Lorentzian function was used for peak fitting of the background-subtracted data.
Samples for Transmission Electron Microscopy (TEM) were prepared by Focussed Ion Beam sectioning (FEI Helios NanoLab 600i). The 30keV Ga + ions have been used for lamellae lift-out and shaping and 5keV Ga + ions have been used for final thinning of lamellae. According to SRIM calculation, the final energy should create approximately 10 nm deep surface damaged layers and as it will be shown later, this greatly hindered displacement damage effect from helium ions implantation.
Observation of the irradiated layer was performed with the JEOL JEM-2100 TEM in the Materials Department of the University of Oxford.

Microstructure of the implanted layer
The implanted layer was investigated by TEM. On bright-field images radiation damage appeared in the form of "black dots" (less than 10 nm diameter) arranged in lines parallel to the surface of samples, as shown in Figure 2 for the S-65 sample irradiated at 50°C. Figure 2(a) shows that the dark lines correspond to the peaks in "dpa" and helium implantation ( It should be also noted that in Figure 2, only 10 out of 11 implanted layers are present since the first layer was polished-off during colloidal silica polishing of the aluminium coating after the implantation, but, as will be discussed later, this was not the case for all the implanted samples.

Figure 2. TEM image (bright field) of the S-65 irradiated at 50°C. The dark lines in (a) correspond to the peaks in "dpa" and helium implantation. Radiation damage appeared in the form of "black dots" with higher number density inside dark lines (b).
Use of over-/under-focus imaging [53] did not reveal any objects which could be identified as cavities or helium bubbles in any of the investigated samples, including in grain boundary areas.
Helium is therefore likely to be either in solid solution in the beryllium matrix or in a form of subnanometric bubbles too small to be resolved in TEM [53]. The observation are in agreement with other experimental studies of beryllium after neutron irradiation [11,20,[54][55][56] or helium implantation [24] where, due to the low diffusive mobility of separate helium atoms or helium-based complexes, gas bubbles were not visible in TEM at temperatures below 400°C. It is also known that at very high helium contents (> 25000 appm) helium bubbles may be observed even at low temperatures in neutron irradiated [57] or helium implanted samples [58], but this level is well above the helium concentrations investigated in this work.

Topography of indentation prints and contact area correction
The elastic, plastic and strain-hardening properties of a material can affect whether material will either plastically sink-in or pile-up around the indenter [50,59,60]; such topographical changes can significantly change the contact area between the indenter tip and surface. This deformation is not accounted for in the standard method of contact area determination proposed by Oliver and Pharr [46], so with this method the calculated indentation hardness would be too high if pile-up occurs (as the real contact area is greater than that calculated from the indenter profile and penetration depth) and too low when sink-in occurs (when the real contact area will be less than the calculated one). To estimate the out-of-plane plastic deformation of beryllium and hence to improve the analysis of the load-   Figure 4 shows changes in contact area between the indenter and the sample due to pile-ups and sink-ins, as a function of indentation angle, as a contact area correction coefficient, , defined as: Here, Acorr is the measured projected contact area and AO-P is the "ideal" contact area calculated from the indenter penetration using the Oliver-Pharr method.
The effects of pile-ups and sink-ins in non-irradiated beryllium have been reported by us previously [35]. It was shown that the pile-up / sink-in behaviour depends highly on the crystallographic orientation of the indented grain. For indentations into grains with surface normal close to [0001] ( < 15° ) (hard orientations) sink-ins were predominant, as shown by the yellow "halo" around the indenter print in Figure 3(a). The corresponding correction coefficient C was measured to be in the range of approximately -0.1 to -0.2. Indentation into grains with higher produced both sink-ins and pile-ups, appearing as hills with height up to 250 nm on opposite sides of the indentation pit (see Figure 3(c-d)). The contribution of pile-ups in determining the real contact area increased with , and so the correction coefficient C gradually increases with indentation angle, reaching +0.25 for indents into grains with surface normal orientated perpendicular to [0001]. This variation of C with grain / indenter orientation for unirradiated Be is indicated by the shaded area in Figure 4. and decreased to -0.1 for close to 90°. In the second area ("Zone 2" in Figure 4), the correction coefficient distribution was very similar to that observed for the S-200-F grade irradiated at the same temperature, with a somewhat lower local maximum in C of about +0.1 for between 30° and 60°.
The selection of the area for the second array was based on study of post-irradiation microstructures and their dependence on sample preparation, and is discussed later in section 3.4.

Nanoindentation hardness
Hardness values for all of the investigated samples were calculated accounting for the contact area correction, using the measured pile-up and sink-in contribution coefficient from Figure 4, combining equations ( 1 ) and ( 2 ): where − was extracted from the nanoindentation test data using the standard Oliver-Pharr method [46]. As was shows in our previous work [35], for unirradiated materials, the averaged correction coefficient as a function of indentation angle (in degrees) between the c axis of the grain and the loading direction can be fitted to a linear function: where coefficients 1 =0.0035 and 0 = −0.123 for the chosen indentation depth range [35]. To account for the more complex variation of C with orientation in the irradiated materials, a fourth degree polynomial function was used: Best-fit values of the coefficients and the coefficients of determination R 2 are given in Table 1.      ( 4 ) and ( 3 );

) and coefficients of determination R 2 for averaged correction coefficient as a function of indentation angle between the c axis of the grain and the loading direction for S-65 and S-200-F grades in as-received and irradiated states
(b) Oliver-Phar method [46]. 0° corresponds to indentation into the basal plane parallel to the [0001] direction, 90° implies indentation perpendicular to [0001]. Only data from grains with the assigned crystallographic orientation are shown (no data from indents on or close to grain boundaries). Before irradiation the lower purity S-200-F grade had slightly higher hardness for the same crystallographic orientations than the high purity grade S-65. This was especially noticeable for the "soft" orientations (see Table 2). Since only limited data were available for indentations close to [0001] and given the observed scatter, the "hard" grain hardness may not be accurately compared. Aldinger [25] found that the critical resolved shear stresses (CRSS) for both basal and prismatic slip in beryllium is proportional to the impurity concentration as 2/3 , ( where c is the impurity concentration in wt%), so the less pure S-200-F would be expected to have increased in CRSS and thus higher hardness.
In both grades and for the both implantation temperatures the largest radiation hardening effect  Figure 5).
After the 50ºC irradiation, the average hardness of the S-200-F was higher than of the S-65, but the crystallographic analysis demonstrated that this was due to a larger number of grains with "hard" orientation being tested in the S-200-F. These effects demonstrate the importance of the use of crystallographic information in studies of the hardness of beryllium and other highly anisotropic materials. The "hard" orientation hardness values were less affected by radiation, therefore, implantation led to decrease of the anisotropy of beryllium hardness.
The high hardness anisotropy of beryllium originates from the very high anisotropy of the CRSS required for activation of different slip systems [35,39,40]. The CRSS is relatively low for the <a> basal and prismatic systems which both provide slip perpendicular to the c-axis, but it is very high for the secondary <c + a> pyramidal systems [37,38] [46]. Table 2. Nanoindentation, microstructural and chemical composition data from two beryllium grades before and after helium implantation The majority of the curves exhibit discrete displacement bursts, the so-called "pop-in" effect, at different stages of indentation, as highlighted by black arrows Figure 7. This pop-in behaviour has been widely observed in different materials [27,[61][62][63]; in the early stages of indentation (< 50 nm indentation depth) pop-ins are typically associated with plasticity bursts caused by dislocation source nucleation in previously defect-free material [27,[61][62][63]. In our experiments these early-stage pop-ins were observed in non-irradiated samples in both grades (Figure 7(b)), but were absent after helium implantation at both temperatures Figure 7(d, f).

Indentation pop-ins
This could be explained by the high number density of defects (in our case point defect cluster, He-vacancy complexes or He bubbles) which can potentially serve as nuclei for dislocation sources; similar effects have been found in indentation of ion-irradiated Fe-Cr [27] and W [64] and polycrystalline Fe after plastic deformation [63].

Figure 7. Typical load-indentation depth data produced from the as-received S-65 beryllium grade (a-b) and after helium implantation (c-f); black arrows indicate pop-in events. Curves in (b), (d) and (f) are initial stage fragments of curves in (a), (c) and (e) correspondently
Sometimes large pop-ins were observed at relatively high indentation depths (>100 nm); see Figure 7 (a, c, e,). This behaviour was strongly affected by the sample condition: while only about 7% of load displacement curves in as-received materials had any of these pop-ins, the behaviour was observed 3 to 6 times more often in helium-implanted samples. Such large pop-ins at relatively high indentation loads have also been observed in polycrystalline iron [63]; it was concluded that these events were closely related to the presence of a grain boundary in the stress zone of the indentation and has been rationalised in terms of preferred emission of dislocations from the grain boundaries. Such grain-boundary related pop-ins have also been observed in Au/Si samples [62] and Fe-C alloy [65]. Our investigations in beryllium consider indentations made in grain centres as observed at the surface by EBSD. It is however, possible that the indentations are interacting with grain boundaries under the surface, invisible to SEM/EBSD, which might possibly introduce some pop-ins observed in beryllium in this work.
However, in that case, the pop-in behaviour should be a statistical process dependent on presence of grain boundaries (i.e. grain size), and not be dependent on irradiation condition, since the grain size did not change during the implantations. Currently this behaviour is not understood. It is possible to propose that the substructure of the implanted layer with sequence of "softer" and "harder" bands can cause avalanche-like deformation when the stress zone reaches deeper "soft" layers.

3.4.Investigating the complex behaviour of S-65
It was observed that irradiation highly affected pile-up/sink-in behavior of beryllium: it become less crystallographically dependent with increased domination of pile-up (see Figure   3 and Figure 4).

Pt (b) (a)
This may be the origin of the observed differences in pile-up/sink-in behaviour after 200°C irradiation. It is known that pile-up / sink-in can be more pronounced for indentations in materials with a thin coating of different mechanical behaviour to the underlying substrate [50].
For indentation into hard coatings on a soft substrate promotes surface sinking-in, as observed by Tayebi et al during indentation of SiO2 films on Al substrate [66], whereas for indentation in a soft coating on a hard substrate promotes surface pile-ups, as, for example, was demonstrated in the aluminium-on-glass system by Tsui et al [67]. The behaviour will be more  Figure   8a, corresponds to the converse "hard coating on the soft substrate", promoting sink-in, as observed. However, it is difficult to make a definitive conclusion on this, since only a few liftouts were prepared and analysed with TEM from these samples.
To further investigate this assumption, Raman spectroscopy analysis was performed on both materials in the as-received state and after implantation at temperatures of 50ºC and 200ºC. A high-frequency acoustic mode is characteristic to beryllium, which results in a highly intense Raman peak (stretching band) [52,68]. Figure 9 shows the Raman spectra of Be samples in as-received state and after helium implantation. A single band was observed in both cases, which suggests the absence of a parasitic signal from impurities, e.g. oxygen. This band is have been observed in deuterium-implanted beryllium [52]. According to [69] downshifts of the stretching band can be attributed to a tensile stress in the material, that can be due to defects or impurities changing the bond lengths. Therefore, it is possible to assume that the surface tensile stresses exist in the irradiated samples studied here, but at a higher level in the samples implantation. Note, that since the nature of the pressure in our samples differs from the compression experiments reported in [68], further numeric modelling is required to quantify and confirm the level of stresses is samples due to the helium implantation . shallow angle to the initial surface plane, as schematically shown in Figure 10(b). The "Zone 1" indentation array in the S-65 sample implanted at 200ºC was located in the "brownish" band in Figure 10(a) with a high defect density, as seen by TEM in Figure 8(a). As discussed earlier, this corresponds to the "hard coating on soft substrate" scenario favouring the observed sink-ins. The "Zone 2" array was made in the "yellowish" area in Figure 10

4.1.Pile up/ sink-in behaviour
In bulk materials it is generally observed that a high yield stress to elastic modulus (σy/E) ratio and a high strain-hardening rate both promote a spread of deformation deeper into the specimen and the surface around the indenter sinks-in; while the converse (low (σy/E) values and low strain hardening) favour localisation of deformation around the indents and the creation of pile-ups [50,59]. The pile-up/sink-in behaviour in anisotropic hcp metals like Be [35] or Ti [70] is highly sensitive to the crystallographic orientation of the indented grains. The observed crystallographic dependence of pile-up/sink-ins in non-irradiated beryllium is in good agreement with crystal plasticity finite element modelling (CPFEM) results [35], which demonstrated that the stress field and geometrically necessary dislocations (GND) penetrate deeper into the material in the "hard" orientation relative to the "soft" orientation, consistent with a higher (σy/E) ratio, also the plastic strain under the "hard" indent is also higher consistent with high GND hardening. Irradiation can induce hardening or softening, changing the (σy/E) ratio, and can change a material's strain hardening rate. Hence, as observed in this study and in other studies on different materials, the pile-up behaviour can be changed by irradiation.
Similar to our observations, many studies report increased domination of pile-ups after irradiation. For example, Hardie et al [27] observed an increase in pile-ups heights after selfion irradiation of iron-chromium alloys. Similarly, a substantial increase in pile-up heights was observed in a helium-implanted single crystal tungsten by Das et al. [71] and in tungstenrhenium alloys by Beck et al [33]. However, in contrast, Armstrong et al. [72] observed a suppression of pile-ups in tungsten-tantalum alloys after self-ion implantation. It is clear that irradiation can change both the (σy/E) ratio and the strain hardening rate (through radiation induced hardening or softening), however the underlying mechanism of interaction of dislocations with radiation induced defects and implanted ions is not clear and is yet to be examined. Before a reliable model for prediction of pile-up behaviour is developed, measurements of contribution of this artefact are highly desirable for accurate measurements of irradiation induced hardening.

4.2.Hardening
Although no obvious differences were noticed by TEM investigation between samples irradiated at 50°C and 200°C, that is highly likely the consequence of the FIB damage, the nanoindentation results demonstrated much higher hardening for the lower temperature irradiation. This is in agreement with other hardness and yield strength data of beryllium irradiated at different temperatures (see [24,73] and review [74]). For example, Snead [73] performed Vickers hardness measurements of the S65-C grade neutron irradiated in high flux beam reactor (HFBR) up to 0.34 dpa and 250 appm He and observed about 480MPa (converted from HV) hardness increase after irradiation at 95°C, about 335MPa increase for Tirr=205°C and about 265MPa increase for Tirr=278°C. These data was in good agreement with the yield strength increase which was 75 MPa, 55 MPa and 40 MPa correspondently [73]. Kesternich and Ullmaier [24] investigated tensile properties of high purity beryllium implanted by helium (220 appm, 0.02dpa) and observed 175 MPa increase in the yield strength after implantation at 100°C, which was about 90MPa after 300°C implantation.
Both helium and lattice defects introduced during the implantation are likely to act as obstacles to dislocation motion. While helium solubility in beryllium has not been measured, this inert gas generally has an extremely low solubility in metals [75]. Therefore, He is likely to be either trapped by vacancies, increasing the CRSS for dislocation movement by solid solution strengthening; or precipitated into sub-nanometric helium bubbles (too small to be resolved by our TEM), where a precipitation hardening mechanism will govern dislocation pinning and CRSS increase. Kesternich et al [24] investigated the influence of helium on yield strength of beryllium and concluded that for low temperature irradiation (100°C), even if bubbles are not visible, the Orowan mechanism for dislocation pinning at precipitates (small bubbles) should be applied for description of radiation induced hardening. Kesternich et al [24] did not include hardening analysis from the observed dislocation loops.
Comparison with the radiation induced hardening data from  ( 8 ) and ( 7 ) yield 1 . This is one order of magnitude less than the anticipated contribution from helium; therefore, the observed hardening cannot be attributed to solid solution hardening by helium atoms.
The dispersed barrier hardening model (equation ( 6 ) Bronson [84] as this approach demonstrated good results in predicting helium bubbles behaviour in tungsten [81]. The equation has a form of: where is a volume of a spherical bubble, -number of helium atoms in a helium bubble, and 1 ( ), 2 ( ) and 3 ( ) are the functions dependent on the temperature, the details and parameters of which can be found in [84]. The volume of the bubble may be approximated as [81]: where = 8.09 × 10 3 −3 is the atomic volume of beryllium and -number of vacancies in a bubble.
Combination of equations ( 9 ), ( 10 ) and ( 11 ) ( 6 ) and ( 7 )). The results of the expected helium bubbles contribution into hardening ∆ are summarised by the ed area in Figure 11b for the case the obstacle strength factor α = 0.2 (weak obstacles after Lucas [77]). As between 1.4 and 2.4 GPa is expected from these objects, it is possible to note that with the chosen calculation parameters, this corresponds to relatively large bubbles with the radius of 0.8 -1.5nm. However, as demonstrated in [78], the strength factor of cavities is highly dependent on the size and for the case of stainless steel may decrees to the values of α = 0.02…0.1 if the cavities are smaller than 2 nm in diameter. The value of α = 0.1 was also deduced for the helium bubbles with diameters of 1-1.5 nm in irradiated ferritic-martensitic steels [85]. The results of the expected helium bubbles contribution into hardening ∆ for α = 0.1 are demonstrated by the grey area in Figure  11c, and the observed 1.4 and 2.4 GPa hardening will correspond to the helium bubbles with the radii in the range between about 0.4 and 1 nm.
Considering the obtained distribution of hardening for the bubbles of different sizes, it is possible to suggest that the differences between hardening at 50 and 200°C observed in our work could be explained by differences in spacing between the bubbles (i.e. number density of bubbles and their size), i.e. bubbles after 200°C implantation are larger than after the 50°C implantation, but still below the TEM resolution (1-2 nm in diameter). It is also should be added that the ratio between vacancies and helium atoms may alter the bubbles strengthening factor α. For example, as it was shown by Schäublin and Chiu [86] in the case of bcc iron, a 2 nm He bubble is a weak obstacle when the He content is low, at 1 to 2 He atoms per vacancy, but beyond 2 He atoms per vacancy, the resistance of the He bubble increases with increasing He content (and pressure) and at 5 He atoms per vacancy, the He bubble becomes a much stronger obstacle. While, no detailed investigation of helium bubbles strengthening factors in beryllium is known, the scenario that increase of helium to vacancy ratio may increase strength factor of bubbles is likely. Due to this, it is possible to suggest that bubbles in the samples irradiated at 50°C may have higher H/V ratio (and potentially, higher strengthening factor) that the bubbles in the samples irradiated at 200C. This suggestion also complies with our Raman spectroscopy measurements. We observed that the samples after helium implantation at 50°C have higher surface tensile strains than the samples after 200°C implantation. Similarly, surface tensile strains have been observed synchrotron X-ray micro-diffraction by Hofmann et al. [87] in helium-implanted tungsten (RT, 3000 appm, 0.2 dpa). Using the relaxation volumes of defects derived from DFT, authors [87] demonstrated that the tensile strains were caused by self-interstitial atom defects which did not recombine with vacancies due to the presence of helium and creation of He2V. The relaxation volumes of vacancies, interstitials and their complexes with helium in beryllium are unknown and require additional numerical modelling investigation, however following the results in [87], it is possible to assume that the SIA have positive relaxation volume (relaxed volume of the supercell containing SIA is larger than the volume of the perfect cell), vacancies have negative relaxation volume and He-V complexes may have either positive or negative relaxation volume and increase of H/V ratio should lead to increase the relaxation volume. In the case of our experiment, after helium implantation at both temperatures the downshift may indicate dominance of defects with the positive relaxation volume: SIA and their complexes, or, following the example of tungsten [87], complexes of vacancies with helium with high He to vacancy ratio (more than 3 in the case of tungsten [87]).
Therefore, it is possible to suggest that the difference in magnitude of the band shift may indicate a larger concentration of crystal defects with the positive relaxation volume (complexes of SIA) and/or higher He to vacancy ratio in bubbles. Both scenario agrees with the observed increased irradiation induced hardening of the lower temperature irradiated samples.
The increased radiation induced hardening of the lower-purity grade after exposure at 200°C may possible originate from an influence of impurities on redistribution of helium or point defects, however, creation of impurity clusters which act as obstacles for dislocations could be also considered. The chemistry effect on radiation resistance of beryllium was never systematically studied, and the available data is scattered. For example, similarly to our observation, Kupriyanov et al [55] demonstrated that after neutron irradiation in CM-3 reactor at 130-180ºC the lower purity TIP-30 and TE-30 beryllium grades experienced a higher radiation-induced increase of the yield strength than the higher purity TshG-56. However, the results of Snead [73] demonstrated lower irradiation induced hardening of the lower purity P0 grade than of the higher purity S-65C grade irradiated in the high flux beam reactor (BNL) at temperatures between 105 and 275ºC.
Currently, the high purity S-65C grade has been selected as the reference plasma-facing material in ITER on the basis of its excellent thermal fatigue and thermal shock behavior, and a good available database on its properties [7]. Our results support this choice, demonstrating that at a helium content that corresponds to the end-of-life of plasma-facing beryllium in ITER [16], the properties of this grade exhibit higher radiation stability, e.g. smaller radiationinduced hardening than the lower purity grade. However, the mechanism(s) responsible for this are not known. A detailed fine scale high resolution TEM analysis, complimented by fine scale chemical analysis, for example with Atom Probe Tomography, and preferably Positron Annihilation Spectroscopy to elucidate the helium atom distribution, would be required to understand the hardening mechanisms from individual features in the irradiated microstructure.
The results are also important for the particle accelerator community. The irradiation conditions and hardening effects used is this study represent the average operating temperatures of beryllium windows in the operating NuMI beamline (50ºC) [18] and the currently-designed LBNF (200ºC) [9,10]. So far, the high purity PF-60 grade (similar purity to S-65, but with different grain size and texture) has been successfully used as a beam window material in the NuMI beamline, accumulating during operation up to 2000 appm of helium [18]. The result of this work demonstrate that in the future LBNF, the similar pure grade exposed to comparable doses to NuMI, but at higher temperature (200ºC), should undergo smaller radiation-induced hardening and that accumulation of up to 3000 appm of helium at this temperature does not lead to creation of cavities or bubbles (including at grain boundaries) which are known be often responsible for the severe loss of ductility in beryllium.

Conclusion
Two industrial beryllium grades, higher-purity S-65 and lower-purity S-200-F, were investigated in the as-received state and after helium implantation at 50 and 200ºC. The main conclusions that can be drawn from this work can be summarized as: • At both temperatures in both grades, under TEM investigation the radiation damage appears as "black dots" (<10 nm in diameter) which are likely to be small dislocation loops with the number density of ~ 10 22 m -3 . No bubbles were observed inside grains and at grain boundaries.
• The lower-purity S-200-F grade has higher average hardness (3.7±0.8 GPa) than the S-65 grade (3.4±0.8 GPa) in the as-received state and remained harder after helium implantations. In both grades the implantation at 50ºC almost doubled the average hardness, and at 200ºC led to a 60% hardness increase, compared to the average hardness of the unimplanted material.
• By combining nanoindentation with EBSD it was possible to separate out the influence of crystal orientation on hardness. In the as-received materials the hardness is about 2.5 times higher when the indentation direction is close to the [0001] c-axis of beryllium, compared to indentation perpendicular to [0001].
Grades with "soft orientation" were most sensitive to radiation induced hardening, which increased almost 3 times after 50ºC implantation and 2 times after exposure at 200ºC. The higher purity grade S-65 had a 12% smaller increase of the "soft orientation" hardness at 200ºC then the less S-200-F. Hardness anisotropy significantly decreased after helium implantation as "hard orientation" hardness values were less affected by radiation .
• Near-indentation surface topography, which is not accounted for in the standard "Oliver-Phaar" analysis, can have a strong influence on the extraction of true hardness values from nanoindentation data. A combination of nanoindentation with EBSD, to give a contact area correction, is an important procedure in such cases, for example here in the investigation of radiation damage effects in different beryllium grades.
• Topography mapping of indentation prints demonstrated that localised surface deformation around indents (pile-up and sink-in) is highly crystallographically dependent in the as-received state (sink-in behaviour dominated in hard grains close to the c-axis while pile-ups dominated in soft grains orthogonal to the c axis).
After implantation this crystallographic dependence of topography is less prominent and either pile-ups or sink-ins may be observed. TEM and Raman microscopy indicate that whether either pile-ups or sink-ins occur in a particular specimen can be related to the detailed layered substructure of the implanted material: when the surface layer is softer (with smaller helium content) pile-ups are mainly observed, while in the opposite case (harder surface layer with higher helium content) sink-ins are dominant.
• Analysis of the possible hardening contribution demonstrated that the "back dots" at calculated number densities could be responsible for up to half of the observed hardening, while the rest should originate from helium bubbles with the size below the TEM resolution (<1…1.5 nm). It is expected that the bubbles are relatively weak obstacles with the strengthening factor < 0.2. Solid solution hardening by helium atoms is unlikely to significantly contribute to the observed hardness increase.

Acknowledgments
This work was supported by the Fermi Research Alliance, LLC under Contract No. DE-AC02-07CH11359 with the United States Department of Energy, and by the UK Science and